diff --git a/cub/cmake/CubBuildCompilerTargets.cmake b/cub/cmake/CubBuildCompilerTargets.cmake
index 2cef662dcaf..4d2b4547048 100644
--- a/cub/cmake/CubBuildCompilerTargets.cmake
+++ b/cub/cmake/CubBuildCompilerTargets.cmake
@@ -11,6 +11,9 @@ function(cub_build_compiler_targets)
   set(cxx_compile_options)
   set(cuda_compile_options)
 
+  # Ensure that we build our tests without treating ourself as system header
+  list(APPEND cxx_compile_definitions "_CCCL_NO_SYSTEM_HEADER")
+
   if ("MSVC" STREQUAL "${CMAKE_CXX_COMPILER_ID}")
     list(APPEND cxx_compile_definitions _ENABLE_EXTENDED_ALIGNED_STORAGE)
     list(APPEND cuda_compile_options "--use-local-env")
diff --git a/cub/cub/agent/agent_adjacent_difference.cuh b/cub/cub/agent/agent_adjacent_difference.cuh
index b135fbbf53b..f5b89daecfe 100644
--- a/cub/cub/agent/agent_adjacent_difference.cuh
+++ b/cub/cub/agent/agent_adjacent_difference.cuh
@@ -28,6 +28,13 @@
 #pragma once
 
 #include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include "../util_type.cuh"
 #include "../util_namespace.cuh"
 #include "../block/block_load.cuh"
@@ -159,7 +166,7 @@ struct AgentDifference
       }
       else
       {
-        InputT tile_prev_input = MayAlias 
+        InputT tile_prev_input = MayAlias
                                ? first_tile_previous[tile_idx]
                                : *(input_it + tile_base - 1);
 
diff --git a/cub/cub/agent/agent_batch_memcpy.cuh b/cub/cub/agent/agent_batch_memcpy.cuh
index 20db5bb6aec..78f89c61823 100644
--- a/cub/cub/agent/agent_batch_memcpy.cuh
+++ b/cub/cub/agent/agent_batch_memcpy.cuh
@@ -33,6 +33,14 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/agent/single_pass_scan_operators.cuh>
 #include <cub/block/block_exchange.cuh>
 #include <cub/block/block_load.cuh>
diff --git a/cub/cub/agent/agent_histogram.cuh b/cub/cub/agent/agent_histogram.cuh
index 94c2efcaf4e..c4226ba913e 100644
--- a/cub/cub/agent/agent_histogram.cuh
+++ b/cub/cub/agent/agent_histogram.cuh
@@ -33,11 +33,18 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <iterator>
 
 #include "../util_type.cuh"
 #include "../block/block_load.cuh"
-#include "../config.cuh"
 #include "../grid/grid_queue.cuh"
 #include "../iterator/cache_modified_input_iterator.cuh"
 
@@ -301,8 +308,8 @@ struct AgentHistogram
         for (int CHANNEL = 0; CHANNEL < NUM_ACTIVE_CHANNELS; ++CHANNEL)
         {
             int channel_bins = num_privatized_bins[CHANNEL];
-            for (int privatized_bin = threadIdx.x; 
-                    privatized_bin < channel_bins;  
+            for (int privatized_bin = threadIdx.x;
+                    privatized_bin < channel_bins;
                     privatized_bin += BLOCK_THREADS)
             {
                 int         output_bin  = -1;
@@ -631,7 +638,7 @@ struct AgentHistogram
                 // Consume a partially-full tile at the end of the row
                 OffsetT num_remaining = (num_row_pixels * NUM_CHANNELS) - col_offset;
                 ConsumeTile<IS_ALIGNED, false>(tile_offset, num_remaining);
-            } 
+            }
             else
             {
                 // Consume full tile
diff --git a/cub/cub/agent/agent_merge_sort.cuh b/cub/cub/agent/agent_merge_sort.cuh
index 2e994dd97e1..6a50f3347a4 100644
--- a/cub/cub/agent/agent_merge_sort.cuh
+++ b/cub/cub/agent/agent_merge_sort.cuh
@@ -28,6 +28,13 @@
 #pragma once
 
 #include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include "../util_type.cuh"
 #include "../util_namespace.cuh"
 #include "../block/block_load.cuh"
diff --git a/cub/cub/agent/agent_radix_sort_downsweep.cuh b/cub/cub/agent/agent_radix_sort_downsweep.cuh
index d7e77bb882c..a28bbdfa597 100644
--- a/cub/cub/agent/agent_radix_sort_downsweep.cuh
+++ b/cub/cub/agent/agent_radix_sort_downsweep.cuh
@@ -35,6 +35,14 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <stdint.h>
 #include <type_traits>
 
@@ -44,7 +52,6 @@
 #include <cub/block/block_radix_rank.cuh>
 #include <cub/block/block_exchange.cuh>
 #include <cub/block/radix_rank_sort_operations.cuh>
-#include <cub/config.cuh>
 #include <cub/util_type.cuh>
 #include <cub/iterator/cache_modified_input_iterator.cuh>
 
@@ -135,7 +142,7 @@ struct AgentRadixSortDownsweep
     using ValuesItr = CacheModifiedInputIterator<LOAD_MODIFIER, ValueT, OffsetT>;
 
     // Radix ranking type to use
-    using BlockRadixRankT = 
+    using BlockRadixRankT =
       cub::detail::block_radix_rank_t<
         RANK_ALGORITHM, BLOCK_THREADS, RADIX_BITS, IS_DESCENDING, SCAN_ALGORITHM>;
 
@@ -202,7 +209,7 @@ struct AgentRadixSortDownsweep
     // The global scatter base offset for each digit (valid in the first RADIX_DIGITS threads)
     OffsetT bin_offset[BINS_TRACKED_PER_THREAD];
 
-    std::uint32_t current_bit; 
+    std::uint32_t current_bit;
     std::uint32_t num_bits;
 
     // Whether to short-cirucit
@@ -488,15 +495,15 @@ struct AgentRadixSortDownsweep
         OffsetT           relative_bin_offsets[ITEMS_PER_THREAD];
 
         // Assign default (min/max) value to all keys
-        bit_ordered_type default_key = IS_DESCENDING 
-                                     ? traits::min_raw_binary_key(decomposer) 
+        bit_ordered_type default_key = IS_DESCENDING
+                                     ? traits::min_raw_binary_key(decomposer)
                                      : traits::max_raw_binary_key(decomposer);
 
         // Load tile of keys
         LoadKeys(
             keys,
             block_offset,
-            valid_items, 
+            valid_items,
             default_key,
             Int2Type<FULL_TILE>(),
             Int2Type<LOAD_WARP_STRIPED>());
diff --git a/cub/cub/agent/agent_radix_sort_histogram.cuh b/cub/cub/agent/agent_radix_sort_histogram.cuh
index 9e895b8d0d2..5d571011482 100644
--- a/cub/cub/agent/agent_radix_sort_histogram.cuh
+++ b/cub/cub/agent/agent_radix_sort_histogram.cuh
@@ -34,9 +34,16 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include "../block/block_load.cuh"
 #include "../block/radix_rank_sort_operations.cuh"
-#include "../config.cuh"
 #include "../thread/thread_reduce.cuh"
 #include "../util_math.cuh"
 #include "../util_type.cuh"
@@ -120,7 +127,7 @@ struct AgentRadixSortHistogram
     // thread fields
     // shared memory storage
     _TempStorage& s;
-  
+
     // bins for the histogram
     OffsetT* d_bins_out;
 
@@ -175,7 +182,7 @@ struct AgentRadixSortHistogram
     }
 
     __device__ __forceinline__
-    void LoadTileKeys(OffsetT tile_offset, bit_ordered_type (&keys)[ITEMS_PER_THREAD])    
+    void LoadTileKeys(OffsetT tile_offset, bit_ordered_type (&keys)[ITEMS_PER_THREAD])
     {
         // tile_offset < num_items always, hence the line below works
         bool full_tile = num_items - tile_offset >= TILE_ITEMS;
@@ -264,7 +271,7 @@ struct AgentRadixSortHistogram
                 AccumulateSharedHistograms(tile_offset, keys);
             }
             CTA_SYNC();
-            
+
             // Accumulate the result in global memory.
             AccumulateGlobalHistograms();
             CTA_SYNC();
diff --git a/cub/cub/agent/agent_radix_sort_onesweep.cuh b/cub/cub/agent/agent_radix_sort_onesweep.cuh
index 6c57d1f0215..190c6c29eb8 100644
--- a/cub/cub/agent/agent_radix_sort_onesweep.cuh
+++ b/cub/cub/agent/agent_radix_sort_onesweep.cuh
@@ -33,10 +33,17 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include "../block/block_radix_rank.cuh"
 #include "../block/radix_rank_sort_operations.cuh"
 #include "../block/block_store.cuh"
-#include "../config.cuh"
 #include "../util_ptx.cuh"
 #include "../util_type.cuh"
 
@@ -64,7 +71,7 @@ template <
     int NOMINAL_BLOCK_THREADS_4B,
     int NOMINAL_ITEMS_PER_THREAD_4B,
     typename ComputeT,
-    /** \brief Number of private histograms to use in the ranker; 
+    /** \brief Number of private histograms to use in the ranker;
         ignored if the ranking algorithm is not one of RADIX_RANK_MATCH_EARLY_COUNTS_* */
     int _RANK_NUM_PARTS,
     /** \brief Ranking algorithm used in the onesweep kernel. Only algorithms that
@@ -106,7 +113,7 @@ struct AgentRadixSortOnesweep
         RANK_NUM_PARTS = AgentRadixSortOnesweepPolicy::RANK_NUM_PARTS,
         TILE_ITEMS = BLOCK_THREADS * ITEMS_PER_THREAD,
         RADIX_BITS = AgentRadixSortOnesweepPolicy::RADIX_BITS,
-        RADIX_DIGITS = 1 << RADIX_BITS,        
+        RADIX_DIGITS = 1 << RADIX_BITS,
         BINS_PER_THREAD = (RADIX_DIGITS + BLOCK_THREADS - 1) / BLOCK_THREADS,
         FULL_BINS = BINS_PER_THREAD * BLOCK_THREADS == RADIX_DIGITS,
         WARP_THREADS = CUB_PTX_WARP_THREADS,
@@ -127,7 +134,7 @@ struct AgentRadixSortOnesweep
       typename traits::template digit_extractor_t<fundamental_digit_extractor_t, DecomposerT>;
 
     typedef PortionOffsetT AtomicOffsetT;
-  
+
     static constexpr RadixRankAlgorithm RANK_ALGORITHM =
                                     AgentRadixSortOnesweepPolicy::RANK_ALGORITHM;
     static constexpr BlockScanAlgorithm SCAN_ALGORITHM =
@@ -224,7 +231,7 @@ struct AgentRadixSortOnesweep
     __device__ __forceinline__ void LookbackPartial(int (&bins)[BINS_PER_THREAD])
     {
         #pragma unroll
-        for (int u = 0; u < BINS_PER_THREAD; ++u) 
+        for (int u = 0; u < BINS_PER_THREAD; ++u)
         {
             int bin = ThreadBin(u);
             if (FULL_BINS || bin < RADIX_DIGITS)
@@ -260,7 +267,7 @@ struct AgentRadixSortOnesweep
             agent.TryShortCircuit(keys, bins);
         }
     };
-  
+
     __device__ __forceinline__ void LookbackGlobal(int (&bins)[BINS_PER_THREAD])
     {
         #pragma unroll
@@ -452,7 +459,7 @@ struct AgentRadixSortOnesweep
             {
                 s.global_offsets[bin] = d_bins_in[bin] - offsets[u];
             }
-        }        
+        }
     }
 
     __device__ __forceinline__ void UpdateBinsGlobal(int (&bins)[BINS_PER_THREAD],
@@ -512,7 +519,7 @@ struct AgentRadixSortOnesweep
         constexpr int ITEMS_PER_WARP = TILE_ITEMS / BLOCK_WARPS;
         constexpr int ALIGN = 8;
         constexpr auto CACHE_MODIFIER = STORE_CG;
-        
+
         int warp_start = warp * ITEMS_PER_WARP;
         int warp_end = (warp + 1) * ITEMS_PER_WARP;
         int warp_offset = warp_start;
@@ -596,11 +603,11 @@ struct AgentRadixSortOnesweep
         // compute digits corresponding to the keys
         int digits[ITEMS_PER_THREAD];
         ComputeKeyDigits(digits);
-        
+
         // load values
         ValueT values[ITEMS_PER_THREAD];
         LoadValues(block_idx * TILE_ITEMS, values);
-        
+
         // scatter values
         CTA_SYNC();
         ScatterValuesShared(values, ranks);
@@ -608,7 +615,7 @@ struct AgentRadixSortOnesweep
         CTA_SYNC();
         ScatterValuesGlobal(digits);
     }
-        
+
 
     __device__ __forceinline__ void GatherScatterValues(
         int (&ranks)[ITEMS_PER_THREAD], Int2Type<true> keys_only) {}
@@ -628,7 +635,7 @@ struct AgentRadixSortOnesweep
         BlockRadixRankT(s.rank_temp_storage).RankKeys(
             keys, ranks, digit_extractor(), exclusive_digit_prefix,
             CountsCallback(*this, bins, keys));
-        
+
         // scatter keys in shared memory
         CTA_SYNC();
         ScatterKeysShared(keys, ranks);
@@ -637,7 +644,7 @@ struct AgentRadixSortOnesweep
         LoadBinsToOffsetsGlobal(exclusive_digit_prefix);
         LookbackGlobal(bins);
         UpdateBinsGlobal(bins, exclusive_digit_prefix);
-                
+
         // scatter keys in global memory
         CTA_SYNC();
         ScatterKeysGlobal();
diff --git a/cub/cub/agent/agent_radix_sort_upsweep.cuh b/cub/cub/agent/agent_radix_sort_upsweep.cuh
index 1a534454627..da048835da9 100644
--- a/cub/cub/agent/agent_radix_sort_upsweep.cuh
+++ b/cub/cub/agent/agent_radix_sort_upsweep.cuh
@@ -33,12 +33,19 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include "../thread/thread_reduce.cuh"
 #include "../thread/thread_load.cuh"
 #include "../warp/warp_reduce.cuh"
 #include "../block/block_load.cuh"
 #include "../block/radix_rank_sort_operations.cuh"
-#include "../config.cuh"
 #include "../util_type.cuh"
 #include "../iterator/cache_modified_input_iterator.cuh"
 
@@ -321,7 +328,7 @@ struct AgentRadixSortUpsweep
         const OffsetT &block_end)
     {
         // Process partial tile if necessary using single loads
-        for (OffsetT offset = threadIdx.x; offset < block_end - block_offset; offset += BLOCK_THREADS) 
+        for (OffsetT offset = threadIdx.x; offset < block_end - block_offset; offset += BLOCK_THREADS)
         {
             // Load and bucket key
             bit_ordered_type key = d_keys_in[block_offset + offset];
@@ -346,7 +353,7 @@ struct AgentRadixSortUpsweep
     :
         temp_storage(temp_storage.Alias()),
         d_keys_in(reinterpret_cast<const bit_ordered_type*>(d_keys_in)),
-        current_bit(current_bit), 
+        current_bit(current_bit),
         num_bits(num_bits),
         decomposer(decomposer)
     {}
diff --git a/cub/cub/agent/agent_reduce.cuh b/cub/cub/agent/agent_reduce.cuh
index 4a29d707b30..697f427eda5 100644
--- a/cub/cub/agent/agent_reduce.cuh
+++ b/cub/cub/agent/agent_reduce.cuh
@@ -13,9 +13,9 @@
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
  *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 
- * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 
- * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  * ARE DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE FOR ANY
  * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
@@ -27,17 +27,24 @@
  ******************************************************************************/
 
 /**
- * @file cub::AgentReduce implements a stateful abstraction of CUDA thread 
+ * @file cub::AgentReduce implements a stateful abstraction of CUDA thread
  *       blocks for participating in device-wide reduction.
  */
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <iterator>
 
 #include <cub/block/block_load.cuh>
 #include <cub/block/block_reduce.cuh>
-#include <cub/config.cuh>
 #include <cub/detail/type_traits.cuh>
 #include <cub/grid/grid_even_share.cuh>
 #include <cub/grid/grid_mapping.cuh>
@@ -368,7 +375,7 @@ struct AgentReduce
         .Reduce(thread_aggregate, reduction_op, valid_items);
     }
 
-    // Extracting this into a function saves 8% of generated kernel size by allowing to reuse 
+    // Extracting this into a function saves 8% of generated kernel size by allowing to reuse
     // the block reduction below. This also workaround hang in nvcc.
     ConsumeFullTileRange(thread_aggregate, even_share, can_vectorize);
 
@@ -439,7 +446,7 @@ private:
 
     even_share.block_offset += even_share.block_stride;
 
-    // Consume subsequent full tiles of input, at least one full tile was processed, so 
+    // Consume subsequent full tiles of input, at least one full tile was processed, so
     // `even_share.block_end >= TILE_ITEMS`
     while (even_share.block_offset <= even_share.block_end - TILE_ITEMS)
     {
diff --git a/cub/cub/agent/agent_reduce_by_key.cuh b/cub/cub/agent/agent_reduce_by_key.cuh
index f04fb73f053..79ee04dd85c 100644
--- a/cub/cub/agent/agent_reduce_by_key.cuh
+++ b/cub/cub/agent/agent_reduce_by_key.cuh
@@ -33,12 +33,19 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/agent/single_pass_scan_operators.cuh>
 #include <cub/block/block_discontinuity.cuh>
 #include <cub/block/block_load.cuh>
 #include <cub/block/block_scan.cuh>
 #include <cub/block/block_store.cuh>
-#include <cub/config.cuh>
 #include <cub/iterator/cache_modified_input_iterator.cuh>
 #include <cub/iterator/constant_input_iterator.cuh>
 
@@ -68,8 +75,8 @@ CUB_NAMESPACE_BEGIN
  * @tparam _SCAN_ALGORITHM
  *   The BlockScan algorithm to use
  *
- * @tparam DelayConstructorT 
- *   Implementation detail, do not specify directly, requirements on the 
+ * @tparam DelayConstructorT
+ *   Implementation detail, do not specify directly, requirements on the
  *   content of this type are subject to breaking change.
  */
 template <int _BLOCK_THREADS,
@@ -95,7 +102,7 @@ struct AgentReduceByKeyPolicy
   ///< The BlockScan algorithm to use
   static constexpr BlockScanAlgorithm SCAN_ALGORITHM = _SCAN_ALGORITHM;
 
-  struct detail 
+  struct detail
   {
     using delay_constructor_t = DelayConstructorT;
   };
diff --git a/cub/cub/agent/agent_rle.cuh b/cub/cub/agent/agent_rle.cuh
index ea09b1e02c7..15f27bd4c53 100644
--- a/cub/cub/agent/agent_rle.cuh
+++ b/cub/cub/agent/agent_rle.cuh
@@ -33,6 +33,14 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <iterator>
 
 #include "single_pass_scan_operators.cuh"
@@ -41,7 +49,6 @@
 #include "../block/block_scan.cuh"
 #include "../block/block_exchange.cuh"
 #include "../block/block_discontinuity.cuh"
-#include "../config.cuh"
 #include "../grid/grid_queue.cuh"
 #include "../iterator/cache_modified_input_iterator.cuh"
 #include "../iterator/constant_input_iterator.cuh"
@@ -56,28 +63,28 @@ CUB_NAMESPACE_BEGIN
 /**
  * Parameterizable tuning policy type for AgentRle
  *
- * @tparam _BLOCK_THREADS 
+ * @tparam _BLOCK_THREADS
  *   Threads per thread block
  *
- * @tparam _ITEMS_PER_THREAD 
+ * @tparam _ITEMS_PER_THREAD
  *   Items per thread (per tile of input)
  *
- * @tparam _LOAD_ALGORITHM 
+ * @tparam _LOAD_ALGORITHM
  *   The BlockLoad algorithm to use
  *
- * @tparam _LOAD_MODIFIER 
+ * @tparam _LOAD_MODIFIER
  *   Cache load modifier for reading input elements
  *
- * @tparam _STORE_WARP_TIME_SLICING 
- *   Whether or not only one warp's worth of shared memory should be allocated and time-sliced among 
- *   block-warps during any store-related data transpositions 
+ * @tparam _STORE_WARP_TIME_SLICING
+ *   Whether or not only one warp's worth of shared memory should be allocated and time-sliced among
+ *   block-warps during any store-related data transpositions
  *   (versus each warp having its own storage)
  *
- * @tparam _SCAN_ALGORITHM 
+ * @tparam _SCAN_ALGORITHM
  *   The BlockScan algorithm to use
  *
- * @tparam DelayConstructorT 
- *   Implementation detail, do not specify directly, requirements on the 
+ * @tparam DelayConstructorT
+ *   Implementation detail, do not specify directly, requirements on the
  *   content of this type are subject to breaking change.
  */
 template <int _BLOCK_THREADS,
@@ -100,7 +107,7 @@ struct AgentRlePolicy
     static constexpr CacheLoadModifier      LOAD_MODIFIER           = _LOAD_MODIFIER;       ///< Cache load modifier for reading input elements
     static constexpr BlockScanAlgorithm     SCAN_ALGORITHM          = _SCAN_ALGORITHM;      ///< The BlockScan algorithm to use
 
-    struct detail 
+    struct detail
     {
         using delay_constructor_t = DelayConstructorT;
     };
@@ -115,7 +122,7 @@ struct AgentRlePolicy
  ******************************************************************************/
 
 /**
- * \brief AgentRle implements a stateful abstraction of CUDA thread blocks for participating in device-wide run-length-encode 
+ * \brief AgentRle implements a stateful abstraction of CUDA thread blocks for participating in device-wide run-length-encode
  */
 template <
     typename    AgentRlePolicyT,        ///< Parameterized AgentRlePolicyT tuning policy type
@@ -413,12 +420,12 @@ struct AgentRle
 
         // `thread_exclusive_in_warp.key`:
         //      number of non-trivial runs starts in previous threads
-        // `thread_exclusive_in_warp.val`: 
+        // `thread_exclusive_in_warp.val`:
         //      number of items in the last non-trivial run in previous threads
 
         // `thread_aggregate.key`:
         //      number of non-trivial runs starts in this thread
-        // `thread_aggregate.val`: 
+        // `thread_aggregate.val`:
         //      number of items in the last non-trivial run in this thread
         LengthOffsetPair thread_aggregate = internal::ThreadReduce(lengths_and_num_runs, scan_op);
         WarpScanPairs(temp_storage.aliasable.scan_storage.warp_scan[warp_id]).Scan(
@@ -430,7 +437,7 @@ struct AgentRle
 
         // `thread_inclusive.key`:
         //      number of non-trivial runs starts in this and previous warp threads
-        // `thread_inclusive.val`: 
+        // `thread_inclusive.val`:
         //      number of items in the last non-trivial run in this or previous warp threads
 
         // Last lane in each warp shares its warp-aggregate
@@ -733,7 +740,7 @@ struct AgentRle
             if (thread_exclusive_in_warp.key == 0)
             {
                 // If there are no non-trivial runs starts in the previous warp threads, then
-                // `thread_exclusive_in_warp.val` denotes the number of items in the last 
+                // `thread_exclusive_in_warp.val` denotes the number of items in the last
                 // non-trivial run of the previous CTA threads, so the better name for it is
                 // `thread_exclusive_in_tile`.
                 thread_exclusive_in_warp.value += warp_exclusive_in_tile.value;
@@ -830,7 +837,7 @@ struct AgentRle
             if (thread_exclusive_in_warp.key == 0)
             {
                 // If there are no non-trivial runs starts in the previous warp threads, then
-                // `thread_exclusive_in_warp.val` denotes the number of items in the last 
+                // `thread_exclusive_in_warp.val` denotes the number of items in the last
                 // non-trivial run of the previous grid threads, so the better name for it is
                 // `thread_exclusive_in_grid`.
                 thread_exclusive_in_warp.value += thread_exclusive.value;
diff --git a/cub/cub/agent/agent_scan.cuh b/cub/cub/agent/agent_scan.cuh
index c076fa42106..86ffc781f65 100644
--- a/cub/cub/agent/agent_scan.cuh
+++ b/cub/cub/agent/agent_scan.cuh
@@ -13,9 +13,9 @@
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
  *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 
- * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 
- * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  * ARE DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE FOR ANY
  * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
@@ -27,19 +27,26 @@
  ******************************************************************************/
 
 /**
- * @file cub::AgentScan implements a stateful abstraction of CUDA thread blocks 
+ * @file cub::AgentScan implements a stateful abstraction of CUDA thread blocks
  *       for participating in device-wide prefix scan .
  */
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <iterator>
 
 #include <cub/agent/single_pass_scan_operators.cuh>
 #include <cub/block/block_load.cuh>
 #include <cub/block/block_scan.cuh>
 #include <cub/block/block_store.cuh>
-#include <cub/config.cuh>
 #include <cub/grid/grid_queue.cuh>
 #include <cub/iterator/cache_modified_input_iterator.cuh>
 
@@ -73,8 +80,8 @@ CUB_NAMESPACE_BEGIN
  * @tparam _SCAN_ALGORITHM
  *   The BlockScan algorithm to use
  *
- * @tparam DelayConstructorT 
- *   Implementation detail, do not specify directly, requirements on the 
+ * @tparam DelayConstructorT
+ *   Implementation detail, do not specify directly, requirements on the
  *   content of this type are subject to breaking change.
  */
 template <int NOMINAL_BLOCK_THREADS_4B,
@@ -95,7 +102,7 @@ struct AgentScanPolicy : ScalingType
   static constexpr BlockStoreAlgorithm STORE_ALGORITHM = _STORE_ALGORITHM;
   static constexpr BlockScanAlgorithm SCAN_ALGORITHM   = _SCAN_ALGORITHM;
 
-  struct detail 
+  struct detail
   {
     using delay_constructor_t = DelayConstructorT;
   };
diff --git a/cub/cub/agent/agent_scan_by_key.cuh b/cub/cub/agent/agent_scan_by_key.cuh
index dffe38f8648..9155beb5faa 100644
--- a/cub/cub/agent/agent_scan_by_key.cuh
+++ b/cub/cub/agent/agent_scan_by_key.cuh
@@ -12,9 +12,9 @@
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
  *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 
- * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 
- * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  * ARE DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE FOR ANY
  * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
@@ -26,18 +26,25 @@
  ******************************************************************************/
 
 /**
- * @file AgentScanByKey implements a stateful abstraction of CUDA thread blocks 
+ * @file AgentScanByKey implements a stateful abstraction of CUDA thread blocks
  *       for participating in device-wide prefix scan by key.
  */
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/agent/single_pass_scan_operators.cuh>
 #include <cub/block/block_discontinuity.cuh>
 #include <cub/block/block_load.cuh>
 #include <cub/block/block_scan.cuh>
 #include <cub/block/block_store.cuh>
-#include <cub/config.cuh>
 #include <cub/iterator/cache_modified_input_iterator.cuh>
 #include <cub/util_type.cuh>
 
@@ -52,8 +59,8 @@ CUB_NAMESPACE_BEGIN
 /**
  * Parameterizable tuning policy type for AgentScanByKey
  *
- * @tparam DelayConstructorT 
- *   Implementation detail, do not specify directly, requirements on the 
+ * @tparam DelayConstructorT
+ *   Implementation detail, do not specify directly, requirements on the
  *   content of this type are subject to breaking change.
  */
 template <int _BLOCK_THREADS,
@@ -73,7 +80,7 @@ struct AgentScanByKeyPolicy
   static constexpr BlockScanAlgorithm SCAN_ALGORITHM   = _SCAN_ALGORITHM;
   static constexpr BlockStoreAlgorithm STORE_ALGORITHM = _STORE_ALGORITHM;
 
-  struct detail 
+  struct detail
   {
     using delay_constructor_t = DelayConstructorT;
   };
diff --git a/cub/cub/agent/agent_segment_fixup.cuh b/cub/cub/agent/agent_segment_fixup.cuh
index a0802fd6032..84d8af5aa6a 100644
--- a/cub/cub/agent/agent_segment_fixup.cuh
+++ b/cub/cub/agent/agent_segment_fixup.cuh
@@ -33,6 +33,14 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <iterator>
 
 #include "single_pass_scan_operators.cuh"
@@ -40,7 +48,6 @@
 #include "../block/block_store.cuh"
 #include "../block/block_scan.cuh"
 #include "../block/block_discontinuity.cuh"
-#include "../config.cuh"
 #include "../iterator/cache_modified_input_iterator.cuh"
 #include "../iterator/constant_input_iterator.cuh"
 
@@ -240,7 +247,7 @@ struct AgentSegmentFixup
         else
             BlockLoadPairs(temp_storage.load_pairs).Load(d_pairs_in + tile_offset, pairs);
 
-        // RLE 
+        // RLE
         #pragma unroll
         for (int ITEM = 1; ITEM < ITEMS_PER_THREAD; ++ITEM)
         {
diff --git a/cub/cub/agent/agent_segmented_radix_sort.cuh b/cub/cub/agent/agent_segmented_radix_sort.cuh
index a629771120e..b869ec60908 100644
--- a/cub/cub/agent/agent_segmented_radix_sort.cuh
+++ b/cub/cub/agent/agent_segmented_radix_sort.cuh
@@ -27,10 +27,17 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/agent/agent_radix_sort_downsweep.cuh>
 #include <cub/agent/agent_radix_sort_upsweep.cuh>
 #include <cub/block/block_radix_sort.cuh>
-#include <cub/config.cuh>
 #include <cub/util_namespace.cuh>
 #include <cub/util_type.cuh>
 
@@ -152,7 +159,7 @@ struct AgentSegmentedRadixSort
     // Lowest() -> -1.79769e+308 = 00...00b -> TwiddleIn -> -0 = 10...00b
     // LOWEST   -> -nan          = 11...11b -> TwiddleIn ->  0 = 00...00b
 
-    bit_ordered_type default_key_bits = IS_DESCENDING 
+    bit_ordered_type default_key_bits = IS_DESCENDING
                                       ? traits::min_raw_binary_key(decomposer)
                                       : traits::max_raw_binary_key(decomposer);
     KeyT oob_default = reinterpret_cast<KeyT &>(default_key_bits);
diff --git a/cub/cub/agent/agent_select_if.cuh b/cub/cub/agent/agent_select_if.cuh
index fcccffe7f6a..564316b30dd 100644
--- a/cub/cub/agent/agent_select_if.cuh
+++ b/cub/cub/agent/agent_select_if.cuh
@@ -33,6 +33,14 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <iterator>
 
 #include "single_pass_scan_operators.cuh"
@@ -41,7 +49,6 @@
 #include "../block/block_scan.cuh"
 #include "../block/block_exchange.cuh"
 #include "../block/block_discontinuity.cuh"
-#include "../config.cuh"
 #include "../grid/grid_queue.cuh"
 #include "../iterator/cache_modified_input_iterator.cuh"
 
@@ -55,23 +62,23 @@ CUB_NAMESPACE_BEGIN
 /**
  * Parameterizable tuning policy type for AgentSelectIf
  *
- * @tparam _BLOCK_THREADS 
+ * @tparam _BLOCK_THREADS
  *   Threads per thread block
  *
- * @tparam _ITEMS_PER_THREAD 
+ * @tparam _ITEMS_PER_THREAD
  *   Items per thread (per tile of input)
  *
- * @tparam _LOAD_ALGORITHM 
+ * @tparam _LOAD_ALGORITHM
  *   The BlockLoad algorithm to use
  *
- * @tparam _LOAD_MODIFIER 
+ * @tparam _LOAD_MODIFIER
  *   Cache load modifier for reading input elements
  *
- * @tparam _SCAN_ALGORITHM 
+ * @tparam _SCAN_ALGORITHM
  *   The BlockScan algorithm to use
  *
- * @tparam DelayConstructorT 
- *   Implementation detail, do not specify directly, requirements on the 
+ * @tparam DelayConstructorT
+ *   Implementation detail, do not specify directly, requirements on the
  *   content of this type are subject to breaking change.
  */
 template <int _BLOCK_THREADS,
@@ -92,7 +99,7 @@ struct AgentSelectIfPolicy
     static constexpr CacheLoadModifier      LOAD_MODIFIER           = _LOAD_MODIFIER;       ///< Cache load modifier for reading input elements
     static constexpr BlockScanAlgorithm     SCAN_ALGORITHM          = _SCAN_ALGORITHM;      ///< The BlockScan algorithm to use
 
-    struct detail 
+    struct detail
     {
         using delay_constructor_t = DelayConstructorT;
     };
diff --git a/cub/cub/agent/agent_spmv_orig.cuh b/cub/cub/agent/agent_spmv_orig.cuh
index 4409436f9d0..4d9ce6ce582 100644
--- a/cub/cub/agent/agent_spmv_orig.cuh
+++ b/cub/cub/agent/agent_spmv_orig.cuh
@@ -33,13 +33,20 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <iterator>
 
 #include "../util_type.cuh"
 #include "../block/block_reduce.cuh"
 #include "../block/block_scan.cuh"
 #include "../block/block_exchange.cuh"
-#include "../config.cuh"
 #include "../thread/thread_search.cuh"
 #include "../thread/thread_operators.cuh"
 #include "../iterator/cache_modified_input_iterator.cuh"
diff --git a/cub/cub/agent/agent_sub_warp_merge_sort.cuh b/cub/cub/agent/agent_sub_warp_merge_sort.cuh
index 21ca2e5030c..e53ec5e6f7e 100644
--- a/cub/cub/agent/agent_sub_warp_merge_sort.cuh
+++ b/cub/cub/agent/agent_sub_warp_merge_sort.cuh
@@ -27,8 +27,15 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/block/radix_rank_sort_operations.cuh>
-#include <cub/config.cuh>
 #include <cub/util_type.cuh>
 #include <cub/warp/warp_load.cuh>
 #include <cub/warp/warp_merge_sort.cuh>
@@ -159,21 +166,21 @@ class AgentSubWarpSort
     return lhs == rhs;
   }
 
-  __device__ static bool get_oob_default(Int2Type<true> /* is bool */) 
+  __device__ static bool get_oob_default(Int2Type<true> /* is bool */)
   {
     // Traits<KeyT>::MAX_KEY for `bool` is 0xFF which is different from `true` and makes
     // comparison with oob unreliable.
     return !IS_DESCENDING;
   }
 
-  __device__ static KeyT get_oob_default(Int2Type<false> /* is bool */) 
+  __device__ static KeyT get_oob_default(Int2Type<false> /* is bool */)
   {
     // For FP64 the difference is:
     // Lowest() -> -1.79769e+308 = 00...00b -> TwiddleIn -> -0 = 10...00b
     // LOWEST   -> -nan          = 11...11b -> TwiddleIn ->  0 = 00...00b
 
     // Segmented sort doesn't support custom types at the moment.
-    bit_ordered_type default_key_bits = IS_DESCENDING 
+    bit_ordered_type default_key_bits = IS_DESCENDING
                                       ? traits::min_raw_binary_key(detail::identity_decomposer_t{})
                                       : traits::max_raw_binary_key(detail::identity_decomposer_t{});
     return reinterpret_cast<KeyT &>(default_key_bits);
@@ -253,7 +260,7 @@ public:
       KeyT keys[PolicyT::ITEMS_PER_THREAD];
       ValueT values[PolicyT::ITEMS_PER_THREAD];
 
-      KeyT oob_default = 
+      KeyT oob_default =
         AgentSubWarpSort::get_oob_default(Int2Type<std::is_same<bool, KeyT>::value>{});
 
       WarpLoadKeysT(storage.load_keys)
diff --git a/cub/cub/agent/agent_three_way_partition.cuh b/cub/cub/agent/agent_three_way_partition.cuh
index 93cc8d3966e..17ec5aa0fe2 100644
--- a/cub/cub/agent/agent_three_way_partition.cuh
+++ b/cub/cub/agent/agent_three_way_partition.cuh
@@ -27,6 +27,14 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <iterator>
 #include <type_traits>
 
@@ -36,7 +44,6 @@
 #include <cub/block/block_load.cuh>
 #include <cub/block/block_scan.cuh>
 #include <cub/block/block_store.cuh>
-#include <cub/config.cuh>
 #include <cub/iterator/cache_modified_input_iterator.cuh>
 
 
diff --git a/cub/cub/agent/agent_unique_by_key.cuh b/cub/cub/agent/agent_unique_by_key.cuh
index bf943a00b09..a7b9e0a367b 100644
--- a/cub/cub/agent/agent_unique_by_key.cuh
+++ b/cub/cub/agent/agent_unique_by_key.cuh
@@ -32,6 +32,14 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <iterator>
 #include <type_traits>
 
@@ -51,8 +59,8 @@ CUB_NAMESPACE_BEGIN
 /**
  * Parameterizable tuning policy type for AgentUniqueByKey
  *
- * @tparam DelayConstructorT 
- *   Implementation detail, do not specify directly, requirements on the 
+ * @tparam DelayConstructorT
+ *   Implementation detail, do not specify directly, requirements on the
  *   content of this type are subject to breaking change.
  */
 template <int                     _BLOCK_THREADS,
@@ -72,7 +80,7 @@ struct AgentUniqueByKeyPolicy
     static constexpr cub::CacheLoadModifier  LOAD_MODIFIER  = _LOAD_MODIFIER;
     static constexpr cub::BlockScanAlgorithm SCAN_ALGORITHM = _SCAN_ALGORITHM;
 
-    struct detail 
+    struct detail
     {
         using delay_constructor_t = DelayConstructorT;
     };
@@ -211,7 +219,7 @@ struct AgentUniqueByKey
         ValueOutputIteratorT         d_values_out_,
         EqualityOpT                  equality_op_,
         OffsetT                      num_items_)
-    : 
+    :
         temp_storage(temp_storage_.Alias()),
         d_keys_in(d_keys_in_),
         d_values_in(d_values_in_),
@@ -567,10 +575,10 @@ struct AgentUniqueByKey
         {
             int  num_remaining  = static_cast<int>(num_items - tile_offset);
             OffsetT num_selections = ConsumeTile<true>(num_remaining,
-                                                       tile_idx,                                    
+                                                       tile_idx,
                                                        tile_offset,
                                                        tile_state);
-            if (threadIdx.x == 0)                                                               
+            if (threadIdx.x == 0)
             {
                 *d_num_selected_out = num_selections;
             }
diff --git a/cub/cub/agent/single_pass_scan_operators.cuh b/cub/cub/agent/single_pass_scan_operators.cuh
index 342e859246c..ff8a4e18347 100644
--- a/cub/cub/agent/single_pass_scan_operators.cuh
+++ b/cub/cub/agent/single_pass_scan_operators.cuh
@@ -33,9 +33,16 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <iterator>
 
-#include <cub/config.cuh>
 #include <cub/detail/strong_load.cuh>
 #include <cub/detail/strong_store.cuh>
 #include <cub/detail/uninitialized_copy.cuh>
@@ -110,22 +117,22 @@ enum ScanTileStatus
     SCAN_TILE_INCLUSIVE,    // Inclusive tile prefix is available
 };
 
-namespace detail 
+namespace detail
 {
 
 template <int Delay, unsigned int GridThreshold = 500>
 __device__ __forceinline__ void delay()
 {
   NV_IF_TARGET(NV_PROVIDES_SM_70,
-               (if (Delay > 0) 
+               (if (Delay > 0)
                 {
-                  if (gridDim.x < GridThreshold) 
+                  if (gridDim.x < GridThreshold)
                   {
                     __threadfence_block();
                   }
-                  else 
+                  else
                   {
-                    __nanosleep(Delay); 
+                    __nanosleep(Delay);
                   }
                 }));
 }
@@ -134,15 +141,15 @@ template <unsigned int GridThreshold = 500>
 __device__ __forceinline__ void delay(int ns)
 {
   NV_IF_TARGET(NV_PROVIDES_SM_70,
-               (if (ns > 0) 
+               (if (ns > 0)
                 {
-                  if (gridDim.x < GridThreshold) 
+                  if (gridDim.x < GridThreshold)
                   {
                     __threadfence_block();
                   }
-                  else 
+                  else
                   {
-                    __nanosleep(ns); 
+                    __nanosleep(ns);
                   }
                 }));
 }
@@ -194,7 +201,7 @@ struct no_delay_constructor_t
 {
   struct delay_t
   {
-    __device__ __forceinline__ void operator()() 
+    __device__ __forceinline__ void operator()()
     {
       NV_IF_TARGET(NV_PROVIDES_SM_70,
                   (),
@@ -215,7 +222,7 @@ struct reduce_by_key_delay_constructor_t
 {
   struct delay_t
   {
-    __device__ __forceinline__ void operator()() 
+    __device__ __forceinline__ void operator()()
     {
       NV_DISPATCH_TARGET(
         NV_IS_EXACTLY_SM_80, (delay<Delay, GridThreshold>();),
@@ -262,7 +269,7 @@ struct exponential_backoff_constructor_t
     }
   };
 
-  __device__ __forceinline__ exponential_backoff_constructor_t(unsigned int /* seed */) 
+  __device__ __forceinline__ exponential_backoff_constructor_t(unsigned int /* seed */)
   {
     always_delay<L2WriteLatency>();
   }
@@ -437,7 +444,7 @@ struct exponential_backon_constructor_t
 
   unsigned int max_delay = InitialDelay;
 
-  __device__ __forceinline__ exponential_backon_constructor_t(unsigned int /* seed */) 
+  __device__ __forceinline__ exponential_backon_constructor_t(unsigned int /* seed */)
   {
     always_delay<L2WriteLatency>();
   }
@@ -613,7 +620,7 @@ struct ScanTileState<T, true>
     /**
      * Wait for the corresponding tile to become non-invalid
      */
-    template <class DelayT = detail::default_delay_t<T>> 
+    template <class DelayT = detail::default_delay_t<T>>
     __device__ __forceinline__ void WaitForValid(
         int             tile_idx,
         StatusWord      &status,
@@ -628,7 +635,7 @@ struct ScanTileState<T, true>
         }
 
         while (WARP_ANY((tile_descriptor.status == SCAN_TILE_INVALID), 0xffffffff))
-        {   
+        {
           delay_or_prevent_hoisting();
           TxnWord alias = detail::load_relaxed(d_tile_descriptors + TILE_STATUS_PADDING + tile_idx);
           tile_descriptor = reinterpret_cast<TileDescriptor&>(alias);
@@ -642,11 +649,11 @@ struct ScanTileState<T, true>
      * Loads and returns the tile's value. The returned value is undefined if either (a) the tile's status is invalid or
      * (b) there is no memory fence between reading a non-invalid status and the call to LoadValid.
      */
-     __device__ __forceinline__ T LoadValid(int tile_idx)                        
-    {                                                                           
+     __device__ __forceinline__ T LoadValid(int tile_idx)
+    {
         TxnWord alias = d_tile_descriptors[TILE_STATUS_PADDING + tile_idx];
         TileDescriptor tile_descriptor = reinterpret_cast<TileDescriptor&>(alias);
-        return tile_descriptor.value;                                           
+        return tile_descriptor.value;
     }
 };
 
@@ -704,7 +711,7 @@ struct ScanTileState<T, false>
             error = CubDebug(
               AliasTemporaries(d_temp_storage, temp_storage_bytes, allocations, allocation_sizes));
 
-            if (cudaSuccess != error) 
+            if (cudaSuccess != error)
             {
               break;
             }
@@ -784,7 +791,7 @@ struct ScanTileState<T, false>
     /**
      * Wait for the corresponding tile to become non-invalid
      */
-    template <class DelayT = detail::default_no_delay_t> 
+    template <class DelayT = detail::default_no_delay_t>
     __device__ __forceinline__ void WaitForValid(
         int             tile_idx,
         StatusWord      &status,
@@ -798,7 +805,7 @@ struct ScanTileState<T, false>
           __threadfence();
         } while (WARP_ANY((status == SCAN_TILE_INVALID), 0xffffffff));
 
-        if (status == StatusWord(SCAN_TILE_PARTIAL)) 
+        if (status == StatusWord(SCAN_TILE_PARTIAL))
         {
           value = ThreadLoad<LOAD_CG>(d_tile_partial + TILE_STATUS_PADDING + tile_idx);
         }
@@ -812,9 +819,9 @@ struct ScanTileState<T, false>
      * Loads and returns the tile's value. The returned value is undefined if either (a) the tile's status is invalid or
      * (b) there is no memory fence between reading a non-invalid status and the call to LoadValid.
      */
-    __device__ __forceinline__ T LoadValid(int tile_idx)                        
-    {                                                                           
-        return d_tile_inclusive[TILE_STATUS_PADDING + tile_idx];                                          
+    __device__ __forceinline__ T LoadValid(int tile_idx)
+    {
+        return d_tile_inclusive[TILE_STATUS_PADDING + tile_idx];
     }
 };
 
@@ -1008,7 +1015,7 @@ struct ReduceByKeyScanTileState<ValueT, KeyT, true>
     /**
      * Wait for the corresponding tile to become non-invalid
      */
-    template <class DelayT = detail::fixed_delay_constructor_t<350, 450>::delay_t> 
+    template <class DelayT = detail::fixed_delay_constructor_t<350, 450>::delay_t>
     __device__ __forceinline__ void WaitForValid(
         int                     tile_idx,
         StatusWord              &status,
@@ -1058,8 +1065,8 @@ struct ReduceByKeyScanTileState<ValueT, KeyT, true>
  * the current tile by using the call-back warp to wait on on
  * aggregates/prefixes from predecessor tiles to become available.
  *
- * @tparam DelayConstructorT 
- *   Implementation detail, do not specify directly, requirements on the 
+ * @tparam DelayConstructorT
+ *   Implementation detail, do not specify directly, requirements on the
  *   content of this type are subject to breaking change.
  */
 template <
@@ -1096,7 +1103,7 @@ struct TilePrefixCallbackOp
     T                           exclusive_prefix;   ///< Exclusive prefix for the tile
     T                           inclusive_prefix;   ///< Inclusive prefix for the tile
 
-    // Constructs prefix functor for a given tile index. 
+    // Constructs prefix functor for a given tile index.
     // Precondition: thread blocks processing all of the predecessor tiles were scheduled.
     __device__ __forceinline__ TilePrefixCallbackOp(ScanTileStateT &tile_status,
                                                     TempStorage &temp_storage,
@@ -1117,7 +1124,7 @@ struct TilePrefixCallbackOp
     {}
 
     // Block until all predecessors within the warp-wide window have non-invalid status
-    template <class DelayT = detail::default_delay_t<T>> 
+    template <class DelayT = detail::default_delay_t<T>>
     __device__ __forceinline__
     void ProcessWindow(
         int         predecessor_idx,        ///< Preceding tile index to inspect
diff --git a/cub/cub/block/block_adjacent_difference.cuh b/cub/cub/block/block_adjacent_difference.cuh
index b62c7fb23ec..b24b4bf5c5d 100644
--- a/cub/cub/block/block_adjacent_difference.cuh
+++ b/cub/cub/block/block_adjacent_difference.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2021, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -36,6 +36,13 @@
 #pragma once
 
 #include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include "../util_type.cuh"
 #include "../util_ptx.cuh"
 
@@ -489,7 +496,7 @@ public:
     }
 
     /**
-     * @brief Subtracts the left element of each adjacent pair of elements 
+     * @brief Subtracts the left element of each adjacent pair of elements
      *        partitioned across a CUDA thread block.
      *
      * @par
@@ -497,7 +504,7 @@ public:
      * - \smemreuse
      *
      * @par Snippet
-     * The code snippet below illustrates how to use @p BlockAdjacentDifference 
+     * The code snippet below illustrates how to use @p BlockAdjacentDifference
      * to compute the left difference between adjacent elements.
      *
      * @par
@@ -516,7 +523,7 @@ public:
      *
      * __global__ void ExampleKernel(...)
      * {
-     *   // Specialize BlockAdjacentDifference for a 1D block of 
+     *   // Specialize BlockAdjacentDifference for a 1D block of
      *   // 128 threads of type int
      *   using BlockAdjacentDifferenceT =
      *      cub::BlockAdjacentDifference<int, 128>;
@@ -607,7 +614,7 @@ public:
     }
 
     /**
-     * @brief Subtracts the left element of each adjacent pair of elements 
+     * @brief Subtracts the left element of each adjacent pair of elements
      *        partitioned across a CUDA thread block.
      *
      * @par
@@ -615,7 +622,7 @@ public:
      * - \smemreuse
      *
      * @par Snippet
-     * The code snippet below illustrates how to use @p BlockAdjacentDifference 
+     * The code snippet below illustrates how to use @p BlockAdjacentDifference
      * to compute the left difference between adjacent elements.
      *
      * @par
@@ -634,7 +641,7 @@ public:
      *
      * __global__ void ExampleKernel(...)
      * {
-     *   // Specialize BlockAdjacentDifference for a 1D block of 
+     *   // Specialize BlockAdjacentDifference for a 1D block of
      *   // 128 threads of type int
      *   using BlockAdjacentDifferenceT =
      *      cub::BlockAdjacentDifference<int, 128>;
@@ -725,9 +732,9 @@ public:
       {
         output[0] = input[0];
       }
-      else if (linear_tid == 0) 
+      else if (linear_tid == 0)
       {
-        output[0] = difference_op(input[0], 
+        output[0] = difference_op(input[0],
                                   tile_predecessor_item);
       }
       else
diff --git a/cub/cub/block/block_discontinuity.cuh b/cub/cub/block/block_discontinuity.cuh
index a3bf17f3190..b94e125e93c 100644
--- a/cub/cub/block/block_discontinuity.cuh
+++ b/cub/cub/block/block_discontinuity.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -34,6 +34,13 @@
 #pragma once
 
 #include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include "../util_type.cuh"
 #include "../util_ptx.cuh"
 
diff --git a/cub/cub/block/block_exchange.cuh b/cub/cub/block/block_exchange.cuh
index 5f4bc2e84a8..d6a7cfbbbe8 100644
--- a/cub/cub/block/block_exchange.cuh
+++ b/cub/cub/block/block_exchange.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -33,7 +33,14 @@
 
 #pragma once
 
-#include <cub/config.cuh>
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/detail/uninitialized_copy.cuh>
 #include <cub/util_ptx.cuh>
 #include <cub/util_type.cuh>
diff --git a/cub/cub/block/block_histogram.cuh b/cub/cub/block/block_histogram.cuh
index b9ab759607d..949bd2bcb34 100644
--- a/cub/cub/block/block_histogram.cuh
+++ b/cub/cub/block/block_histogram.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -33,9 +33,16 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include "specializations/block_histogram_sort.cuh"
 #include "specializations/block_histogram_atomic.cuh"
-#include "../config.cuh"
 #include "../util_ptx.cuh"
 
 CUB_NAMESPACE_BEGIN
diff --git a/cub/cub/block/block_load.cuh b/cub/cub/block/block_load.cuh
index b419ab6e726..5af8abca9e1 100644
--- a/cub/cub/block/block_load.cuh
+++ b/cub/cub/block/block_load.cuh
@@ -33,12 +33,19 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <iterator>
 #include <type_traits>
 
 #include "../block/block_exchange.cuh"
 #include "../iterator/cache_modified_input_iterator.cuh"
-#include "../config.cuh"
 #include "../util_ptx.cuh"
 #include "../util_type.cuh"
 
diff --git a/cub/cub/block/block_merge_sort.cuh b/cub/cub/block/block_merge_sort.cuh
index dc07ef6c294..7e414cdc4be 100644
--- a/cub/cub/block/block_merge_sort.cuh
+++ b/cub/cub/block/block_merge_sort.cuh
@@ -27,6 +27,14 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/thread/thread_sort.cuh>
 #include <cub/util_math.cuh>
 #include <cub/util_namespace.cuh>
diff --git a/cub/cub/block/block_radix_rank.cuh b/cub/cub/block/block_radix_rank.cuh
index 18655af4f0b..8495947789c 100644
--- a/cub/cub/block/block_radix_rank.cuh
+++ b/cub/cub/block/block_radix_rank.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -33,13 +33,20 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <stdint.h>
 
 #include "../thread/thread_reduce.cuh"
 #include "../thread/thread_scan.cuh"
 #include "../block/block_scan.cuh"
 #include "../block/radix_rank_sort_operations.cuh"
-#include "../config.cuh"
 #include "../util_ptx.cuh"
 #include "../util_type.cuh"
 
@@ -102,7 +109,7 @@ struct warp_in_block_matcher_t
 {
   static __device__ std::uint32_t match_any(std::uint32_t label, std::uint32_t warp_id)
   {
-    if (warp_id == static_cast<std::uint32_t>(PartialWarpId)) 
+    if (warp_id == static_cast<std::uint32_t>(PartialWarpId))
     {
       return MatchAny<Bits, PartialWarpThreads>(label);
     }
@@ -155,7 +162,7 @@ struct warp_in_block_matcher_t<Bits, 0, PartialWarpId>
  *   constexpr int block_threads = 2;
  *   constexpr int radix_bits = 5;
  *
- *   // Specialize BlockRadixRank for a 1D block of 2 threads 
+ *   // Specialize BlockRadixRank for a 1D block of 2 threads
  *   using block_radix_rank = cub::BlockRadixRank<block_threads, radix_bits>;
  *   using storage_t = typename block_radix_rank::TempStorage;
  *
@@ -172,7 +179,7 @@ struct warp_in_block_matcher_t<Bits, 0, PartialWarpId>
  *
  *   ...
  * \endcode
- * Suppose the set of input `keys` across the block of threads is `{ [16,10], [9,11] }`.  
+ * Suppose the set of input `keys` across the block of threads is `{ [16,10], [9,11] }`.
  * The corresponding output `ranks` in those threads will be `{ [3,1], [0,2] }`.
  *
  * \par Re-using dynamically allocating shared memory
@@ -758,8 +765,8 @@ public:
             // Mask of peers who have same digit as me
             uint32_t peer_mask =
               detail::warp_in_block_matcher_t<
-                RADIX_BITS, 
-                PARTIAL_WARP_THREADS, 
+                RADIX_BITS,
+                PARTIAL_WARP_THREADS,
                 WARPS - 1>::match_any(digit, warp_id);
 
             // Pointer to smem digit counter for this key
@@ -918,7 +925,7 @@ struct BlockRadixRankMatchEarlyCounts
     // types
     typedef cub::BlockScan<int, BLOCK_THREADS, INNER_SCAN_ALGORITHM> BlockScan;
 
-    
+
 
     // temporary storage
     struct TempStorage
@@ -981,7 +988,7 @@ struct BlockRadixRankMatchEarlyCounts
                 for (int bin = lane; bin < RADIX_DIGITS; bin += WARP_THREADS)
                 {
                     match_masks[bin] = 0;
-                }                    
+                }
             }
             WARP_SYNC(WARP_MASK);
 
@@ -992,7 +999,7 @@ struct BlockRadixRankMatchEarlyCounts
             {
                 atomicAdd(&warp_histograms[Digit(keys[u])][part], 1);
             }
-            
+
             // sum different parts;
             // no extra work is necessary if NUM_PARTS == 1
             if (NUM_PARTS > 1)
@@ -1025,7 +1032,7 @@ struct BlockRadixRankMatchEarlyCounts
         {
             // sum up warp-private histograms
             #pragma unroll
-            for (int u = 0; u < BINS_PER_THREAD; ++u) 
+            for (int u = 0; u < BINS_PER_THREAD; ++u)
             {
                 bins[u] = 0;
                 int bin = ThreadBin(u);
@@ -1127,12 +1134,12 @@ struct BlockRadixRankMatchEarlyCounts
             int (&exclusive_digit_prefix)[BINS_PER_THREAD])
         {
             ComputeHistogramsWarp(keys);
-            
+
             CTA_SYNC();
             int bins[BINS_PER_THREAD];
             ComputeOffsetsWarpUpsweep(bins);
             callback(bins);
-            
+
             BlockScan(s.prefix_tmp).ExclusiveSum(bins, exclusive_digit_prefix);
 
             ComputeOffsetsWarpDownsweep(exclusive_digit_prefix);
@@ -1164,7 +1171,7 @@ struct BlockRadixRankMatchEarlyCounts
     {
         BlockRadixRankMatchInternal<UnsignedBits, KEYS_PER_THREAD, DigitExtractorT, CountsCallback>
             internal(temp_storage, digit_extractor, callback);
-        internal.RankKeys(keys, ranks, exclusive_digit_prefix);        
+        internal.RankKeys(keys, ranks, exclusive_digit_prefix);
     }
 
     template <typename UnsignedBits, int KEYS_PER_THREAD, typename DigitExtractorT>
@@ -1193,13 +1200,13 @@ struct BlockRadixRankMatchEarlyCounts
 
 
 #ifndef DOXYGEN_SHOULD_SKIP_THIS    // Do not document
-namespace detail 
+namespace detail
 {
 
-// `BlockRadixRank` doesn't conform to the typical pattern, not exposing the algorithm 
-// template parameter. Other algorithms don't provide the same template parameters, not allowing 
-// multi-dimensional thread block specializations. 
-// 
+// `BlockRadixRank` doesn't conform to the typical pattern, not exposing the algorithm
+// template parameter. Other algorithms don't provide the same template parameters, not allowing
+// multi-dimensional thread block specializations.
+//
 // TODO(senior-zero) for 3.0:
 // - Put existing implementations into the detail namespace
 // - Support multi-dimensional thread blocks in the rest of implementations
diff --git a/cub/cub/block/block_radix_sort.cuh b/cub/cub/block/block_radix_sort.cuh
index e275d6f611e..1a98a66f646 100644
--- a/cub/cub/block/block_radix_sort.cuh
+++ b/cub/cub/block/block_radix_sort.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -34,18 +34,25 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include "block_exchange.cuh"
 #include "block_radix_rank.cuh"
 #include "radix_rank_sort_operations.cuh"
-#include "../config.cuh"
 #include "../util_ptx.cuh"
 #include "../util_type.cuh"
 
 CUB_NAMESPACE_BEGIN
 
 //! @rst
-//! BlockRadixSort class provides :ref:`collective <collective-primitives>` methods for sorting 
-//! items partitioned across a CUDA thread block using a radix sorting method.  
+//! BlockRadixSort class provides :ref:`collective <collective-primitives>` methods for sorting
+//! items partitioned across a CUDA thread block using a radix sorting method.
 //!
 //! .. image:: ../img/sorting_logo.png
 //!     :align: center
@@ -156,7 +163,7 @@ CUB_NAMESPACE_BEGIN
 //!         ...
 //!
 //! Suppose the set of input ``thread_keys`` across the block of threads is
-//! ``{ [0,511,1,510], [2,509,3,508], [4,507,5,506], ..., [254,257,255,256] }``.  
+//! ``{ [0,511,1,510], [2,509,3,508], [4,507,5,506], ..., [254,257,255,256] }``.
 //! The corresponding output ``thread_keys`` in those threads will be
 //! ``{ [0,1,2,3], [4,5,6,7], [8,9,10,11], ..., [508,509,510,511] }``.
 //!
@@ -173,40 +180,40 @@ CUB_NAMESPACE_BEGIN
 //!
 //! @ingroup BlockModule
 //!
-//! @tparam KeyT                 
+//! @tparam KeyT
 //!   KeyT type
 //!
-//! @tparam BLOCK_DIM_X          
+//! @tparam BLOCK_DIM_X
 //!   The thread block length in threads along the X dimension
 //!
-//! @tparam ITEMS_PER_THREAD     
+//! @tparam ITEMS_PER_THREAD
 //!   The number of items per thread
 //!
-//! @tparam ValueT               
+//! @tparam ValueT
 //!   **[optional]** ValueT type (default: cub::NullType, which indicates a keys-only sort)
 //!
-//! @tparam RADIX_BITS           
+//! @tparam RADIX_BITS
 //!   **[optional]** The number of radix bits per digit place (default: 4 bits)
 //!
-//! @tparam MEMOIZE_OUTER_SCAN   
-//!  **[optional]** Whether or not to buffer outer raking scan partials to incur fewer shared memory 
-//!  reads at the expense of higher register pressure (default: true for architectures SM35 and 
+//! @tparam MEMOIZE_OUTER_SCAN
+//!  **[optional]** Whether or not to buffer outer raking scan partials to incur fewer shared memory
+//!  reads at the expense of higher register pressure (default: true for architectures SM35 and
 //!  newer, false otherwise).
 //!
-//! @tparam INNER_SCAN_ALGORITHM 
-//!   **[optional]** The cub::BlockScanAlgorithm algorithm to use 
+//! @tparam INNER_SCAN_ALGORITHM
+//!   **[optional]** The cub::BlockScanAlgorithm algorithm to use
 //!   (default: cub::BLOCK_SCAN_WARP_SCANS)
 //!
-//! @tparam SMEM_CONFIG          
+//! @tparam SMEM_CONFIG
 //!   **[optional]*8 Shared memory bank mode (default: `cudaSharedMemBankSizeFourByte`)
 //!
-//! @tparam BLOCK_DIM_Y          
+//! @tparam BLOCK_DIM_Y
 //!   **[optional]** The thread block length in threads along the Y dimension (default: 1)
 //!
-//! @tparam BLOCK_DIM_Z          
+//! @tparam BLOCK_DIM_Z
 //!   **[optional]** The thread block length in threads along the Z dimension (default: 1)
 //!
-//! @tparam LEGACY_PTX_ARCH      
+//! @tparam LEGACY_PTX_ARCH
 //!   **[optional]** Unused
 template <
     typename                KeyT,
@@ -575,7 +582,7 @@ public:
     }
 
     //! @rst
-    //! Performs an ascending block-wide radix sort over a 
+    //! Performs an ascending block-wide radix sort over a
     //! :ref:`blocked arrangement <flexible-data-arrangement>` of keys.
     //!
     //! * @granularity
@@ -608,27 +615,27 @@ public:
     //! @endrst
     //!
     //! @tparam DecomposerT
-    //!   **[inferred]** Type of a callable object responsible for decomposing a 
+    //!   **[inferred]** Type of a callable object responsible for decomposing a
     //!   ``KeyT`` into a tuple of references to its constituent arithmetic types:
-    //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``. 
-    //!   The leftmost element of the tuple is considered the most significant. 
+    //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``.
+    //!   The leftmost element of the tuple is considered the most significant.
     //!   The call operator must not modify members of the key.
     //!
-    //! @param[in,out] keys 
+    //! @param[in,out] keys
     //!   Keys to sort
     //!
     //! @param decomposer
     //!   Callable object responsible for decomposing a ``KeyT`` into a tuple of
-    //!   references to its constituent arithmetic types. The leftmost element of 
-    //!   the tuple is considered the most significant. The call operator must not 
+    //!   references to its constituent arithmetic types. The leftmost element of
+    //!   the tuple is considered the most significant. The call operator must not
     //!   modify members of the key.
     //!
-    //! @param[in] begin_bit 
-    //!   The least-significant bit index (inclusive) needed for 
+    //! @param[in] begin_bit
+    //!   The least-significant bit index (inclusive) needed for
     //!   key comparison
     //!
-    //! @param[in] end_bit 
-    //!   The most-significant bit index (exclusive) needed for key 
+    //! @param[in] end_bit
+    //!   The most-significant bit index (exclusive) needed for key
     //!   comparison (e.g., `(sizeof(float) + sizeof(long long int)) * 8`)
     template <class DecomposerT>
     __device__ __forceinline__         //
@@ -648,7 +655,7 @@ public:
     }
 
     //! @rst
-    //! Performs an ascending block-wide radix sort over a 
+    //! Performs an ascending block-wide radix sort over a
     //! :ref:`blocked arrangement <flexible-data-arrangement>` of keys.
     //!
     //! * @granularity
@@ -681,19 +688,19 @@ public:
     //! @endrst
     //!
     //! @tparam DecomposerT
-    //!   **[inferred]** Type of a callable object responsible for decomposing a 
+    //!   **[inferred]** Type of a callable object responsible for decomposing a
     //!   ``KeyT`` into a tuple of references to its constituent arithmetic types:
-    //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``. 
-    //!   The leftmost element of the tuple is considered the most significant. 
+    //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``.
+    //!   The leftmost element of the tuple is considered the most significant.
     //!   The call operator must not modify members of the key.
     //!
-    //! @param[in,out] keys 
+    //! @param[in,out] keys
     //!   Keys to sort
     //!
     //! @param decomposer
     //!   Callable object responsible for decomposing a ``KeyT`` into a tuple of
-    //!   references to its constituent arithmetic types. The leftmost element of 
-    //!   the tuple is considered the most significant. The call operator must not 
+    //!   references to its constituent arithmetic types. The leftmost element of
+    //!   the tuple is considered the most significant. The call operator must not
     //!   modify members of the key.
     template <class DecomposerT>
     __device__ __forceinline__         //
@@ -758,7 +765,7 @@ public:
     }
 
     //! @rst
-    //! Performs an ascending block-wide radix sort over a 
+    //! Performs an ascending block-wide radix sort over a
     //! :ref:`blocked arrangement <flexible-data-arrangement>` of keys and values.
     //!
     //! * BlockRadixSort can only accommodate one associated tile of values. To "truck along"
@@ -796,13 +803,13 @@ public:
     //! @endrst
     //!
     //! @tparam DecomposerT
-    //!   **[inferred]** Type of a callable object responsible for decomposing a 
+    //!   **[inferred]** Type of a callable object responsible for decomposing a
     //!   ``KeyT`` into a tuple of references to its constituent arithmetic types:
-    //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``. 
-    //!   The leftmost element of the tuple is considered the most significant. 
+    //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``.
+    //!   The leftmost element of the tuple is considered the most significant.
     //!   The call operator must not modify members of the key.
     //!
-    //! @param[in,out] keys 
+    //! @param[in,out] keys
     //!   Keys to sort
     //!
     //! @param[in,out] values
@@ -810,16 +817,16 @@ public:
     //!
     //! @param decomposer
     //!   Callable object responsible for decomposing a ``KeyT`` into a tuple of
-    //!   references to its constituent arithmetic types. The leftmost element of 
-    //!   the tuple is considered the most significant. The call operator must not 
+    //!   references to its constituent arithmetic types. The leftmost element of
+    //!   the tuple is considered the most significant. The call operator must not
     //!   modify members of the key.
     //!
-    //! @param[in] begin_bit 
-    //!   The least-significant bit index (inclusive) needed for 
+    //! @param[in] begin_bit
+    //!   The least-significant bit index (inclusive) needed for
     //!   key comparison
     //!
-    //! @param[in] end_bit 
-    //!   The most-significant bit index (exclusive) needed for key 
+    //! @param[in] end_bit
+    //!   The most-significant bit index (exclusive) needed for key
     //!   comparison (e.g., `(sizeof(float) + sizeof(long long int)) * 8`)
     template <class DecomposerT>
     __device__ __forceinline__         //
@@ -841,7 +848,7 @@ public:
     }
 
     //! @rst
-    //! Performs an ascending block-wide radix sort over a 
+    //! Performs an ascending block-wide radix sort over a
     //! :ref:`blocked arrangement <flexible-data-arrangement>` of keys and values.
     //!
     //! * BlockRadixSort can only accommodate one associated tile of values. To "truck along"
@@ -879,13 +886,13 @@ public:
     //! @endrst
     //!
     //! @tparam DecomposerT
-    //!   **[inferred]** Type of a callable object responsible for decomposing a 
+    //!   **[inferred]** Type of a callable object responsible for decomposing a
     //!   ``KeyT`` into a tuple of references to its constituent arithmetic types:
-    //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``. 
-    //!   The leftmost element of the tuple is considered the most significant. 
+    //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``.
+    //!   The leftmost element of the tuple is considered the most significant.
     //!   The call operator must not modify members of the key.
     //!
-    //! @param[in,out] keys 
+    //! @param[in,out] keys
     //!   Keys to sort
     //!
     //! @param[in,out] values
@@ -893,8 +900,8 @@ public:
     //!
     //! @param decomposer
     //!   Callable object responsible for decomposing a ``KeyT`` into a tuple of
-    //!   references to its constituent arithmetic types. The leftmost element of 
-    //!   the tuple is considered the most significant. The call operator must not 
+    //!   references to its constituent arithmetic types. The leftmost element of
+    //!   the tuple is considered the most significant. The call operator must not
     //!   modify members of the key.
     template <class DecomposerT>
     __device__ __forceinline__         //
@@ -959,7 +966,7 @@ public:
     }
 
     //! @rst
-    //! Performs a descending block-wide radix sort over a 
+    //! Performs a descending block-wide radix sort over a
     //! :ref:`blocked arrangement <flexible-data-arrangement>` of keys.
     //!
     //! * @granularity
@@ -992,27 +999,27 @@ public:
     //! @endrst
     //!
     //! @tparam DecomposerT
-    //!   **[inferred]** Type of a callable object responsible for decomposing a 
+    //!   **[inferred]** Type of a callable object responsible for decomposing a
     //!   ``KeyT`` into a tuple of references to its constituent arithmetic types:
-    //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``. 
-    //!   The leftmost element of the tuple is considered the most significant. 
+    //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``.
+    //!   The leftmost element of the tuple is considered the most significant.
     //!   The call operator must not modify members of the key.
     //!
-    //! @param[in,out] keys 
+    //! @param[in,out] keys
     //!   Keys to sort
     //!
     //! @param decomposer
     //!   Callable object responsible for decomposing a ``KeyT`` into a tuple of
-    //!   references to its constituent arithmetic types. The leftmost element of 
-    //!   the tuple is considered the most significant. The call operator must not 
+    //!   references to its constituent arithmetic types. The leftmost element of
+    //!   the tuple is considered the most significant. The call operator must not
     //!   modify members of the key.
     //!
-    //! @param[in] begin_bit 
-    //!   The least-significant bit index (inclusive) needed for 
+    //! @param[in] begin_bit
+    //!   The least-significant bit index (inclusive) needed for
     //!   key comparison
     //!
-    //! @param[in] end_bit 
-    //!   The most-significant bit index (exclusive) needed for key 
+    //! @param[in] end_bit
+    //!   The most-significant bit index (exclusive) needed for key
     //!   comparison (e.g., `(sizeof(float) + sizeof(long long int)) * 8`)
     template <class DecomposerT>
     __device__ __forceinline__         //
@@ -1035,7 +1042,7 @@ public:
     }
 
     //! @rst
-    //! Performs a descending block-wide radix sort over a 
+    //! Performs a descending block-wide radix sort over a
     //! :ref:`blocked arrangement <flexible-data-arrangement>` of keys.
     //!
     //! * @granularity
@@ -1068,19 +1075,19 @@ public:
     //! @endrst
     //!
     //! @tparam DecomposerT
-    //!   **[inferred]** Type of a callable object responsible for decomposing a 
+    //!   **[inferred]** Type of a callable object responsible for decomposing a
     //!   ``KeyT`` into a tuple of references to its constituent arithmetic types:
-    //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``. 
-    //!   The leftmost element of the tuple is considered the most significant. 
+    //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``.
+    //!   The leftmost element of the tuple is considered the most significant.
     //!   The call operator must not modify members of the key.
     //!
-    //! @param[in,out] keys 
+    //! @param[in,out] keys
     //!   Keys to sort
     //!
     //! @param decomposer
     //!   Callable object responsible for decomposing a ``KeyT`` into a tuple of
-    //!   references to its constituent arithmetic types. The leftmost element of 
-    //!   the tuple is considered the most significant. The call operator must not 
+    //!   references to its constituent arithmetic types. The leftmost element of
+    //!   the tuple is considered the most significant. The call operator must not
     //!   modify members of the key.
     template <class DecomposerT>
     __device__ __forceinline__         //
@@ -1153,7 +1160,7 @@ public:
     }
 
     //! @rst
-    //! Performs a descending block-wide radix sort over a 
+    //! Performs a descending block-wide radix sort over a
     //! :ref:`blocked arrangement <flexible-data-arrangement>` of keys and values.
     //!
     //! * BlockRadixSort can only accommodate one associated tile of values. To "truck along"
@@ -1191,13 +1198,13 @@ public:
     //! @endrst
     //!
     //! @tparam DecomposerT
-    //!   **[inferred]** Type of a callable object responsible for decomposing a 
+    //!   **[inferred]** Type of a callable object responsible for decomposing a
     //!   ``KeyT`` into a tuple of references to its constituent arithmetic types:
-    //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``. 
-    //!   The leftmost element of the tuple is considered the most significant. 
+    //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``.
+    //!   The leftmost element of the tuple is considered the most significant.
     //!   The call operator must not modify members of the key.
     //!
-    //! @param[in,out] keys 
+    //! @param[in,out] keys
     //!   Keys to sort
     //!
     //! @param[in,out] values
@@ -1205,16 +1212,16 @@ public:
     //!
     //! @param decomposer
     //!   Callable object responsible for decomposing a ``KeyT`` into a tuple of
-    //!   references to its constituent arithmetic types. The leftmost element of 
-    //!   the tuple is considered the most significant. The call operator must not 
+    //!   references to its constituent arithmetic types. The leftmost element of
+    //!   the tuple is considered the most significant. The call operator must not
     //!   modify members of the key.
     //!
-    //! @param[in] begin_bit 
-    //!   The least-significant bit index (inclusive) needed for 
+    //! @param[in] begin_bit
+    //!   The least-significant bit index (inclusive) needed for
     //!   key comparison
     //!
-    //! @param[in] end_bit 
-    //!   The most-significant bit index (exclusive) needed for key 
+    //! @param[in] end_bit
+    //!   The most-significant bit index (exclusive) needed for key
     //!   comparison (e.g., `(sizeof(float) + sizeof(long long int)) * 8`)
     template <class DecomposerT>
     __device__ __forceinline__         //
@@ -1236,7 +1243,7 @@ public:
     }
 
     //! @rst
-    //! Performs a descending block-wide radix sort over a 
+    //! Performs a descending block-wide radix sort over a
     //! :ref:`blocked arrangement <flexible-data-arrangement>` of keys and values.
     //!
     //! * BlockRadixSort can only accommodate one associated tile of values. To "truck along"
@@ -1274,13 +1281,13 @@ public:
     //! @endrst
     //!
     //! @tparam DecomposerT
-    //!   **[inferred]** Type of a callable object responsible for decomposing a 
+    //!   **[inferred]** Type of a callable object responsible for decomposing a
     //!   ``KeyT`` into a tuple of references to its constituent arithmetic types:
-    //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``. 
-    //!   The leftmost element of the tuple is considered the most significant. 
+    //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``.
+    //!   The leftmost element of the tuple is considered the most significant.
     //!   The call operator must not modify members of the key.
     //!
-    //! @param[in,out] keys 
+    //! @param[in,out] keys
     //!   Keys to sort
     //!
     //! @param[in,out] values
@@ -1288,8 +1295,8 @@ public:
     //!
     //! @param decomposer
     //!   Callable object responsible for decomposing a ``KeyT`` into a tuple of
-    //!   references to its constituent arithmetic types. The leftmost element of 
-    //!   the tuple is considered the most significant. The call operator must not 
+    //!   references to its constituent arithmetic types. The leftmost element of
+    //!   the tuple is considered the most significant. The call operator must not
     //!   modify members of the key.
     template <class DecomposerT>
     __device__ __forceinline__         //
@@ -1364,8 +1371,8 @@ public:
     }
 
     //! @rst
-    //! Performs an ascending block-wide radix sort over a 
-    //! :ref:`blocked arrangement <flexible-data-arrangement>` of keys, leaving them in a 
+    //! Performs an ascending block-wide radix sort over a
+    //! :ref:`blocked arrangement <flexible-data-arrangement>` of keys, leaving them in a
     //! :ref:`striped arrangement <flexible-data-arrangement>`.
     //!
     //! * @granularity
@@ -1377,7 +1384,7 @@ public:
     //! Let's consider a user-defined ``custom_t`` type below. To sort an array of
     //! ``custom_t`` objects, we have to tell CUB about relevant members of the
     //! ``custom_t`` type. We do this by providing a decomposer that returns a
-    //! tuple of references to relevant members of the key. 
+    //! tuple of references to relevant members of the key.
     //!
     //! .. literalinclude:: ../../test/catch2_test_block_radix_sort_custom.cu
     //!     :language: c++
@@ -1398,27 +1405,27 @@ public:
     //! @endrst
     //!
     //! @tparam DecomposerT
-    //!   **[inferred]** Type of a callable object responsible for decomposing a 
+    //!   **[inferred]** Type of a callable object responsible for decomposing a
     //!   ``KeyT`` into a tuple of references to its constituent arithmetic types:
-    //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``. 
-    //!   The leftmost element of the tuple is considered the most significant. 
+    //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``.
+    //!   The leftmost element of the tuple is considered the most significant.
     //!   The call operator must not modify members of the key.
     //!
-    //! @param[in,out] keys 
+    //! @param[in,out] keys
     //!   Keys to sort
     //!
     //! @param decomposer
     //!   Callable object responsible for decomposing a ``KeyT`` into a tuple of
-    //!   references to its constituent arithmetic types. The leftmost element of 
-    //!   the tuple is considered the most significant. The call operator must not 
+    //!   references to its constituent arithmetic types. The leftmost element of
+    //!   the tuple is considered the most significant. The call operator must not
     //!   modify members of the key.
     //!
-    //! @param[in] begin_bit 
-    //!   The least-significant bit index (inclusive) needed for 
+    //! @param[in] begin_bit
+    //!   The least-significant bit index (inclusive) needed for
     //!   key comparison
     //!
-    //! @param[in] end_bit 
-    //!   The most-significant bit index (exclusive) needed for key 
+    //! @param[in] end_bit
+    //!   The most-significant bit index (exclusive) needed for key
     //!   comparison (e.g., `(sizeof(float) + sizeof(long long int)) * 8`)
     template <class DecomposerT>
     __device__ __forceinline__         //
@@ -1441,8 +1448,8 @@ public:
     }
 
     //! @rst
-    //! Performs an ascending block-wide radix sort over a 
-    //! :ref:`blocked arrangement <flexible-data-arrangement>` of keys, leaving them in a 
+    //! Performs an ascending block-wide radix sort over a
+    //! :ref:`blocked arrangement <flexible-data-arrangement>` of keys, leaving them in a
     //! :ref:`striped arrangement <flexible-data-arrangement>`.
     //!
     //! * @granularity
@@ -1454,7 +1461,7 @@ public:
     //! Let's consider a user-defined ``custom_t`` type below. To sort an array of
     //! ``custom_t`` objects, we have to tell CUB about relevant members of the
     //! ``custom_t`` type. We do this by providing a decomposer that returns a
-    //! tuple of references to relevant members of the key. 
+    //! tuple of references to relevant members of the key.
     //!
     //! .. literalinclude:: ../../test/catch2_test_block_radix_sort_custom.cu
     //!     :language: c++
@@ -1475,19 +1482,19 @@ public:
     //! @endrst
     //!
     //! @tparam DecomposerT
-    //!   **[inferred]** Type of a callable object responsible for decomposing a 
+    //!   **[inferred]** Type of a callable object responsible for decomposing a
     //!   ``KeyT`` into a tuple of references to its constituent arithmetic types:
-    //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``. 
-    //!   The leftmost element of the tuple is considered the most significant. 
+    //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``.
+    //!   The leftmost element of the tuple is considered the most significant.
     //!   The call operator must not modify members of the key.
     //!
-    //! @param[in,out] keys 
+    //! @param[in,out] keys
     //!   Keys to sort
     //!
     //! @param decomposer
     //!   Callable object responsible for decomposing a ``KeyT`` into a tuple of
-    //!   references to its constituent arithmetic types. The leftmost element of 
-    //!   the tuple is considered the most significant. The call operator must not 
+    //!   references to its constituent arithmetic types. The leftmost element of
+    //!   the tuple is considered the most significant. The call operator must not
     //!   modify members of the key.
     template <class DecomposerT>
     __device__ __forceinline__         //
@@ -1560,8 +1567,8 @@ public:
     }
 
     //! @rst
-    //! Performs an ascending block-wide radix sort over a 
-    //! :ref:`blocked arrangement <flexible-data-arrangement>` of keys and values, leaving them in a 
+    //! Performs an ascending block-wide radix sort over a
+    //! :ref:`blocked arrangement <flexible-data-arrangement>` of keys and values, leaving them in a
     //! :ref:`striped arrangement <flexible-data-arrangement>`.
     //!
     //! * @granularity
@@ -1573,7 +1580,7 @@ public:
     //! Let's consider a user-defined ``custom_t`` type below. To sort an array of
     //! ``custom_t`` objects, we have to tell CUB about relevant members of the
     //! ``custom_t`` type. We do this by providing a decomposer that returns a
-    //! tuple of references to relevant members of the key. 
+    //! tuple of references to relevant members of the key.
     //!
     //! .. literalinclude:: ../../test/catch2_test_block_radix_sort_custom.cu
     //!     :language: c++
@@ -1594,13 +1601,13 @@ public:
     //! @endrst
     //!
     //! @tparam DecomposerT
-    //!   **[inferred]** Type of a callable object responsible for decomposing a 
+    //!   **[inferred]** Type of a callable object responsible for decomposing a
     //!   ``KeyT`` into a tuple of references to its constituent arithmetic types:
-    //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``. 
-    //!   The leftmost element of the tuple is considered the most significant. 
+    //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``.
+    //!   The leftmost element of the tuple is considered the most significant.
     //!   The call operator must not modify members of the key.
     //!
-    //! @param[in,out] keys 
+    //! @param[in,out] keys
     //!   Keys to sort
     //!
     //! @param[in,out] values
@@ -1608,16 +1615,16 @@ public:
     //!
     //! @param decomposer
     //!   Callable object responsible for decomposing a ``KeyT`` into a tuple of
-    //!   references to its constituent arithmetic types. The leftmost element of 
-    //!   the tuple is considered the most significant. The call operator must not 
+    //!   references to its constituent arithmetic types. The leftmost element of
+    //!   the tuple is considered the most significant. The call operator must not
     //!   modify members of the key.
     //!
-    //! @param[in] begin_bit 
-    //!   The least-significant bit index (inclusive) needed for 
+    //! @param[in] begin_bit
+    //!   The least-significant bit index (inclusive) needed for
     //!   key comparison
     //!
-    //! @param[in] end_bit 
-    //!   The most-significant bit index (exclusive) needed for key 
+    //! @param[in] end_bit
+    //!   The most-significant bit index (exclusive) needed for key
     //!   comparison (e.g., `(sizeof(float) + sizeof(long long int)) * 8`)
     template <class DecomposerT>
     __device__ __forceinline__         //
@@ -1639,8 +1646,8 @@ public:
     }
 
     //! @rst
-    //! Performs an ascending block-wide radix sort over a 
-    //! :ref:`blocked arrangement <flexible-data-arrangement>` of keys and values, leaving them in a 
+    //! Performs an ascending block-wide radix sort over a
+    //! :ref:`blocked arrangement <flexible-data-arrangement>` of keys and values, leaving them in a
     //! :ref:`striped arrangement <flexible-data-arrangement>`.
     //!
     //! * @granularity
@@ -1652,7 +1659,7 @@ public:
     //! Let's consider a user-defined ``custom_t`` type below. To sort an array of
     //! ``custom_t`` objects, we have to tell CUB about relevant members of the
     //! ``custom_t`` type. We do this by providing a decomposer that returns a
-    //! tuple of references to relevant members of the key. 
+    //! tuple of references to relevant members of the key.
     //!
     //! .. literalinclude:: ../../test/catch2_test_block_radix_sort_custom.cu
     //!     :language: c++
@@ -1673,13 +1680,13 @@ public:
     //! @endrst
     //!
     //! @tparam DecomposerT
-    //!   **[inferred]** Type of a callable object responsible for decomposing a 
+    //!   **[inferred]** Type of a callable object responsible for decomposing a
     //!   ``KeyT`` into a tuple of references to its constituent arithmetic types:
-    //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``. 
-    //!   The leftmost element of the tuple is considered the most significant. 
+    //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``.
+    //!   The leftmost element of the tuple is considered the most significant.
     //!   The call operator must not modify members of the key.
     //!
-    //! @param[in,out] keys 
+    //! @param[in,out] keys
     //!   Keys to sort
     //!
     //! @param[in,out] values
@@ -1687,8 +1694,8 @@ public:
     //!
     //! @param decomposer
     //!   Callable object responsible for decomposing a ``KeyT`` into a tuple of
-    //!   references to its constituent arithmetic types. The leftmost element of 
-    //!   the tuple is considered the most significant. The call operator must not 
+    //!   references to its constituent arithmetic types. The leftmost element of
+    //!   the tuple is considered the most significant. The call operator must not
     //!   modify members of the key.
     template <class DecomposerT>
     __device__ __forceinline__         //
@@ -1756,8 +1763,8 @@ public:
     }
 
     //! @rst
-    //! Performs a descending block-wide radix sort over a 
-    //! :ref:`blocked arrangement <flexible-data-arrangement>` of keys, leaving them in a 
+    //! Performs a descending block-wide radix sort over a
+    //! :ref:`blocked arrangement <flexible-data-arrangement>` of keys, leaving them in a
     //! :ref:`striped arrangement <flexible-data-arrangement>`.
     //!
     //! * @granularity
@@ -1769,7 +1776,7 @@ public:
     //! Let's consider a user-defined ``custom_t`` type below. To sort an array of
     //! ``custom_t`` objects, we have to tell CUB about relevant members of the
     //! ``custom_t`` type. We do this by providing a decomposer that returns a
-    //! tuple of references to relevant members of the key. 
+    //! tuple of references to relevant members of the key.
     //!
     //! .. literalinclude:: ../../test/catch2_test_block_radix_sort_custom.cu
     //!     :language: c++
@@ -1790,27 +1797,27 @@ public:
     //! @endrst
     //!
     //! @tparam DecomposerT
-    //!   **[inferred]** Type of a callable object responsible for decomposing a 
+    //!   **[inferred]** Type of a callable object responsible for decomposing a
     //!   ``KeyT`` into a tuple of references to its constituent arithmetic types:
-    //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``. 
-    //!   The leftmost element of the tuple is considered the most significant. 
+    //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``.
+    //!   The leftmost element of the tuple is considered the most significant.
     //!   The call operator must not modify members of the key.
     //!
-    //! @param[in,out] keys 
+    //! @param[in,out] keys
     //!   Keys to sort
     //!
     //! @param decomposer
     //!   Callable object responsible for decomposing a ``KeyT`` into a tuple of
-    //!   references to its constituent arithmetic types. The leftmost element of 
-    //!   the tuple is considered the most significant. The call operator must not 
+    //!   references to its constituent arithmetic types. The leftmost element of
+    //!   the tuple is considered the most significant. The call operator must not
     //!   modify members of the key.
     //!
-    //! @param[in] begin_bit 
-    //!   The least-significant bit index (inclusive) needed for 
+    //! @param[in] begin_bit
+    //!   The least-significant bit index (inclusive) needed for
     //!   key comparison
     //!
-    //! @param[in] end_bit 
-    //!   The most-significant bit index (exclusive) needed for key 
+    //! @param[in] end_bit
+    //!   The most-significant bit index (exclusive) needed for key
     //!   comparison (e.g., `(sizeof(float) + sizeof(long long int)) * 8`)
     template <class DecomposerT>
     __device__ __forceinline__         //
@@ -1833,8 +1840,8 @@ public:
     }
 
     //! @rst
-    //! Performs a descending block-wide radix sort over a 
-    //! :ref:`blocked arrangement <flexible-data-arrangement>` of keys, leaving them in a 
+    //! Performs a descending block-wide radix sort over a
+    //! :ref:`blocked arrangement <flexible-data-arrangement>` of keys, leaving them in a
     //! :ref:`striped arrangement <flexible-data-arrangement>`.
     //!
     //! * @granularity
@@ -1846,7 +1853,7 @@ public:
     //! Let's consider a user-defined ``custom_t`` type below. To sort an array of
     //! ``custom_t`` objects, we have to tell CUB about relevant members of the
     //! ``custom_t`` type. We do this by providing a decomposer that returns a
-    //! tuple of references to relevant members of the key. 
+    //! tuple of references to relevant members of the key.
     //!
     //! .. literalinclude:: ../../test/catch2_test_block_radix_sort_custom.cu
     //!     :language: c++
@@ -1867,19 +1874,19 @@ public:
     //! @endrst
     //!
     //! @tparam DecomposerT
-    //!   **[inferred]** Type of a callable object responsible for decomposing a 
+    //!   **[inferred]** Type of a callable object responsible for decomposing a
     //!   ``KeyT`` into a tuple of references to its constituent arithmetic types:
-    //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``. 
-    //!   The leftmost element of the tuple is considered the most significant. 
+    //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``.
+    //!   The leftmost element of the tuple is considered the most significant.
     //!   The call operator must not modify members of the key.
     //!
-    //! @param[in,out] keys 
+    //! @param[in,out] keys
     //!   Keys to sort
     //!
     //! @param decomposer
     //!   Callable object responsible for decomposing a ``KeyT`` into a tuple of
-    //!   references to its constituent arithmetic types. The leftmost element of 
-    //!   the tuple is considered the most significant. The call operator must not 
+    //!   references to its constituent arithmetic types. The leftmost element of
+    //!   the tuple is considered the most significant. The call operator must not
     //!   modify members of the key.
     template <class DecomposerT>
     __device__ __forceinline__         //
@@ -1952,8 +1959,8 @@ public:
     }
 
     //! @rst
-    //! Performs a descending block-wide radix sort over a 
-    //! :ref:`blocked arrangement <flexible-data-arrangement>` of keys and values, leaving them in a 
+    //! Performs a descending block-wide radix sort over a
+    //! :ref:`blocked arrangement <flexible-data-arrangement>` of keys and values, leaving them in a
     //! :ref:`striped arrangement <flexible-data-arrangement>`.
     //!
     //! * @granularity
@@ -1965,7 +1972,7 @@ public:
     //! Let's consider a user-defined ``custom_t`` type below. To sort an array of
     //! ``custom_t`` objects, we have to tell CUB about relevant members of the
     //! ``custom_t`` type. We do this by providing a decomposer that returns a
-    //! tuple of references to relevant members of the key. 
+    //! tuple of references to relevant members of the key.
     //!
     //! .. literalinclude:: ../../test/catch2_test_block_radix_sort_custom.cu
     //!     :language: c++
@@ -1986,13 +1993,13 @@ public:
     //! @endrst
     //!
     //! @tparam DecomposerT
-    //!   **[inferred]** Type of a callable object responsible for decomposing a 
+    //!   **[inferred]** Type of a callable object responsible for decomposing a
     //!   ``KeyT`` into a tuple of references to its constituent arithmetic types:
-    //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``. 
-    //!   The leftmost element of the tuple is considered the most significant. 
+    //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``.
+    //!   The leftmost element of the tuple is considered the most significant.
     //!   The call operator must not modify members of the key.
     //!
-    //! @param[in,out] keys 
+    //! @param[in,out] keys
     //!   Keys to sort
     //!
     //! @param[in,out] values
@@ -2000,16 +2007,16 @@ public:
     //!
     //! @param decomposer
     //!   Callable object responsible for decomposing a ``KeyT`` into a tuple of
-    //!   references to its constituent arithmetic types. The leftmost element of 
-    //!   the tuple is considered the most significant. The call operator must not 
+    //!   references to its constituent arithmetic types. The leftmost element of
+    //!   the tuple is considered the most significant. The call operator must not
     //!   modify members of the key.
     //!
-    //! @param[in] begin_bit 
-    //!   The least-significant bit index (inclusive) needed for 
+    //! @param[in] begin_bit
+    //!   The least-significant bit index (inclusive) needed for
     //!   key comparison
     //!
-    //! @param[in] end_bit 
-    //!   The most-significant bit index (exclusive) needed for key 
+    //! @param[in] end_bit
+    //!   The most-significant bit index (exclusive) needed for key
     //!   comparison (e.g., `(sizeof(float) + sizeof(long long int)) * 8`)
     template <class DecomposerT>
     __device__ __forceinline__         //
@@ -2031,8 +2038,8 @@ public:
     }
 
     //! @rst
-    //! Performs a descending block-wide radix sort over a 
-    //! :ref:`blocked arrangement <flexible-data-arrangement>` of keys and values, leaving them in a 
+    //! Performs a descending block-wide radix sort over a
+    //! :ref:`blocked arrangement <flexible-data-arrangement>` of keys and values, leaving them in a
     //! :ref:`striped arrangement <flexible-data-arrangement>`.
     //!
     //! * @granularity
@@ -2044,7 +2051,7 @@ public:
     //! Let's consider a user-defined ``custom_t`` type below. To sort an array of
     //! ``custom_t`` objects, we have to tell CUB about relevant members of the
     //! ``custom_t`` type. We do this by providing a decomposer that returns a
-    //! tuple of references to relevant members of the key. 
+    //! tuple of references to relevant members of the key.
     //!
     //! .. literalinclude:: ../../test/catch2_test_block_radix_sort_custom.cu
     //!     :language: c++
@@ -2065,13 +2072,13 @@ public:
     //! @endrst
     //!
     //! @tparam DecomposerT
-    //!   **[inferred]** Type of a callable object responsible for decomposing a 
+    //!   **[inferred]** Type of a callable object responsible for decomposing a
     //!   ``KeyT`` into a tuple of references to its constituent arithmetic types:
-    //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``. 
-    //!   The leftmost element of the tuple is considered the most significant. 
+    //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``.
+    //!   The leftmost element of the tuple is considered the most significant.
     //!   The call operator must not modify members of the key.
     //!
-    //! @param[in,out] keys 
+    //! @param[in,out] keys
     //!   Keys to sort
     //!
     //! @param[in,out] values
@@ -2079,8 +2086,8 @@ public:
     //!
     //! @param decomposer
     //!   Callable object responsible for decomposing a ``KeyT`` into a tuple of
-    //!   references to its constituent arithmetic types. The leftmost element of 
-    //!   the tuple is considered the most significant. The call operator must not 
+    //!   references to its constituent arithmetic types. The leftmost element of
+    //!   the tuple is considered the most significant. The call operator must not
     //!   modify members of the key.
     template <class DecomposerT>
     __device__ __forceinline__         //
diff --git a/cub/cub/block/block_raking_layout.cuh b/cub/cub/block/block_raking_layout.cuh
index 4d49f54f2c8..fbe332c9e07 100644
--- a/cub/cub/block/block_raking_layout.cuh
+++ b/cub/cub/block/block_raking_layout.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -35,6 +35,13 @@
 #pragma once
 
 #include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include "../util_type.cuh"
 
 CUB_NAMESPACE_BEGIN
diff --git a/cub/cub/block/block_reduce.cuh b/cub/cub/block/block_reduce.cuh
index 5a9db703db9..8e0814604e0 100644
--- a/cub/cub/block/block_reduce.cuh
+++ b/cub/cub/block/block_reduce.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -33,10 +33,17 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include "specializations/block_reduce_raking.cuh"
 #include "specializations/block_reduce_raking_commutative_only.cuh"
 #include "specializations/block_reduce_warp_reductions.cuh"
-#include "../config.cuh"
 #include "../util_ptx.cuh"
 #include "../util_type.cuh"
 #include "../thread/thread_operators.cuh"
@@ -349,7 +356,7 @@ public:
     template <typename ReductionOp>
     __device__ __forceinline__ T Reduce(
         T               input,                      ///< [in] Calling thread's input
-        ReductionOp     reduction_op)               ///< [in] Binary reduction functor 
+        ReductionOp     reduction_op)               ///< [in] Binary reduction functor
     {
         return InternalBlockReduce(temp_storage).template Reduce<true>(input, BLOCK_THREADS, reduction_op);
     }
@@ -396,7 +403,7 @@ public:
         typename ReductionOp>
     __device__ __forceinline__ T Reduce(
         T               (&inputs)[ITEMS_PER_THREAD],    ///< [in] Calling thread's input segment
-        ReductionOp     reduction_op)                   ///< [in] Binary reduction functor 
+        ReductionOp     reduction_op)                   ///< [in] Binary reduction functor
     {
         // Reduce partials
         T partial = internal::ThreadReduce(inputs, reduction_op);
@@ -441,7 +448,7 @@ public:
     template <typename ReductionOp>
     __device__ __forceinline__ T Reduce(
         T                   input,                  ///< [in] Calling thread's input
-        ReductionOp         reduction_op,           ///< [in] Binary reduction functor 
+        ReductionOp         reduction_op,           ///< [in] Binary reduction functor
         int                 num_valid)              ///< [in] Number of threads containing valid elements (may be less than BLOCK_THREADS)
     {
         // Determine if we skip bounds checking
diff --git a/cub/cub/block/block_run_length_decode.cuh b/cub/cub/block/block_run_length_decode.cuh
index 41a3ab22b15..c14d36fa015 100644
--- a/cub/cub/block/block_run_length_decode.cuh
+++ b/cub/cub/block/block_run_length_decode.cuh
@@ -28,6 +28,13 @@
 #pragma once
 
 #include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include "../thread/thread_search.cuh"
 #include "../util_math.cuh"
 #include "../util_namespace.cuh"
diff --git a/cub/cub/block/block_scan.cuh b/cub/cub/block/block_scan.cuh
index 544c15f0dbb..6629fc8b087 100644
--- a/cub/cub/block/block_scan.cuh
+++ b/cub/cub/block/block_scan.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -33,9 +33,16 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include "specializations/block_scan_raking.cuh"
 #include "specializations/block_scan_warp_scans.cuh"
-#include "../config.cuh"
 #include "../util_type.cuh"
 #include "../util_ptx.cuh"
 
@@ -735,7 +742,7 @@ public:
         T               input,                          ///< [in] Calling thread's input item
         T               &output,                        ///< [out] Calling thread's output item (may be aliased to \p input)
         T               initial_value,                  ///< [in] Initial value to seed the exclusive scan (and is assigned to \p output[0] in <em>thread</em><sub>0</sub>)
-        ScanOp          scan_op)                        ///< [in] Binary scan functor 
+        ScanOp          scan_op)                        ///< [in] Binary scan functor
     {
         InternalBlockScan(temp_storage).ExclusiveScan(input, output, initial_value, scan_op);
     }
@@ -785,7 +792,7 @@ public:
         T               input,              ///< [in] Calling thread's input items
         T               &output,            ///< [out] Calling thread's output items (may be aliased to \p input)
         T               initial_value,      ///< [in] Initial value to seed the exclusive scan (and is assigned to \p output[0] in <em>thread</em><sub>0</sub>)
-        ScanOp          scan_op,            ///< [in] Binary scan functor 
+        ScanOp          scan_op,            ///< [in] Binary scan functor
         T               &block_aggregate)   ///< [out] block-wide aggregate reduction of input items
     {
         InternalBlockScan(temp_storage).ExclusiveScan(input, output, initial_value, scan_op, block_aggregate);
@@ -873,7 +880,7 @@ public:
     __device__ __forceinline__ void ExclusiveScan(
         T                       input,                          ///< [in] Calling thread's input item
         T                       &output,                        ///< [out] Calling thread's output item (may be aliased to \p input)
-        ScanOp                  scan_op,                        ///< [in] Binary scan functor 
+        ScanOp                  scan_op,                        ///< [in] Binary scan functor
         BlockPrefixCallbackOp   &block_prefix_callback_op)      ///< [in-out] <b>[<em>warp</em><sub>0</sub> only]</b> Call-back functor for specifying a block-wide prefix to be applied to the logical input sequence.
     {
         InternalBlockScan(temp_storage).ExclusiveScan(input, output, scan_op, block_prefix_callback_op);
@@ -1717,7 +1724,7 @@ public:
     __device__ __forceinline__ void InclusiveScan(
         T               input,                          ///< [in] Calling thread's input item
         T               &output,                        ///< [out] Calling thread's output item (may be aliased to \p input)
-        ScanOp          scan_op)                        ///< [in] Binary scan functor 
+        ScanOp          scan_op)                        ///< [in] Binary scan functor
     {
         InternalBlockScan(temp_storage).InclusiveScan(input, output, scan_op);
     }
@@ -1766,7 +1773,7 @@ public:
     __device__ __forceinline__ void InclusiveScan(
         T               input,                          ///< [in] Calling thread's input item
         T               &output,                        ///< [out] Calling thread's output item (may be aliased to \p input)
-        ScanOp          scan_op,                        ///< [in] Binary scan functor 
+        ScanOp          scan_op,                        ///< [in] Binary scan functor
         T               &block_aggregate)               ///< [out] block-wide aggregate reduction of input items
     {
         InternalBlockScan(temp_storage).InclusiveScan(input, output, scan_op, block_aggregate);
@@ -1854,7 +1861,7 @@ public:
     __device__ __forceinline__ void InclusiveScan(
         T                       input,                          ///< [in] Calling thread's input item
         T                       &output,                        ///< [out] Calling thread's output item (may be aliased to \p input)
-        ScanOp                  scan_op,                        ///< [in] Binary scan functor 
+        ScanOp                  scan_op,                        ///< [in] Binary scan functor
         BlockPrefixCallbackOp   &block_prefix_callback_op)      ///< [in-out] <b>[<em>warp</em><sub>0</sub> only]</b> Call-back functor for specifying a block-wide prefix to be applied to the logical input sequence.
     {
         InternalBlockScan(temp_storage).InclusiveScan(input, output, scan_op, block_prefix_callback_op);
@@ -1914,7 +1921,7 @@ public:
     __device__ __forceinline__ void InclusiveScan(
         T               (&input)[ITEMS_PER_THREAD],     ///< [in] Calling thread's input items
         T               (&output)[ITEMS_PER_THREAD],    ///< [out] Calling thread's output items (may be aliased to \p input)
-        ScanOp          scan_op)                        ///< [in] Binary scan functor 
+        ScanOp          scan_op)                        ///< [in] Binary scan functor
     {
         if (ITEMS_PER_THREAD == 1)
         {
@@ -1984,7 +1991,7 @@ public:
     __device__ __forceinline__ void InclusiveScan(
         T               (&input)[ITEMS_PER_THREAD],     ///< [in] Calling thread's input items
         T               (&output)[ITEMS_PER_THREAD],    ///< [out] Calling thread's output items (may be aliased to \p input)
-        ScanOp          scan_op,                        ///< [in] Binary scan functor 
+        ScanOp          scan_op,                        ///< [in] Binary scan functor
         T               &block_aggregate)               ///< [out] block-wide aggregate reduction of input items
     {
         if (ITEMS_PER_THREAD == 1)
@@ -2098,7 +2105,7 @@ public:
     __device__ __forceinline__ void InclusiveScan(
         T                       (&input)[ITEMS_PER_THREAD],     ///< [in] Calling thread's input items
         T                       (&output)[ITEMS_PER_THREAD],    ///< [out] Calling thread's output items (may be aliased to \p input)
-        ScanOp                  scan_op,                        ///< [in] Binary scan functor 
+        ScanOp                  scan_op,                        ///< [in] Binary scan functor
         BlockPrefixCallbackOp   &block_prefix_callback_op)      ///< [in-out] <b>[<em>warp</em><sub>0</sub> only]</b> Call-back functor for specifying a block-wide prefix to be applied to the logical input sequence.
     {
         if (ITEMS_PER_THREAD == 1)
diff --git a/cub/cub/block/block_shuffle.cuh b/cub/cub/block/block_shuffle.cuh
index 58938301c11..4118344df7a 100644
--- a/cub/cub/block/block_shuffle.cuh
+++ b/cub/cub/block/block_shuffle.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -34,6 +34,13 @@
 #pragma once
 
 #include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include "../util_ptx.cuh"
 #include "../util_type.cuh"
 
diff --git a/cub/cub/block/block_store.cuh b/cub/cub/block/block_store.cuh
index 2cb6bee4337..2c7c93347ea 100644
--- a/cub/cub/block/block_store.cuh
+++ b/cub/cub/block/block_store.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -33,11 +33,18 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <iterator>
 #include <type_traits>
 
 #include "block_exchange.cuh"
-#include "../config.cuh"
 #include "../util_ptx.cuh"
 #include "../util_type.cuh"
 
@@ -878,7 +885,7 @@ private:
             StoreDirectWarpStriped(linear_tid, block_itr, items, temp_storage.valid_items);
         }
     };
- 
+
 
     /******************************************************************************
      * Type definitions
diff --git a/cub/cub/block/radix_rank_sort_operations.cuh b/cub/cub/block/radix_rank_sort_operations.cuh
index 679dfd4230e..6567cbfac0f 100644
--- a/cub/cub/block/radix_rank_sort_operations.cuh
+++ b/cub/cub/block/radix_rank_sort_operations.cuh
@@ -33,6 +33,14 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/type_traits/integer_sequence.h>
 
 #include <cuda/std/tuple>
@@ -40,7 +48,6 @@
 
 #include <type_traits>
 
-#include "../config.cuh"
 #include "../util_ptx.cuh"
 #include "../util_type.cuh"
 #include "cub/detail/cpp_compatibility.cuh"
@@ -96,14 +103,14 @@ struct BaseDigitExtractor<KeyT, FLOATING_POINT>
  * key from a digit. */
 template <typename KeyT>
 struct BFEDigitExtractor : BaseDigitExtractor<KeyT>
-{   
+{
     using typename BaseDigitExtractor<KeyT>::UnsignedBits;
 
-    std::uint32_t bit_start; 
+    std::uint32_t bit_start;
     std::uint32_t num_bits;
 
     explicit __device__ __forceinline__ BFEDigitExtractor(
-        std::uint32_t bit_start = 0, 
+        std::uint32_t bit_start = 0,
         std::uint32_t num_bits = 0)
       : bit_start(bit_start)
       , num_bits(num_bits)
@@ -139,14 +146,14 @@ struct ShiftDigitExtractor : BaseDigitExtractor<KeyT>
 
 
 #ifndef DOXYGEN_SHOULD_SKIP_THIS    // Do not document
-namespace detail 
+namespace detail
 {
 
 template <bool... Bs>
 struct logic_helper_t;
 
 template <bool>
-struct true_t 
+struct true_t
 {
   static constexpr bool value = true;
 };
@@ -181,9 +188,9 @@ for_each_member_impl(F f, const ::cuda::std::tuple<Ts&...>& tpl)
   static_assert(sizeof...(Ts), "Empty aggregates are not supported");
 
   // Most radix operations are indifferent to the order of operations.
-  // Conversely, the digit extractor traverses fields from the least significant 
-  // to the most significant to imitate bitset printing where higher bits are on 
-  // the left. It also maps to intuition, where something coming first is more 
+  // Conversely, the digit extractor traverses fields from the least significant
+  // to the most significant to imitate bitset printing where higher bits are on
+  // the left. It also maps to intuition, where something coming first is more
   // important. Therefore, we traverse fields on the opposite order.
   for_each_member_impl_helper(f, tpl, THRUST_NS_QUALIFIER::make_reversed_index_sequence<sizeof...(Ts)>{});
 }
@@ -578,7 +585,7 @@ struct traits_t<T, false /* is_fundamental */>
 } // namespace detail
 #endif // DOXYGEN_SHOULD_SKIP_THIS
 
-//! Twiddling keys for radix sort 
+//! Twiddling keys for radix sort
 template <bool IS_DESCENDING, typename KeyT>
 struct RadixSortTwiddle
 {
diff --git a/cub/cub/block/specializations/block_histogram_atomic.cuh b/cub/cub/block/specializations/block_histogram_atomic.cuh
index 93299fa7192..367599b3dc2 100644
--- a/cub/cub/block/specializations/block_histogram_atomic.cuh
+++ b/cub/cub/block/specializations/block_histogram_atomic.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -35,6 +35,12 @@
 
 #include "../../config.cuh"
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 CUB_NAMESPACE_BEGIN
 
 
@@ -57,7 +63,7 @@ struct BlockHistogramAtomic
     /// Composite data onto an existing histogram
     template <
         typename            T,
-        typename            CounterT,     
+        typename            CounterT,
         int                 ITEMS_PER_THREAD>
     __device__ __forceinline__ void Composite(
         T                   (&items)[ITEMS_PER_THREAD],     ///< [in] Calling thread's input values to histogram
diff --git a/cub/cub/block/specializations/block_histogram_sort.cuh b/cub/cub/block/specializations/block_histogram_sort.cuh
index 79659ae106d..4ae46fc9af4 100644
--- a/cub/cub/block/specializations/block_histogram_sort.cuh
+++ b/cub/cub/block/specializations/block_histogram_sort.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -33,9 +33,16 @@
 
 #pragma once
 
+#include "../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include "../../block/block_radix_sort.cuh"
 #include "../../block/block_discontinuity.cuh"
-#include "../../config.cuh"
 #include "../../util_ptx.cuh"
 
 CUB_NAMESPACE_BEGIN
diff --git a/cub/cub/block/specializations/block_reduce_raking.cuh b/cub/cub/block/specializations/block_reduce_raking.cuh
index 423f18f8679..98ab45c794d 100644
--- a/cub/cub/block/specializations/block_reduce_raking.cuh
+++ b/cub/cub/block/specializations/block_reduce_raking.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -33,10 +33,17 @@
 
 #pragma once
 
+#include "../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include "../../block/block_raking_layout.cuh"
 #include "../../warp/warp_reduce.cuh"
 #include "../../thread/thread_reduce.cuh"
-#include "../../config.cuh"
 #include "../../util_ptx.cuh"
 
 CUB_NAMESPACE_BEGIN
diff --git a/cub/cub/block/specializations/block_reduce_raking_commutative_only.cuh b/cub/cub/block/specializations/block_reduce_raking_commutative_only.cuh
index 1fc1caab15a..f4178f31edd 100644
--- a/cub/cub/block/specializations/block_reduce_raking_commutative_only.cuh
+++ b/cub/cub/block/specializations/block_reduce_raking_commutative_only.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -33,10 +33,17 @@
 
 #pragma once
 
+#include "../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include "block_reduce_raking.cuh"
 #include "../../warp/warp_reduce.cuh"
 #include "../../thread/thread_reduce.cuh"
-#include "../../config.cuh"
 #include "../../util_ptx.cuh"
 
 CUB_NAMESPACE_BEGIN
diff --git a/cub/cub/block/specializations/block_reduce_warp_reductions.cuh b/cub/cub/block/specializations/block_reduce_warp_reductions.cuh
index 1e51a9fcfe1..3b68a283bcb 100644
--- a/cub/cub/block/specializations/block_reduce_warp_reductions.cuh
+++ b/cub/cub/block/specializations/block_reduce_warp_reductions.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -33,7 +33,14 @@
 
 #pragma once
 
-#include <cub/config.cuh>
+#include "../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/detail/uninitialized_copy.cuh>
 #include <cub/util_ptx.cuh>
 #include <cub/warp/warp_reduce.cuh>
diff --git a/cub/cub/block/specializations/block_scan_raking.cuh b/cub/cub/block/specializations/block_scan_raking.cuh
index 4891dad141d..b1aad44e04e 100644
--- a/cub/cub/block/specializations/block_scan_raking.cuh
+++ b/cub/cub/block/specializations/block_scan_raking.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -34,7 +34,13 @@
 
 #pragma once
 
-#include <cub/config.cuh>
+#include "../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #include <cub/block/block_raking_layout.cuh>
 #include <cub/detail/uninitialized_copy.cuh>
@@ -156,7 +162,7 @@ struct BlockScanRaking
         CopySegment(out, in, Int2Type<ITERATION + 1>());
     }
 
- 
+
     /// Templated copy (base case)
     __device__ __forceinline__ void CopySegment(
         T*                  /*out*/,            ///< [out] Out array
diff --git a/cub/cub/block/specializations/block_scan_warp_scans.cuh b/cub/cub/block/specializations/block_scan_warp_scans.cuh
index f76131a7856..f48b6dec0a4 100644
--- a/cub/cub/block/specializations/block_scan_warp_scans.cuh
+++ b/cub/cub/block/specializations/block_scan_warp_scans.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -33,7 +33,14 @@
 
 #pragma once
 
-#include <cub/config.cuh>
+#include "../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/detail/uninitialized_copy.cuh>
 #include <cub/util_ptx.cuh>
 #include <cub/warp/warp_scan.cuh>
diff --git a/cub/cub/config.cuh b/cub/cub/config.cuh
index b909bbf7237..ea2591b4447 100644
--- a/cub/cub/config.cuh
+++ b/cub/cub/config.cuh
@@ -32,6 +32,15 @@
 
 #pragma once
 
+// For _CCCL_IMPLICIT_SYSTEM_HEADER
+#include <cuda/__cccl_config>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include "util_arch.cuh"
 #include "util_compiler.cuh"
 #include "util_cpp_dialect.cuh"
diff --git a/cub/cub/cub.cuh b/cub/cub/cub.cuh
index c5cac22cf14..edb7ea0b256 100644
--- a/cub/cub/cub.cuh
+++ b/cub/cub/cub.cuh
@@ -36,6 +36,12 @@
 // Static configuration
 #include "config.cuh"
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // Block
 #include "block/block_adjacent_difference.cuh"
 #include "block/block_discontinuity.cuh"
@@ -101,10 +107,8 @@
 
 // Util
 #include "util_allocator.cuh"
-#include "util_arch.cuh"
 #include "util_debug.cuh"
 #include "util_device.cuh"
-#include "util_macro.cuh"
 #include "util_ptx.cuh"
 #include "util_temporary_storage.cuh"
 #include "util_type.cuh"
diff --git a/cub/cub/detail/choose_offset.cuh b/cub/cub/detail/choose_offset.cuh
index 4154123973a..2894abe01e7 100644
--- a/cub/cub/detail/choose_offset.cuh
+++ b/cub/cub/detail/choose_offset.cuh
@@ -27,7 +27,14 @@
 
 #pragma once
 
-#include <cub/config.cuh>
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 
 #include <cuda/std/iterator>
 #include <cuda/std/type_traits>
diff --git a/cub/cub/detail/cpp_compatibility.cuh b/cub/cub/detail/cpp_compatibility.cuh
index c4fbe649692..ce29f92fa31 100644
--- a/cub/cub/detail/cpp_compatibility.cuh
+++ b/cub/cub/detail/cpp_compatibility.cuh
@@ -17,7 +17,13 @@
 
 #pragma once
 
-#include <cub/util_cpp_dialect.cuh>
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #if CUB_CPP_DIALECT >= 2017 && __cpp_if_constexpr
 #  define CUB_IF_CONSTEXPR if constexpr
diff --git a/cub/cub/detail/detect_cuda_runtime.cuh b/cub/cub/detail/detect_cuda_runtime.cuh
index b8e776db748..0707182bb96 100644
--- a/cub/cub/detail/detect_cuda_runtime.cuh
+++ b/cub/cub/detail/detect_cuda_runtime.cuh
@@ -33,6 +33,15 @@
 
 #pragma once
 
+// We cannot use `cub/config.cuh` here due to circular dependencies
+#include <cuda/__cccl_config>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cuda_runtime_api.h>
 
 #ifdef DOXYGEN_SHOULD_SKIP_THIS // Only parse this during doxygen passes:
diff --git a/cub/cub/detail/device_double_buffer.cuh b/cub/cub/detail/device_double_buffer.cuh
index c427dcb438b..382d0c7bf02 100644
--- a/cub/cub/detail/device_double_buffer.cuh
+++ b/cub/cub/detail/device_double_buffer.cuh
@@ -16,6 +16,14 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/util_namespace.cuh>
 
 
diff --git a/cub/cub/detail/device_synchronize.cuh b/cub/cub/detail/device_synchronize.cuh
index 9da0a361aff..8f015f7a41e 100644
--- a/cub/cub/detail/device_synchronize.cuh
+++ b/cub/cub/detail/device_synchronize.cuh
@@ -16,10 +16,16 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/detail/detect_cuda_runtime.cuh>
 #include <cub/detail/exec_check_disable.cuh>
-#include <cub/util_arch.cuh>
-#include <cub/util_namespace.cuh>
 
 #include <nv/target>
 
diff --git a/cub/cub/detail/exec_check_disable.cuh b/cub/cub/detail/exec_check_disable.cuh
index c5f4b4572c7..bd7d6704d66 100644
--- a/cub/cub/detail/exec_check_disable.cuh
+++ b/cub/cub/detail/exec_check_disable.cuh
@@ -16,7 +16,13 @@
 
 #pragma once
 
-#include <cub/util_compiler.cuh>
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 /**
  * @def CUB_EXEC_CHECK_DISABLE
diff --git a/cub/cub/detail/strong_load.cuh b/cub/cub/detail/strong_load.cuh
index 12e6672b9eb..8d80eaa3466 100644
--- a/cub/cub/detail/strong_load.cuh
+++ b/cub/cub/detail/strong_load.cuh
@@ -32,7 +32,14 @@
 
 #pragma once
 
-#include <cub/config.cuh>
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/util_ptx.cuh>
 #include <cub/util_type.cuh>
 
diff --git a/cub/cub/detail/strong_store.cuh b/cub/cub/detail/strong_store.cuh
index fd293519a38..59530983726 100644
--- a/cub/cub/detail/strong_store.cuh
+++ b/cub/cub/detail/strong_store.cuh
@@ -32,7 +32,14 @@
 
 #pragma once
 
-#include <cub/config.cuh>
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/detail/cpp_compatibility.cuh>
 #include <cub/util_ptx.cuh>
 #include <cub/util_type.cuh>
diff --git a/cub/cub/detail/temporary_storage.cuh b/cub/cub/detail/temporary_storage.cuh
index 51cb3cc855c..22696beb46f 100644
--- a/cub/cub/detail/temporary_storage.cuh
+++ b/cub/cub/detail/temporary_storage.cuh
@@ -16,6 +16,14 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/util_namespace.cuh>
 #include <cub/util_temporary_storage.cuh>
 
diff --git a/cub/cub/detail/type_traits.cuh b/cub/cub/detail/type_traits.cuh
index 77903306250..cc0722c81ae 100644
--- a/cub/cub/detail/type_traits.cuh
+++ b/cub/cub/detail/type_traits.cuh
@@ -32,6 +32,14 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/util_cpp_dialect.cuh>
 #include <cub/util_namespace.cuh>
 
@@ -51,7 +59,7 @@ using invoke_result_t =
 
 /// The type of intermediate accumulator (according to P2322R6)
 template <typename Invokable, typename InitT, typename InputT>
-using accumulator_t = 
+using accumulator_t =
   typename ::cuda::std::decay<invoke_result_t<Invokable, InitT, InputT>>::type;
 
 } // namespace detail
diff --git a/cub/cub/detail/uninitialized_copy.cuh b/cub/cub/detail/uninitialized_copy.cuh
index 2b3e4b1da26..257e5d38639 100644
--- a/cub/cub/detail/uninitialized_copy.cuh
+++ b/cub/cub/detail/uninitialized_copy.cuh
@@ -12,9 +12,9 @@
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
  *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 
- * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 
- * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  * ARE DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE FOR ANY
  * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
@@ -27,7 +27,14 @@
 
 #pragma once
 
-#include <cub/config.cuh>
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 
 #include <cuda/std/type_traits>
 
@@ -48,7 +55,7 @@ __host__ __device__ void uninitialized_copy(T *ptr, U &&val)
 template <typename T,
           typename U,
           typename ::cuda::std::enable_if<
-            ::cuda::std::is_trivially_copyable<T>::value, 
+            ::cuda::std::is_trivially_copyable<T>::value,
             int
           >::type = 0>
 __host__ __device__ void uninitialized_copy(T *ptr, U &&val)
@@ -56,7 +63,7 @@ __host__ __device__ void uninitialized_copy(T *ptr, U &&val)
   *ptr = ::cuda::std::forward<U>(val);
 }
 
-template <typename T, 
+template <typename T,
          typename U,
          typename ::cuda::std::enable_if<
            !::cuda::std::is_trivially_copyable<T>::value,
diff --git a/cub/cub/device/device_adjacent_difference.cuh b/cub/cub/device/device_adjacent_difference.cuh
index 0f614d1876d..e007e3bc9ae 100644
--- a/cub/cub/device/device_adjacent_difference.cuh
+++ b/cub/cub/device/device_adjacent_difference.cuh
@@ -27,7 +27,14 @@
 
 #pragma once
 
-#include <cub/config.cuh>
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/detail/choose_offset.cuh>
 #include <cub/detail/type_traits.cuh>
 #include <cub/device/dispatch/dispatch_adjacent_difference.cuh>
diff --git a/cub/cub/device/device_copy.cuh b/cub/cub/device/device_copy.cuh
index 445f5d862bc..d3d67b5bc5d 100644
--- a/cub/cub/device/device_copy.cuh
+++ b/cub/cub/device/device_copy.cuh
@@ -32,7 +32,14 @@
 
 #pragma once
 
-#include <cub/config.cuh>
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/device/dispatch/dispatch_batch_memcpy.cuh>
 
 #include <thrust/system/cuda/detail/core/triple_chevron_launch.h>
diff --git a/cub/cub/device/device_histogram.cuh b/cub/cub/device/device_histogram.cuh
index 4e71c04ed82..6a8e7323370 100644
--- a/cub/cub/device/device_histogram.cuh
+++ b/cub/cub/device/device_histogram.cuh
@@ -13,9 +13,9 @@
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
  *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 
- * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 
- * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  * ARE DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE FOR ANY
  * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
@@ -27,18 +27,25 @@
  ******************************************************************************/
 
 /**
- * @file cub::DeviceHistogram provides device-wide parallel operations for 
- *       constructing histogram(s) from a sequence of samples data residing 
+ * @file cub::DeviceHistogram provides device-wide parallel operations for
+ *       constructing histogram(s) from a sequence of samples data residing
  *       within device-accessible memory.
  */
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <stdio.h>
 #include <iterator>
 #include <limits>
 
-#include <cub/config.cuh>
 #include <cub/device/dispatch/dispatch_histogram.cuh>
 #include <cub/util_deprecated.cuh>
 
@@ -46,8 +53,8 @@ CUB_NAMESPACE_BEGIN
 
 
 /**
- * @brief DeviceHistogram provides device-wide parallel operations for 
- *        constructing histogram(s) from a sequence of samples data residing 
+ * @brief DeviceHistogram provides device-wide parallel operations for
+ *        constructing histogram(s) from a sequence of samples data residing
  *        within device-accessible memory. ![](histogram_logo.png)
  * @ingroup SingleModule
  *
@@ -67,7 +74,7 @@ struct DeviceHistogram
   //@{
 
   /**
-   * @brief Computes an intensity histogram from a sequence of data samples 
+   * @brief Computes an intensity histogram from a sequence of data samples
    *        using equal-width bins.
    *
    * @par
@@ -97,7 +104,7 @@ struct DeviceHistogram
    * @code
    * #include <cub/cub.cuh> // or equivalently <cub/device/device_histogram.cuh>
    *
-   * // Declare, allocate, and initialize device-accessible pointers for 
+   * // Declare, allocate, and initialize device-accessible pointers for
    * // input samples and output histogram
    * int      num_samples;    // e.g., 10
    * float*   d_samples;      // e.g., [2.2, 6.1, 7.1, 2.9, 3.5, 0.3, 2.9, 2.1, 6.1, 999.5]
@@ -112,7 +119,7 @@ struct DeviceHistogram
    * size_t   temp_storage_bytes = 0;
    * cub::DeviceHistogram::HistogramEven(
    *   d_temp_storage, temp_storage_bytes,
-   *   d_samples, d_histogram, num_levels, 
+   *   d_samples, d_histogram, num_levels,
    *   lower_level, upper_level, num_samples);
    *
    * // Allocate temporary storage
@@ -121,56 +128,56 @@ struct DeviceHistogram
    * // Compute histograms
    * cub::DeviceHistogram::HistogramEven(
    *   d_temp_storage, temp_storage_bytes,
-   *   d_samples, d_histogram, num_levels, 
+   *   d_samples, d_histogram, num_levels,
    *   lower_level, upper_level, num_samples);
    *
    * // d_histogram   <-- [1, 5, 0, 3, 0, 0];
    * @endcode
    *
-   * @tparam SampleIteratorT          
-   *   **[inferred]** Random-access input iterator type for reading input 
+   * @tparam SampleIteratorT
+   *   **[inferred]** Random-access input iterator type for reading input
    *   samples \iterator
    *
-   * @tparam CounterT                 
+   * @tparam CounterT
    *   **[inferred]** Integer type for histogram bin counters
    *
-   * @tparam LevelT                   
+   * @tparam LevelT
    *   **[inferred]** Type for specifying boundaries (levels)
    *
-   * @tparam OffsetT                  
-   *   **[inferred]** Signed integer type for sequence offsets, list lengths, 
+   * @tparam OffsetT
+   *   **[inferred]** Signed integer type for sequence offsets, list lengths,
    *   pointer differences, etc.  \offset_size1
    *
-   * @param[in] d_temp_storage 
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no
    *   work is done.
    *
-   * @param[in,out] temp_storage_bytes 
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of `d_temp_storage` allocation
    *
-   * @param[in] d_samples 
+   * @param[in] d_samples
    *   The pointer to the input sequence of data samples.
    *
-   * @param[out] d_histogram 
-   *   The pointer to the histogram counter output array of length 
+   * @param[out] d_histogram
+   *   The pointer to the histogram counter output array of length
    *   `num_levels - 1`.
    *
-   * @param[in] num_levels 
-   *   The number of boundaries (levels) for delineating histogram samples.  
+   * @param[in] num_levels
+   *   The number of boundaries (levels) for delineating histogram samples.
    *   Implies that the number of bins is `num_levels - 1`.
    *
-   * @param[in] lower_level 
+   * @param[in] lower_level
    *   The lower sample value bound (inclusive) for the lowest histogram bin.
    *
-   * @param[in] upper_level 
+   * @param[in] upper_level
    *   The upper sample value bound (exclusive) for the highest histogram bin.
    *
-   * @param[in] num_samples 
+   * @param[in] num_samples
    *   The number of input samples (i.e., the length of `d_samples`)
    *
-   * @param[in] stream 
-   *   **[optional]** CUDA stream to launch kernels within.  
+   * @param[in] stream
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    */
   template <typename SampleIteratorT,
@@ -240,12 +247,12 @@ struct DeviceHistogram
   }
 
   /**
-   * @brief Computes an intensity histogram from a sequence of data samples 
+   * @brief Computes an intensity histogram from a sequence of data samples
    *        using equal-width bins.
    *
    * @par
-   * - A two-dimensional *region of interest* within `d_samples` can be 
-   *   specified using the `num_row_samples`, `num_rows`, and 
+   * - A two-dimensional *region of interest* within `d_samples` can be
+   *   specified using the `num_row_samples`, `num_rows`, and
    *   `row_stride_bytes` parameters.
    * - The row stride must be a whole multiple of the sample data type
    *   size, i.e., `(row_stride_bytes % sizeof(SampleT)) == 0`.
@@ -259,10 +266,10 @@ struct DeviceHistogram
    *   `uint64_t`, the cuda error `cudaErrorInvalidValue` is returned. If the common type is 128
    *   bits wide, bin computation will use 128-bit arithmetic and `cudaErrorInvalidValue` will only
    *   be returned if bin computation would overflow for 128-bit arithmetic.
-   * - For a given row `r` in `[0, num_rows)`, let 
-   *   `row_begin = d_samples + r * row_stride_bytes / sizeof(SampleT)` and 
+   * - For a given row `r` in `[0, num_rows)`, let
+   *   `row_begin = d_samples + r * row_stride_bytes / sizeof(SampleT)` and
    *   `row_end = row_begin + num_row_samples`. The ranges
-   *   `[row_begin, row_end)` and `[d_histogram, d_histogram + num_levels - 1)` 
+   *   `[row_begin, row_end)` and `[d_histogram, d_histogram + num_levels - 1)`
    *   shall not overlap in any way.
    * - `cuda::std::common_type<LevelT, SampleT>` must be valid, and both LevelT
    *   and SampleT must be valid arithmetic types. The common type must be
@@ -277,7 +284,7 @@ struct DeviceHistogram
    * @code
    * #include <cub/cub.cuh> // or equivalently <cub/device/device_histogram.cuh>
    *
-   * // Declare, allocate, and initialize device-accessible pointers for 
+   * // Declare, allocate, and initialize device-accessible pointers for
    * // input samples and output histogram
    * int      num_row_samples;    // e.g., 5
    * int      num_rows;           // e.g., 2;
@@ -310,57 +317,57 @@ struct DeviceHistogram
    * // d_histogram   <-- [1, 5, 0, 3, 0, 0];
    * @endcode
    *
-   * @tparam SampleIteratorT          
-   *   **[inferred]** Random-access input iterator type for reading 
+   * @tparam SampleIteratorT
+   *   **[inferred]** Random-access input iterator type for reading
    *   input samples. \iterator
    *
-   * @tparam CounterT                 
+   * @tparam CounterT
    *   **[inferred]** Integer type for histogram bin counters
    *
-   * @tparam LevelT                   
+   * @tparam LevelT
    *   **[inferred]** Type for specifying boundaries (levels)
    *
-   * @tparam OffsetT                  
+   * @tparam OffsetT
    *   **[inferred]** Signed integer type for sequence offsets, list lengths,
    *   pointer differences, etc. \offset_size1
 
-   * @param[in] d_temp_storage 
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no
    *   work is done.
    *
-   * @param[in,out] temp_storage_bytes 
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of `d_temp_storage` allocation
    *
-   * @param[in] d_samples 
+   * @param[in] d_samples
    *   The pointer to the input sequence of data samples.
    *
-   * @param[out] d_histogram 
-   *   The pointer to the histogram counter output array of 
+   * @param[out] d_histogram
+   *   The pointer to the histogram counter output array of
    *   length `num_levels - 1`.
    *
-   * @param[in] num_levels 
-   *   The number of boundaries (levels) for delineating histogram samples.  
+   * @param[in] num_levels
+   *   The number of boundaries (levels) for delineating histogram samples.
    *   Implies that the number of bins is `num_levels - 1`.
    *
-   * @param[in] lower_level 
+   * @param[in] lower_level
    *   The lower sample value bound (inclusive) for the lowest histogram bin.
    *
-   * @param[in] upper_level 
+   * @param[in] upper_level
    *   The upper sample value bound (exclusive) for the highest histogram bin.
    *
-   * @param[in] num_row_samples 
+   * @param[in] num_row_samples
    *   The number of data samples per row in the region of interest
    *
-   * @param[in] num_rows 
+   * @param[in] num_rows
    *   The number of rows in the region of interest
    *
-   * @param[in] row_stride_bytes 
-   *   The number of bytes between starts of consecutive rows in 
+   * @param[in] row_stride_bytes
+   *   The number of bytes between starts of consecutive rows in
    *   the region of interest
    *
-   * @param[in] stream 
-   *   **[optional]** CUDA stream to launch kernels within.  
+   * @param[in] stream
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    */
   template <typename SampleIteratorT,
@@ -433,20 +440,20 @@ struct DeviceHistogram
   }
 
   /**
-   * @brief Computes per-channel intensity histograms from a sequence of 
+   * @brief Computes per-channel intensity histograms from a sequence of
    *        multi-channel "pixel" data samples using equal-width bins.
    *
    * @par
    * - The input is a sequence of *pixel* structures, where each pixel comprises
-   *   a record of `NUM_CHANNELS` consecutive data samples 
+   *   a record of `NUM_CHANNELS` consecutive data samples
    *   (e.g., an *RGBA* pixel).
-   * - Of the `NUM_CHANNELS` specified, the function will only compute 
-   *   histograms for the first `NUM_ACTIVE_CHANNELS` 
+   * - Of the `NUM_CHANNELS` specified, the function will only compute
+   *   histograms for the first `NUM_ACTIVE_CHANNELS`
    *   (e.g., only *RGB* histograms from *RGBA* pixel samples).
-   * - The number of histogram bins for channel<sub><em>i</em></sub> is 
+   * - The number of histogram bins for channel<sub><em>i</em></sub> is
    *   `num_levels[i] - 1`.
    * - For channel<sub><em>i</em></sub>, the range of values for all histogram bins
-   *   have the same width: 
+   *   have the same width:
    *   `(upper_level[i] - lower_level[i]) / (num_levels[i] - 1)`
    * - If the common type of sample and level is of integral type, the bin for a sample is
    *   computed as `(sample - lower_level[i]) * (num_levels - 1) / (upper_level[i] -
@@ -473,7 +480,7 @@ struct DeviceHistogram
    * @code
    * #include <cub/cub.cuh> // or equivalently <cub/device/device_histogram.cuh>
    *
-   * // Declare, allocate, and initialize device-accessible pointers for 
+   * // Declare, allocate, and initialize device-accessible pointers for
    * // input samples and output histograms
    * int              num_pixels;         // e.g., 5
    * unsigned char*   d_samples;          // e.g., [(2, 6, 7, 5), (3, 0, 2, 1), (7, 0, 6, 2),
@@ -490,7 +497,7 @@ struct DeviceHistogram
    * size_t   temp_storage_bytes = 0;
    * cub::DeviceHistogram::MultiHistogramEven<4, 3>(
    *   d_temp_storage, temp_storage_bytes,
-   *   d_samples, d_histogram, num_levels, 
+   *   d_samples, d_histogram, num_levels,
    *   lower_level, upper_level, num_pixels);
    *
    * // Allocate temporary storage
@@ -499,7 +506,7 @@ struct DeviceHistogram
    * // Compute histograms
    * cub::DeviceHistogram::MultiHistogramEven<4, 3>(
    *   d_temp_storage, temp_storage_bytes,
-   *   d_samples, d_histogram, num_levels, 
+   *   d_samples, d_histogram, num_levels,
    *   lower_level, upper_level, num_pixels);
    *
    * // d_histogram   <-- [ [1, 0, 1, 2, 0, 0, 0, 1, 0, 0, 0, ..., 0],
@@ -507,65 +514,65 @@ struct DeviceHistogram
    * //                     [0, 0, 2, 0, 0, 0, 1, 2, 0, 0, 0, ..., 0] ]
    * @endcode
    *
-   * @tparam NUM_CHANNELS             
-   *   Number of channels interleaved in the input data (may be greater than 
+   * @tparam NUM_CHANNELS
+   *   Number of channels interleaved in the input data (may be greater than
    *   the number of channels being actively histogrammed)
    *
-   * @tparam NUM_ACTIVE_CHANNELS      
+   * @tparam NUM_ACTIVE_CHANNELS
    *   **[inferred]** Number of channels actively being histogrammed
    *
-   * @tparam SampleIteratorT          
-   *   **[inferred]** Random-access input iterator type for reading 
+   * @tparam SampleIteratorT
+   *   **[inferred]** Random-access input iterator type for reading
    *   input samples. \iterator
    *
-   * @tparam CounterT                 
+   * @tparam CounterT
    *   **[inferred]** Integer type for histogram bin counters
    *
-   * @tparam LevelT                   
+   * @tparam LevelT
    *   **[inferred]** Type for specifying boundaries (levels)
    *
-   * @tparam OffsetT                  
-   *   **[inferred]** Signed integer type for sequence offsets, list lengths, 
+   * @tparam OffsetT
+   *   **[inferred]** Signed integer type for sequence offsets, list lengths,
    *   pointer differences, etc. \offset_size1
    *
-   * @param[in] d_temp_storage 
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no
    *   work is done.
    *
-   * @param[in,out] temp_storage_bytes 
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of `d_temp_storage` allocation
    *
-   * @param[in] d_samples 
-   *   The pointer to the multi-channel input sequence of data samples. 
-   *   The samples from different channels are assumed to be interleaved 
-   *   (e.g., an array of 32-bit pixels where each pixel consists of four 
+   * @param[in] d_samples
+   *   The pointer to the multi-channel input sequence of data samples.
+   *   The samples from different channels are assumed to be interleaved
+   *   (e.g., an array of 32-bit pixels where each pixel consists of four
    *   *RGBA* 8-bit samples).
    *
    * @param[out] d_histogram
-   *   The pointers to the histogram counter output arrays, one for each active 
-   *   channel. For channel<sub><em>i</em></sub>, the allocation length of 
+   *   The pointers to the histogram counter output arrays, one for each active
+   *   channel. For channel<sub><em>i</em></sub>, the allocation length of
    *   `d_histogram[i]` should be `num_levels[i] - 1`.
    *
    * @param[in] num_levels
-   *   The number of boundaries (levels) for delineating histogram samples in 
-   *   each active channel. Implies that the number of bins for 
+   *   The number of boundaries (levels) for delineating histogram samples in
+   *   each active channel. Implies that the number of bins for
    *   channel<sub><em>i</em></sub> is `num_levels[i] - 1`.
    *
    * @param[in] lower_level
-   *   The lower sample value bound (inclusive) for the lowest histogram bin in 
+   *   The lower sample value bound (inclusive) for the lowest histogram bin in
    *   each active channel.
    *
    * @param[in] upper_level
-   *   The upper sample value bound (exclusive) for the highest histogram bin 
+   *   The upper sample value bound (exclusive) for the highest histogram bin
    *   in each active channel.
    *
-   * @param[in] num_pixels 
-   *   The number of multi-channel pixels 
+   * @param[in] num_pixels
+   *   The number of multi-channel pixels
    *   (i.e., the length of `d_samples / NUM_CHANNELS`)
    *
-   * @param[in] stream 
-   *   **[optional]** CUDA stream to launch kernels within.  
+   * @param[in] stream
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    */
   template <int NUM_CHANNELS,
@@ -635,25 +642,25 @@ struct DeviceHistogram
   }
 
   /**
-   * @brief Computes per-channel intensity histograms from a sequence of 
+   * @brief Computes per-channel intensity histograms from a sequence of
    *        multi-channel "pixel" data samples using equal-width bins.
    *
    * @par
-   * - The input is a sequence of *pixel* structures, where each pixel 
-   *   comprises a record of `NUM_CHANNELS` consecutive data samples 
+   * - The input is a sequence of *pixel* structures, where each pixel
+   *   comprises a record of `NUM_CHANNELS` consecutive data samples
    *   (e.g., an *RGBA* pixel).
-   * - Of the `NUM_CHANNELS` specified, the function will only compute 
-   *   histograms for the first `NUM_ACTIVE_CHANNELS` (e.g., only *RGB* 
+   * - Of the `NUM_CHANNELS` specified, the function will only compute
+   *   histograms for the first `NUM_ACTIVE_CHANNELS` (e.g., only *RGB*
    *   histograms from *RGBA* pixel samples).
-   * - A two-dimensional *region of interest* within `d_samples` can be 
-   *   specified using the `num_row_samples`, `num_rows`, and 
+   * - A two-dimensional *region of interest* within `d_samples` can be
+   *   specified using the `num_row_samples`, `num_rows`, and
    *   `row_stride_bytes` parameters.
    * - The row stride must be a whole multiple of the sample data type
    *   size, i.e., `(row_stride_bytes % sizeof(SampleT)) == 0`.
-   * - The number of histogram bins for channel<sub><em>i</em></sub> is 
+   * - The number of histogram bins for channel<sub><em>i</em></sub> is
    *   `num_levels[i] - 1`.
-   * - For channel<sub><em>i</em></sub>, the range of values for all histogram 
-   *   bins have the same width: 
+   * - For channel<sub><em>i</em></sub>, the range of values for all histogram
+   *   bins have the same width:
    *   `(upper_level[i] - lower_level[i]) / (num_levels[i] - 1)`
    * - If the common type of sample and level is of integral type, the bin for a sample is
    *   computed as `(sample - lower_level[i]) * (num_levels - 1) / (upper_level[i] -
@@ -663,14 +670,14 @@ struct DeviceHistogram
    *   `cudaErrorInvalidValue` is returned. If the common type is 128 bits wide, bin computation
    *   will use 128-bit arithmetic and `cudaErrorInvalidValue` will only be returned if bin
    *   computation would overflow for 128-bit arithmetic.
-   * - For a given row `r` in `[0, num_rows)`, and sample `s` in 
-   *   `[0, num_row_pixels)`, let 
-   *   `row_begin = d_samples + r * row_stride_bytes / sizeof(SampleT)`, 
+   * - For a given row `r` in `[0, num_rows)`, and sample `s` in
+   *   `[0, num_row_pixels)`, let
+   *   `row_begin = d_samples + r * row_stride_bytes / sizeof(SampleT)`,
    *   `sample_begin = row_begin + s * NUM_CHANNELS`, and
    *   `sample_end = sample_begin + NUM_ACTIVE_CHANNELS`. For a given channel
-   *    `c` in `[0, NUM_ACTIVE_CHANNELS)`, the ranges 
-   *   `[sample_begin, sample_end)` and 
-   *   `[d_histogram[c], d_histogram[c] + num_levels[c] - 1)` shall not overlap 
+   *    `c` in `[0, NUM_ACTIVE_CHANNELS)`, the ranges
+   *   `[sample_begin, sample_end)` and
+   *   `[d_histogram[c], d_histogram[c] + num_levels[c] - 1)` shall not overlap
    *   in any way.
    * - `cuda::std::common_type<LevelT, SampleT>` must be valid, and both LevelT
    *   and SampleT must be valid arithmetic types. The common type must be
@@ -678,15 +685,15 @@ struct DeviceHistogram
    * - @devicestorage
    *
    * @par Snippet
-   * The code snippet below illustrates the computation of three 256-bin 
-   * *RGB* histograms from a 2x3 region of interest of within a flattened 2x4 
+   * The code snippet below illustrates the computation of three 256-bin
+   * *RGB* histograms from a 2x3 region of interest of within a flattened 2x4
    * array of quad-channel *RGBA* pixels (8 bits per channel per pixel).
    *
    * @par
    * @code
    * #include <cub/cub.cuh> // or equivalently <cub/device/device_histogram.cuh>
    *
-   * // Declare, allocate, and initialize device-accessible pointers for input 
+   * // Declare, allocate, and initialize device-accessible pointers for input
    * // samples and output histograms
    * int              num_row_pixels;     // e.g., 3
    * int              num_rows;           // e.g., 2
@@ -722,71 +729,71 @@ struct DeviceHistogram
    * //                     [0, 1, 2, 0, 0, 0, 1, 2, 0, 0, 0, ..., 0] ]
    * @endcode
    *
-   * @tparam NUM_CHANNELS             
-   *   Number of channels interleaved in the input data (may be greater than 
+   * @tparam NUM_CHANNELS
+   *   Number of channels interleaved in the input data (may be greater than
    *   the number of channels being actively histogrammed)
    *
-   * @tparam NUM_ACTIVE_CHANNELS      
+   * @tparam NUM_ACTIVE_CHANNELS
    *   **[inferred]** Number of channels actively being histogrammed
    *
-   * @tparam SampleIteratorT          
-   *   **[inferred]** Random-access input iterator type for reading input 
+   * @tparam SampleIteratorT
+   *   **[inferred]** Random-access input iterator type for reading input
    *   samples. \iterator
    *
-   * @tparam CounterT                 
+   * @tparam CounterT
    *   **[inferred]** Integer type for histogram bin counters
    *
-   * @tparam LevelT                   
+   * @tparam LevelT
    *   **[inferred]** Type for specifying boundaries (levels)
    *
-   * @tparam OffsetT                  
-   *   **[inferred]** Signed integer type for sequence offsets, list lengths, 
+   * @tparam OffsetT
+   *   **[inferred]** Signed integer type for sequence offsets, list lengths,
    *   pointer differences, etc. \offset_size1
    *
-   * @param[in] d_temp_storage 
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no
    *   work is done.
    *
-   * @param[in,out] temp_storage_bytes 
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of `d_temp_storage` allocation
    *
-   * @param[in] d_samples 
-   *   The pointer to the multi-channel input sequence of data samples. The 
-   *   samples from different channels are assumed to be interleaved (e.g., 
-   *   an array of 32-bit pixels where each pixel consists of four 
+   * @param[in] d_samples
+   *   The pointer to the multi-channel input sequence of data samples. The
+   *   samples from different channels are assumed to be interleaved (e.g.,
+   *   an array of 32-bit pixels where each pixel consists of four
    *   *RGBA* 8-bit samples).
    *
-   * @param[out] d_histogram 
-   *   The pointers to the histogram counter output arrays, one for each 
-   *   active channel. For channel<sub><em>i</em></sub>, the allocation length 
+   * @param[out] d_histogram
+   *   The pointers to the histogram counter output arrays, one for each
+   *   active channel. For channel<sub><em>i</em></sub>, the allocation length
    *   of `d_histogram[i]` should be `num_levels[i] - 1`.
    *
-   * @param[in] num_levels 
-   *   The number of boundaries (levels) for delineating histogram samples in 
-   *   each active channel. Implies that the number of bins for 
+   * @param[in] num_levels
+   *   The number of boundaries (levels) for delineating histogram samples in
+   *   each active channel. Implies that the number of bins for
    *   channel<sub><em>i</em></sub> is `num_levels[i] - 1`.
    *
-   * @param[in] lower_level 
-   *   The lower sample value bound (inclusive) for the lowest histogram bin in 
+   * @param[in] lower_level
+   *   The lower sample value bound (inclusive) for the lowest histogram bin in
    *   each active channel.
    *
-   * @param[in] upper_level 
-   *   The upper sample value bound (exclusive) for the highest histogram bin 
+   * @param[in] upper_level
+   *   The upper sample value bound (exclusive) for the highest histogram bin
    *   in each active channel.
    *
-   * @param[in] num_row_pixels 
+   * @param[in] num_row_pixels
    *   The number of multi-channel pixels per row in the region of interest
    *
-   * @param[in] num_rows 
+   * @param[in] num_rows
    *   The number of rows in the region of interest
    *
-   * @param[in] row_stride_bytes 
-   *   The number of bytes between starts of consecutive rows in the region of 
+   * @param[in] row_stride_bytes
+   *   The number of bytes between starts of consecutive rows in the region of
    *   interest
    *
-   * @param[in] stream 
-   *   **[optional]** CUDA stream to launch kernels within.  
+   * @param[in] stream
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    */
   template <int NUM_CHANNELS,
@@ -900,16 +907,16 @@ struct DeviceHistogram
   //@{
 
   /**
-   * @brief Computes an intensity histogram from a sequence of data samples 
+   * @brief Computes an intensity histogram from a sequence of data samples
    *        using the specified bin boundary levels.
    *
    * @par
    * - The number of histogram bins is (`num_levels - 1`)
    * - The value range for bin<sub><em>i</em></sub> is `[level[i], level[i+1])`
-   * - The range `[d_histogram, d_histogram + num_levels - 1)` shall not 
-   *   overlap `[d_samples, d_samples + num_samples)` nor 
-   *   `[d_levels, d_levels + num_levels)` in any way. The ranges 
-   *   `[d_levels, d_levels + num_levels)` and 
+   * - The range `[d_histogram, d_histogram + num_levels - 1)` shall not
+   *   overlap `[d_samples, d_samples + num_samples)` nor
+   *   `[d_levels, d_levels + num_levels)` in any way. The ranges
+   *   `[d_levels, d_levels + num_levels)` and
    *   `[d_samples, d_samples + num_samples)` may overlap.
    * - @devicestorage
    *
@@ -921,7 +928,7 @@ struct DeviceHistogram
    * @code
    * #include <cub/cub.cuh> // or equivalently <cub/device/device_histogram.cuh>
    *
-   * // Declare, allocate, and initialize device-accessible pointers for input 
+   * // Declare, allocate, and initialize device-accessible pointers for input
    * // samples and output histogram
    * int      num_samples;    // e.g., 10
    * float*   d_samples;      // e.g., [2.2, 6.0, 7.1, 2.9, 3.5, 0.3, 2.9, 2.0, 6.1, 999.5]
@@ -949,49 +956,49 @@ struct DeviceHistogram
    *
    * @endcode
    *
-   * @tparam SampleIteratorT          
-   *   **[inferred]** Random-access input iterator type for reading 
+   * @tparam SampleIteratorT
+   *   **[inferred]** Random-access input iterator type for reading
    *   input samples.\iterator
    *
-   * @tparam CounterT                 
+   * @tparam CounterT
    *   **[inferred]** Integer type for histogram bin counters
    *
-   * @tparam LevelT                   
+   * @tparam LevelT
    *   **[inferred]** Type for specifying boundaries (levels)
    *
-   * @tparam OffsetT                  
-   *   **[inferred]** Signed integer type for sequence offsets, list lengths, 
+   * @tparam OffsetT
+   *   **[inferred]** Signed integer type for sequence offsets, list lengths,
    *   pointer differences, etc. \offset_size1
    *
-   * @param[in] d_temp_storage 
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no work 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no work
    *   is done.
    *
-   * @param[in,out] temp_storage_bytes 
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of `d_temp_storage` allocation
    *
-   * @param[in] d_samples 
+   * @param[in] d_samples
    *   The pointer to the input sequence of data samples.
    *
-   * @param[out] d_histogram 
-   *   The pointer to the histogram counter output array of length 
+   * @param[out] d_histogram
+   *   The pointer to the histogram counter output array of length
    *   `num_levels - 1`.
    *
-   * @param[in] num_levels 
-   *   The number of boundaries (levels) for delineating histogram samples.  
+   * @param[in] num_levels
+   *   The number of boundaries (levels) for delineating histogram samples.
    *   Implies that the number of bins is `num_levels - 1`.
    *
-   * @param[in] d_levels 
-   *   The pointer to the array of boundaries (levels). Bin ranges are defined 
-   *   by consecutive boundary pairings: lower sample value boundaries are 
+   * @param[in] d_levels
+   *   The pointer to the array of boundaries (levels). Bin ranges are defined
+   *   by consecutive boundary pairings: lower sample value boundaries are
    *   inclusive and upper sample value boundaries are exclusive.
    *
-   * @param[in] num_samples 
+   * @param[in] num_samples
    *   The number of data samples per row in the region of interest
    *
-   * @param[in] stream 
-   *   **[optional]** CUDA stream to launch kernels within.  
+   * @param[in] stream
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    */
   template <typename SampleIteratorT,
@@ -1056,19 +1063,19 @@ struct DeviceHistogram
   }
 
   /**
-   * @brief Computes an intensity histogram from a sequence of data samples 
+   * @brief Computes an intensity histogram from a sequence of data samples
    *        using the specified bin boundary levels.
    *
    * @par
-   * - A two-dimensional *region of interest* within `d_samples` can be 
-   *   specified using the `num_row_samples`, `num_rows`, and 
+   * - A two-dimensional *region of interest* within `d_samples` can be
+   *   specified using the `num_row_samples`, `num_rows`, and
    *   `row_stride_bytes` parameters.
    * - The row stride must be a whole multiple of the sample data type
    *   size, i.e., `(row_stride_bytes % sizeof(SampleT)) == 0`.
    * - The number of histogram bins is (`num_levels - 1`)
    * - The value range for bin<sub><em>i</em></sub> is `[level[i], level[i+1])`
-   * - For a given row `r` in `[0, num_rows)`, let 
-   *   `row_begin = d_samples + r * row_stride_bytes / sizeof(SampleT)` and 
+   * - For a given row `r` in `[0, num_rows)`, let
+   *   `row_begin = d_samples + r * row_stride_bytes / sizeof(SampleT)` and
    *   `row_end = row_begin + num_row_samples`. The range
    *   `[d_histogram, d_histogram + num_levels - 1)` shall not overlap
    *   `[row_begin, row_end)` nor `[d_levels, d_levels + num_levels)`.
@@ -1116,55 +1123,55 @@ struct DeviceHistogram
    * // d_histogram   <-- [1, 5, 0, 3, 0, 0];
    * @endcode
    *
-   * @tparam SampleIteratorT          
-   *   **[inferred]** Random-access input iterator type for reading 
+   * @tparam SampleIteratorT
+   *   **[inferred]** Random-access input iterator type for reading
    *   input samples. \iterator
-   * 
-   * @tparam CounterT                 
+   *
+   * @tparam CounterT
    *   **[inferred]** Integer type for histogram bin counters
-   * 
-   * @tparam LevelT                   
+   *
+   * @tparam LevelT
    *   **[inferred]** Type for specifying boundaries (levels)
-   * 
-   * @tparam OffsetT                  
-   *   **[inferred]** Signed integer type for sequence offsets, list lengths, 
+   *
+   * @tparam OffsetT
+   *   **[inferred]** Signed integer type for sequence offsets, list lengths,
    *   pointer differences, etc. \offset_size1
    *
-   * @param[in] d_temp_storage 
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no
    *   work is done.
    *
-   * @param[in,out] temp_storage_bytes 
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of `d_temp_storage` allocation
    *
-   * @param[in] d_samples 
+   * @param[in] d_samples
    *   The pointer to the input sequence of data samples.
    *
-   * @param[out] d_histogram 
-   *   The pointer to the histogram counter output array of length 
+   * @param[out] d_histogram
+   *   The pointer to the histogram counter output array of length
    *   `num_levels - 1`.
    *
-   * @param[in] num_levels 
-   *   The number of boundaries (levels) for delineating histogram samples.  
+   * @param[in] num_levels
+   *   The number of boundaries (levels) for delineating histogram samples.
    *   Implies that the number of bins is `num_levels - 1`.
    *
-   * @param[in] d_levels 
-   *   The pointer to the array of boundaries (levels). Bin ranges are defined 
-   *   by consecutive boundary pairings: lower sample value boundaries are 
+   * @param[in] d_levels
+   *   The pointer to the array of boundaries (levels). Bin ranges are defined
+   *   by consecutive boundary pairings: lower sample value boundaries are
    *   inclusive and upper sample value boundaries are exclusive.
    *
-   * @param[in] num_row_samples 
+   * @param[in] num_row_samples
    *   The number of data samples per row in the region of interest
    *
-   * @param[in] num_rows 
+   * @param[in] num_rows
    *   The number of rows in the region of interest
    *
-   * @param[in] row_stride_bytes 
-   *   The number of bytes between starts of consecutive rows in the region 
+   * @param[in] row_stride_bytes
+   *   The number of bytes between starts of consecutive rows in the region
    *   of interest
    *
-   * @param[in] stream 
+   * @param[in] stream
    *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    */
@@ -1233,24 +1240,24 @@ struct DeviceHistogram
   }
 
   /**
-   * @brief Computes per-channel intensity histograms from a sequence of 
-   *        multi-channel "pixel" data samples using the specified bin 
+   * @brief Computes per-channel intensity histograms from a sequence of
+   *        multi-channel "pixel" data samples using the specified bin
    *        boundary levels.
    *
    * @par
-   * - The input is a sequence of *pixel* structures, where each pixel 
-   *   comprises a record of `NUM_CHANNELS` consecutive data samples 
+   * - The input is a sequence of *pixel* structures, where each pixel
+   *   comprises a record of `NUM_CHANNELS` consecutive data samples
    *   (e.g., an *RGBA* pixel).
-   * - Of the `NUM_CHANNELS` specified, the function will only compute 
-   *   histograms for the first `NUM_ACTIVE_CHANNELS` (e.g., *RGB* histograms 
+   * - Of the `NUM_CHANNELS` specified, the function will only compute
+   *   histograms for the first `NUM_ACTIVE_CHANNELS` (e.g., *RGB* histograms
    *   from *RGBA* pixel samples).
-   * - The number of histogram bins for channel<sub><em>i</em></sub> is 
+   * - The number of histogram bins for channel<sub><em>i</em></sub> is
    *   `num_levels[i] - 1`.
-   * - For channel<sub><em>i</em></sub>, the range of values for all histogram 
-   *   bins have the same width: 
+   * - For channel<sub><em>i</em></sub>, the range of values for all histogram
+   *   bins have the same width:
    *   `(upper_level[i] - lower_level[i]) / (num_levels[i] - 1)`
-   * - For given channels `c1` and `c2` in `[0, NUM_ACTIVE_CHANNELS)`, the 
-   *   range `[d_histogram[c1], d_histogram[c1] + num_levels[c1] - 1)` shall 
+   * - For given channels `c1` and `c2` in `[0, NUM_ACTIVE_CHANNELS)`, the
+   *   range `[d_histogram[c1], d_histogram[c1] + num_levels[c1] - 1)` shall
    *   not overlap `[d_samples, d_samples + NUM_CHANNELS * num_pixels)` nor
    *   `[d_levels[c2], d_levels[c2] + num_levels[c2])` in any way.
    *   The ranges `[d_levels[c2], d_levels[c2] + num_levels[c2])` and
@@ -1258,15 +1265,15 @@ struct DeviceHistogram
    * - @devicestorage
    *
    * @par Snippet
-   * The code snippet below illustrates the computation of three 4-bin *RGB* 
-   * histograms from a quad-channel sequence of *RGBA* pixels 
+   * The code snippet below illustrates the computation of three 4-bin *RGB*
+   * histograms from a quad-channel sequence of *RGBA* pixels
    * (8 bits per channel per pixel)
    *
    * @par
    * @code
    * #include <cub/cub.cuh> // or equivalently <cub/device/device_histogram.cuh>
    *
-   * // Declare, allocate, and initialize device-accessible pointers for 
+   * // Declare, allocate, and initialize device-accessible pointers for
    * // input samples and output histograms
    * int            num_pixels;       // e.g., 5
    * unsigned char  *d_samples;       // e.g., [(2, 6, 7, 5),(3, 0, 2, 1),(7, 0, 6, 2),
@@ -1299,63 +1306,63 @@ struct DeviceHistogram
    *
    * @endcode
    *
-   * @tparam NUM_CHANNELS             
-   *   Number of channels interleaved in the input data (may be greater than 
+   * @tparam NUM_CHANNELS
+   *   Number of channels interleaved in the input data (may be greater than
    *   the number of channels being actively histogrammed)
-   * 
-   * @tparam NUM_ACTIVE_CHANNELS      
+   *
+   * @tparam NUM_ACTIVE_CHANNELS
    *   **[inferred]** Number of channels actively being histogrammed
-   * 
-   * @tparam SampleIteratorT          
-   *   **[inferred]** Random-access input iterator type for reading 
+   *
+   * @tparam SampleIteratorT
+   *   **[inferred]** Random-access input iterator type for reading
    *   input samples. \iterator
-   * 
-   * @tparam CounterT                 
+   *
+   * @tparam CounterT
    *   **[inferred]** Integer type for histogram bin counters
-   * 
-   * @tparam LevelT                   
+   *
+   * @tparam LevelT
    *   **[inferred]** Type for specifying boundaries (levels)
-   * 
-   * @tparam OffsetT                  
-   *   **[inferred]** Signed integer type for sequence offsets, list lengths, 
+   *
+   * @tparam OffsetT
+   *   **[inferred]** Signed integer type for sequence offsets, list lengths,
    *   pointer differences, etc. \offset_size1
    *
-   * @param[in] d_temp_storage 
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no
    *   work is done.
    *
-   * @param[in,out] temp_storage_bytes 
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of `d_temp_storage` allocation
    *
-   * @param[in] d_samples 
-   *   The pointer to the multi-channel input sequence of data samples. 
-   *   The samples from different channels are assumed to be interleaved (e.g., 
-   *   an array of 32-bit pixels where each pixel consists of four *RGBA* 
+   * @param[in] d_samples
+   *   The pointer to the multi-channel input sequence of data samples.
+   *   The samples from different channels are assumed to be interleaved (e.g.,
+   *   an array of 32-bit pixels where each pixel consists of four *RGBA*
    *   8-bit samples).
    *
-   * @param[out] d_histogram 
-   *   The pointers to the histogram counter output arrays, one for each active 
-   *   channel. For channel<sub><em>i</em></sub>, the allocation length of 
+   * @param[out] d_histogram
+   *   The pointers to the histogram counter output arrays, one for each active
+   *   channel. For channel<sub><em>i</em></sub>, the allocation length of
    *   `d_histogram[i]` should be `num_levels[i] - 1`.
    *
-   * @param[in] num_levels 
-   *   The number of boundaries (levels) for delineating histogram samples in 
-   *   each active channel. Implies that the number of bins for 
+   * @param[in] num_levels
+   *   The number of boundaries (levels) for delineating histogram samples in
+   *   each active channel. Implies that the number of bins for
    *   channel<sub><em>i</em></sub> is `num_levels[i] - 1`.
    *
-   * @param[in] d_levels 
-   *   The pointers to the arrays of boundaries (levels), one for each active 
-   *   channel. Bin ranges are defined by consecutive boundary pairings: lower 
-   *   sample value boundaries are inclusive and upper sample value boundaries 
+   * @param[in] d_levels
+   *   The pointers to the arrays of boundaries (levels), one for each active
+   *   channel. Bin ranges are defined by consecutive boundary pairings: lower
+   *   sample value boundaries are inclusive and upper sample value boundaries
    *   are exclusive.
    *
-   * @param[in] num_pixels 
-   *   The number of multi-channel pixels 
+   * @param[in] num_pixels
+   *   The number of multi-channel pixels
    *   (i.e., the length of `d_samples / NUM_CHANNELS`)
    *
-   * @param[in] stream 
-   *   **[optional]** CUDA stream to launch kernels within.  
+   * @param[in] stream
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    */
   template <int NUM_CHANNELS,
@@ -1421,50 +1428,50 @@ struct DeviceHistogram
   }
 
   /**
-   * @brief Computes per-channel intensity histograms from a sequence of 
-   *        multi-channel "pixel" data samples using the specified bin boundary 
+   * @brief Computes per-channel intensity histograms from a sequence of
+   *        multi-channel "pixel" data samples using the specified bin boundary
    *        levels.
    *
    * @par
    * - The input is a sequence of *pixel* structures, where each pixel comprises
-   *   a record of `NUM_CHANNELS` consecutive data samples 
+   *   a record of `NUM_CHANNELS` consecutive data samples
    *   (e.g., an *RGBA* pixel).
-   * - Of the `NUM_CHANNELS` specified, the function will only compute 
-   *   histograms for the first `NUM_ACTIVE_CHANNELS` (e.g., *RGB* histograms 
+   * - Of the `NUM_CHANNELS` specified, the function will only compute
+   *   histograms for the first `NUM_ACTIVE_CHANNELS` (e.g., *RGB* histograms
    *   from *RGBA* pixel samples).
-   * - A two-dimensional *region of interest* within `d_samples` can be 
-   *   specified using the `num_row_samples`, `num_rows`, and `row_stride_bytes` 
+   * - A two-dimensional *region of interest* within `d_samples` can be
+   *   specified using the `num_row_samples`, `num_rows`, and `row_stride_bytes`
    *   parameters.
    * - The row stride must be a whole multiple of the sample data type
    *   size, i.e., `(row_stride_bytes % sizeof(SampleT)) == 0`.
-   * - The number of histogram bins for channel<sub><em>i</em></sub> is 
+   * - The number of histogram bins for channel<sub><em>i</em></sub> is
    *   `num_levels[i] - 1`.
-   * - For channel<sub><em>i</em></sub>, the range of values for all histogram 
-   *   bins have the same width: 
+   * - For channel<sub><em>i</em></sub>, the range of values for all histogram
+   *   bins have the same width:
    *   `(upper_level[i] - lower_level[i]) / (num_levels[i] - 1)`
-   * - For a given row `r` in `[0, num_rows)`, and sample `s` in 
-   *   `[0, num_row_pixels)`, let 
-   *   `row_begin = d_samples + r * row_stride_bytes / sizeof(SampleT)`, 
+   * - For a given row `r` in `[0, num_rows)`, and sample `s` in
+   *   `[0, num_row_pixels)`, let
+   *   `row_begin = d_samples + r * row_stride_bytes / sizeof(SampleT)`,
    *   `sample_begin = row_begin + s * NUM_CHANNELS`, and
    *   `sample_end = sample_begin + NUM_ACTIVE_CHANNELS`. For given channels
    *    `c1` and `c2` in `[0, NUM_ACTIVE_CHANNELS)`, the range
-   *   `[d_histogram[c1], d_histogram[c1] + num_levels[c1] - 1)` shall not 
+   *   `[d_histogram[c1], d_histogram[c1] + num_levels[c1] - 1)` shall not
    *   overlap `[sample_begin, sample_end)` nor
    *   `[d_levels[c2], d_levels[c2] + num_levels[c2])` in any way. The ranges
-   *   `[d_levels[c2], d_levels[c2] + num_levels[c2])` and 
+   *   `[d_levels[c2], d_levels[c2] + num_levels[c2])` and
    *   `[sample_begin, sample_end)` may overlap.
    * - @devicestorage
    *
    * @par Snippet
-   * The code snippet below illustrates the computation of three 4-bin *RGB* 
-   * histograms from a 2x3 region of interest of within a flattened 2x4 array 
+   * The code snippet below illustrates the computation of three 4-bin *RGB*
+   * histograms from a 2x3 region of interest of within a flattened 2x4 array
    * of quad-channel *RGBA* pixels (8 bits per channel per pixel).
    *
    * @par
    * @code
    * #include <cub/cub.cuh> // or equivalently <cub/device/device_histogram.cuh>
    *
-   * // Declare, allocate, and initialize device-accessible pointers for input 
+   * // Declare, allocate, and initialize device-accessible pointers for input
    * // samples and output histograms
    * int              num_row_pixels;     // e.g., 3
    * int              num_rows;           // e.g., 2
@@ -1483,7 +1490,7 @@ struct DeviceHistogram
    * size_t   temp_storage_bytes = 0;
    * cub::DeviceHistogram::MultiHistogramRange<4, 3>(
    *   d_temp_storage, temp_storage_bytes,
-   *   d_samples, d_histogram, num_levels, d_levels, 
+   *   d_samples, d_histogram, num_levels, d_levels,
    *   num_row_pixels, num_rows, row_stride_bytes);
    *
    * // Allocate temporary storage
@@ -1492,7 +1499,7 @@ struct DeviceHistogram
    * // Compute histograms
    * cub::DeviceHistogram::MultiHistogramRange<4, 3>(
    *   d_temp_storage, temp_storage_bytes,
-   *   d_samples, d_histogram, num_levels, 
+   *   d_samples, d_histogram, num_levels,
    *   d_levels, num_row_pixels, num_rows, row_stride_bytes);
    *
    * // d_histogram   <-- [ [2, 3, 0, 1],
@@ -1501,68 +1508,68 @@ struct DeviceHistogram
    *
    * @endcode
    *
-   * @tparam NUM_CHANNELS             
-   *   Number of channels interleaved in the input data (may be greater than 
+   * @tparam NUM_CHANNELS
+   *   Number of channels interleaved in the input data (may be greater than
    *   the number of channels being actively histogrammed)
-   * 
-   * @tparam NUM_ACTIVE_CHANNELS      
+   *
+   * @tparam NUM_ACTIVE_CHANNELS
    *   **[inferred]** Number of channels actively being histogrammed
-   * 
-   * @tparam SampleIteratorT          
-   *   **[inferred]** Random-access input iterator type for reading input 
+   *
+   * @tparam SampleIteratorT
+   *   **[inferred]** Random-access input iterator type for reading input
    *   samples. \iterator
-   * 
-   * @tparam CounterT                 
+   *
+   * @tparam CounterT
    *   **[inferred]** Integer type for histogram bin counters
-   * 
-   * @tparam LevelT                   
+   *
+   * @tparam LevelT
    *   **[inferred]** Type for specifying boundaries (levels)
-   * 
-   * @tparam OffsetT                  
-   *   **[inferred]** Signed integer type for sequence offsets, list lengths, 
+   *
+   * @tparam OffsetT
+   *   **[inferred]** Signed integer type for sequence offsets, list lengths,
    *   pointer differences, etc.  \offset_size1
    *
-   * @param[in] d_temp_storage 
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
    *   required allocation size is written to \p temp_storage_bytes and no work is done.
    *
-   * @param[in,out] temp_storage_bytes 
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of `d_temp_storage` allocation
    *
-   * @param[in] d_samples 
-   *   The pointer to the multi-channel input sequence of data samples. The 
-   *   samples from different channels are assumed to be interleaved (e.g., an 
-   *   array of 32-bit pixels where each pixel consists of four 
+   * @param[in] d_samples
+   *   The pointer to the multi-channel input sequence of data samples. The
+   *   samples from different channels are assumed to be interleaved (e.g., an
+   *   array of 32-bit pixels where each pixel consists of four
    *   *RGBA* 8-bit samples).
    *
-   * @param[out] d_histogram 
-   *   The pointers to the histogram counter output arrays, one for each active 
-   *   channel. For channel<sub><em>i</em></sub>, the allocation length of 
+   * @param[out] d_histogram
+   *   The pointers to the histogram counter output arrays, one for each active
+   *   channel. For channel<sub><em>i</em></sub>, the allocation length of
    *   `d_histogram[i]` should be `num_levels[i] - 1`.
    *
-   * @param[in] num_levels 
-   *   The number of boundaries (levels) for delineating histogram samples in 
-   *   each active channel. Implies that the number of bins for 
+   * @param[in] num_levels
+   *   The number of boundaries (levels) for delineating histogram samples in
+   *   each active channel. Implies that the number of bins for
    *   channel<sub><em>i</em></sub> is `num_levels[i] - 1`.
    *
-   * @param[in] d_levels 
-   *   The pointers to the arrays of boundaries (levels), one for each active 
-   *   channel. Bin ranges are defined by consecutive boundary pairings: lower 
-   *   sample value boundaries are inclusive and upper sample value boundaries 
+   * @param[in] d_levels
+   *   The pointers to the arrays of boundaries (levels), one for each active
+   *   channel. Bin ranges are defined by consecutive boundary pairings: lower
+   *   sample value boundaries are inclusive and upper sample value boundaries
    *   are exclusive.
    *
-   * @param[in] num_row_pixels 
+   * @param[in] num_row_pixels
    *   The number of multi-channel pixels per row in the region of interest
    *
-   * @param[in] num_rows 
+   * @param[in] num_rows
    *   The number of rows in the region of interest
    *
-   * @param[in] row_stride_bytes 
-   *   The number of bytes between starts of consecutive rows in the 
+   * @param[in] row_stride_bytes
+   *   The number of bytes between starts of consecutive rows in the
    *   region of interest
    *
-   * @param[in] stream 
-   *   **[optional]** CUDA stream to launch kernels within.  
+   * @param[in] stream
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    */
   template <int NUM_CHANNELS,
diff --git a/cub/cub/device/device_memcpy.cuh b/cub/cub/device/device_memcpy.cuh
index a0f8fe5415f..bba6b0e001b 100644
--- a/cub/cub/device/device_memcpy.cuh
+++ b/cub/cub/device/device_memcpy.cuh
@@ -32,7 +32,14 @@
 
 #pragma once
 
-#include <cub/config.cuh>
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/device/dispatch/dispatch_batch_memcpy.cuh>
 
 #include <cstdint>
diff --git a/cub/cub/device/device_merge_sort.cuh b/cub/cub/device/device_merge_sort.cuh
index 9b17ac39125..cc9865bce9b 100644
--- a/cub/cub/device/device_merge_sort.cuh
+++ b/cub/cub/device/device_merge_sort.cuh
@@ -27,7 +27,14 @@
 
 #pragma once
 
-#include <cub/config.cuh>
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/device/dispatch/dispatch_merge_sort.cuh>
 #include <cub/util_deprecated.cuh>
 #include <cub/util_namespace.cuh>
diff --git a/cub/cub/device/device_partition.cuh b/cub/cub/device/device_partition.cuh
index 6692a489b20..4c311d39e3f 100644
--- a/cub/cub/device/device_partition.cuh
+++ b/cub/cub/device/device_partition.cuh
@@ -34,10 +34,17 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <stdio.h>
 #include <iterator>
 
-#include <cub/config.cuh>
 #include <cub/device/dispatch/dispatch_select_if.cuh>
 #include <cub/device/dispatch/dispatch_three_way_partition.cuh>
 #include <cub/util_deprecated.cuh>
@@ -85,9 +92,9 @@ struct DevicePartition
      * - Copies of the selected items are compacted into @p d_out and maintain
      *   their original relative ordering, however copies of the unselected
      *   items are compacted into the rear of @p d_out in reverse order.
-     * - The range `[d_out, d_out + num_items)` shall not overlap 
-     *   `[d_in, d_in + num_items)` nor `[d_flags, d_flags + num_items)` in any 
-     *   way. The range `[d_in, d_in + num_items)` may overlap 
+     * - The range `[d_out, d_out + num_items)` shall not overlap
+     *   `[d_in, d_in + num_items)` nor `[d_flags, d_flags + num_items)` in any
+     *   way. The range `[d_in, d_in + num_items)` may overlap
      *  `[d_flags, d_flags + num_items)`.
      * - \devicestorage
      *
@@ -251,8 +258,8 @@ struct DevicePartition
      * - Copies of the selected items are compacted into @p d_out and maintain
      *   their original relative ordering, however copies of the unselected
      *   items are compacted into the rear of @p d_out in reverse order.
-     * - The range `[d_out, d_out + num_items)` shall not overlap 
-     *   `[d_in, d_in + num_items)` in any way. 
+     * - The range `[d_out, d_out + num_items)` shall not overlap
+     *   `[d_in, d_in + num_items)` in any way.
      * - \devicestorage
      *
      * @par Performance
@@ -451,10 +458,10 @@ struct DevicePartition
      * - Copies of the unselected items are compacted into the
      *   @p d_unselected_out in reverse order.
      * - The ranges `[d_out, d_out + num_items)`,
-     *   `[d_first_part_out, d_first_part_out + d_num_selected_out[0])`, 
-     *   `[d_second_part_out, d_second_part_out + d_num_selected_out[1])`, 
-     *   `[d_unselected_out, d_unselected_out + num_items - d_num_selected_out[0] - d_num_selected_out[1])`, 
-     *   shall not overlap in any way. 
+     *   `[d_first_part_out, d_first_part_out + d_num_selected_out[0])`,
+     *   `[d_second_part_out, d_second_part_out + d_num_selected_out[1])`,
+     *   `[d_unselected_out, d_unselected_out + num_items - d_num_selected_out[0] - d_num_selected_out[1])`,
+     *   shall not overlap in any way.
      *
      * @par Snippet
      * The code snippet below illustrates how this algorithm can partition an
diff --git a/cub/cub/device/device_radix_sort.cuh b/cub/cub/device/device_radix_sort.cuh
index e9d1b4d1d91..adf2385ab6d 100644
--- a/cub/cub/device/device_radix_sort.cuh
+++ b/cub/cub/device/device_radix_sort.cuh
@@ -13,9 +13,9 @@
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
  *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 
- * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 
- * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  * ARE DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE FOR ANY
  * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
@@ -26,13 +26,20 @@
  *
  ******************************************************************************/
 
-//! @file cub::DeviceRadixSort provides device-wide, parallel operations for 
-//!       computing a radix sort across a sequence of data items residing within 
+//! @file cub::DeviceRadixSort provides device-wide, parallel operations for
+//!       computing a radix sort across a sequence of data items residing within
 //!       device-accessible memory.
 
 #pragma once
 
-#include <cub/config.cuh>
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/detail/choose_offset.cuh>
 #include <cub/device/dispatch/dispatch_radix_sort.cuh>
 #include <cub/util_deprecated.cuh>
@@ -43,33 +50,33 @@
 CUB_NAMESPACE_BEGIN
 
 
-//! @brief DeviceRadixSort provides device-wide, parallel operations for 
-//!        computing a radix sort across a sequence of data items residing 
+//! @brief DeviceRadixSort provides device-wide, parallel operations for
+//!        computing a radix sort across a sequence of data items residing
 //!        within device-accessible memory. ![](sorting_logo.png)
 //! @ingroup SingleModule
-//! 
+//!
 //! @par Overview
-//! The [*radix sorting method*](http://en.wikipedia.org/wiki/Radix_sort) 
-//! arranges items into ascending (or descending) order. The algorithm relies 
-//! upon a positional representation for keys, i.e., each key is comprised of an 
-//! ordered sequence of symbols (e.g., digits, characters, etc.) specified from 
-//! least-significant to most-significant. For a given input sequence of keys 
-//! and a set of rules specifying a total ordering of the symbolic alphabet, the 
+//! The [*radix sorting method*](http://en.wikipedia.org/wiki/Radix_sort)
+//! arranges items into ascending (or descending) order. The algorithm relies
+//! upon a positional representation for keys, i.e., each key is comprised of an
+//! ordered sequence of symbols (e.g., digits, characters, etc.) specified from
+//! least-significant to most-significant. For a given input sequence of keys
+//! and a set of rules specifying a total ordering of the symbolic alphabet, the
 //! radix sorting method produces a lexicographic ordering of those keys.
-//! 
+//!
 //! @par Supported Types
 //! DeviceRadixSort can sort all of the built-in C++ numeric primitive types
 //! (`unsigned char`, `int`, `double`, etc.) as well as CUDA's `__half`
-//! and `__nv_bfloat16` 16-bit floating-point types. User-defined types are 
+//! and `__nv_bfloat16` 16-bit floating-point types. User-defined types are
 //! supported as long as decomposer object is provided.
-//! 
+//!
 //! @par Floating-Point Special Cases
-//! 
+//!
 //! - Positive and negative zeros are considered equivalent, and will be treated
 //!   as such in the output.
 //! - No special handling is implemented for NaN values; these are sorted
 //!   according to their bit representations after any transformations.
-//! 
+//!
 //! @par Transformations
 //! Although the direct radix sorting method can only be applied to unsigned
 //! integral types, DeviceRadixSort is able to sort signed and floating-point
@@ -78,41 +85,41 @@ CUB_NAMESPACE_BEGIN
 //! transformations must be considered when restricting the
 //! `[begin_bit, end_bit)` range, as the bitwise transformations will occur
 //! before the bit-range truncation.
-//! 
+//!
 //! Any transformations applied to the keys prior to sorting are reversed
 //! while writing to the final output buffer.
-//! 
+//!
 //! @par Type Specific Bitwise Transformations
 //! To convert the input values into a radix-sortable bitwise representation,
 //! the following transformations take place prior to sorting:
-//! 
+//!
 //! - For unsigned integral values, the keys are used directly.
 //! - For signed integral values, the sign bit is inverted.
 //! - For positive floating point values, the sign bit is inverted.
 //! - For negative floating point values, the full key is inverted.
-//! 
+//!
 //! For floating point types, positive and negative zero are a special case and
 //! will be considered equivalent during sorting.
-//! 
+//!
 //! @par Descending Sort Bitwise Transformations
 //! If descending sort is used, the keys are inverted after performing any
 //! type-specific transformations, and the resulting keys are sorted in ascending
 //! order.
-//! 
+//!
 //! @par Stability
 //! DeviceRadixSort is stable. For floating-point types, `-0.0` and `+0.0` are
 //! considered equal and appear in the result in the same order as they appear in
 //! the input.
-//! 
+//!
 //! @par Usage Considerations
 //! @cdp_class{DeviceRadixSort}
-//! 
+//!
 //! @par Performance
-//! @linear_performance{radix sort} The following chart illustrates 
-//! DeviceRadixSort::SortKeys performance across different CUDA architectures 
+//! @linear_performance{radix sort} The following chart illustrates
+//! DeviceRadixSort::SortKeys performance across different CUDA architectures
 //! for uniform-random `uint32` keys.
 //! @plots_below
-//! 
+//!
 //! @image html lsb_radix_sort_int32_keys.png
 struct DeviceRadixSort
 {
@@ -203,7 +210,7 @@ public:
   //! @name KeyT-value pairs
   //@{
 
-  //! @brief Sorts key-value pairs into ascending order. 
+  //! @brief Sorts key-value pairs into ascending order.
   //!        (`~2N` auxiliary storage required)
   //!
   //! @par
@@ -216,15 +223,15 @@ public:
   //!   - `[d_keys_out,   d_keys_out   + num_items)`
   //!   - `[d_values_in,  d_values_in  + num_items)`
   //!   - `[d_values_out, d_values_out + num_items)`
-  //! - An optional bit subrange `[begin_bit, end_bit)` of differentiating key 
-  //!   bits can be specified. This can reduce overall sorting overhead and 
+  //! - An optional bit subrange `[begin_bit, end_bit)` of differentiating key
+  //!   bits can be specified. This can reduce overall sorting overhead and
   //!   yield a corresponding performance improvement.
-  //! - @devicestorageNP For sorting using only `O(P)` temporary storage, see 
+  //! - @devicestorageNP For sorting using only `O(P)` temporary storage, see
   //!   the sorting interface using DoubleBuffer wrappers below.
   //! - @devicestorage
   //!
   //! @par Performance
-  //! The following charts illustrate saturated sorting performance across 
+  //! The following charts illustrate saturated sorting performance across
   //! different CUDA architectures for uniform-random `uint32, uint32` and
   //! `uint64, uint64` pairs, respectively.
   //!
@@ -236,10 +243,10 @@ public:
   //! keys with associated vector of `int` values.
   //! @par
   //! @code
-  //! #include <cub/cub.cuh>   
+  //! #include <cub/cub.cuh>
   //! // or equivalently <cub/device/device_radix_sort.cuh>
   //!
-  //! // Declare, allocate, and initialize device-accessible pointers 
+  //! // Declare, allocate, and initialize device-accessible pointers
   //! // for sorting data
   //! int  num_items;          // e.g., 7
   //! int  *d_keys_in;         // e.g., [8, 6, 7, 5, 3, 0, 9]
@@ -265,49 +272,49 @@ public:
   //! // d_values_out          <-- [5, 4, 3, 1, 2, 0, 6]
   //! @endcode
   //!
-  //! @tparam KeyT      
+  //! @tparam KeyT
   //!   **[inferred]** KeyT type
   //!
-  //! @tparam ValueT    
+  //! @tparam ValueT
   //!   **[inferred]** ValueT type
   //!
-  //! @tparam NumItemsT 
+  //! @tparam NumItemsT
   //!   **[inferred]** Type of num_items
   //!
-  //! @param[in] d_temp_storage 
-  //!   Device-accessible allocation of temporary storage. When `nullptr`, the 
-  //!   required allocation size is written to `temp_storage_bytes` and no work 
+  //! @param[in] d_temp_storage
+  //!   Device-accessible allocation of temporary storage. When `nullptr`, the
+  //!   required allocation size is written to `temp_storage_bytes` and no work
   //!   is done.
   //!
-  //! @param[in,out] temp_storage_bytes 
+  //! @param[in,out] temp_storage_bytes
   //!   Reference to size in bytes of `d_temp_storage` allocation
   //!
-  //! @param[in] d_keys_in 
+  //! @param[in] d_keys_in
   //!   Pointer to the input data of key data to sort
   //!
-  //! @param[out] d_keys_out 
+  //! @param[out] d_keys_out
   //!   Pointer to the sorted output sequence of key data
   //!
-  //! @param[in] d_values_in 
+  //! @param[in] d_values_in
   //!   Pointer to the corresponding input sequence of associated value items
   //!
-  //! @param[out] d_values_out 
-  //!   Pointer to the correspondingly-reordered output sequence of associated 
+  //! @param[out] d_values_out
+  //!   Pointer to the correspondingly-reordered output sequence of associated
   //!   value items
   //!
-  //! @param[in] num_items 
+  //! @param[in] num_items
   //!   Number of items to sort
   //!
-  //! @param[in] begin_bit 
-  //!   **[optional]** The least-significant bit index (inclusive) needed for 
+  //! @param[in] begin_bit
+  //!   **[optional]** The least-significant bit index (inclusive) needed for
   //!   key comparison
   //!
-  //! @param[in] end_bit 
-  //!   **[optional]** The most-significant bit index (exclusive) needed for key 
+  //! @param[in] end_bit
+  //!   **[optional]** The most-significant bit index (exclusive) needed for key
   //!   comparison (e.g., sizeof(unsigned int) * 8)
   //!
-  //! @param[in] stream 
-  //!   **[optional]** CUDA stream to launch kernels within. 
+  //! @param[in] stream
+  //!   **[optional]** CUDA stream to launch kernels within.
   //!   Default is stream<sub>0</sub>.
   template <typename KeyT, typename ValueT, typename NumItemsT>
   CUB_RUNTIME_FUNCTION static cudaError_t
@@ -380,9 +387,9 @@ public:
   }
   #endif
 
-  //! @rst 
+  //! @rst
   //! Sorts key-value pairs into ascending order using :math:`\approx 2N` auxiliary storage.
-  //! 
+  //!
   //! * The contents of the input data are not altered by the sorting operation.
   //! * Pointers to contiguous memory must be used; iterators are not currently
   //!   supported.
@@ -394,10 +401,10 @@ public:
   //!   * ``[d_values_in,  d_values_in  + num_items)``
   //!   * ``[d_values_out, d_values_out + num_items)``
   //!
-  //! * A bit subrange ``[begin_bit, end_bit)`` is provided to specify 
-  //!   differentiating key bits. This can reduce overall sorting overhead and 
+  //! * A bit subrange ``[begin_bit, end_bit)`` is provided to specify
+  //!   differentiating key bits. This can reduce overall sorting overhead and
   //!   yield a corresponding performance improvement.
-  //! * @devicestorageNP For sorting using only :math:`O(P)` temporary storage, see 
+  //! * @devicestorageNP For sorting using only :math:`O(P)` temporary storage, see
   //!   the sorting interface using DoubleBuffer wrappers below.
   //! * @devicestorage
   //!
@@ -417,7 +424,7 @@ public:
   //!
   //! The following snippet shows how to sort an array of ``custom_t`` objects
   //! using ``cub::DeviceRadixSort::SortPairs``:
-  //! 
+  //!
   //! .. literalinclude:: ../../test/catch2_test_device_radix_sort_custom.cu
   //!     :language: c++
   //!     :dedent:
@@ -426,62 +433,62 @@ public:
   //!
   //! @endrst
   //!
-  //! @tparam KeyT      
+  //! @tparam KeyT
   //!   **[inferred]** KeyT type
   //!
-  //! @tparam ValueT    
+  //! @tparam ValueT
   //!   **[inferred]** ValueT type
   //!
-  //! @tparam NumItemsT 
+  //! @tparam NumItemsT
   //!   **[inferred]** Type of num_items
   //!
   //! @tparam DecomposerT
-  //!   **[inferred]** Type of a callable object responsible for decomposing a 
+  //!   **[inferred]** Type of a callable object responsible for decomposing a
   //!   ``KeyT`` into a tuple of references to its constituent arithmetic types:
-  //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``. 
-  //!   The leftmost element of the tuple is considered the most significant. 
+  //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``.
+  //!   The leftmost element of the tuple is considered the most significant.
   //!   The call operator must not modify members of the key.
   //!
-  //! @param[in] d_temp_storage 
-  //!   Device-accessible allocation of temporary storage. When `nullptr`, the 
-  //!   required allocation size is written to `temp_storage_bytes` and no work 
+  //! @param[in] d_temp_storage
+  //!   Device-accessible allocation of temporary storage. When `nullptr`, the
+  //!   required allocation size is written to `temp_storage_bytes` and no work
   //!   is done.
   //!
-  //! @param[in,out] temp_storage_bytes 
+  //! @param[in,out] temp_storage_bytes
   //!   Reference to size in bytes of `d_temp_storage` allocation
   //!
-  //! @param[in] d_keys_in 
+  //! @param[in] d_keys_in
   //!   Pointer to the input data of key data to sort
   //!
-  //! @param[out] d_keys_out 
+  //! @param[out] d_keys_out
   //!   Pointer to the sorted output sequence of key data
   //!
-  //! @param[in] d_values_in 
+  //! @param[in] d_values_in
   //!   Pointer to the corresponding input sequence of associated value items
   //!
-  //! @param[out] d_values_out 
-  //!   Pointer to the correspondingly-reordered output sequence of associated 
+  //! @param[out] d_values_out
+  //!   Pointer to the correspondingly-reordered output sequence of associated
   //!   value items
   //!
-  //! @param[in] num_items 
+  //! @param[in] num_items
   //!   Number of items to sort
   //!
   //! @param decomposer
   //!   Callable object responsible for decomposing a ``KeyT`` into a tuple of
-  //!   references to its constituent arithmetic types. The leftmost element of 
-  //!   the tuple is considered the most significant. The call operator must not 
+  //!   references to its constituent arithmetic types. The leftmost element of
+  //!   the tuple is considered the most significant. The call operator must not
   //!   modify members of the key.
   //!
-  //! @param[in] begin_bit 
-  //!   **[optional]** The least-significant bit index (inclusive) needed for 
+  //! @param[in] begin_bit
+  //!   **[optional]** The least-significant bit index (inclusive) needed for
   //!   key comparison
   //!
-  //! @param[in] end_bit 
-  //!   **[optional]** The most-significant bit index (exclusive) needed for key 
+  //! @param[in] end_bit
+  //!   **[optional]** The most-significant bit index (exclusive) needed for key
   //!   comparison (e.g., `(sizeof(float) + sizeof(long long int)) * 8`)
   //!
-  //! @param[in] stream 
-  //!   **[optional]** CUDA stream to launch kernels within. 
+  //! @param[in] stream
+  //!   **[optional]** CUDA stream to launch kernels within.
   //!   Default is stream<sub>0</sub>.
   template <typename KeyT, typename ValueT, typename NumItemsT, typename DecomposerT>
   CUB_RUNTIME_FUNCTION static                                //
@@ -530,9 +537,9 @@ public:
                                                              stream);
   }
 
-  //! @rst 
+  //! @rst
   //! Sorts key-value pairs into ascending order using :math:`\approx 2N` auxiliary storage.
-  //! 
+  //!
   //! * The contents of the input data are not altered by the sorting operation.
   //! * Pointers to contiguous memory must be used; iterators are not currently
   //!   supported.
@@ -544,7 +551,7 @@ public:
   //!   * ``[d_values_in,  d_values_in  + num_items)``
   //!   * ``[d_values_out, d_values_out + num_items)``
   //!
-  //! * @devicestorageNP For sorting using only :math:`O(P)` temporary storage, see 
+  //! * @devicestorageNP For sorting using only :math:`O(P)` temporary storage, see
   //!   the sorting interface using DoubleBuffer wrappers below.
   //! * @devicestorage
   //!
@@ -564,7 +571,7 @@ public:
   //!
   //! The following snippet shows how to sort an array of ``custom_t`` objects
   //! using ``cub::DeviceRadixSort::SortPairs``:
-  //! 
+  //!
   //! .. literalinclude:: ../../test/catch2_test_device_radix_sort_custom.cu
   //!     :language: c++
   //!     :dedent:
@@ -573,54 +580,54 @@ public:
   //!
   //! @endrst
   //!
-  //! @tparam KeyT      
+  //! @tparam KeyT
   //!   **[inferred]** KeyT type
   //!
-  //! @tparam ValueT    
+  //! @tparam ValueT
   //!   **[inferred]** ValueT type
   //!
-  //! @tparam NumItemsT 
+  //! @tparam NumItemsT
   //!   **[inferred]** Type of num_items
   //!
   //! @tparam DecomposerT
-  //!   **[inferred]** Type of a callable object responsible for decomposing a 
+  //!   **[inferred]** Type of a callable object responsible for decomposing a
   //!   ``KeyT`` into a tuple of references to its constituent arithmetic types:
-  //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``. 
-  //!   The leftmost element of the tuple is considered the most significant. 
+  //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``.
+  //!   The leftmost element of the tuple is considered the most significant.
   //!   The call operator must not modify members of the key.
   //!
-  //! @param[in] d_temp_storage 
-  //!   Device-accessible allocation of temporary storage. When `nullptr`, the 
-  //!   required allocation size is written to `temp_storage_bytes` and no work 
+  //! @param[in] d_temp_storage
+  //!   Device-accessible allocation of temporary storage. When `nullptr`, the
+  //!   required allocation size is written to `temp_storage_bytes` and no work
   //!   is done.
   //!
-  //! @param[in,out] temp_storage_bytes 
+  //! @param[in,out] temp_storage_bytes
   //!   Reference to size in bytes of `d_temp_storage` allocation
   //!
-  //! @param[in] d_keys_in 
+  //! @param[in] d_keys_in
   //!   Pointer to the input data of key data to sort
   //!
-  //! @param[out] d_keys_out 
+  //! @param[out] d_keys_out
   //!   Pointer to the sorted output sequence of key data
   //!
-  //! @param[in] d_values_in 
+  //! @param[in] d_values_in
   //!   Pointer to the corresponding input sequence of associated value items
   //!
-  //! @param[out] d_values_out 
-  //!   Pointer to the correspondingly-reordered output sequence of associated 
+  //! @param[out] d_values_out
+  //!   Pointer to the correspondingly-reordered output sequence of associated
   //!   value items
   //!
-  //! @param[in] num_items 
+  //! @param[in] num_items
   //!   Number of items to sort
   //!
   //! @param decomposer
   //!   Callable object responsible for decomposing a ``KeyT`` into a tuple of
-  //!   references to its constituent arithmetic types. The leftmost element of 
-  //!   the tuple is considered the most significant. The call operator must not 
+  //!   references to its constituent arithmetic types. The leftmost element of
+  //!   the tuple is considered the most significant. The call operator must not
   //!   modify members of the key.
   //!
-  //! @param[in] stream 
-  //!   **[optional]** CUDA stream to launch kernels within. 
+  //! @param[in] stream
+  //!   **[optional]** CUDA stream to launch kernels within.
   //!   Default is stream<sub>0</sub>.
   template <typename KeyT, typename ValueT, typename NumItemsT, typename DecomposerT>
   CUB_RUNTIME_FUNCTION static                                //
@@ -665,7 +672,7 @@ public:
                                                              stream);
   }
 
-  //! @brief Sorts key-value pairs into ascending order. 
+  //! @brief Sorts key-value pairs into ascending order.
   //!        (`~N` auxiliary storage required)
   //!
   //! @par
@@ -673,7 +680,7 @@ public:
   //!   pair of associated value buffers.  Each pair is managed by a DoubleBuffer
   //!   structure that indicates which of the two buffers is "current" (and thus
   //!   contains the input data to be sorted).
-  //! - The contents of both buffers within each pair may be altered by the 
+  //! - The contents of both buffers within each pair may be altered by the
   //!   sorting operation.
   //! - In-place operations are not supported. There must be no overlap between
   //!   any of the provided ranges:
@@ -681,18 +688,18 @@ public:
   //!   - `[d_keys.Alternate(),   d_keys.Alternate()   + num_items)`
   //!   - `[d_values.Current(),   d_values.Current()   + num_items)`
   //!   - `[d_values.Alternate(), d_values.Alternate() + num_items)`
-  //! - Upon completion, the sorting operation will update the "current" 
-  //!   indicator within each DoubleBuffer wrapper to reference which of the two 
-  //!   buffers now contains the sorted output sequence (a function of the 
+  //! - Upon completion, the sorting operation will update the "current"
+  //!   indicator within each DoubleBuffer wrapper to reference which of the two
+  //!   buffers now contains the sorted output sequence (a function of the
   //!   number of key bits specified and the targeted device architecture).
-  //! - An optional bit subrange `[begin_bit, end_bit)` of differentiating key 
-  //!   bits can be specified. This can reduce overall sorting overhead and 
+  //! - An optional bit subrange `[begin_bit, end_bit)` of differentiating key
+  //!   bits can be specified. This can reduce overall sorting overhead and
   //!   yield a corresponding performance improvement.
   //! - @devicestorageP
   //! - @devicestorage
   //!
   //! @par Performance
-  //! The following charts illustrate saturated sorting performance across 
+  //! The following charts illustrate saturated sorting performance across
   //! different CUDA architectures for uniform-random `uint32, uint32` and
   //! `uint64, uint64` pairs, respectively.
   //!
@@ -700,14 +707,14 @@ public:
   //! @image html lsb_radix_sort_int64_pairs.png
   //!
   //! @par Snippet
-  //! The code snippet below illustrates the sorting of a device vector of `int` 
+  //! The code snippet below illustrates the sorting of a device vector of `int`
   //! keys with associated vector of `int` values.
   //! @par
   //! @code
-  //! #include <cub/cub.cuh>   
+  //! #include <cub/cub.cuh>
   //! // or equivalently <cub/device/device_radix_sort.cuh>
   //!
-  //! // Declare, allocate, and initialize device-accessible pointers for 
+  //! // Declare, allocate, and initialize device-accessible pointers for
   //! // sorting data
   //! int  num_items;          // e.g., 7
   //! int  *d_key_buf;         // e.g., [8, 6, 7, 5, 3, 0, 9]
@@ -738,45 +745,45 @@ public:
   //!
   //! @endcode
   //!
-  //! @tparam KeyT      
+  //! @tparam KeyT
   //!   **[inferred]** KeyT type
   //!
-  //! @tparam ValueT    
+  //! @tparam ValueT
   //!   **[inferred]** ValueT type
   //!
-  //! @tparam NumItemsT 
+  //! @tparam NumItemsT
   //!   **[inferred]** Type of num_items
   //!
-  //! @param[in] d_temp_storage 
-  //!   Device-accessible allocation of temporary storage. When `nullptr`, the 
+  //! @param[in] d_temp_storage
+  //!   Device-accessible allocation of temporary storage. When `nullptr`, the
   //!   required allocation size is written to @p temp_storage_bytes and no work is done.
   //!
-  //! @param[in,out] temp_storage_bytes 
+  //! @param[in,out] temp_storage_bytes
   //!   Reference to size in bytes of `d_temp_storage` allocation
   //!
-  //! @param[in,out] d_keys 
-  //!   Reference to the double-buffer of keys whose "current" device-accessible 
-  //!   buffer contains the unsorted input keys and, upon return, is updated to 
+  //! @param[in,out] d_keys
+  //!   Reference to the double-buffer of keys whose "current" device-accessible
+  //!   buffer contains the unsorted input keys and, upon return, is updated to
   //!   point to the sorted output keys
   //!
-  //! @param[in,out] d_values 
-  //!   Double-buffer of values whose "current" device-accessible buffer 
-  //!   contains the unsorted input values and, upon return, is updated to point 
+  //! @param[in,out] d_values
+  //!   Double-buffer of values whose "current" device-accessible buffer
+  //!   contains the unsorted input values and, upon return, is updated to point
   //!   to the sorted output values
   //!
-  //! @param[in] num_items 
+  //! @param[in] num_items
   //!   Number of items to sort
   //!
-  //! @param[in] begin_bit 
-  //!   **[optional]** The least-significant bit index (inclusive) needed for 
+  //! @param[in] begin_bit
+  //!   **[optional]** The least-significant bit index (inclusive) needed for
   //!   key comparison
   //!
-  //! @param[in] end_bit 
-  //!   **[optional]** The most-significant bit index (exclusive) needed for key 
+  //! @param[in] end_bit
+  //!   **[optional]** The most-significant bit index (exclusive) needed for key
   //!   comparison (e.g., `sizeof(unsigned int) * 8`)
   //!
-  //! @param[in] stream 
-  //!   **[optional]** CUDA stream to launch kernels within. 
+  //! @param[in] stream
+  //!   **[optional]** CUDA stream to launch kernels within.
   //!   Default is stream<sub>0</sub>.
   template <typename KeyT, typename ValueT, typename NumItemsT>
   CUB_RUNTIME_FUNCTION static cudaError_t
@@ -833,14 +840,14 @@ public:
   }
   #endif
 
-  //! @rst 
+  //! @rst
   //! Sorts key-value pairs into ascending order using :math:`\approx N` auxiliary storage.
-  //! 
+  //!
   //! * The sorting operation is given a pair of key buffers and a corresponding
   //!   pair of associated value buffers.  Each pair is managed by a DoubleBuffer
   //!   structure that indicates which of the two buffers is "current" (and thus
   //!   contains the input data to be sorted).
-  //! * The contents of both buffers within each pair may be altered by the 
+  //! * The contents of both buffers within each pair may be altered by the
   //!   sorting operation.
   //! * In-place operations are not supported. There must be no overlap between
   //!   any of the provided ranges:
@@ -850,9 +857,9 @@ public:
   //!   - ``[d_values.Current(),   d_values.Current()   + num_items)``
   //!   - ``[d_values.Alternate(), d_values.Alternate() + num_items)``
   //!
-  //! - Upon completion, the sorting operation will update the "current" 
-  //!   indicator within each DoubleBuffer wrapper to reference which of the two 
-  //!   buffers now contains the sorted output sequence (a function of the 
+  //! - Upon completion, the sorting operation will update the "current"
+  //!   indicator within each DoubleBuffer wrapper to reference which of the two
+  //!   buffers now contains the sorted output sequence (a function of the
   //!   number of key bits specified and the targeted device architecture).
   //! - @devicestorageP
   //! - @devicestorage
@@ -873,7 +880,7 @@ public:
   //!
   //! The following snippet shows how to sort an array of ``custom_t`` objects
   //! using ``cub::DeviceRadixSort::SortPairs``:
-  //! 
+  //!
   //! .. literalinclude:: ../../test/catch2_test_device_radix_sort_custom.cu
   //!     :language: c++
   //!     :dedent:
@@ -882,51 +889,51 @@ public:
   //!
   //! @endrst
   //!
-  //! @tparam KeyT      
+  //! @tparam KeyT
   //!   **[inferred]** KeyT type
   //!
-  //! @tparam ValueT    
+  //! @tparam ValueT
   //!   **[inferred]** ValueT type
   //!
-  //! @tparam NumItemsT 
+  //! @tparam NumItemsT
   //!   **[inferred]** Type of num_items
   //!
   //! @tparam DecomposerT
-  //!   **[inferred]** Type of a callable object responsible for decomposing a 
+  //!   **[inferred]** Type of a callable object responsible for decomposing a
   //!   ``KeyT`` into a tuple of references to its constituent arithmetic types:
-  //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``. 
-  //!   The leftmost element of the tuple is considered the most significant. 
+  //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``.
+  //!   The leftmost element of the tuple is considered the most significant.
   //!   The call operator must not modify members of the key.
   //!
-  //! @param[in] d_temp_storage 
-  //!   Device-accessible allocation of temporary storage. When `nullptr`, the 
-  //!   required allocation size is written to `temp_storage_bytes` and no work 
+  //! @param[in] d_temp_storage
+  //!   Device-accessible allocation of temporary storage. When `nullptr`, the
+  //!   required allocation size is written to `temp_storage_bytes` and no work
   //!   is done.
   //!
-  //! @param[in,out] temp_storage_bytes 
+  //! @param[in,out] temp_storage_bytes
   //!   Reference to size in bytes of `d_temp_storage` allocation
   //!
-  //! @param[in,out] d_keys 
-  //!   Reference to the double-buffer of keys whose "current" device-accessible 
-  //!   buffer contains the unsorted input keys and, upon return, is updated to 
+  //! @param[in,out] d_keys
+  //!   Reference to the double-buffer of keys whose "current" device-accessible
+  //!   buffer contains the unsorted input keys and, upon return, is updated to
   //!   point to the sorted output keys
   //!
-  //! @param[in,out] d_values 
-  //!   Double-buffer of values whose "current" device-accessible buffer 
-  //!   contains the unsorted input values and, upon return, is updated to point 
+  //! @param[in,out] d_values
+  //!   Double-buffer of values whose "current" device-accessible buffer
+  //!   contains the unsorted input values and, upon return, is updated to point
   //!   to the sorted output values
   //!
-  //! @param[in] num_items 
+  //! @param[in] num_items
   //!   Number of items to sort
   //!
   //! @param decomposer
   //!   Callable object responsible for decomposing a ``KeyT`` into a tuple of
-  //!   references to its constituent arithmetic types. The leftmost element of 
-  //!   the tuple is considered the most significant. The call operator must not 
+  //!   references to its constituent arithmetic types. The leftmost element of
+  //!   the tuple is considered the most significant. The call operator must not
   //!   modify members of the key.
   //!
-  //! @param[in] stream 
-  //!   **[optional]** CUDA stream to launch kernels within. 
+  //! @param[in] stream
+  //!   **[optional]** CUDA stream to launch kernels within.
   //!   Default is stream<sub>0</sub>.
   template <typename KeyT, typename ValueT, typename NumItemsT, typename DecomposerT>
   CUB_RUNTIME_FUNCTION static                                //
@@ -963,14 +970,14 @@ public:
                                                              stream);
   }
 
-  //! @rst 
+  //! @rst
   //! Sorts key-value pairs into ascending order using :math:`\approx N` auxiliary storage.
-  //! 
+  //!
   //! * The sorting operation is given a pair of key buffers and a corresponding
   //!   pair of associated value buffers.  Each pair is managed by a DoubleBuffer
   //!   structure that indicates which of the two buffers is "current" (and thus
   //!   contains the input data to be sorted).
-  //! * The contents of both buffers within each pair may be altered by the 
+  //! * The contents of both buffers within each pair may be altered by the
   //!   sorting operation.
   //! * In-place operations are not supported. There must be no overlap between
   //!   any of the provided ranges:
@@ -980,12 +987,12 @@ public:
   //!   - ``[d_values.Current(),   d_values.Current()   + num_items)``
   //!   - ``[d_values.Alternate(), d_values.Alternate() + num_items)``
   //!
-  //! - Upon completion, the sorting operation will update the "current" 
-  //!   indicator within each DoubleBuffer wrapper to reference which of the two 
-  //!   buffers now contains the sorted output sequence (a function of the 
+  //! - Upon completion, the sorting operation will update the "current"
+  //!   indicator within each DoubleBuffer wrapper to reference which of the two
+  //!   buffers now contains the sorted output sequence (a function of the
   //!   number of key bits specified and the targeted device architecture).
-  //! - An optional bit subrange ``[begin_bit, end_bit)`` of differentiating key 
-  //!   bits can be specified. This can reduce overall sorting overhead and 
+  //! - An optional bit subrange ``[begin_bit, end_bit)`` of differentiating key
+  //!   bits can be specified. This can reduce overall sorting overhead and
   //!   yield a corresponding performance improvement.
   //! - @devicestorageP
   //! - @devicestorage
@@ -1006,7 +1013,7 @@ public:
   //!
   //! The following snippet shows how to sort an array of ``custom_t`` objects
   //! using ``cub::DeviceRadixSort::SortPairs``:
-  //! 
+  //!
   //! .. literalinclude:: ../../test/catch2_test_device_radix_sort_custom.cu
   //!     :language: c++
   //!     :dedent:
@@ -1015,59 +1022,59 @@ public:
   //!
   //! @endrst
   //!
-  //! @tparam KeyT      
+  //! @tparam KeyT
   //!   **[inferred]** KeyT type
   //!
-  //! @tparam ValueT    
+  //! @tparam ValueT
   //!   **[inferred]** ValueT type
   //!
-  //! @tparam NumItemsT 
+  //! @tparam NumItemsT
   //!   **[inferred]** Type of num_items
   //!
   //! @tparam DecomposerT
-  //!   **[inferred]** Type of a callable object responsible for decomposing a 
+  //!   **[inferred]** Type of a callable object responsible for decomposing a
   //!   ``KeyT`` into a tuple of references to its constituent arithmetic types:
-  //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``. 
-  //!   The leftmost element of the tuple is considered the most significant. 
+  //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``.
+  //!   The leftmost element of the tuple is considered the most significant.
   //!   The call operator must not modify members of the key.
   //!
-  //! @param[in] d_temp_storage 
-  //!   Device-accessible allocation of temporary storage. When `nullptr`, the 
-  //!   required allocation size is written to `temp_storage_bytes` and no work 
+  //! @param[in] d_temp_storage
+  //!   Device-accessible allocation of temporary storage. When `nullptr`, the
+  //!   required allocation size is written to `temp_storage_bytes` and no work
   //!   is done.
   //!
-  //! @param[in,out] temp_storage_bytes 
+  //! @param[in,out] temp_storage_bytes
   //!   Reference to size in bytes of `d_temp_storage` allocation
   //!
-  //! @param[in,out] d_keys 
-  //!   Reference to the double-buffer of keys whose "current" device-accessible 
-  //!   buffer contains the unsorted input keys and, upon return, is updated to 
+  //! @param[in,out] d_keys
+  //!   Reference to the double-buffer of keys whose "current" device-accessible
+  //!   buffer contains the unsorted input keys and, upon return, is updated to
   //!   point to the sorted output keys
   //!
-  //! @param[in,out] d_values 
-  //!   Double-buffer of values whose "current" device-accessible buffer 
-  //!   contains the unsorted input values and, upon return, is updated to point 
+  //! @param[in,out] d_values
+  //!   Double-buffer of values whose "current" device-accessible buffer
+  //!   contains the unsorted input values and, upon return, is updated to point
   //!   to the sorted output values
   //!
-  //! @param[in] num_items 
+  //! @param[in] num_items
   //!   Number of items to sort
   //!
   //! @param decomposer
   //!   Callable object responsible for decomposing a ``KeyT`` into a tuple of
-  //!   references to its constituent arithmetic types. The leftmost element of 
-  //!   the tuple is considered the most significant. The call operator must not 
+  //!   references to its constituent arithmetic types. The leftmost element of
+  //!   the tuple is considered the most significant. The call operator must not
   //!   modify members of the key.
   //!
-  //! @param[in] begin_bit 
-  //!   **[optional]** The least-significant bit index (inclusive) needed for 
+  //! @param[in] begin_bit
+  //!   **[optional]** The least-significant bit index (inclusive) needed for
   //!   key comparison
   //!
-  //! @param[in] end_bit 
-  //!   **[optional]** The most-significant bit index (exclusive) needed for key 
+  //! @param[in] end_bit
+  //!   **[optional]** The most-significant bit index (exclusive) needed for key
   //!   comparison (e.g., `(sizeof(float) + sizeof(long long int)) * 8`)
   //!
-  //! @param[in] stream 
-  //!   **[optional]** CUDA stream to launch kernels within. 
+  //! @param[in] stream
+  //!   **[optional]** CUDA stream to launch kernels within.
   //!   Default is stream<sub>0</sub>.
   template <typename KeyT, typename ValueT, typename NumItemsT, typename DecomposerT>
   CUB_RUNTIME_FUNCTION static                                //
@@ -1108,7 +1115,7 @@ public:
                                                              stream);
   }
 
-  //! @brief Sorts key-value pairs into descending order. 
+  //! @brief Sorts key-value pairs into descending order.
   //!        (`~2N` auxiliary storage required).
   //!
   //! @par
@@ -1121,10 +1128,10 @@ public:
   //!   - `[d_keys_out,   d_keys_out   + num_items)`
   //!   - `[d_values_in,  d_values_in  + num_items)`
   //!   - `[d_values_out, d_values_out + num_items)`
-  //! - An optional bit subrange `[begin_bit, end_bit)` of differentiating key 
-  //!   bits can be specified. This can reduce overall sorting overhead and 
+  //! - An optional bit subrange `[begin_bit, end_bit)` of differentiating key
+  //!   bits can be specified. This can reduce overall sorting overhead and
   //!   yield a corresponding performance improvement.
-  //! - @devicestorageNP  For sorting using only `O(P)` temporary storage, see 
+  //! - @devicestorageNP  For sorting using only `O(P)` temporary storage, see
   //!   the sorting interface using DoubleBuffer wrappers below.
   //! - @devicestorage
   //!
@@ -1132,14 +1139,14 @@ public:
   //! Performance is similar to DeviceRadixSort::SortPairs.
   //!
   //! @par Snippet
-  //! The code snippet below illustrates the sorting of a device vector of `int` 
+  //! The code snippet below illustrates the sorting of a device vector of `int`
   //! keys with associated vector of `int` values.
   //! @par
   //! @code
-  //! #include <cub/cub.cuh>   
+  //! #include <cub/cub.cuh>
   //! // or equivalently <cub/device/device_radix_sort.cuh>
   //!
-  //! // Declare, allocate, and initialize device-accessible pointers 
+  //! // Declare, allocate, and initialize device-accessible pointers
   //! // for sorting data
   //! int  num_items;          // e.g., 7
   //! int  *d_keys_in;         // e.g., [8, 6, 7, 5, 3, 0, 9]
@@ -1167,49 +1174,49 @@ public:
   //! // d_values_out          <-- [6, 0, 2, 1, 3, 4, 5]
   //! @endcode
   //!
-  //! @tparam KeyT      
+  //! @tparam KeyT
   //!   **[inferred]** KeyT type
   //!
-  //! @tparam ValueT    
+  //! @tparam ValueT
   //!   **[inferred]** ValueT type
   //!
-  //! @tparam NumItemsT 
+  //! @tparam NumItemsT
   //!   **[inferred]** Type of num_items
   //!
-  //! @param[in] d_temp_storage 
-  //!   Device-accessible allocation of temporary storage. When `nullptr`, the 
-  //!   required allocation size is written to `temp_storage_bytes` and no work 
+  //! @param[in] d_temp_storage
+  //!   Device-accessible allocation of temporary storage. When `nullptr`, the
+  //!   required allocation size is written to `temp_storage_bytes` and no work
   //!   is done.
   //!
-  //! @param[in,out] temp_storage_bytes 
+  //! @param[in,out] temp_storage_bytes
   //!   Reference to size in bytes of @p d_temp_storage allocation
   //!
-  //! @param[in] d_keys_in 
+  //! @param[in] d_keys_in
   //!   Pointer to the input data of key data to sort
   //!
-  //! @param[out] d_keys_out 
+  //! @param[out] d_keys_out
   //!   Pointer to the sorted output sequence of key data
   //!
-  //! @param[in] d_values_in 
+  //! @param[in] d_values_in
   //!   Pointer to the corresponding input sequence of associated value items
   //!
-  //! @param[out] d_values_out 
-  //!   Pointer to the correspondingly-reordered output sequence of associated 
+  //! @param[out] d_values_out
+  //!   Pointer to the correspondingly-reordered output sequence of associated
   //!   value items
   //!
-  //! @param[in] num_items 
+  //! @param[in] num_items
   //!   Number of items to sort
   //!
-  //! @param[in] begin_bit 
-  //!   **[optional]** The least-significant bit index (inclusive) needed for 
+  //! @param[in] begin_bit
+  //!   **[optional]** The least-significant bit index (inclusive) needed for
   //!   key comparison
   //!
-  //! @param[in] end_bit 
-  //!   **[optional]** The most-significant bit index (exclusive) needed for key 
+  //! @param[in] end_bit
+  //!   **[optional]** The most-significant bit index (exclusive) needed for key
   //!   comparison (e.g., `sizeof(unsigned int) * 8`)
   //!
-  //! @param[in] stream 
-  //!   **[optional]** CUDA stream to launch kernels within. 
+  //! @param[in] stream
+  //!   **[optional]** CUDA stream to launch kernels within.
   //!   Default is stream<sub>0</sub>.
   template <typename KeyT, typename ValueT, typename NumItemsT>
   CUB_RUNTIME_FUNCTION static cudaError_t
@@ -1279,9 +1286,9 @@ public:
   }
   #endif
 
-  //! @rst 
+  //! @rst
   //! Sorts key-value pairs into descending order using :math:`\approx 2N` auxiliary storage.
-  //! 
+  //!
   //! * The contents of the input data are not altered by the sorting operation.
   //! * Pointers to contiguous memory must be used; iterators are not currently
   //!   supported.
@@ -1293,10 +1300,10 @@ public:
   //!   * ``[d_values_in,  d_values_in  + num_items)``
   //!   * ``[d_values_out, d_values_out + num_items)``
   //!
-  //! * A bit subrange ``[begin_bit, end_bit)`` is provided to specify 
-  //!   differentiating key bits. This can reduce overall sorting overhead and 
+  //! * A bit subrange ``[begin_bit, end_bit)`` is provided to specify
+  //!   differentiating key bits. This can reduce overall sorting overhead and
   //!   yield a corresponding performance improvement.
-  //! * @devicestorageNP For sorting using only :math:`O(P)` temporary storage, see 
+  //! * @devicestorageNP For sorting using only :math:`O(P)` temporary storage, see
   //!   the sorting interface using DoubleBuffer wrappers below.
   //! * @devicestorage
   //!
@@ -1316,7 +1323,7 @@ public:
   //!
   //! The following snippet shows how to sort an array of ``custom_t`` objects
   //! using ``cub::DeviceRadixSort::SortPairsDescending``:
-  //! 
+  //!
   //! .. literalinclude:: ../../test/catch2_test_device_radix_sort_custom.cu
   //!     :language: c++
   //!     :dedent:
@@ -1325,62 +1332,62 @@ public:
   //!
   //! @endrst
   //!
-  //! @tparam KeyT      
+  //! @tparam KeyT
   //!   **[inferred]** KeyT type
   //!
-  //! @tparam ValueT    
+  //! @tparam ValueT
   //!   **[inferred]** ValueT type
   //!
-  //! @tparam NumItemsT 
+  //! @tparam NumItemsT
   //!   **[inferred]** Type of num_items
   //!
   //! @tparam DecomposerT
-  //!   **[inferred]** Type of a callable object responsible for decomposing a 
+  //!   **[inferred]** Type of a callable object responsible for decomposing a
   //!   ``KeyT`` into a tuple of references to its constituent arithmetic types:
-  //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``. 
-  //!   The leftmost element of the tuple is considered the most significant. 
+  //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``.
+  //!   The leftmost element of the tuple is considered the most significant.
   //!   The call operator must not modify members of the key.
   //!
-  //! @param[in] d_temp_storage 
-  //!   Device-accessible allocation of temporary storage. When `nullptr`, the 
-  //!   required allocation size is written to `temp_storage_bytes` and no work 
+  //! @param[in] d_temp_storage
+  //!   Device-accessible allocation of temporary storage. When `nullptr`, the
+  //!   required allocation size is written to `temp_storage_bytes` and no work
   //!   is done.
   //!
-  //! @param[in,out] temp_storage_bytes 
+  //! @param[in,out] temp_storage_bytes
   //!   Reference to size in bytes of `d_temp_storage` allocation
   //!
-  //! @param[in] d_keys_in 
+  //! @param[in] d_keys_in
   //!   Pointer to the input data of key data to sort
   //!
-  //! @param[out] d_keys_out 
+  //! @param[out] d_keys_out
   //!   Pointer to the sorted output sequence of key data
   //!
-  //! @param[in] d_values_in 
+  //! @param[in] d_values_in
   //!   Pointer to the corresponding input sequence of associated value items
   //!
-  //! @param[out] d_values_out 
-  //!   Pointer to the correspondingly-reordered output sequence of associated 
+  //! @param[out] d_values_out
+  //!   Pointer to the correspondingly-reordered output sequence of associated
   //!   value items
   //!
-  //! @param[in] num_items 
+  //! @param[in] num_items
   //!   Number of items to sort
   //!
   //! @param decomposer
   //!   Callable object responsible for decomposing a ``KeyT`` into a tuple of
-  //!   references to its constituent arithmetic types. The leftmost element of 
-  //!   the tuple is considered the most significant. The call operator must not 
+  //!   references to its constituent arithmetic types. The leftmost element of
+  //!   the tuple is considered the most significant. The call operator must not
   //!   modify members of the key.
   //!
-  //! @param[in] begin_bit 
-  //!   **[optional]** The least-significant bit index (inclusive) needed for 
+  //! @param[in] begin_bit
+  //!   **[optional]** The least-significant bit index (inclusive) needed for
   //!   key comparison
   //!
-  //! @param[in] end_bit 
-  //!   **[optional]** The most-significant bit index (exclusive) needed for key 
+  //! @param[in] end_bit
+  //!   **[optional]** The most-significant bit index (exclusive) needed for key
   //!   comparison (e.g., `(sizeof(float) + sizeof(long long int)) * 8`)
   //!
-  //! @param[in] stream 
-  //!   **[optional]** CUDA stream to launch kernels within. 
+  //! @param[in] stream
+  //!   **[optional]** CUDA stream to launch kernels within.
   //!   Default is stream<sub>0</sub>.
   template <typename KeyT, typename ValueT, typename NumItemsT, typename DecomposerT>
   CUB_RUNTIME_FUNCTION static                                //
@@ -1429,9 +1436,9 @@ public:
                                                              stream);
   }
 
-  //! @rst 
+  //! @rst
   //! Sorts key-value pairs into descending order using :math:`\approx 2N` auxiliary storage.
-  //! 
+  //!
   //! * The contents of the input data are not altered by the sorting operation.
   //! * Pointers to contiguous memory must be used; iterators are not currently
   //!   supported.
@@ -1443,7 +1450,7 @@ public:
   //!   * ``[d_values_in,  d_values_in  + num_items)``
   //!   * ``[d_values_out, d_values_out + num_items)``
   //!
-  //! * @devicestorageNP For sorting using only :math:`O(P)` temporary storage, see 
+  //! * @devicestorageNP For sorting using only :math:`O(P)` temporary storage, see
   //!   the sorting interface using DoubleBuffer wrappers below.
   //! * @devicestorage
   //!
@@ -1463,7 +1470,7 @@ public:
   //!
   //! The following snippet shows how to sort an array of ``custom_t`` objects
   //! using ``cub::DeviceRadixSort::SortPairsDescending``:
-  //! 
+  //!
   //! .. literalinclude:: ../../test/catch2_test_device_radix_sort_custom.cu
   //!     :language: c++
   //!     :dedent:
@@ -1472,54 +1479,54 @@ public:
   //!
   //! @endrst
   //!
-  //! @tparam KeyT      
+  //! @tparam KeyT
   //!   **[inferred]** KeyT type
   //!
-  //! @tparam ValueT    
+  //! @tparam ValueT
   //!   **[inferred]** ValueT type
   //!
-  //! @tparam NumItemsT 
+  //! @tparam NumItemsT
   //!   **[inferred]** Type of num_items
   //!
   //! @tparam DecomposerT
-  //!   **[inferred]** Type of a callable object responsible for decomposing a 
+  //!   **[inferred]** Type of a callable object responsible for decomposing a
   //!   ``KeyT`` into a tuple of references to its constituent arithmetic types:
-  //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``. 
-  //!   The leftmost element of the tuple is considered the most significant. 
+  //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``.
+  //!   The leftmost element of the tuple is considered the most significant.
   //!   The call operator must not modify members of the key.
   //!
-  //! @param[in] d_temp_storage 
-  //!   Device-accessible allocation of temporary storage. When `nullptr`, the 
-  //!   required allocation size is written to `temp_storage_bytes` and no work 
+  //! @param[in] d_temp_storage
+  //!   Device-accessible allocation of temporary storage. When `nullptr`, the
+  //!   required allocation size is written to `temp_storage_bytes` and no work
   //!   is done.
   //!
-  //! @param[in,out] temp_storage_bytes 
+  //! @param[in,out] temp_storage_bytes
   //!   Reference to size in bytes of `d_temp_storage` allocation
   //!
-  //! @param[in] d_keys_in 
+  //! @param[in] d_keys_in
   //!   Pointer to the input data of key data to sort
   //!
-  //! @param[out] d_keys_out 
+  //! @param[out] d_keys_out
   //!   Pointer to the sorted output sequence of key data
   //!
-  //! @param[in] d_values_in 
+  //! @param[in] d_values_in
   //!   Pointer to the corresponding input sequence of associated value items
   //!
-  //! @param[out] d_values_out 
-  //!   Pointer to the correspondingly-reordered output sequence of associated 
+  //! @param[out] d_values_out
+  //!   Pointer to the correspondingly-reordered output sequence of associated
   //!   value items
   //!
-  //! @param[in] num_items 
+  //! @param[in] num_items
   //!   Number of items to sort
   //!
   //! @param decomposer
   //!   Callable object responsible for decomposing a ``KeyT`` into a tuple of
-  //!   references to its constituent arithmetic types. The leftmost element of 
-  //!   the tuple is considered the most significant. The call operator must not 
+  //!   references to its constituent arithmetic types. The leftmost element of
+  //!   the tuple is considered the most significant. The call operator must not
   //!   modify members of the key.
   //!
-  //! @param[in] stream 
-  //!   **[optional]** CUDA stream to launch kernels within. 
+  //! @param[in] stream
+  //!   **[optional]** CUDA stream to launch kernels within.
   //!   Default is stream<sub>0</sub>.
   template <typename KeyT, typename ValueT, typename NumItemsT, typename DecomposerT>
   CUB_RUNTIME_FUNCTION static                                //
@@ -1564,7 +1571,7 @@ public:
                                                              stream);
   }
 
-  //! @brief Sorts key-value pairs into descending order. 
+  //! @brief Sorts key-value pairs into descending order.
   //!        (`~N` auxiliary storage required).
   //!
   //! @par
@@ -1572,7 +1579,7 @@ public:
   //!   pair of associated value buffers.  Each pair is managed by a DoubleBuffer
   //!   structure that indicates which of the two buffers is "current" (and thus
   //!   contains the input data to be sorted).
-  //! - The contents of both buffers within each pair may be altered by the 
+  //! - The contents of both buffers within each pair may be altered by the
   //!   sorting operation.
   //! - In-place operations are not supported. There must be no overlap between
   //!   any of the provided ranges:
@@ -1580,12 +1587,12 @@ public:
   //!   - `[d_keys.Alternate(),   d_keys.Alternate()   + num_items)`
   //!   - `[d_values.Current(),   d_values.Current()   + num_items)`
   //!   - `[d_values.Alternate(), d_values.Alternate() + num_items)`
-  //! - Upon completion, the sorting operation will update the "current" 
-  //!   indicator within each DoubleBuffer wrapper to reference which of the two 
-  //!   buffers now contains the sorted output sequence (a function of the number 
+  //! - Upon completion, the sorting operation will update the "current"
+  //!   indicator within each DoubleBuffer wrapper to reference which of the two
+  //!   buffers now contains the sorted output sequence (a function of the number
   //!   of key bits specified and the targeted device architecture).
-  //! - An optional bit subrange `[begin_bit, end_bit)` of differentiating key 
-  //!   bits can be specified. This can reduce overall sorting overhead and 
+  //! - An optional bit subrange `[begin_bit, end_bit)` of differentiating key
+  //!   bits can be specified. This can reduce overall sorting overhead and
   //!   yield a corresponding performance improvement.
   //! - @devicestorageP
   //! - @devicestorage
@@ -1594,14 +1601,14 @@ public:
   //! Performance is similar to DeviceRadixSort::SortPairs.
   //!
   //! @par Snippet
-  //! The code snippet below illustrates the sorting of a device vector of `int` 
+  //! The code snippet below illustrates the sorting of a device vector of `int`
   //! keys with associated vector of `int` values.
   //! @par
   //! @code
-  //! #include <cub/cub.cuh>   
+  //! #include <cub/cub.cuh>
   //! // or equivalently <cub/device/device_radix_sort.cuh>
   //!
-  //! // Declare, allocate, and initialize device-accessible pointers 
+  //! // Declare, allocate, and initialize device-accessible pointers
   //! // for sorting data
   //! int  num_items;          // e.g., 7
   //! int  *d_key_buf;         // e.g., [8, 6, 7, 5, 3, 0, 9]
@@ -1631,46 +1638,46 @@ public:
   //! // d_values.Current()    <-- [6, 0, 2, 1, 3, 4, 5]
   //! @endcode
   //!
-  //! @tparam KeyT      
+  //! @tparam KeyT
   //!   **[inferred]** KeyT type
   //!
-  //! @tparam ValueT    
+  //! @tparam ValueT
   //!   **[inferred]** ValueT type
   //!
-  //! @tparam NumItemsT 
+  //! @tparam NumItemsT
   //!   **[inferred]** Type of num_items
   //!
-  //! @param[in] d_temp_storage 
-  //!   Device-accessible allocation of temporary storage. When `nullptr`, the 
-  //!   required allocation size is written to `temp_storage_bytes` and no work 
+  //! @param[in] d_temp_storage
+  //!   Device-accessible allocation of temporary storage. When `nullptr`, the
+  //!   required allocation size is written to `temp_storage_bytes` and no work
   //!   is done.
   //!
-  //! @param[in,out] temp_storage_bytes 
+  //! @param[in,out] temp_storage_bytes
   //!   Reference to size in bytes of `d_temp_storage` allocation
   //!
-  //! @param[in,out] d_keys 
-  //!   Reference to the double-buffer of keys whose "current" device-accessible 
-  //!   buffer contains the unsorted input keys and, upon return, is updated to 
+  //! @param[in,out] d_keys
+  //!   Reference to the double-buffer of keys whose "current" device-accessible
+  //!   buffer contains the unsorted input keys and, upon return, is updated to
   //!   point to the sorted output keys
   //!
-  //! @param[in,out] d_values 
-  //!   Double-buffer of values whose "current" device-accessible buffer 
-  //!   contains the unsorted input values and, upon return, is updated to point 
+  //! @param[in,out] d_values
+  //!   Double-buffer of values whose "current" device-accessible buffer
+  //!   contains the unsorted input values and, upon return, is updated to point
   //!   to the sorted output values
   //!
-  //! @param[in] num_items 
+  //! @param[in] num_items
   //!   Number of items to sort
   //!
-  //! @param[in] begin_bit 
-  //!   **[optional]** The least-significant bit index (inclusive) needed for 
+  //! @param[in] begin_bit
+  //!   **[optional]** The least-significant bit index (inclusive) needed for
   //!   key comparison
   //!
-  //! @param[in] end_bit 
-  //!   **[optional]** The most-significant bit index (exclusive) needed for key 
+  //! @param[in] end_bit
+  //!   **[optional]** The most-significant bit index (exclusive) needed for key
   //!   comparison (e.g., `sizeof(unsigned int) * 8`)
   //!
-  //! @param[in] stream 
-  //!   **[optional]** CUDA stream to launch kernels within.  
+  //! @param[in] stream
+  //!   **[optional]** CUDA stream to launch kernels within.
   //!   Default is stream<sub>0</sub>.
   template <typename KeyT, typename ValueT, typename NumItemsT>
   CUB_RUNTIME_FUNCTION static cudaError_t
@@ -1727,14 +1734,14 @@ public:
   }
   #endif
 
-  //! @rst 
+  //! @rst
   //! Sorts key-value pairs into descending order using :math:`\approx N` auxiliary storage.
-  //! 
+  //!
   //! * The sorting operation is given a pair of key buffers and a corresponding
   //!   pair of associated value buffers.  Each pair is managed by a DoubleBuffer
   //!   structure that indicates which of the two buffers is "current" (and thus
   //!   contains the input data to be sorted).
-  //! * The contents of both buffers within each pair may be altered by the 
+  //! * The contents of both buffers within each pair may be altered by the
   //!   sorting operation.
   //! * In-place operations are not supported. There must be no overlap between
   //!   any of the provided ranges:
@@ -1744,9 +1751,9 @@ public:
   //!   - ``[d_values.Current(),   d_values.Current()   + num_items)``
   //!   - ``[d_values.Alternate(), d_values.Alternate() + num_items)``
   //!
-  //! - Upon completion, the sorting operation will update the "current" 
-  //!   indicator within each DoubleBuffer wrapper to reference which of the two 
-  //!   buffers now contains the sorted output sequence (a function of the 
+  //! - Upon completion, the sorting operation will update the "current"
+  //!   indicator within each DoubleBuffer wrapper to reference which of the two
+  //!   buffers now contains the sorted output sequence (a function of the
   //!   number of key bits specified and the targeted device architecture).
   //! - @devicestorageP
   //! - @devicestorage
@@ -1767,7 +1774,7 @@ public:
   //!
   //! The following snippet shows how to sort an array of ``custom_t`` objects
   //! using ``cub::DeviceRadixSort::SortPairsDescending``:
-  //! 
+  //!
   //! .. literalinclude:: ../../test/catch2_test_device_radix_sort_custom.cu
   //!     :language: c++
   //!     :dedent:
@@ -1776,51 +1783,51 @@ public:
   //!
   //! @endrst
   //!
-  //! @tparam KeyT      
+  //! @tparam KeyT
   //!   **[inferred]** KeyT type
   //!
-  //! @tparam ValueT    
+  //! @tparam ValueT
   //!   **[inferred]** ValueT type
   //!
-  //! @tparam NumItemsT 
+  //! @tparam NumItemsT
   //!   **[inferred]** Type of num_items
   //!
   //! @tparam DecomposerT
-  //!   **[inferred]** Type of a callable object responsible for decomposing a 
+  //!   **[inferred]** Type of a callable object responsible for decomposing a
   //!   ``KeyT`` into a tuple of references to its constituent arithmetic types:
-  //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``. 
-  //!   The leftmost element of the tuple is considered the most significant. 
+  //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``.
+  //!   The leftmost element of the tuple is considered the most significant.
   //!   The call operator must not modify members of the key.
   //!
-  //! @param[in] d_temp_storage 
-  //!   Device-accessible allocation of temporary storage. When `nullptr`, the 
-  //!   required allocation size is written to `temp_storage_bytes` and no work 
+  //! @param[in] d_temp_storage
+  //!   Device-accessible allocation of temporary storage. When `nullptr`, the
+  //!   required allocation size is written to `temp_storage_bytes` and no work
   //!   is done.
   //!
-  //! @param[in,out] temp_storage_bytes 
+  //! @param[in,out] temp_storage_bytes
   //!   Reference to size in bytes of `d_temp_storage` allocation
   //!
-  //! @param[in,out] d_keys 
-  //!   Reference to the double-buffer of keys whose "current" device-accessible 
-  //!   buffer contains the unsorted input keys and, upon return, is updated to 
+  //! @param[in,out] d_keys
+  //!   Reference to the double-buffer of keys whose "current" device-accessible
+  //!   buffer contains the unsorted input keys and, upon return, is updated to
   //!   point to the sorted output keys
   //!
-  //! @param[in,out] d_values 
-  //!   Double-buffer of values whose "current" device-accessible buffer 
-  //!   contains the unsorted input values and, upon return, is updated to point 
+  //! @param[in,out] d_values
+  //!   Double-buffer of values whose "current" device-accessible buffer
+  //!   contains the unsorted input values and, upon return, is updated to point
   //!   to the sorted output values
   //!
-  //! @param[in] num_items 
+  //! @param[in] num_items
   //!   Number of items to sort
   //!
   //! @param decomposer
   //!   Callable object responsible for decomposing a ``KeyT`` into a tuple of
-  //!   references to its constituent arithmetic types. The leftmost element of 
-  //!   the tuple is considered the most significant. The call operator must not 
+  //!   references to its constituent arithmetic types. The leftmost element of
+  //!   the tuple is considered the most significant. The call operator must not
   //!   modify members of the key.
   //!
-  //! @param[in] stream 
-  //!   **[optional]** CUDA stream to launch kernels within. 
+  //! @param[in] stream
+  //!   **[optional]** CUDA stream to launch kernels within.
   //!   Default is stream<sub>0</sub>.
   template <typename KeyT, typename ValueT, typename NumItemsT, typename DecomposerT>
   CUB_RUNTIME_FUNCTION static                                //
@@ -1857,14 +1864,14 @@ public:
                                                              stream);
   }
 
-  //! @rst 
+  //! @rst
   //! Sorts key-value pairs into descending order using :math:`\approx N` auxiliary storage.
-  //! 
+  //!
   //! * The sorting operation is given a pair of key buffers and a corresponding
   //!   pair of associated value buffers.  Each pair is managed by a DoubleBuffer
   //!   structure that indicates which of the two buffers is "current" (and thus
   //!   contains the input data to be sorted).
-  //! * The contents of both buffers within each pair may be altered by the 
+  //! * The contents of both buffers within each pair may be altered by the
   //!   sorting operation.
   //! * In-place operations are not supported. There must be no overlap between
   //!   any of the provided ranges:
@@ -1874,12 +1881,12 @@ public:
   //!   - ``[d_values.Current(),   d_values.Current()   + num_items)``
   //!   - ``[d_values.Alternate(), d_values.Alternate() + num_items)``
   //!
-  //! - Upon completion, the sorting operation will update the "current" 
-  //!   indicator within each DoubleBuffer wrapper to reference which of the two 
-  //!   buffers now contains the sorted output sequence (a function of the 
+  //! - Upon completion, the sorting operation will update the "current"
+  //!   indicator within each DoubleBuffer wrapper to reference which of the two
+  //!   buffers now contains the sorted output sequence (a function of the
   //!   number of key bits specified and the targeted device architecture).
-  //! - An optional bit subrange ``[begin_bit, end_bit)`` of differentiating key 
-  //!   bits can be specified. This can reduce overall sorting overhead and 
+  //! - An optional bit subrange ``[begin_bit, end_bit)`` of differentiating key
+  //!   bits can be specified. This can reduce overall sorting overhead and
   //!   yield a corresponding performance improvement.
   //! - @devicestorageP
   //! - @devicestorage
@@ -1900,7 +1907,7 @@ public:
   //!
   //! The following snippet shows how to sort an array of ``custom_t`` objects
   //! using ``cub::DeviceRadixSort::SortPairsDescending``:
-  //! 
+  //!
   //! .. literalinclude:: ../../test/catch2_test_device_radix_sort_custom.cu
   //!     :language: c++
   //!     :dedent:
@@ -1909,59 +1916,59 @@ public:
   //!
   //! @endrst
   //!
-  //! @tparam KeyT      
+  //! @tparam KeyT
   //!   **[inferred]** KeyT type
   //!
-  //! @tparam ValueT    
+  //! @tparam ValueT
   //!   **[inferred]** ValueT type
   //!
-  //! @tparam NumItemsT 
+  //! @tparam NumItemsT
   //!   **[inferred]** Type of num_items
   //!
   //! @tparam DecomposerT
-  //!   **[inferred]** Type of a callable object responsible for decomposing a 
+  //!   **[inferred]** Type of a callable object responsible for decomposing a
   //!   ``KeyT`` into a tuple of references to its constituent arithmetic types:
-  //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``. 
-  //!   The leftmost element of the tuple is considered the most significant. 
+  //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``.
+  //!   The leftmost element of the tuple is considered the most significant.
   //!   The call operator must not modify members of the key.
   //!
-  //! @param[in] d_temp_storage 
-  //!   Device-accessible allocation of temporary storage. When `nullptr`, the 
-  //!   required allocation size is written to `temp_storage_bytes` and no work 
+  //! @param[in] d_temp_storage
+  //!   Device-accessible allocation of temporary storage. When `nullptr`, the
+  //!   required allocation size is written to `temp_storage_bytes` and no work
   //!   is done.
   //!
-  //! @param[in,out] temp_storage_bytes 
+  //! @param[in,out] temp_storage_bytes
   //!   Reference to size in bytes of `d_temp_storage` allocation
   //!
-  //! @param[in,out] d_keys 
-  //!   Reference to the double-buffer of keys whose "current" device-accessible 
-  //!   buffer contains the unsorted input keys and, upon return, is updated to 
+  //! @param[in,out] d_keys
+  //!   Reference to the double-buffer of keys whose "current" device-accessible
+  //!   buffer contains the unsorted input keys and, upon return, is updated to
   //!   point to the sorted output keys
   //!
-  //! @param[in,out] d_values 
-  //!   Double-buffer of values whose "current" device-accessible buffer 
-  //!   contains the unsorted input values and, upon return, is updated to point 
+  //! @param[in,out] d_values
+  //!   Double-buffer of values whose "current" device-accessible buffer
+  //!   contains the unsorted input values and, upon return, is updated to point
   //!   to the sorted output values
   //!
-  //! @param[in] num_items 
+  //! @param[in] num_items
   //!   Number of items to sort
   //!
   //! @param decomposer
   //!   Callable object responsible for decomposing a ``KeyT`` into a tuple of
-  //!   references to its constituent arithmetic types. The leftmost element of 
-  //!   the tuple is considered the most significant. The call operator must not 
+  //!   references to its constituent arithmetic types. The leftmost element of
+  //!   the tuple is considered the most significant. The call operator must not
   //!   modify members of the key.
   //!
-  //! @param[in] begin_bit 
-  //!   **[optional]** The least-significant bit index (inclusive) needed for 
+  //! @param[in] begin_bit
+  //!   **[optional]** The least-significant bit index (inclusive) needed for
   //!   key comparison
   //!
-  //! @param[in] end_bit 
-  //!   **[optional]** The most-significant bit index (exclusive) needed for key 
+  //! @param[in] end_bit
+  //!   **[optional]** The most-significant bit index (exclusive) needed for key
   //!   comparison (e.g., `(sizeof(float) + sizeof(long long int)) * 8`)
   //!
-  //! @param[in] stream 
-  //!   **[optional]** CUDA stream to launch kernels within. 
+  //! @param[in] stream
+  //!   **[optional]** CUDA stream to launch kernels within.
   //!   Default is stream<sub>0</sub>.
   template <typename KeyT, typename ValueT, typename NumItemsT, typename DecomposerT>
   CUB_RUNTIME_FUNCTION static                                //
@@ -2009,7 +2016,7 @@ public:
   //@{
 
 
-  //! @brief Sorts keys into ascending order. 
+  //! @brief Sorts keys into ascending order.
   //!        (`~2N` auxiliary storage required)
   //!
   //! @par
@@ -2020,30 +2027,30 @@ public:
   //!   any of the provided ranges:
   //!   - `[d_keys_in,    d_keys_in    + num_items)`
   //!   - `[d_keys_out,   d_keys_out   + num_items)`
-  //! - An optional bit subrange `[begin_bit, end_bit)` of differentiating key 
-  //!   bits can be specified. This can reduce overall sorting overhead and 
+  //! - An optional bit subrange `[begin_bit, end_bit)` of differentiating key
+  //!   bits can be specified. This can reduce overall sorting overhead and
   //!   yield a corresponding performance improvement.
-  //! - @devicestorageNP  For sorting using only `O(P)` temporary storage, see 
+  //! - @devicestorageNP  For sorting using only `O(P)` temporary storage, see
   //!   the sorting interface using DoubleBuffer wrappers below.
   //! - @devicestorage
   //!
   //! @par Performance
-  //! The following charts illustrate saturated sorting performance across 
-  //! different CUDA architectures for uniform-random `uint32` and `uint64` 
+  //! The following charts illustrate saturated sorting performance across
+  //! different CUDA architectures for uniform-random `uint32` and `uint64`
   //! keys, respectively.
   //!
   //! @image html lsb_radix_sort_int32_keys.png
   //! @image html lsb_radix_sort_int64_keys.png
   //!
   //! @par Snippet
-  //! The code snippet below illustrates the sorting of a device vector of 
+  //! The code snippet below illustrates the sorting of a device vector of
   //! `int` keys.
   //! @par
   //! @code
-  //! #include <cub/cub.cuh>   
+  //! #include <cub/cub.cuh>
   //! // or equivalently <cub/device/device_radix_sort.cuh>
   //!
-  //! // Declare, allocate, and initialize device-accessible pointers 
+  //! // Declare, allocate, and initialize device-accessible pointers
   //! // for sorting data
   //! int  num_items;          // e.g., 7
   //! int  *d_keys_in;         // e.g., [8, 6, 7, 5, 3, 0, 9]
@@ -2066,42 +2073,42 @@ public:
   //! // d_keys_out            <-- [0, 3, 5, 6, 7, 8, 9]
   //! @endcode
   //!
-  //! @tparam KeyT      
+  //! @tparam KeyT
   //!   **[inferred]** KeyT type
   //!
-  //! @tparam NumItemsT 
+  //! @tparam NumItemsT
   //!   **[inferred]** Type of num_items
   //!
-  //! @tparam NumItemsT 
+  //! @tparam NumItemsT
   //!   **[inferred]** Type of num_items
   //!
-  //! @param[in] d_temp_storage 
-  //!   Device-accessible allocation of temporary storage. When `nullptr`, the 
-  //!   required allocation size is written to `temp_storage_bytes` and no work 
+  //! @param[in] d_temp_storage
+  //!   Device-accessible allocation of temporary storage. When `nullptr`, the
+  //!   required allocation size is written to `temp_storage_bytes` and no work
   //!   is done.
   //!
-  //! @param[in,out] temp_storage_bytes 
+  //! @param[in,out] temp_storage_bytes
   //!   Reference to size in bytes of `d_temp_storage` allocation
   //!
-  //! @param[in] d_keys_in 
+  //! @param[in] d_keys_in
   //!   Pointer to the input data of key data to sort
   //!
-  //! @param[out] d_keys_out 
+  //! @param[out] d_keys_out
   //!   Pointer to the sorted output sequence of key data
   //!
-  //! @param[in] num_items 
+  //! @param[in] num_items
   //!   Number of items to sort
   //!
-  //! @param[in] begin_bit 
-  //!   **[optional]** The least-significant bit index (inclusive) needed for 
+  //! @param[in] begin_bit
+  //!   **[optional]** The least-significant bit index (inclusive) needed for
   //!   key comparison
   //!
-  //! @param[in] end_bit 
-  //!   **[optional]** The most-significant bit index (exclusive) needed for key 
+  //! @param[in] end_bit
+  //!   **[optional]** The most-significant bit index (exclusive) needed for key
   //!   comparison (e.g., `sizeof(unsigned int) * 8`)
   //!
-  //! @param[in] stream 
-  //!   **[optional]** CUDA stream to launch kernels within. 
+  //! @param[in] stream
+  //!   **[optional]** CUDA stream to launch kernels within.
   //!   Default is stream<sub>0</sub>.
   template <typename KeyT, typename NumItemsT>
   CUB_RUNTIME_FUNCTION static cudaError_t
@@ -2138,9 +2145,9 @@ public:
       stream);
   }
 
-  //! @rst 
+  //! @rst
   //! Sorts keys into ascending order using :math:`\approx 2N` auxiliary storage.
-  //! 
+  //!
   //! * The contents of the input data are not altered by the sorting operation.
   //! * Pointers to contiguous memory must be used; iterators are not currently
   //!   supported.
@@ -2150,10 +2157,10 @@ public:
   //!   * ``[d_keys_in,  d_keys_in  + num_items)``
   //!   * ``[d_keys_out, d_keys_out + num_items)``
   //!
-  //! * A bit subrange ``[begin_bit, end_bit)`` is provided to specify 
-  //!   differentiating key bits. This can reduce overall sorting overhead and 
+  //! * A bit subrange ``[begin_bit, end_bit)`` is provided to specify
+  //!   differentiating key bits. This can reduce overall sorting overhead and
   //!   yield a corresponding performance improvement.
-  //! * @devicestorageNP For sorting using only :math:`O(P)` temporary storage, see 
+  //! * @devicestorageNP For sorting using only :math:`O(P)` temporary storage, see
   //!   the sorting interface using DoubleBuffer wrappers below.
   //! * @devicestorage
   //!
@@ -2173,7 +2180,7 @@ public:
   //!
   //! The following snippet shows how to sort an array of ``custom_t`` objects
   //! using ``cub::DeviceRadixSort::SortKeys``:
-  //! 
+  //!
   //! .. literalinclude:: ../../test/catch2_test_device_radix_sort_custom.cu
   //!     :language: c++
   //!     :dedent:
@@ -2182,52 +2189,52 @@ public:
   //!
   //! @endrst
   //!
-  //! @tparam KeyT      
+  //! @tparam KeyT
   //!   **[inferred]** KeyT type
   //!
-  //! @tparam NumItemsT 
+  //! @tparam NumItemsT
   //!   **[inferred]** Type of num_items
   //!
   //! @tparam DecomposerT
-  //!   **[inferred]** Type of a callable object responsible for decomposing a 
+  //!   **[inferred]** Type of a callable object responsible for decomposing a
   //!   ``KeyT`` into a tuple of references to its constituent arithmetic types:
-  //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``. 
-  //!   The leftmost element of the tuple is considered the most significant. 
+  //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``.
+  //!   The leftmost element of the tuple is considered the most significant.
   //!   The call operator must not modify members of the key.
   //!
-  //! @param[in] d_temp_storage 
-  //!   Device-accessible allocation of temporary storage. When `nullptr`, the 
-  //!   required allocation size is written to `temp_storage_bytes` and no work 
+  //! @param[in] d_temp_storage
+  //!   Device-accessible allocation of temporary storage. When `nullptr`, the
+  //!   required allocation size is written to `temp_storage_bytes` and no work
   //!   is done.
   //!
-  //! @param[in,out] temp_storage_bytes 
+  //! @param[in,out] temp_storage_bytes
   //!   Reference to size in bytes of `d_temp_storage` allocation
   //!
-  //! @param[in] d_keys_in 
+  //! @param[in] d_keys_in
   //!   Pointer to the input data of key data to sort
   //!
-  //! @param[out] d_keys_out 
+  //! @param[out] d_keys_out
   //!   Pointer to the sorted output sequence of key data
   //!
-  //! @param[in] num_items 
+  //! @param[in] num_items
   //!   Number of items to sort
   //!
   //! @param decomposer
   //!   Callable object responsible for decomposing a ``KeyT`` into a tuple of
-  //!   references to its constituent arithmetic types. The leftmost element of 
-  //!   the tuple is considered the most significant. The call operator must not 
+  //!   references to its constituent arithmetic types. The leftmost element of
+  //!   the tuple is considered the most significant. The call operator must not
   //!   modify members of the key.
   //!
-  //! @param[in] begin_bit 
-  //!   **[optional]** The least-significant bit index (inclusive) needed for 
+  //! @param[in] begin_bit
+  //!   **[optional]** The least-significant bit index (inclusive) needed for
   //!   key comparison
   //!
-  //! @param[in] end_bit 
-  //!   **[optional]** The most-significant bit index (exclusive) needed for key 
+  //! @param[in] end_bit
+  //!   **[optional]** The most-significant bit index (exclusive) needed for key
   //!   comparison (e.g., `(sizeof(float) + sizeof(long long int)) * 8`)
   //!
-  //! @param[in] stream 
-  //!   **[optional]** CUDA stream to launch kernels within. 
+  //! @param[in] stream
+  //!   **[optional]** CUDA stream to launch kernels within.
   //!   Default is stream<sub>0</sub>.
   template <typename KeyT, typename NumItemsT, typename DecomposerT>
   CUB_RUNTIME_FUNCTION static                                //
@@ -2274,9 +2281,9 @@ public:
                                                              stream);
   }
 
-  //! @rst 
+  //! @rst
   //! Sorts keys into ascending order using :math:`\approx 2N` auxiliary storage.
-  //! 
+  //!
   //! * The contents of the input data are not altered by the sorting operation.
   //! * Pointers to contiguous memory must be used; iterators are not currently
   //!   supported.
@@ -2286,10 +2293,10 @@ public:
   //!   * ``[d_keys_in,  d_keys_in  + num_items)``
   //!   * ``[d_keys_out, d_keys_out + num_items)``
   //!
-  //! * An optional bit subrange ``[begin_bit, end_bit)`` of differentiating key 
-  //!   bits can be specified. This can reduce overall sorting overhead and 
+  //! * An optional bit subrange ``[begin_bit, end_bit)`` of differentiating key
+  //!   bits can be specified. This can reduce overall sorting overhead and
   //!   yield a corresponding performance improvement.
-  //! * @devicestorageNP For sorting using only :math:`O(P)` temporary storage, see 
+  //! * @devicestorageNP For sorting using only :math:`O(P)` temporary storage, see
   //!   the sorting interface using DoubleBuffer wrappers below.
   //! * @devicestorage
   //!
@@ -2309,7 +2316,7 @@ public:
   //!
   //! The following snippet shows how to sort an array of ``custom_t`` objects
   //! using ``cub::DeviceRadixSort::SortKeys``:
-  //! 
+  //!
   //! .. literalinclude:: ../../test/catch2_test_device_radix_sort_custom.cu
   //!     :language: c++
   //!     :dedent:
@@ -2318,44 +2325,44 @@ public:
   //!
   //! @endrst
   //!
-  //! @tparam KeyT      
+  //! @tparam KeyT
   //!   **[inferred]** KeyT type
   //!
-  //! @tparam NumItemsT 
+  //! @tparam NumItemsT
   //!   **[inferred]** Type of num_items
   //!
   //! @tparam DecomposerT
-  //!   **[inferred]** Type of a callable object responsible for decomposing a 
+  //!   **[inferred]** Type of a callable object responsible for decomposing a
   //!   ``KeyT`` into a tuple of references to its constituent arithmetic types:
-  //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``. 
-  //!   The leftmost element of the tuple is considered the most significant. 
+  //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``.
+  //!   The leftmost element of the tuple is considered the most significant.
   //!   The call operator must not modify members of the key.
   //!
-  //! @param[in] d_temp_storage 
-  //!   Device-accessible allocation of temporary storage. When `nullptr`, the 
-  //!   required allocation size is written to `temp_storage_bytes` and no work 
+  //! @param[in] d_temp_storage
+  //!   Device-accessible allocation of temporary storage. When `nullptr`, the
+  //!   required allocation size is written to `temp_storage_bytes` and no work
   //!   is done.
   //!
-  //! @param[in,out] temp_storage_bytes 
+  //! @param[in,out] temp_storage_bytes
   //!   Reference to size in bytes of `d_temp_storage` allocation
   //!
-  //! @param[in] d_keys_in 
+  //! @param[in] d_keys_in
   //!   Pointer to the input data of key data to sort
   //!
-  //! @param[out] d_keys_out 
+  //! @param[out] d_keys_out
   //!   Pointer to the sorted output sequence of key data
   //!
-  //! @param[in] num_items 
+  //! @param[in] num_items
   //!   Number of items to sort
   //!
   //! @param decomposer
   //!   Callable object responsible for decomposing a ``KeyT`` into a tuple of
-  //!   references to its constituent arithmetic types. The leftmost element of 
-  //!   the tuple is considered the most significant. The call operator must not 
+  //!   references to its constituent arithmetic types. The leftmost element of
+  //!   the tuple is considered the most significant. The call operator must not
   //!   modify members of the key.
   //!
-  //! @param[in] stream 
-  //!   **[optional]** CUDA stream to launch kernels within. 
+  //! @param[in] stream
+  //!   **[optional]** CUDA stream to launch kernels within.
   //!   Default is stream<sub>0</sub>.
   template <typename KeyT, typename NumItemsT, typename DecomposerT>
   CUB_RUNTIME_FUNCTION static                                //
@@ -2437,33 +2444,33 @@ public:
   //!   any of the provided ranges:
   //!   - `[d_keys.Current(),     d_keys.Current()     + num_items)`
   //!   - `[d_keys.Alternate(),   d_keys.Alternate()   + num_items)`
-  //! - Upon completion, the sorting operation will update the "current" 
-  //!   indicator within the DoubleBuffer wrapper to reference which of the two 
-  //!   buffers now contains the sorted output sequence (a function of the 
+  //! - Upon completion, the sorting operation will update the "current"
+  //!   indicator within the DoubleBuffer wrapper to reference which of the two
+  //!   buffers now contains the sorted output sequence (a function of the
   //!   number of key bits specified and the targeted device architecture).
-  //! - An optional bit subrange `[begin_bit, end_bit)` of differentiating key 
-  //!   bits can be specified. This can reduce overall sorting overhead and 
+  //! - An optional bit subrange `[begin_bit, end_bit)` of differentiating key
+  //!   bits can be specified. This can reduce overall sorting overhead and
   //!   yield a corresponding performance improvement.
   //! - @devicestorageP
   //! - @devicestorage
   //!
   //! @par Performance
-  //! The following charts illustrate saturated sorting performance across 
-  //! different CUDA architectures for uniform-random `uint32` and `uint64` 
+  //! The following charts illustrate saturated sorting performance across
+  //! different CUDA architectures for uniform-random `uint32` and `uint64`
   //! keys, respectively.
   //!
   //! @image html lsb_radix_sort_int32_keys.png
   //! @image html lsb_radix_sort_int64_keys.png
   //!
   //! @par Snippet
-  //! The code snippet below illustrates the sorting of a device vector of 
+  //! The code snippet below illustrates the sorting of a device vector of
   //! `int` keys.
   //! @par
   //! @code
-  //! #include <cub/cub.cuh>   
+  //! #include <cub/cub.cuh>
   //! // or equivalently <cub/device/device_radix_sort.cuh>
   //!
-  //! // Declare, allocate, and initialize device-accessible pointers 
+  //! // Declare, allocate, and initialize device-accessible pointers
   //! // for sorting data
   //! int  num_items;          // e.g., 7
   //! int  *d_key_buf;         // e.g., [8, 6, 7, 5, 3, 0, 9]
@@ -2489,38 +2496,38 @@ public:
   //! // d_keys.Current()      <-- [0, 3, 5, 6, 7, 8, 9]
   //! @endcode
   //!
-  //! @tparam KeyT      
+  //! @tparam KeyT
   //!   **[inferred]** KeyT type
   //!
-  //! @tparam NumItemsT 
+  //! @tparam NumItemsT
   //!   **[inferred]** Type of num_items
   //!
-  //! @param[in] d_temp_storage 
-  //!   Device-accessible allocation of temporary storage. When `nullptr`, the 
-  //!   required allocation size is written to `temp_storage_bytes` and no work 
+  //! @param[in] d_temp_storage
+  //!   Device-accessible allocation of temporary storage. When `nullptr`, the
+  //!   required allocation size is written to `temp_storage_bytes` and no work
   //!   is done.
   //!
-  //! @param[in,out] temp_storage_bytes 
+  //! @param[in,out] temp_storage_bytes
   //!   Reference to size in bytes of `d_temp_storage` allocation
   //!
-  //! @param[in,out] d_keys 
-  //!   Reference to the double-buffer of keys whose "current" device-accessible 
-  //!   buffer contains the unsorted input keys and, upon return, is updated to 
+  //! @param[in,out] d_keys
+  //!   Reference to the double-buffer of keys whose "current" device-accessible
+  //!   buffer contains the unsorted input keys and, upon return, is updated to
   //!   point to the sorted output keys
   //!
-  //! @param[in] num_items 
+  //! @param[in] num_items
   //!   Number of items to sort
   //!
-  //! @param[in] begin_bit 
-  //!   **[optional]** The least-significant bit index (inclusive) needed for 
+  //! @param[in] begin_bit
+  //!   **[optional]** The least-significant bit index (inclusive) needed for
   //!   key comparison
   //!
-  //! @param[in] end_bit 
-  //!   **[optional]** The most-significant bit index (exclusive) needed for key 
+  //! @param[in] end_bit
+  //!   **[optional]** The most-significant bit index (exclusive) needed for key
   //!   comparison (e.g., `sizeof(unsigned int) * 8`)
   //!
-  //! @param[in] stream 
-  //!   **[optional]** CUDA stream to launch kernels within. 
+  //! @param[in] stream
+  //!   **[optional]** CUDA stream to launch kernels within.
   //!   Default is stream<sub>0</sub>.
   template <typename KeyT, typename NumItemsT>
   CUB_RUNTIME_FUNCTION static cudaError_t
@@ -2577,9 +2584,9 @@ public:
   }
   #endif
 
-  //! @rst 
+  //! @rst
   //! Sorts keys into ascending order using :math:`\approx N` auxiliary storage.
-  //! 
+  //!
   //! * The sorting operation is given a pair of key buffers managed by a
   //!   DoubleBuffer structure that indicates which of the two buffers is
   //!   "current" (and thus contains the input data to be sorted).
@@ -2590,9 +2597,9 @@ public:
   //!   * ``[d_keys.Current(),     d_keys.Current()     + num_items)``
   //!   * ``[d_keys.Alternate(),   d_keys.Alternate()   + num_items)`
   //!
-  //! * Upon completion, the sorting operation will update the "current" 
-  //!   indicator within the DoubleBuffer wrapper to reference which of the two 
-  //!   buffers now contains the sorted output sequence (a function of the 
+  //! * Upon completion, the sorting operation will update the "current"
+  //!   indicator within the DoubleBuffer wrapper to reference which of the two
+  //!   buffers now contains the sorted output sequence (a function of the
   //!   number of key bits specified and the targeted device architecture).
   //! * @devicestorageP
   //! * @devicestorage
@@ -2613,7 +2620,7 @@ public:
   //!
   //! The following snippet shows how to sort an array of ``custom_t`` objects
   //! using ``cub::DeviceRadixSort::SortKeys``:
-  //! 
+  //!
   //! .. literalinclude:: ../../test/catch2_test_device_radix_sort_custom.cu
   //!     :language: c++
   //!     :dedent:
@@ -2622,43 +2629,43 @@ public:
   //!
   //! @endrst
   //!
-  //! @tparam KeyT      
+  //! @tparam KeyT
   //!   **[inferred]** KeyT type
   //!
-  //! @tparam NumItemsT 
+  //! @tparam NumItemsT
   //!   **[inferred]** Type of num_items
   //!
   //! @tparam DecomposerT
-  //!   **[inferred]** Type of a callable object responsible for decomposing a 
+  //!   **[inferred]** Type of a callable object responsible for decomposing a
   //!   ``KeyT`` into a tuple of references to its constituent arithmetic types:
-  //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``. 
-  //!   The leftmost element of the tuple is considered the most significant. 
+  //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``.
+  //!   The leftmost element of the tuple is considered the most significant.
   //!   The call operator must not modify members of the key.
   //!
-  //! @param[in] d_temp_storage 
-  //!   Device-accessible allocation of temporary storage. When `nullptr`, the 
-  //!   required allocation size is written to `temp_storage_bytes` and no work 
+  //! @param[in] d_temp_storage
+  //!   Device-accessible allocation of temporary storage. When `nullptr`, the
+  //!   required allocation size is written to `temp_storage_bytes` and no work
   //!   is done.
   //!
-  //! @param[in,out] temp_storage_bytes 
+  //! @param[in,out] temp_storage_bytes
   //!   Reference to size in bytes of `d_temp_storage` allocation
   //!
-  //! @param[in,out] d_keys 
-  //!   Reference to the double-buffer of keys whose "current" device-accessible 
-  //!   buffer contains the unsorted input keys and, upon return, is updated to 
+  //! @param[in,out] d_keys
+  //!   Reference to the double-buffer of keys whose "current" device-accessible
+  //!   buffer contains the unsorted input keys and, upon return, is updated to
   //!   point to the sorted output keys
   //!
-  //! @param[in] num_items 
+  //! @param[in] num_items
   //!   Number of items to sort
   //!
   //! @param decomposer
   //!   Callable object responsible for decomposing a ``KeyT`` into a tuple of
-  //!   references to its constituent arithmetic types. The leftmost element of 
-  //!   the tuple is considered the most significant. The call operator must not 
+  //!   references to its constituent arithmetic types. The leftmost element of
+  //!   the tuple is considered the most significant. The call operator must not
   //!   modify members of the key.
   //!
-  //! @param[in] stream 
-  //!   **[optional]** CUDA stream to launch kernels within. 
+  //! @param[in] stream
+  //!   **[optional]** CUDA stream to launch kernels within.
   //!   Default is stream<sub>0</sub>.
   template <typename KeyT, typename NumItemsT, typename DecomposerT>
   CUB_RUNTIME_FUNCTION static                                //
@@ -2695,9 +2702,9 @@ public:
                                                              stream);
   }
 
-  //! @rst 
+  //! @rst
   //! Sorts keys into ascending order using :math:`\approx N` auxiliary storage.
-  //! 
+  //!
   //! * The sorting operation is given a pair of key buffers managed by a
   //!   DoubleBuffer structure that indicates which of the two buffers is
   //!   "current" (and thus contains the input data to be sorted).
@@ -2708,12 +2715,12 @@ public:
   //!   * ``[d_keys.Current(),     d_keys.Current()     + num_items)``
   //!   * ``[d_keys.Alternate(),   d_keys.Alternate()   + num_items)`
   //!
-  //! * A bit subrange ``[begin_bit, end_bit)`` is provided to specify 
-  //!   differentiating key bits. This can reduce overall sorting overhead and 
+  //! * A bit subrange ``[begin_bit, end_bit)`` is provided to specify
+  //!   differentiating key bits. This can reduce overall sorting overhead and
   //!   yield a corresponding performance improvement.
-  //! * Upon completion, the sorting operation will update the "current" 
-  //!   indicator within the DoubleBuffer wrapper to reference which of the two 
-  //!   buffers now contains the sorted output sequence (a function of the 
+  //! * Upon completion, the sorting operation will update the "current"
+  //!   indicator within the DoubleBuffer wrapper to reference which of the two
+  //!   buffers now contains the sorted output sequence (a function of the
   //!   number of key bits specified and the targeted device architecture).
   //! * @devicestorageP
   //! * @devicestorage
@@ -2734,7 +2741,7 @@ public:
   //!
   //! The following snippet shows how to sort an array of ``custom_t`` objects
   //! using ``cub::DeviceRadixSort::SortKeys``:
-  //! 
+  //!
   //! .. literalinclude:: ../../test/catch2_test_device_radix_sort_custom.cu
   //!     :language: c++
   //!     :dedent:
@@ -2743,51 +2750,51 @@ public:
   //!
   //! @endrst
   //!
-  //! @tparam KeyT      
+  //! @tparam KeyT
   //!   **[inferred]** KeyT type
   //!
-  //! @tparam NumItemsT 
+  //! @tparam NumItemsT
   //!   **[inferred]** Type of num_items
   //!
   //! @tparam DecomposerT
-  //!   **[inferred]** Type of a callable object responsible for decomposing a 
+  //!   **[inferred]** Type of a callable object responsible for decomposing a
   //!   ``KeyT`` into a tuple of references to its constituent arithmetic types:
-  //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``. 
-  //!   The leftmost element of the tuple is considered the most significant. 
+  //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``.
+  //!   The leftmost element of the tuple is considered the most significant.
   //!   The call operator must not modify members of the key.
   //!
-  //! @param[in] d_temp_storage 
-  //!   Device-accessible allocation of temporary storage. When `nullptr`, the 
-  //!   required allocation size is written to `temp_storage_bytes` and no work 
+  //! @param[in] d_temp_storage
+  //!   Device-accessible allocation of temporary storage. When `nullptr`, the
+  //!   required allocation size is written to `temp_storage_bytes` and no work
   //!   is done.
   //!
-  //! @param[in,out] temp_storage_bytes 
+  //! @param[in,out] temp_storage_bytes
   //!   Reference to size in bytes of `d_temp_storage` allocation
   //!
-  //! @param[in,out] d_keys 
-  //!   Reference to the double-buffer of keys whose "current" device-accessible 
-  //!   buffer contains the unsorted input keys and, upon return, is updated to 
+  //! @param[in,out] d_keys
+  //!   Reference to the double-buffer of keys whose "current" device-accessible
+  //!   buffer contains the unsorted input keys and, upon return, is updated to
   //!   point to the sorted output keys
   //!
-  //! @param[in] num_items 
+  //! @param[in] num_items
   //!   Number of items to sort
   //!
   //! @param decomposer
   //!   Callable object responsible for decomposing a ``KeyT`` into a tuple of
-  //!   references to its constituent arithmetic types. The leftmost element of 
-  //!   the tuple is considered the most significant. The call operator must not 
+  //!   references to its constituent arithmetic types. The leftmost element of
+  //!   the tuple is considered the most significant. The call operator must not
   //!   modify members of the key.
   //!
-  //! @param[in] begin_bit 
-  //!   **[optional]** The least-significant bit index (inclusive) needed for 
+  //! @param[in] begin_bit
+  //!   **[optional]** The least-significant bit index (inclusive) needed for
   //!   key comparison
   //!
-  //! @param[in] end_bit 
-  //!   **[optional]** The most-significant bit index (exclusive) needed for key 
+  //! @param[in] end_bit
+  //!   **[optional]** The most-significant bit index (exclusive) needed for key
   //!   comparison (e.g., `(sizeof(float) + sizeof(long long int)) * 8`)
   //!
-  //! @param[in] stream 
-  //!   **[optional]** CUDA stream to launch kernels within. 
+  //! @param[in] stream
+  //!   **[optional]** CUDA stream to launch kernels within.
   //!   Default is stream<sub>0</sub>.
   template <typename KeyT, typename NumItemsT, typename DecomposerT>
   CUB_RUNTIME_FUNCTION static                                //
@@ -2827,8 +2834,8 @@ public:
                                                              end_bit,
                                                              stream);
   }
-  
-  //! @brief Sorts keys into descending order. 
+
+  //! @brief Sorts keys into descending order.
   //!        (`~2N` auxiliary storage required).
   //!
   //! @par
@@ -2839,10 +2846,10 @@ public:
   //!   any of the provided ranges:
   //!   - `[d_keys_in,    d_keys_in    + num_items)`
   //!   - `[d_keys_out,   d_keys_out   + num_items)`
-  //! - An optional bit subrange `[begin_bit, end_bit)` of differentiating key 
-  //!   bits can be specified. This can reduce overall sorting overhead and 
+  //! - An optional bit subrange `[begin_bit, end_bit)` of differentiating key
+  //!   bits can be specified. This can reduce overall sorting overhead and
   //!   yield a corresponding performance improvement.
-  //! - @devicestorageNP For sorting using only `O(P)` temporary storage, see 
+  //! - @devicestorageNP For sorting using only `O(P)` temporary storage, see
   //!   the sorting interface using DoubleBuffer wrappers below.
   //! - @devicestorage
   //!
@@ -2850,14 +2857,14 @@ public:
   //! Performance is similar to DeviceRadixSort::SortKeys.
   //!
   //! @par Snippet
-  //! The code snippet below illustrates the sorting of a device vector of 
+  //! The code snippet below illustrates the sorting of a device vector of
   //! `int` keys.
   //! @par
   //! @code
-  //! #include <cub/cub.cuh>   
+  //! #include <cub/cub.cuh>
   //! // or equivalently <cub/device/device_radix_sort.cuh>
   //!
-  //! // Declare, allocate, and initialize device-accessible pointers 
+  //! // Declare, allocate, and initialize device-accessible pointers
   //! // for sorting data
   //! int  num_items;          // e.g., 7
   //! int  *d_keys_in;         // e.g., [8, 6, 7, 5, 3, 0, 9]
@@ -2870,7 +2877,7 @@ public:
   //! // Determine temporary device storage requirements
   //! void     *d_temp_storage = NULL;
   //! size_t   temp_storage_bytes = 0;
-  //! cub::DeviceRadixSort::SortKeysDescending( 
+  //! cub::DeviceRadixSort::SortKeysDescending(
   //!   d_temp_storage, temp_storage_bytes, d_keys_in, d_keys_out, num_items);
   //!
   //! // Allocate temporary storage
@@ -2884,39 +2891,39 @@ public:
   //!
   //! @endcode
   //!
-  //! @tparam KeyT      
+  //! @tparam KeyT
   //!   **[inferred]** KeyT type
   //!
-  //! @tparam NumItemsT 
+  //! @tparam NumItemsT
   //!   **[inferred]** Type of num_items
   //!
-  //! @param[in] d_temp_storage 
-  //!   Device-accessible allocation of temporary storage. When `nullptr`, the 
-  //!   required allocation size is written to `temp_storage_bytes` and no work 
+  //! @param[in] d_temp_storage
+  //!   Device-accessible allocation of temporary storage. When `nullptr`, the
+  //!   required allocation size is written to `temp_storage_bytes` and no work
   //!   is done.
   //!
-  //! @param[in,out] temp_storage_bytes 
+  //! @param[in,out] temp_storage_bytes
   //!   Reference to size in bytes of `d_temp_storage` allocation
   //!
-  //! @param[in] d_keys_in 
+  //! @param[in] d_keys_in
   //!   Pointer to the input data of key data to sort
   //!
-  //! @param[out] d_keys_out 
+  //! @param[out] d_keys_out
   //!   Pointer to the sorted output sequence of key data
   //!
-  //! @param[in] num_items 
+  //! @param[in] num_items
   //!   Number of items to sort
   //!
-  //! @param[in] begin_bit 
-  //!   **[optional]** The least-significant bit index (inclusive) needed for 
+  //! @param[in] begin_bit
+  //!   **[optional]** The least-significant bit index (inclusive) needed for
   //!   key comparison
   //!
-  //! @param[in] end_bit 
-  //!   **[optional]** The most-significant bit index (exclusive) needed for key 
+  //! @param[in] end_bit
+  //!   **[optional]** The most-significant bit index (exclusive) needed for key
   //!   comparison (e.g., `sizeof(unsigned int) * 8`)
   //!
-  //! @param[in] stream 
-  //!   **[optional]** CUDA stream to launch kernels within.  
+  //! @param[in] stream
+  //!   **[optional]** CUDA stream to launch kernels within.
   //!   Default is stream<sub>0</sub>.
   template <typename KeyT, typename NumItemsT>
   CUB_RUNTIME_FUNCTION static cudaError_t
@@ -2979,9 +2986,9 @@ public:
   }
   #endif
 
-  //! @rst 
+  //! @rst
   //! Sorts keys into descending order using :math:`\approx 2N` auxiliary storage.
-  //! 
+  //!
   //! * The contents of the input data are not altered by the sorting operation.
   //! * Pointers to contiguous memory must be used; iterators are not currently
   //!   supported.
@@ -2991,10 +2998,10 @@ public:
   //!   * ``[d_keys_in,  d_keys_in  + num_items)``
   //!   * ``[d_keys_out, d_keys_out + num_items)``
   //!
-  //! * An optional bit subrange ``[begin_bit, end_bit)`` of differentiating key 
-  //!   bits can be specified. This can reduce overall sorting overhead and 
+  //! * An optional bit subrange ``[begin_bit, end_bit)`` of differentiating key
+  //!   bits can be specified. This can reduce overall sorting overhead and
   //!   yield a corresponding performance improvement.
-  //! * @devicestorageNP For sorting using only :math:`O(P)` temporary storage, see 
+  //! * @devicestorageNP For sorting using only :math:`O(P)` temporary storage, see
   //!   the sorting interface using DoubleBuffer wrappers below.
   //! * @devicestorage
   //!
@@ -3014,7 +3021,7 @@ public:
   //!
   //! The following snippet shows how to sort an array of ``custom_t`` objects
   //! using ``cub::DeviceRadixSort::SortKeysDescending``:
-  //! 
+  //!
   //! .. literalinclude:: ../../test/catch2_test_device_radix_sort_custom.cu
   //!     :language: c++
   //!     :dedent:
@@ -3023,52 +3030,52 @@ public:
   //!
   //! @endrst
   //!
-  //! @tparam KeyT      
+  //! @tparam KeyT
   //!   **[inferred]** KeyT type
   //!
-  //! @tparam NumItemsT 
+  //! @tparam NumItemsT
   //!   **[inferred]** Type of num_items
   //!
   //! @tparam DecomposerT
-  //!   **[inferred]** Type of a callable object responsible for decomposing a 
+  //!   **[inferred]** Type of a callable object responsible for decomposing a
   //!   ``KeyT`` into a tuple of references to its constituent arithmetic types:
-  //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``. 
-  //!   The leftmost element of the tuple is considered the most significant. 
+  //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``.
+  //!   The leftmost element of the tuple is considered the most significant.
   //!   The call operator must not modify members of the key.
   //!
-  //! @param[in] d_temp_storage 
-  //!   Device-accessible allocation of temporary storage. When `nullptr`, the 
-  //!   required allocation size is written to `temp_storage_bytes` and no work 
+  //! @param[in] d_temp_storage
+  //!   Device-accessible allocation of temporary storage. When `nullptr`, the
+  //!   required allocation size is written to `temp_storage_bytes` and no work
   //!   is done.
   //!
-  //! @param[in,out] temp_storage_bytes 
+  //! @param[in,out] temp_storage_bytes
   //!   Reference to size in bytes of `d_temp_storage` allocation
   //!
-  //! @param[in] d_keys_in 
+  //! @param[in] d_keys_in
   //!   Pointer to the input data of key data to sort
   //!
-  //! @param[out] d_keys_out 
+  //! @param[out] d_keys_out
   //!   Pointer to the sorted output sequence of key data
   //!
-  //! @param[in] num_items 
+  //! @param[in] num_items
   //!   Number of items to sort
   //!
   //! @param decomposer
   //!   Callable object responsible for decomposing a ``KeyT`` into a tuple of
-  //!   references to its constituent arithmetic types. The leftmost element of 
-  //!   the tuple is considered the most significant. The call operator must not 
+  //!   references to its constituent arithmetic types. The leftmost element of
+  //!   the tuple is considered the most significant. The call operator must not
   //!   modify members of the key.
   //!
-  //! @param[in] begin_bit 
-  //!   **[optional]** The least-significant bit index (inclusive) needed for 
+  //! @param[in] begin_bit
+  //!   **[optional]** The least-significant bit index (inclusive) needed for
   //!   key comparison
   //!
-  //! @param[in] end_bit 
-  //!   **[optional]** The most-significant bit index (exclusive) needed for key 
+  //! @param[in] end_bit
+  //!   **[optional]** The most-significant bit index (exclusive) needed for key
   //!   comparison (e.g., `(sizeof(float) + sizeof(long long int)) * 8`)
   //!
-  //! @param[in] stream 
-  //!   **[optional]** CUDA stream to launch kernels within. 
+  //! @param[in] stream
+  //!   **[optional]** CUDA stream to launch kernels within.
   //!   Default is stream<sub>0</sub>.
   template <typename KeyT, typename NumItemsT, typename DecomposerT>
   CUB_RUNTIME_FUNCTION static                                //
@@ -3115,9 +3122,9 @@ public:
                                                              stream);
   }
 
-  //! @rst 
+  //! @rst
   //! Sorts keys into descending order using :math:`\approx 2N` auxiliary storage.
-  //! 
+  //!
   //! * The contents of the input data are not altered by the sorting operation.
   //! * Pointers to contiguous memory must be used; iterators are not currently
   //!   supported.
@@ -3127,7 +3134,7 @@ public:
   //!   * ``[d_keys_in,  d_keys_in  + num_items)``
   //!   * ``[d_keys_out, d_keys_out + num_items)``
   //!
-  //! * @devicestorageNP For sorting using only :math:`O(P)` temporary storage, see 
+  //! * @devicestorageNP For sorting using only :math:`O(P)` temporary storage, see
   //!   the sorting interface using DoubleBuffer wrappers below.
   //! * @devicestorage
   //!
@@ -3147,7 +3154,7 @@ public:
   //!
   //! The following snippet shows how to sort an array of ``custom_t`` objects
   //! using ``cub::DeviceRadixSort::SortKeysDescending``:
-  //! 
+  //!
   //! .. literalinclude:: ../../test/catch2_test_device_radix_sort_custom.cu
   //!     :language: c++
   //!     :dedent:
@@ -3156,44 +3163,44 @@ public:
   //!
   //! @endrst
   //!
-  //! @tparam KeyT      
+  //! @tparam KeyT
   //!   **[inferred]** KeyT type
   //!
-  //! @tparam NumItemsT 
+  //! @tparam NumItemsT
   //!   **[inferred]** Type of num_items
   //!
   //! @tparam DecomposerT
-  //!   **[inferred]** Type of a callable object responsible for decomposing a 
+  //!   **[inferred]** Type of a callable object responsible for decomposing a
   //!   ``KeyT`` into a tuple of references to its constituent arithmetic types:
-  //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``. 
-  //!   The leftmost element of the tuple is considered the most significant. 
+  //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``.
+  //!   The leftmost element of the tuple is considered the most significant.
   //!   The call operator must not modify members of the key.
   //!
-  //! @param[in] d_temp_storage 
-  //!   Device-accessible allocation of temporary storage. When `nullptr`, the 
-  //!   required allocation size is written to `temp_storage_bytes` and no work 
+  //! @param[in] d_temp_storage
+  //!   Device-accessible allocation of temporary storage. When `nullptr`, the
+  //!   required allocation size is written to `temp_storage_bytes` and no work
   //!   is done.
   //!
-  //! @param[in,out] temp_storage_bytes 
+  //! @param[in,out] temp_storage_bytes
   //!   Reference to size in bytes of `d_temp_storage` allocation
   //!
-  //! @param[in] d_keys_in 
+  //! @param[in] d_keys_in
   //!   Pointer to the input data of key data to sort
   //!
-  //! @param[out] d_keys_out 
+  //! @param[out] d_keys_out
   //!   Pointer to the sorted output sequence of key data
   //!
-  //! @param[in] num_items 
+  //! @param[in] num_items
   //!   Number of items to sort
   //!
   //! @param decomposer
   //!   Callable object responsible for decomposing a ``KeyT`` into a tuple of
-  //!   references to its constituent arithmetic types. The leftmost element of 
-  //!   the tuple is considered the most significant. The call operator must not 
+  //!   references to its constituent arithmetic types. The leftmost element of
+  //!   the tuple is considered the most significant. The call operator must not
   //!   modify members of the key.
   //!
-  //! @param[in] stream 
-  //!   **[optional]** CUDA stream to launch kernels within. 
+  //! @param[in] stream
+  //!   **[optional]** CUDA stream to launch kernels within.
   //!   Default is stream<sub>0</sub>.
   template <typename KeyT, typename NumItemsT, typename DecomposerT>
   CUB_RUNTIME_FUNCTION static                                //
@@ -3236,7 +3243,7 @@ public:
                                                              stream);
   }
 
-  //! @brief Sorts keys into descending order. 
+  //! @brief Sorts keys into descending order.
   //!        (`~N` auxiliary storage required).
   //!
   //! @par
@@ -3248,12 +3255,12 @@ public:
   //!   any of the provided ranges:
   //!   - `[d_keys.Current(),     d_keys.Current()     + num_items)`
   //!   - `[d_keys.Alternate(),   d_keys.Alternate()   + num_items)`
-  //! - Upon completion, the sorting operation will update the "current" 
-  //!   indicator within the DoubleBuffer wrapper to reference which of the two 
-  //!   buffers now contains the sorted output sequence (a function of the 
+  //! - Upon completion, the sorting operation will update the "current"
+  //!   indicator within the DoubleBuffer wrapper to reference which of the two
+  //!   buffers now contains the sorted output sequence (a function of the
   //!   number of key bits specified and the targeted device architecture).
-  //! - An optional bit subrange `[begin_bit, end_bit)` of differentiating key 
-  //!   bits can be specified. This can reduce overall sorting overhead and 
+  //! - An optional bit subrange `[begin_bit, end_bit)` of differentiating key
+  //!   bits can be specified. This can reduce overall sorting overhead and
   //!   yield a corresponding performance improvement.
   //! - @devicestorageP
   //! - @devicestorage
@@ -3265,10 +3272,10 @@ public:
   //! The code snippet below illustrates the sorting of a device vector of @p int keys.
   //! @par
   //! @code
-  //! #include <cub/cub.cuh>   
+  //! #include <cub/cub.cuh>
   //! // or equivalently <cub/device/device_radix_sort.cuh>
   //!
-  //! // Declare, allocate, and initialize device-accessible pointers 
+  //! // Declare, allocate, and initialize device-accessible pointers
   //! // for sorting data
   //! int  num_items;          // e.g., 7
   //! int  *d_key_buf;         // e.g., [8, 6, 7, 5, 3, 0, 9]
@@ -3294,38 +3301,38 @@ public:
   //! // d_keys.Current()      <-- [9, 8, 7, 6, 5, 3, 0]
   //! @endcode
   //!
-  //! @tparam KeyT      
+  //! @tparam KeyT
   //!   **[inferred]** KeyT type
   //!
-  //! @tparam NumItemsT 
+  //! @tparam NumItemsT
   //!   **[inferred]** Type of num_items
   //!
-  //! @param[in] d_temp_storage 
-  //!   Device-accessible allocation of temporary storage. When `nullptr`, the 
-  //!   required allocation size is written to `temp_storage_bytes` and no work 
+  //! @param[in] d_temp_storage
+  //!   Device-accessible allocation of temporary storage. When `nullptr`, the
+  //!   required allocation size is written to `temp_storage_bytes` and no work
   //!   is done.
   //!
-  //! @param[in,out] temp_storage_bytes 
+  //! @param[in,out] temp_storage_bytes
   //!   Reference to size in bytes of `d_temp_storage` allocation
   //!
-  //! @param[in,out] d_keys 
-  //!   Reference to the double-buffer of keys whose "current" device-accessible 
-  //!   buffer contains the unsorted input keys and, upon return, is updated to 
+  //! @param[in,out] d_keys
+  //!   Reference to the double-buffer of keys whose "current" device-accessible
+  //!   buffer contains the unsorted input keys and, upon return, is updated to
   //!   point to the sorted output keys
   //!
-  //! @param[in] num_items 
+  //! @param[in] num_items
   //!   Number of items to sort
   //!
-  //! @param[in] begin_bit 
-  //!   **[optional]** The least-significant bit index (inclusive) needed for 
+  //! @param[in] begin_bit
+  //!   **[optional]** The least-significant bit index (inclusive) needed for
   //!   key comparison
   //!
-  //! @param[in] end_bit 
-  //!   **[optional]** The most-significant bit index (exclusive) needed for key 
+  //! @param[in] end_bit
+  //!   **[optional]** The most-significant bit index (exclusive) needed for key
   //!   comparison (e.g., `sizeof(unsigned int) * 8`)
   //!
-  //! @param[in] stream 
-  //!   **[optional]** CUDA stream to launch kernels within. 
+  //! @param[in] stream
+  //!   **[optional]** CUDA stream to launch kernels within.
   //!   Default is stream<sub>0</sub>.
   template <typename KeyT, typename NumItemsT>
   CUB_RUNTIME_FUNCTION static cudaError_t
@@ -3382,9 +3389,9 @@ public:
   }
   #endif
 
-  //! @rst 
+  //! @rst
   //! Sorts keys into descending order using :math:`\approx N` auxiliary storage.
-  //! 
+  //!
   //! * The sorting operation is given a pair of key buffers managed by a
   //!   DoubleBuffer structure that indicates which of the two buffers is
   //!   "current" (and thus contains the input data to be sorted).
@@ -3395,9 +3402,9 @@ public:
   //!   * ``[d_keys.Current(),     d_keys.Current()     + num_items)``
   //!   * ``[d_keys.Alternate(),   d_keys.Alternate()   + num_items)`
   //!
-  //! * Upon completion, the sorting operation will update the "current" 
-  //!   indicator within the DoubleBuffer wrapper to reference which of the two 
-  //!   buffers now contains the sorted output sequence (a function of the 
+  //! * Upon completion, the sorting operation will update the "current"
+  //!   indicator within the DoubleBuffer wrapper to reference which of the two
+  //!   buffers now contains the sorted output sequence (a function of the
   //!   number of key bits specified and the targeted device architecture).
   //! * @devicestorageP
   //! * @devicestorage
@@ -3418,7 +3425,7 @@ public:
   //!
   //! The following snippet shows how to sort an array of ``custom_t`` objects
   //! using ``cub::DeviceRadixSort::SortKeysDescending``:
-  //! 
+  //!
   //! .. literalinclude:: ../../test/catch2_test_device_radix_sort_custom.cu
   //!     :language: c++
   //!     :dedent:
@@ -3427,43 +3434,43 @@ public:
   //!
   //! @endrst
   //!
-  //! @tparam KeyT      
+  //! @tparam KeyT
   //!   **[inferred]** KeyT type
   //!
-  //! @tparam NumItemsT 
+  //! @tparam NumItemsT
   //!   **[inferred]** Type of num_items
   //!
   //! @tparam DecomposerT
-  //!   **[inferred]** Type of a callable object responsible for decomposing a 
+  //!   **[inferred]** Type of a callable object responsible for decomposing a
   //!   ``KeyT`` into a tuple of references to its constituent arithmetic types:
-  //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``. 
-  //!   The leftmost element of the tuple is considered the most significant. 
+  //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``.
+  //!   The leftmost element of the tuple is considered the most significant.
   //!   The call operator must not modify members of the key.
   //!
-  //! @param[in] d_temp_storage 
-  //!   Device-accessible allocation of temporary storage. When `nullptr`, the 
-  //!   required allocation size is written to `temp_storage_bytes` and no work 
+  //! @param[in] d_temp_storage
+  //!   Device-accessible allocation of temporary storage. When `nullptr`, the
+  //!   required allocation size is written to `temp_storage_bytes` and no work
   //!   is done.
   //!
-  //! @param[in,out] temp_storage_bytes 
+  //! @param[in,out] temp_storage_bytes
   //!   Reference to size in bytes of `d_temp_storage` allocation
   //!
-  //! @param[in,out] d_keys 
-  //!   Reference to the double-buffer of keys whose "current" device-accessible 
-  //!   buffer contains the unsorted input keys and, upon return, is updated to 
+  //! @param[in,out] d_keys
+  //!   Reference to the double-buffer of keys whose "current" device-accessible
+  //!   buffer contains the unsorted input keys and, upon return, is updated to
   //!   point to the sorted output keys
   //!
-  //! @param[in] num_items 
+  //! @param[in] num_items
   //!   Number of items to sort
   //!
   //! @param decomposer
   //!   Callable object responsible for decomposing a ``KeyT`` into a tuple of
-  //!   references to its constituent arithmetic types. The leftmost element of 
-  //!   the tuple is considered the most significant. The call operator must not 
+  //!   references to its constituent arithmetic types. The leftmost element of
+  //!   the tuple is considered the most significant. The call operator must not
   //!   modify members of the key.
   //!
-  //! @param[in] stream 
-  //!   **[optional]** CUDA stream to launch kernels within. 
+  //! @param[in] stream
+  //!   **[optional]** CUDA stream to launch kernels within.
   //!   Default is stream<sub>0</sub>.
   template <typename KeyT, typename NumItemsT, typename DecomposerT>
   CUB_RUNTIME_FUNCTION static                                //
@@ -3500,9 +3507,9 @@ public:
                                                              stream);
   }
 
-  //! @rst 
+  //! @rst
   //! Sorts keys into descending order using :math:`\approx N` auxiliary storage.
-  //! 
+  //!
   //! * The sorting operation is given a pair of key buffers managed by a
   //!   DoubleBuffer structure that indicates which of the two buffers is
   //!   "current" (and thus contains the input data to be sorted).
@@ -3513,12 +3520,12 @@ public:
   //!   * ``[d_keys.Current(),     d_keys.Current()     + num_items)``
   //!   * ``[d_keys.Alternate(),   d_keys.Alternate()   + num_items)`
   //!
-  //! * A bit subrange ``[begin_bit, end_bit)`` is provided to specify 
-  //!   differentiating key bits. This can reduce overall sorting overhead and 
+  //! * A bit subrange ``[begin_bit, end_bit)`` is provided to specify
+  //!   differentiating key bits. This can reduce overall sorting overhead and
   //!   yield a corresponding performance improvement.
-  //! * Upon completion, the sorting operation will update the "current" 
-  //!   indicator within the DoubleBuffer wrapper to reference which of the two 
-  //!   buffers now contains the sorted output sequence (a function of the 
+  //! * Upon completion, the sorting operation will update the "current"
+  //!   indicator within the DoubleBuffer wrapper to reference which of the two
+  //!   buffers now contains the sorted output sequence (a function of the
   //!   number of key bits specified and the targeted device architecture).
   //! * @devicestorageP
   //! * @devicestorage
@@ -3539,7 +3546,7 @@ public:
   //!
   //! The following snippet shows how to sort an array of ``custom_t`` objects
   //! using ``cub::DeviceRadixSort::SortKeysDescending``:
-  //! 
+  //!
   //! .. literalinclude:: ../../test/catch2_test_device_radix_sort_custom.cu
   //!     :language: c++
   //!     :dedent:
@@ -3548,51 +3555,51 @@ public:
   //!
   //! @endrst
   //!
-  //! @tparam KeyT      
+  //! @tparam KeyT
   //!   **[inferred]** KeyT type
   //!
-  //! @tparam NumItemsT 
+  //! @tparam NumItemsT
   //!   **[inferred]** Type of num_items
   //!
   //! @tparam DecomposerT
-  //!   **[inferred]** Type of a callable object responsible for decomposing a 
+  //!   **[inferred]** Type of a callable object responsible for decomposing a
   //!   ``KeyT`` into a tuple of references to its constituent arithmetic types:
-  //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``. 
-  //!   The leftmost element of the tuple is considered the most significant. 
+  //!   ``::cuda::std::tuple<ArithmeticTs&...> operator()(KeyT &key)``.
+  //!   The leftmost element of the tuple is considered the most significant.
   //!   The call operator must not modify members of the key.
   //!
-  //! @param[in] d_temp_storage 
-  //!   Device-accessible allocation of temporary storage. When `nullptr`, the 
-  //!   required allocation size is written to `temp_storage_bytes` and no work 
+  //! @param[in] d_temp_storage
+  //!   Device-accessible allocation of temporary storage. When `nullptr`, the
+  //!   required allocation size is written to `temp_storage_bytes` and no work
   //!   is done.
   //!
-  //! @param[in,out] temp_storage_bytes 
+  //! @param[in,out] temp_storage_bytes
   //!   Reference to size in bytes of `d_temp_storage` allocation
   //!
-  //! @param[in,out] d_keys 
-  //!   Reference to the double-buffer of keys whose "current" device-accessible 
-  //!   buffer contains the unsorted input keys and, upon return, is updated to 
+  //! @param[in,out] d_keys
+  //!   Reference to the double-buffer of keys whose "current" device-accessible
+  //!   buffer contains the unsorted input keys and, upon return, is updated to
   //!   point to the sorted output keys
   //!
-  //! @param[in] num_items 
+  //! @param[in] num_items
   //!   Number of items to sort
   //!
   //! @param decomposer
   //!   Callable object responsible for decomposing a ``KeyT`` into a tuple of
-  //!   references to its constituent arithmetic types. The leftmost element of 
-  //!   the tuple is considered the most significant. The call operator must not 
+  //!   references to its constituent arithmetic types. The leftmost element of
+  //!   the tuple is considered the most significant. The call operator must not
   //!   modify members of the key.
   //!
-  //! @param[in] begin_bit 
-  //!   **[optional]** The least-significant bit index (inclusive) needed for 
+  //! @param[in] begin_bit
+  //!   **[optional]** The least-significant bit index (inclusive) needed for
   //!   key comparison
   //!
-  //! @param[in] end_bit 
-  //!   **[optional]** The most-significant bit index (exclusive) needed for key 
+  //! @param[in] end_bit
+  //!   **[optional]** The most-significant bit index (exclusive) needed for key
   //!   comparison (e.g., `(sizeof(float) + sizeof(long long int)) * 8`)
   //!
-  //! @param[in] stream 
-  //!   **[optional]** CUDA stream to launch kernels within. 
+  //! @param[in] stream
+  //!   **[optional]** CUDA stream to launch kernels within.
   //!   Default is stream<sub>0</sub>.
   template <typename KeyT, typename NumItemsT, typename DecomposerT>
   CUB_RUNTIME_FUNCTION static                                //
diff --git a/cub/cub/device/device_reduce.cuh b/cub/cub/device/device_reduce.cuh
index f2c4090f8b5..5bd12bc3c28 100644
--- a/cub/cub/device/device_reduce.cuh
+++ b/cub/cub/device/device_reduce.cuh
@@ -13,9 +13,9 @@
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
  *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 
- * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 
- * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  * ARE DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE FOR ANY
  * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
@@ -27,17 +27,24 @@
  ******************************************************************************/
 
 /**
- * @file cub::DeviceReduce provides device-wide, parallel operations for 
- *       computing a reduction across a sequence of data items residing within 
+ * @file cub::DeviceReduce provides device-wide, parallel operations for
+ *       computing a reduction across a sequence of data items residing within
  *       device-accessible memory.
  */
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <iterator>
 #include <limits>
 
-#include <cub/config.cuh>
 #include <cub/detail/choose_offset.cuh>
 #include <cub/device/dispatch/dispatch_reduce.cuh>
 #include <cub/device/dispatch/dispatch_reduce_by_key.cuh>
@@ -50,9 +57,9 @@ CUB_NAMESPACE_BEGIN
 //! @ingroup SingleModule
 //!
 //! @rst
-//! DeviceReduce provides device-wide, parallel operations for computing 
-//! a reduction across a sequence of data items residing within 
-//! device-accessible memory. 
+//! DeviceReduce provides device-wide, parallel operations for computing
+//! a reduction across a sequence of data items residing within
+//! device-accessible memory.
 //!
 //! .. image:: ../img/reduce_logo.png
 //!     :align: center
@@ -60,7 +67,7 @@ CUB_NAMESPACE_BEGIN
 //! Overview
 //! ====================================
 //! A `reduction <http://en.wikipedia.org/wiki/Reduce_(higher-order_function)>`_
-//! (or *fold*) uses a binary combining operator to compute a single aggregate 
+//! (or *fold*) uses a binary combining operator to compute a single aggregate
 //! from a sequence of input elements.
 //!
 //! Usage Considerations
@@ -79,8 +86,8 @@ CUB_NAMESPACE_BEGIN
 //!
 //! @par
 //! The following chart illustrates DeviceReduce::ReduceByKey (summation)
-//! performance across different CUDA architectures for `fp32` values. Segments 
-//! are identified by `int32` keys, and have lengths uniformly sampled 
+//! performance across different CUDA architectures for `fp32` values. Segments
+//! are identified by `int32` keys, and have lengths uniformly sampled
 //! from `[1, 1000]`.
 //!
 //! .. image:: ../img/reduce_by_key_fp32_len_500.png
@@ -90,7 +97,7 @@ CUB_NAMESPACE_BEGIN
 struct DeviceReduce
 {
   /**
-   * @brief Computes a device-wide reduction using the specified binary 
+   * @brief Computes a device-wide reduction using the specified binary
    *        `reduction_op` functor and initial value `init`.
    *
    * @par
@@ -104,11 +111,11 @@ struct DeviceReduce
    * - @devicestorage
    *
    * @par Snippet
-   * The code snippet below illustrates a user-defined min-reduction of a 
+   * The code snippet below illustrates a user-defined min-reduction of a
    * device vector of `int` data elements.
    * @par
    * @code
-   * #include <cub/cub.cuh>   
+   * #include <cub/cub.cuh>
    * // or equivalently <cub/device/device_radix_sort.cuh>
    *
    * // CustomMin functor
@@ -121,7 +128,7 @@ struct DeviceReduce
    *     }
    * };
    *
-   * // Declare, allocate, and initialize device-accessible pointers for 
+   * // Declare, allocate, and initialize device-accessible pointers for
    * // input and output
    * int          num_items;  // e.g., 7
    * int          *d_in;      // e.g., [8, 6, 7, 5, 3, 0, 9]
@@ -134,7 +141,7 @@ struct DeviceReduce
    * void     *d_temp_storage = NULL;
    * size_t   temp_storage_bytes = 0;
    * cub::DeviceReduce::Reduce(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_in, d_out, num_items, min_op, init);
    *
    * // Allocate temporary storage
@@ -142,55 +149,55 @@ struct DeviceReduce
    *
    * // Run reduction
    * cub::DeviceReduce::Reduce(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_in, d_out, num_items, min_op, init);
    *
    * // d_out <-- [0]
    * @endcode
    *
-   * @tparam InputIteratorT       
-   *   **[inferred]** Random-access input iterator type for reading input 
+   * @tparam InputIteratorT
+   *   **[inferred]** Random-access input iterator type for reading input
    *   items \iterator
    *
-   * @tparam OutputIteratorT      
-   *   **[inferred]** Output iterator type for recording the reduced 
+   * @tparam OutputIteratorT
+   *   **[inferred]** Output iterator type for recording the reduced
    *   aggregate \iterator
    *
-   * @tparam ReductionOpT         
-   *   **[inferred]** Binary reduction functor type having member 
+   * @tparam ReductionOpT
+   *   **[inferred]** Binary reduction functor type having member
    *   `T operator()(const T &a, const T &b)`
    *
-   * @tparam T                    
-   *   **[inferred]** Data element type that is convertible to the `value` type 
+   * @tparam T
+   *   **[inferred]** Data element type that is convertible to the `value` type
    *   of `InputIteratorT`
    *
    * @tparam NumItemsT **[inferred]** Type of num_items
    *
-   * @param[in] d_temp_storage 
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no work 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no work
    *   is done.
    *
-   * @param[in,out] temp_storage_bytes 
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of `d_temp_storage` allocation
    *
-   * @param d_in[in] 
+   * @param d_in[in]
    *   Pointer to the input sequence of data items
    *
-   * @param d_out[out] 
+   * @param d_out[out]
    *   Pointer to the output aggregate
    *
-   * @param num_items[in] 
+   * @param num_items[in]
    *   Total number of input items (i.e., length of `d_in`)
    *
-   * @param reduction_op[in] 
+   * @param reduction_op[in]
    *   Binary reduction functor
    *
-   * @param[in] init  
+   * @param[in] init
    *   Initial value of the reduction
    *
-   * @param[in] stream  
-   *   **[optional]** CUDA stream to launch kernels within.  
+   * @param[in] stream
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    */
   template <typename InputIteratorT,
@@ -267,21 +274,21 @@ struct DeviceReduce
    * - @devicestorage
    *
    * @par Performance
-   * The following charts illustrate saturated sum-reduction performance across 
+   * The following charts illustrate saturated sum-reduction performance across
    * different CUDA architectures for `int32` and `int64` items, respectively.
    *
    * @image html reduce_int32.png
    * @image html reduce_int64.png
    *
    * @par Snippet
-   * The code snippet below illustrates the sum-reduction of a device vector 
+   * The code snippet below illustrates the sum-reduction of a device vector
    * of `int` data elements.
    * @par
    * @code
-   * #include <cub/cub.cuh>   
+   * #include <cub/cub.cuh>
    * // or equivalently <cub/device/device_radix_sort.cuh>
    *
-   * // Declare, allocate, and initialize device-accessible pointers 
+   * // Declare, allocate, and initialize device-accessible pointers
    * // for input and output
    * int  num_items;      // e.g., 7
    * int  *d_in;          // e.g., [8, 6, 7, 5, 3, 0, 9]
@@ -304,39 +311,39 @@ struct DeviceReduce
    * // d_out <-- [38]
    * @endcode
    *
-   * @tparam InputIteratorT     
-   *   **[inferred]** Random-access input iterator type for reading input 
+   * @tparam InputIteratorT
+   *   **[inferred]** Random-access input iterator type for reading input
    *   items \iterator
    *
-   * @tparam OutputIteratorT    
-   *   **[inferred]** Output iterator type for recording the reduced 
+   * @tparam OutputIteratorT
+   *   **[inferred]** Output iterator type for recording the reduced
    *   aggregate \iterator
    *
    * @tparam NumItemsT **[inferred]** Type of num_items
    *
-   * @param[in] d_temp_storage  
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no work 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no work
    *   is done.
    *
-   * @param[in,out] temp_storage_bytes  
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of `d_temp_storage` allocation
    *
-   * @param[in] d_in  
+   * @param[in] d_in
    *   Pointer to the input sequence of data items
    *
-   * @param[out] d_out  
+   * @param[out] d_out
    *   Pointer to the output aggregate
    *
-   * @param[in] num_items  
+   * @param[in] num_items
    *   Total number of input items (i.e., length of `d_in`)
    *
-   * @param[in] stream  
-   *   **[optional]** CUDA stream to launch kernels within.  
+   * @param[in] stream
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    */
-  template <typename InputIteratorT, 
-            typename OutputIteratorT, 
+  template <typename InputIteratorT,
+            typename OutputIteratorT,
             typename NumItemsT>
   CUB_RUNTIME_FUNCTION static cudaError_t Sum(void *d_temp_storage,
                                               size_t &temp_storage_bytes,
@@ -353,12 +360,12 @@ struct DeviceReduce
       cub::detail::non_void_value_t<OutputIteratorT,
                                     cub::detail::value_t<InputIteratorT>>;
 
-    using InitT = OutputT; 
+    using InitT = OutputT;
 
-    return DispatchReduce<InputIteratorT, 
-                          OutputIteratorT,  
-                          OffsetT, 
-                          cub::Sum, 
+    return DispatchReduce<InputIteratorT,
+                          OutputIteratorT,
+                          OffsetT,
+                          cub::Sum,
                           InitT>::Dispatch(d_temp_storage,
                                            temp_storage_bytes,
                                            d_in,
@@ -404,14 +411,14 @@ struct DeviceReduce
    * - @devicestorage
    *
    * @par Snippet
-   * The code snippet below illustrates the min-reduction of a device vector of 
+   * The code snippet below illustrates the min-reduction of a device vector of
    * `int` data elements.
    * @par
    * @code
-   * #include <cub/cub.cuh>   
+   * #include <cub/cub.cuh>
    * // or equivalently <cub/device/device_radix_sort.cuh>
    *
-   * // Declare, allocate, and initialize device-accessible pointers 
+   * // Declare, allocate, and initialize device-accessible pointers
    * // for input and output
    * int  num_items;      // e.g., 7
    * int  *d_in;          // e.g., [8, 6, 7, 5, 3, 0, 9]
@@ -434,35 +441,35 @@ struct DeviceReduce
    * // d_out <-- [0]
    * @endcode
    *
-   * @tparam InputIteratorT     
-   *   **[inferred]** Random-access input iterator type for reading input 
+   * @tparam InputIteratorT
+   *   **[inferred]** Random-access input iterator type for reading input
    *   items \iterator
    *
-   * @tparam OutputIteratorT    
-   *   **[inferred]** Output iterator type for recording the reduced 
+   * @tparam OutputIteratorT
+   *   **[inferred]** Output iterator type for recording the reduced
    *   aggregate \iterator
    *
    * @tparam NumItemsT **[inferred]** Type of num_items
    *
-   * @param[in] d_temp_storage  
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no work 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no work
    *   is done.
    *
-   * @param[in,out] temp_storage_bytes  
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of `d_temp_storage` allocation
    *
-   * @param[in] d_in  
+   * @param[in] d_in
    *   Pointer to the input sequence of data items
    *
-   * @param[out] d_out  
+   * @param[out] d_out
    *   Pointer to the output aggregate
    *
-   * @param[in] num_items  
+   * @param[in] num_items
    *   Total number of input items (i.e., length of `d_in`)
    *
-   * @param[in] stream  
-   *   **[optional]** CUDA stream to launch kernels within.  
+   * @param[in] stream
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    */
   template <typename InputIteratorT,
@@ -483,9 +490,9 @@ struct DeviceReduce
 
     using InitT = InputT;
 
-    return DispatchReduce<InputIteratorT,   
-                          OutputIteratorT,  
-                          OffsetT, 
+    return DispatchReduce<InputIteratorT,
+                          OutputIteratorT,
+                          OffsetT,
                           cub::Min,
                           InitT>::Dispatch(d_temp_storage,
                                            temp_storage_bytes,
@@ -493,10 +500,10 @@ struct DeviceReduce
                                            d_out,
                                            static_cast<OffsetT>(num_items),
                                            cub::Min(),
-                                           // replace with 
+                                           // replace with
                                            // std::numeric_limits<T>::max() when
                                            // C++11 support is more prevalent
-                                           Traits<InitT>::Max(), 
+                                           Traits<InitT>::Max(),
                                            stream);
   }
 
@@ -521,15 +528,15 @@ struct DeviceReduce
   }
 
   /**
-   * @brief Finds the first device-wide minimum using the less-than ('<') 
+   * @brief Finds the first device-wide minimum using the less-than ('<')
    *        operator, also returning the index of that item.
    *
    * @par
-   * - The output value type of `d_out` is cub::KeyValuePair `<int, T>` 
+   * - The output value type of `d_out` is cub::KeyValuePair `<int, T>`
    *   (assuming the value type of `d_in` is `T`)
-   *   - The minimum is written to `d_out.value` and its offset in the input 
+   *   - The minimum is written to `d_out.value` and its offset in the input
    *     array is written to `d_out.key`.
-   *   - The `{1, std::numeric_limits<T>::max()}` tuple is produced for 
+   *   - The `{1, std::numeric_limits<T>::max()}` tuple is produced for
    *     zero-length inputs
    * - Does not support `<` operators that are non-commutative.
    * - Provides "run-to-run" determinism for pseudo-associative reduction
@@ -541,14 +548,14 @@ struct DeviceReduce
    * - @devicestorage
    *
    * @par Snippet
-   * The code snippet below illustrates the argmin-reduction of a device vector 
+   * The code snippet below illustrates the argmin-reduction of a device vector
    * of `int` data elements.
    * @par
    * @code
-   * #include <cub/cub.cuh>   
+   * #include <cub/cub.cuh>
    * // or equivalently <cub/device/device_radix_sort.cuh>
    *
-   * // Declare, allocate, and initialize device-accessible pointers 
+   * // Declare, allocate, and initialize device-accessible pointers
    * // for input and output
    * int                      num_items;      // e.g., 7
    * int                      *d_in;          // e.g., [8, 6, 7, 5, 3, 0, 9]
@@ -572,35 +579,35 @@ struct DeviceReduce
    *
    * @endcode
    *
-   * @tparam InputIteratorT     
-   *   **[inferred]** Random-access input iterator type for reading input items 
+   * @tparam InputIteratorT
+   *   **[inferred]** Random-access input iterator type for reading input items
    *   (of some type `T`) \iterator
    *
-   * @tparam OutputIteratorT    
-   *   **[inferred]** Output iterator type for recording the reduced aggregate 
+   * @tparam OutputIteratorT
+   *   **[inferred]** Output iterator type for recording the reduced aggregate
    *   (having value type `cub::KeyValuePair<int, T>`) \iterator
    *
-   * @param[in] d_temp_storage  
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
    *   required allocation size is written to \p temp_storage_bytes and no work is done.
    *
-   * @param[in,out] temp_storage_bytes  
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of `d_temp_storage` allocation
    *
-   * @param[in] d_in  
+   * @param[in] d_in
    *   Pointer to the input sequence of data items
    *
-   * @param[out] d_out  
+   * @param[out] d_out
    *   Pointer to the output aggregate
    *
-   * @param[in] num_items  
+   * @param[in] num_items
    *   Total number of input items (i.e., length of `d_in`)
    *
-   * @param[in] stream  
-   *   **[optional]** CUDA stream to launch kernels within.  
+   * @param[in] stream
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    */
-  template <typename InputIteratorT, 
+  template <typename InputIteratorT,
             typename OutputIteratorT>
   CUB_RUNTIME_FUNCTION static cudaError_t ArgMin(void *d_temp_storage,
                                                  size_t &temp_storage_bytes,
@@ -620,7 +627,7 @@ struct DeviceReduce
       cub::detail::non_void_value_t<OutputIteratorT, KeyValuePair<OffsetT, InputValueT>>;
 
     using AccumT = OutputTupleT;
-    
+
     using InitT = detail::reduce::empty_problem_init_t<AccumT>;
 
     // The output value type
@@ -634,7 +641,7 @@ struct DeviceReduce
 
     // Initial value
     // TODO Address https://github.com/NVIDIA/cub/issues/651
-    InitT initial_value{AccumT(1, Traits<InputValueT>::Max())}; 
+    InitT initial_value{AccumT(1, Traits<InputValueT>::Max())};
 
     return DispatchReduce<ArgIndexInputIteratorT,
                           OutputIteratorT,
@@ -675,7 +682,7 @@ struct DeviceReduce
    * @brief Computes a device-wide maximum using the greater-than ('>') operator.
    *
    * @par
-   * - Uses `std::numeric_limits<T>::lowest()` as the initial value of the 
+   * - Uses `std::numeric_limits<T>::lowest()` as the initial value of the
    *   reduction.
    * - Does not support `>` operators that are non-commutative.
    * - Provides "run-to-run" determinism for pseudo-associative reduction
@@ -687,14 +694,14 @@ struct DeviceReduce
    * - @devicestorage
    *
    * @par Snippet
-   * The code snippet below illustrates the max-reduction of a device vector of 
+   * The code snippet below illustrates the max-reduction of a device vector of
    * `int` data elements.
    * @par
    * @code
-   * #include <cub/cub.cuh>   
+   * #include <cub/cub.cuh>
    * // or equivalently <cub/device/device_radix_sort.cuh>
    *
-   * // Declare, allocate, and initialize device-accessible pointers 
+   * // Declare, allocate, and initialize device-accessible pointers
    * // for input and output
    * int  num_items;      // e.g., 7
    * int  *d_in;          // e.g., [8, 6, 7, 5, 3, 0, 9]
@@ -717,38 +724,38 @@ struct DeviceReduce
    * // d_out <-- [9]
    * @endcode
    *
-   * @tparam InputIteratorT     
-   *   **[inferred]** Random-access input iterator type for reading input 
+   * @tparam InputIteratorT
+   *   **[inferred]** Random-access input iterator type for reading input
    *   items \iterator
    *
-   * @tparam OutputIteratorT    
-   *   **[inferred]** Output iterator type for recording the reduced 
+   * @tparam OutputIteratorT
+   *   **[inferred]** Output iterator type for recording the reduced
    *   aggregate \iterator
    *
    * @tparam NumItemsT **[inferred]** Type of num_items
    *
-   * @param[in] d_temp_storage  
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no work 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no work
    *   is done.
    *
-   * @param[in,out] temp_storage_bytes  
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of `d_temp_storage` allocation
    *
-   * @param[in] d_in  
+   * @param[in] d_in
    *   Pointer to the input sequence of data items
    *
-   * @param[out] d_out  
+   * @param[out] d_out
    *   Pointer to the output aggregate
    *
-   * @param[in] num_items  
+   * @param[in] num_items
    *   Total number of input items (i.e., length of `d_in`)
    *
-   * @param[in] stream  
-   *   **[optional]** CUDA stream to launch kernels within. 
+   * @param[in] stream
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    */
-  template <typename InputIteratorT, 
+  template <typename InputIteratorT,
             typename OutputIteratorT,
             typename NumItemsT>
   CUB_RUNTIME_FUNCTION static cudaError_t Max(void *d_temp_storage,
@@ -766,9 +773,9 @@ struct DeviceReduce
 
     using InitT = InputT;
 
-    return DispatchReduce<InputIteratorT,   
-                          OutputIteratorT,  
-                          OffsetT, 
+    return DispatchReduce<InputIteratorT,
+                          OutputIteratorT,
+                          OffsetT,
                           cub::Max,
                           InitT>::Dispatch(d_temp_storage,
                                            temp_storage_bytes,
@@ -776,11 +783,11 @@ struct DeviceReduce
                                            d_out,
                                            static_cast<OffsetT>(num_items),
                                            cub::Max(),
-                                           // replace with 
+                                           // replace with
                                            // std::numeric_limits<T>::lowest()
-                                           // when C++11 support is more 
+                                           // when C++11 support is more
                                            // prevalent
-                                           Traits<InitT>::Lowest(), 
+                                           Traits<InitT>::Lowest(),
                                            stream);
   }
 
@@ -805,15 +812,15 @@ struct DeviceReduce
   }
 
   /**
-   * @brief Finds the first device-wide maximum using the greater-than ('>') 
+   * @brief Finds the first device-wide maximum using the greater-than ('>')
    *        operator, also returning the index of that item
    *
    * @par
-   * - The output value type of `d_out` is cub::KeyValuePair `<int, T>` 
+   * - The output value type of `d_out` is cub::KeyValuePair `<int, T>`
    *   (assuming the value type of `d_in` is `T`)
-   *   - The maximum is written to `d_out.value` and its offset in the input 
+   *   - The maximum is written to `d_out.value` and its offset in the input
    *     array is written to `d_out.key`.
-   *   - The `{1, std::numeric_limits<T>::lowest()}` tuple is produced for 
+   *   - The `{1, std::numeric_limits<T>::lowest()}` tuple is produced for
    *     zero-length inputs
    * - Does not support `>` operators that are non-commutative.
    * - Provides "run-to-run" determinism for pseudo-associative reduction
@@ -825,14 +832,14 @@ struct DeviceReduce
    * - @devicestorage
    *
    * @par Snippet
-   * The code snippet below illustrates the argmax-reduction of a device vector 
+   * The code snippet below illustrates the argmax-reduction of a device vector
    * of `int` data elements.
    * @par
    * @code
-   * #include <cub/cub.cuh>   
+   * #include <cub/cub.cuh>
    * // or equivalently <cub/device/device_reduce.cuh>
    *
-   * // Declare, allocate, and initialize device-accessible pointers 
+   * // Declare, allocate, and initialize device-accessible pointers
    * // for input and output
    * int                      num_items;      // e.g., 7
    * int                      *d_in;          // e.g., [8, 6, 7, 5, 3, 0, 9]
@@ -856,36 +863,36 @@ struct DeviceReduce
    *
    * @endcode
    *
-   * @tparam InputIteratorT     
-   *   **[inferred]** Random-access input iterator type for reading input items 
+   * @tparam InputIteratorT
+   *   **[inferred]** Random-access input iterator type for reading input items
    *   (of some type \p T) \iterator
    *
-   * @tparam OutputIteratorT    
-   *   **[inferred]** Output iterator type for recording the reduced aggregate 
+   * @tparam OutputIteratorT
+   *   **[inferred]** Output iterator type for recording the reduced aggregate
    *   (having value type `cub::KeyValuePair<int, T>`) \iterator
    *
-   * @param[in] d_temp_storage  
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no work 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no work
    *   is done.
    *
-   * @param[in,out] temp_storage_bytes  
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of `d_temp_storage` allocation
    *
-   * @param[in] d_in  
+   * @param[in] d_in
    *   Pointer to the input sequence of data items
    *
-   * @param[out] d_out  
+   * @param[out] d_out
    *   Pointer to the output aggregate
    *
-   * @param[in] num_items  
+   * @param[in] num_items
    *   Total number of input items (i.e., length of `d_in`)
    *
-   * @param[in] stream  
-   *   **[optional]** CUDA stream to launch kernels within.  
+   * @param[in] stream
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    */
-  template <typename InputIteratorT, 
+  template <typename InputIteratorT,
             typename OutputIteratorT>
   CUB_RUNTIME_FUNCTION static cudaError_t ArgMax(void *d_temp_storage,
                                                  size_t &temp_storage_bytes,
@@ -920,7 +927,7 @@ struct DeviceReduce
 
     // Initial value
     // TODO Address https://github.com/NVIDIA/cub/issues/651
-    InitT initial_value{AccumT(1, Traits<InputValueT>::Lowest())}; 
+    InitT initial_value{AccumT(1, Traits<InputValueT>::Lowest())};
 
     return DispatchReduce<ArgIndexInputIteratorT,
                           OutputIteratorT,
@@ -958,57 +965,57 @@ struct DeviceReduce
   }
 
   /**
-   * @brief Reduces segments of values, where segments are demarcated by 
+   * @brief Reduces segments of values, where segments are demarcated by
    *        corresponding runs of identical keys.
    *
    * @par
    * This operation computes segmented reductions within `d_values_in` using
-   * the specified binary `reduction_op` functor. The segments are identified 
-   * by "runs" of corresponding keys in `d_keys_in`, where runs are maximal 
-   * ranges of consecutive, identical keys. For the *i*<sup>th</sup> run 
-   * encountered, the first key of the run and the corresponding value 
-   * aggregate of that run are written to `d_unique_out[i] and 
-   * `d_aggregates_out[i]`, respectively. The total number of runs encountered 
+   * the specified binary `reduction_op` functor. The segments are identified
+   * by "runs" of corresponding keys in `d_keys_in`, where runs are maximal
+   * ranges of consecutive, identical keys. For the *i*<sup>th</sup> run
+   * encountered, the first key of the run and the corresponding value
+   * aggregate of that run are written to `d_unique_out[i] and
+   * `d_aggregates_out[i]`, respectively. The total number of runs encountered
    * is written to `d_num_runs_out`.
    *
    * @par
-   * - The `==` equality operator is used to determine whether keys are 
+   * - The `==` equality operator is used to determine whether keys are
    *   equivalent
    * - Provides "run-to-run" determinism for pseudo-associative reduction
    *   (e.g., addition of floating point types) on the same GPU device.
    *   However, results for pseudo-associative reduction may be inconsistent
    *   from one device to a another device of a different compute-capability
    *   because CUB can employ different tile-sizing for different architectures.
-   * - Let `out` be any of 
+   * - Let `out` be any of
    *   `[d_unique_out, d_unique_out + *d_num_runs_out)`
    *   `[d_aggregates_out, d_aggregates_out + *d_num_runs_out)`
-   *   `d_num_runs_out`. The ranges represented by `out` shall not overlap 
+   *   `d_num_runs_out`. The ranges represented by `out` shall not overlap
    *   `[d_keys_in, d_keys_in + num_items)`,
    *   `[d_values_in, d_values_in + num_items)` nor `out` in any way.
    * - @devicestorage
    *
    * @par Performance
    * The following chart illustrates reduction-by-key (sum) performance across
-   * different CUDA architectures for `fp32` and `fp64` values, respectively.  
-   * Segments are identified by `int32` keys, and have lengths uniformly 
+   * different CUDA architectures for `fp32` and `fp64` values, respectively.
+   * Segments are identified by `int32` keys, and have lengths uniformly
    * sampled from `[1, 1000]`.
    *
    * @image html reduce_by_key_fp32_len_500.png
    * @image html reduce_by_key_fp64_len_500.png
    *
    * @par
-   * The following charts are similar, but with segment lengths uniformly 
+   * The following charts are similar, but with segment lengths uniformly
    * sampled from [1,10]:
    *
    * @image html reduce_by_key_fp32_len_5.png
    * @image html reduce_by_key_fp64_len_5.png
    *
    * @par Snippet
-   * The code snippet below illustrates the segmented reduction of `int` values 
+   * The code snippet below illustrates the segmented reduction of `int` values
    * grouped by runs of associated `int` keys.
    * @par
    * @code
-   * #include <cub/cub.cuh>   
+   * #include <cub/cub.cuh>
    * // or equivalently <cub/device/device_reduce.cuh>
    *
    * // CustomMin functor
@@ -1021,7 +1028,7 @@ struct DeviceReduce
    *     }
    * };
    *
-   * // Declare, allocate, and initialize device-accessible pointers 
+   * // Declare, allocate, and initialize device-accessible pointers
    * // for input and output
    * int          num_items;          // e.g., 8
    * int          *d_keys_in;         // e.g., [0, 2, 2, 9, 5, 5, 5, 8]
@@ -1036,8 +1043,8 @@ struct DeviceReduce
    * void     *d_temp_storage = NULL;
    * size_t   temp_storage_bytes = 0;
    * cub::DeviceReduce::ReduceByKey(
-   *   d_temp_storage, temp_storage_bytes, 
-   *   d_keys_in, d_unique_out, d_values_in, 
+   *   d_temp_storage, temp_storage_bytes,
+   *   d_keys_in, d_unique_out, d_values_in,
    *   d_aggregates_out, d_num_runs_out, reduction_op, num_items);
    *
    * // Allocate temporary storage
@@ -1045,8 +1052,8 @@ struct DeviceReduce
    *
    * // Run reduce-by-key
    * cub::DeviceReduce::ReduceByKey(
-   *   d_temp_storage, temp_storage_bytes, 
-   *   d_keys_in, d_unique_out, d_values_in, 
+   *   d_temp_storage, temp_storage_bytes,
+   *   d_keys_in, d_unique_out, d_values_in,
    *   d_aggregates_out, d_num_runs_out, reduction_op, num_items);
    *
    * // d_unique_out      <-- [0, 2, 9, 5, 8]
@@ -1054,66 +1061,66 @@ struct DeviceReduce
    * // d_num_runs_out    <-- [5]
    * @endcode
    *
-   * @tparam KeysInputIteratorT       
-   *   **[inferred]** Random-access input iterator type for reading input 
+   * @tparam KeysInputIteratorT
+   *   **[inferred]** Random-access input iterator type for reading input
    *   keys \iterator
    *
-   * @tparam UniqueOutputIteratorT    
-   *   **[inferred]** Random-access output iterator type for writing unique 
+   * @tparam UniqueOutputIteratorT
+   *   **[inferred]** Random-access output iterator type for writing unique
    *   output keys \iterator
    *
-   * @tparam ValuesInputIteratorT     
-   *   **[inferred]** Random-access input iterator type for reading input 
+   * @tparam ValuesInputIteratorT
+   *   **[inferred]** Random-access input iterator type for reading input
    *   values \iterator
    *
-   * @tparam AggregatesOutputIterator 
-   *   **[inferred]** Random-access output iterator type for writing output 
+   * @tparam AggregatesOutputIterator
+   *   **[inferred]** Random-access output iterator type for writing output
    *   value aggregates \iterator
    *
-   * @tparam NumRunsOutputIteratorT   
-   *   **[inferred]** Output iterator type for recording the number of runs 
+   * @tparam NumRunsOutputIteratorT
+   *   **[inferred]** Output iterator type for recording the number of runs
    *   encountered \iterator
    *
-   * @tparam ReductionOpT              
-   *   **[inferred]*8 Binary reduction functor type having member 
+   * @tparam ReductionOpT
+   *   **[inferred]*8 Binary reduction functor type having member
    *   `T operator()(const T &a, const T &b)`
    *
    * @tparam NumItemsT **[inferred]** Type of num_items
    *
-   * @param[in] d_temp_storage  
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no work 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no work
    *   is done.
    *
-   * @param[in,out] temp_storage_bytes  
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of `d_temp_storage` allocation
    *
-   * @param[in] d_keys_in  
+   * @param[in] d_keys_in
    *   Pointer to the input sequence of keys
    *
-   * @param[out] d_unique_out  
+   * @param[out] d_unique_out
    *   Pointer to the output sequence of unique keys (one key per run)
    *
-   * @param[in] d_values_in  
+   * @param[in] d_values_in
    *   Pointer to the input sequence of corresponding values
    *
-   * @param[out] d_aggregates_out  
-   *   Pointer to the output sequence of value aggregates 
+   * @param[out] d_aggregates_out
+   *   Pointer to the output sequence of value aggregates
    *   (one aggregate per run)
    *
-   * @param[out] d_num_runs_out  
-   *   Pointer to total number of runs encountered 
+   * @param[out] d_num_runs_out
+   *   Pointer to total number of runs encountered
    *   (i.e., the length of `d_unique_out`)
    *
-   * @param[in] reduction_op  
+   * @param[in] reduction_op
    *   Binary reduction functor
    *
-   * @param[in] num_items  
-   *   Total number of associated key+value pairs 
+   * @param[in] num_items
+   *   Total number of associated key+value pairs
    *   (i.e., the length of `d_in_keys` and `d_in_values`)
    *
-   * @param[in] stream  
-   *   **[optional]** CUDA stream to launch kernels within.  
+   * @param[in] stream
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    */
   template <typename KeysInputIteratorT,
diff --git a/cub/cub/device/device_run_length_encode.cuh b/cub/cub/device/device_run_length_encode.cuh
index a299d24fa01..e7dbd21d0fc 100644
--- a/cub/cub/device/device_run_length_encode.cuh
+++ b/cub/cub/device/device_run_length_encode.cuh
@@ -34,7 +34,14 @@
 
 #pragma once
 
-#include <cub/config.cuh>
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/device/dispatch/dispatch_reduce_by_key.cuh>
 #include <cub/device/dispatch/dispatch_rle.cuh>
 #include <cub/device/dispatch/tuning/tuning_run_length_encode.cuh>
diff --git a/cub/cub/device/device_scan.cuh b/cub/cub/device/device_scan.cuh
index 20cb8ba872f..9f7fd725f49 100644
--- a/cub/cub/device/device_scan.cuh
+++ b/cub/cub/device/device_scan.cuh
@@ -13,9 +13,9 @@
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
  *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 
- * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 
- * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  * ARE DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE FOR ANY
  * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
@@ -27,14 +27,21 @@
  ******************************************************************************/
 
 /**
- * @file cub::DeviceScan provides device-wide, parallel operations for 
- *       computing a prefix scan across a sequence of data items residing 
+ * @file cub::DeviceScan provides device-wide, parallel operations for
+ *       computing a prefix scan across a sequence of data items residing
  *       within device-accessible memory.
  */
 
 #pragma once
 
-#include <cub/config.cuh>
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/device/dispatch/dispatch_scan.cuh>
 #include <cub/device/dispatch/dispatch_scan_by_key.cuh>
 #include <cub/thread/thread_operators.cuh>
@@ -44,32 +51,32 @@ CUB_NAMESPACE_BEGIN
 
 
 /**
- * @brief DeviceScan provides device-wide, parallel operations for computing a 
- *   prefix scan across a sequence of data items residing within 
+ * @brief DeviceScan provides device-wide, parallel operations for computing a
+ *   prefix scan across a sequence of data items residing within
  *   device-accessible memory. ![](device_scan.png)
  *
  * @ingroup SingleModule
  *
  * @par Overview
- * Given a sequence of input elements and a binary reduction operator, a 
- * [*prefix scan*](http://en.wikipedia.org/wiki/Prefix_sum) produces an output 
- * sequence where each element is computed to be the reduction of the elements 
- * occurring earlier in the input sequence. *Prefix sum* connotes a prefix scan 
- * with the addition operator. The term *inclusive* indicates that the 
+ * Given a sequence of input elements and a binary reduction operator, a
+ * [*prefix scan*](http://en.wikipedia.org/wiki/Prefix_sum) produces an output
+ * sequence where each element is computed to be the reduction of the elements
+ * occurring earlier in the input sequence. *Prefix sum* connotes a prefix scan
+ * with the addition operator. The term *inclusive* indicates that the
  * *i*<sup>th</sup> output reduction incorporates the *i*<sup>th</sup> input.
- * The term *exclusive* indicates the *i*<sup>th</sup> input is not 
- * incorporated into the *i*<sup>th</sup> output reduction. When the input and 
+ * The term *exclusive* indicates the *i*<sup>th</sup> input is not
+ * incorporated into the *i*<sup>th</sup> output reduction. When the input and
  * output sequences are the same, the scan is performed in-place.
  *
  * @par
- * As of CUB 1.0.1 (2013), CUB's device-wide scan APIs have implemented our 
- * *"decoupled look-back"* algorithm for performing global prefix scan with 
- * only a single pass through the input data, as described in our 2016 technical 
- * report [1]. The central idea is to leverage a small, constant factor of 
- * redundant work in order to overlap the latencies of global prefix 
+ * As of CUB 1.0.1 (2013), CUB's device-wide scan APIs have implemented our
+ * *"decoupled look-back"* algorithm for performing global prefix scan with
+ * only a single pass through the input data, as described in our 2016 technical
+ * report [1]. The central idea is to leverage a small, constant factor of
+ * redundant work in order to overlap the latencies of global prefix
  * propagation with local computation. As such, our algorithm requires only
- * ~2*n* data movement (*n* inputs are read, *n* outputs are written), and 
- * typically proceeds at "memcpy" speeds. Our algorithm supports inplace 
+ * ~2*n* data movement (*n* inputs are read, *n* outputs are written), and
+ * typically proceeds at "memcpy" speeds. Our algorithm supports inplace
  * operations.
  *
  * @par
@@ -82,7 +89,7 @@ CUB_NAMESPACE_BEGIN
  * @linear_performance{prefix scan}
  *
  * @par
- * The following chart illustrates DeviceScan::ExclusiveSum performance across 
+ * The following chart illustrates DeviceScan::ExclusiveSum performance across
  * different CUDA architectures for `int32` keys.
  * @plots_below
  *
@@ -97,7 +104,7 @@ struct DeviceScan
   //@{
 
   /**
-   * @brief Computes a device-wide exclusive prefix sum. The value of `0` is 
+   * @brief Computes a device-wide exclusive prefix sum. The value of `0` is
    *        applied as the initial value, and is assigned to `*d_out`.
    *
    * @par
@@ -106,13 +113,13 @@ struct DeviceScan
    *   addition of floating-point types). Results for pseudo-associative
    *   operators may vary from run to run. Additional details can be found in
    *   the [decoupled look-back] description.
-   * - When `d_in` and `d_out` are equal, the scan is performed in-place. The 
-   *   range `[d_in, d_in + num_items)` and `[d_out, d_out + num_items)` 
+   * - When `d_in` and `d_out` are equal, the scan is performed in-place. The
+   *   range `[d_in, d_in + num_items)` and `[d_out, d_out + num_items)`
    *   shall not overlap in any other way.
    * - @devicestorage
    *
    * @par Performance
-   * The following charts illustrate saturated exclusive sum performance across 
+   * The following charts illustrate saturated exclusive sum performance across
    * different CUDA architectures for `int32` and `int64` items, respectively.
    *
    * @image html scan_int32.png
@@ -125,7 +132,7 @@ struct DeviceScan
    * @code
    * #include <cub/cub.cuh> // or equivalently <cub/device/device_scan.cuh>
    *
-   * // Declare, allocate, and initialize device-accessible pointers for 
+   * // Declare, allocate, and initialize device-accessible pointers for
    * // input and output
    * int  num_items;      // e.g., 7
    * int  *d_in;          // e.g., [8, 6, 7, 5, 3, 0, 9]
@@ -136,7 +143,7 @@ struct DeviceScan
    * void     *d_temp_storage = NULL;
    * size_t   temp_storage_bytes = 0;
    * cub::DeviceScan::ExclusiveSum(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_in, d_out, num_items);
    *
    * // Allocate temporary storage
@@ -144,24 +151,24 @@ struct DeviceScan
    *
    * // Run exclusive prefix sum
    * cub::DeviceScan::ExclusiveSum(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_in, d_out, num_items);
    *
    * // d_out <-- [0, 8, 14, 21, 26, 29, 29]
    *
    * @endcode
    *
-   * @tparam InputIteratorT 
-   *   **[inferred]** Random-access input iterator type for reading scan 
+   * @tparam InputIteratorT
+   *   **[inferred]** Random-access input iterator type for reading scan
    *   inputs \iterator
    *
-   * @tparam OutputIteratorT    
-   *   **[inferred]** Random-access output iterator type for writing scan 
+   * @tparam OutputIteratorT
+   *   **[inferred]** Random-access output iterator type for writing scan
    *   outputs \iterator
    *
    * @param[in] d_temp_storage
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no 
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no
    *   work is done.
    *
    * @param[in,out] temp_storage_bytes
@@ -177,7 +184,7 @@ struct DeviceScan
    *   Total number of input items (i.e., the length of `d_in`)
    *
    * @param[in] stream
-   *   **[optional]** CUDA stream to launch kernels within. 
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    *
    * [decoupled look-back]: https://research.nvidia.com/publication/single-pass-parallel-prefix-scan-decoupled-look-back
@@ -227,7 +234,7 @@ struct DeviceScan
   }
 
   /**
-   * @brief Computes a device-wide exclusive prefix sum in-place. The value of 
+   * @brief Computes a device-wide exclusive prefix sum in-place. The value of
    *        `0` is applied as the initial value, and is assigned to `*d_data`.
    *
    * @par
@@ -239,7 +246,7 @@ struct DeviceScan
    * - @devicestorage
    *
    * @par Performance
-   * The following charts illustrate saturated exclusive sum performance across 
+   * The following charts illustrate saturated exclusive sum performance across
    * different CUDA architectures for `int32` and `int64` items, respectively.
    *
    * @image html scan_int32.png
@@ -252,7 +259,7 @@ struct DeviceScan
    * @code
    * #include <cub/cub.cuh> // or equivalently <cub/device/device_scan.cuh>
    *
-   * // Declare, allocate, and initialize device-accessible pointers for 
+   * // Declare, allocate, and initialize device-accessible pointers for
    * // input and output
    * int  num_items;      // e.g., 7
    * int  *d_data;        // e.g., [8, 6, 7, 5, 3, 0, 9]
@@ -262,7 +269,7 @@ struct DeviceScan
    * void     *d_temp_storage = NULL;
    * size_t   temp_storage_bytes = 0;
    * cub::DeviceScan::ExclusiveSum(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_data, num_items);
    *
    * // Allocate temporary storage
@@ -270,20 +277,20 @@ struct DeviceScan
    *
    * // Run exclusive prefix sum
    * cub::DeviceScan::ExclusiveSum(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_data, num_items);
    *
    * // d_data <-- [0, 8, 14, 21, 26, 29, 29]
    *
    * @endcode
    *
-   * @tparam IteratorT 
-   *   **[inferred]** Random-access iterator type for reading scan 
+   * @tparam IteratorT
+   *   **[inferred]** Random-access iterator type for reading scan
    *   inputs and wrigin scan outputs
    *
    * @param[in] d_temp_storage
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no 
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no
    *   work is done.
    *
    * @param[in,out] temp_storage_bytes
@@ -296,7 +303,7 @@ struct DeviceScan
    *   Total number of input items (i.e., the length of `d_in`)
    *
    * @param[in] stream
-   *   **[optional]** CUDA stream to launch kernels within. 
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    *
    * [decoupled look-back]: https://research.nvidia.com/publication/single-pass-parallel-prefix-scan-decoupled-look-back
@@ -337,8 +344,8 @@ struct DeviceScan
   }
 
   /**
-   * @brief Computes a device-wide exclusive prefix scan using the specified 
-   *        binary `scan_op` functor. The `init_value` value is applied as 
+   * @brief Computes a device-wide exclusive prefix scan using the specified
+   *        binary `scan_op` functor. The `init_value` value is applied as
    *        the initial value, and is assigned to `*d_out`.
    *
    * @par
@@ -347,13 +354,13 @@ struct DeviceScan
    *   addition of floating-point types). Results for pseudo-associative
    *   operators may vary from run to run. Additional details can be found in
    *   the [decoupled look-back] description.
-   * - When `d_in` and `d_out` are equal, the scan is performed in-place. The 
-   *   range `[d_in, d_in + num_items)` and `[d_out, d_out + num_items)` 
+   * - When `d_in` and `d_out` are equal, the scan is performed in-place. The
+   *   range `[d_in, d_in + num_items)` and `[d_out, d_out + num_items)`
    *   shall not overlap in any other way.
    * - @devicestorage
    *
    * @par Snippet
-   * The code snippet below illustrates the exclusive prefix min-scan of an 
+   * The code snippet below illustrates the exclusive prefix min-scan of an
    * `int` device vector
    * @par
    * @code
@@ -370,7 +377,7 @@ struct DeviceScan
    *     }
    * };
    *
-   * // Declare, allocate, and initialize device-accessible pointers for 
+   * // Declare, allocate, and initialize device-accessible pointers for
    * // input and output
    * int          num_items;      // e.g., 7
    * int          *d_in;          // e.g., [8, 6, 7, 5, 3, 0, 9]
@@ -378,12 +385,12 @@ struct DeviceScan
    * CustomMin    min_op;
    * ...
    *
-   * // Determine temporary device storage requirements for exclusive 
+   * // Determine temporary device storage requirements for exclusive
    * // prefix scan
    * void     *d_temp_storage = NULL;
    * size_t   temp_storage_bytes = 0;
    * cub::DeviceScan::ExclusiveScan(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_in, d_out, min_op, (int) INT_MAX, num_items);
    *
    * // Allocate temporary storage for exclusive prefix scan
@@ -391,32 +398,32 @@ struct DeviceScan
    *
    * // Run exclusive prefix min-scan
    * cub::DeviceScan::ExclusiveScan(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_in, d_out, min_op, (int) INT_MAX, num_items);
    *
    * // d_out <-- [2147483647, 8, 6, 6, 5, 3, 0]
    *
    * @endcode
    *
-   * @tparam InputIteratorT   
-   *   **[inferred]** Random-access input iterator type for reading scan 
+   * @tparam InputIteratorT
+   *   **[inferred]** Random-access input iterator type for reading scan
    *   inputs \iterator
    *
-   * @tparam OutputIteratorT  
-   *   **[inferred]** Random-access output iterator type for writing scan 
+   * @tparam OutputIteratorT
+   *   **[inferred]** Random-access output iterator type for writing scan
    *   outputs \iterator
    *
-   * @tparam ScanOp           
-   *   **[inferred]** Binary scan functor type having member 
+   * @tparam ScanOp
+   *   **[inferred]** Binary scan functor type having member
    *   `T operator()(const T &a, const T &b)`
-   * 
-   * @tparam InitValueT       
-   *  **[inferred]** Type of the `init_value` used Binary scan functor type 
+   *
+   * @tparam InitValueT
+   *  **[inferred]** Type of the `init_value` used Binary scan functor type
    *   having member `T operator()(const T &a, const T &b)`
    *
-   * @param[in] d_temp_storage 
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no
    *   work is done.
    *
    * @param[in,out] temp_storage_bytes
@@ -438,7 +445,7 @@ struct DeviceScan
    *   Total number of input items (i.e., the length of \p d_in)
    *
    * @param[in] stream
-   *   **[optional]** CUDA stream to launch kernels within. Default is 
+   *   **[optional]** CUDA stream to launch kernels within. Default is
    *   stream<sub>0</sub>.
    *
    * [decoupled look-back]: https://research.nvidia.com/publication/single-pass-parallel-prefix-scan-decoupled-look-back
@@ -505,8 +512,8 @@ struct DeviceScan
   }
 
   /**
-   * @brief Computes a device-wide exclusive prefix scan using the specified 
-   *        binary `scan_op` functor. The `init_value` value is applied as 
+   * @brief Computes a device-wide exclusive prefix scan using the specified
+   *        binary `scan_op` functor. The `init_value` value is applied as
    *        the initial value, and is assigned to `*d_data`.
    *
    * @par
@@ -518,7 +525,7 @@ struct DeviceScan
    * - @devicestorage
    *
    * @par Snippet
-   * The code snippet below illustrates the exclusive prefix min-scan of an 
+   * The code snippet below illustrates the exclusive prefix min-scan of an
    * `int` device vector
    * @par
    * @code
@@ -535,19 +542,19 @@ struct DeviceScan
    *     }
    * };
    *
-   * // Declare, allocate, and initialize device-accessible pointers for 
+   * // Declare, allocate, and initialize device-accessible pointers for
    * // input and output
    * int          num_items;      // e.g., 7
    * int          *d_data;        // e.g., [8, 6, 7, 5, 3, 0, 9]
    * CustomMin    min_op;
    * ...
    *
-   * // Determine temporary device storage requirements for exclusive 
+   * // Determine temporary device storage requirements for exclusive
    * // prefix scan
    * void     *d_temp_storage = NULL;
    * size_t   temp_storage_bytes = 0;
    * cub::DeviceScan::ExclusiveScan(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_data, min_op, (int) INT_MAX, num_items);
    *
    * // Allocate temporary storage for exclusive prefix scan
@@ -555,28 +562,28 @@ struct DeviceScan
    *
    * // Run exclusive prefix min-scan
    * cub::DeviceScan::ExclusiveScan(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_data, min_op, (int) INT_MAX, num_items);
    *
    * // d_data <-- [2147483647, 8, 6, 6, 5, 3, 0]
    *
    * @endcode
    *
-   * @tparam IteratorT   
-   *   **[inferred]** Random-access input iterator type for reading scan 
+   * @tparam IteratorT
+   *   **[inferred]** Random-access input iterator type for reading scan
    *   inputs and writing scan outputs
    *
-   * @tparam ScanOp           
-   *   **[inferred]** Binary scan functor type having member 
+   * @tparam ScanOp
+   *   **[inferred]** Binary scan functor type having member
    *   `T operator()(const T &a, const T &b)`
-   * 
-   * @tparam InitValueT       
-   *  **[inferred]** Type of the `init_value` used Binary scan functor type 
+   *
+   * @tparam InitValueT
+   *  **[inferred]** Type of the `init_value` used Binary scan functor type
    *   having member `T operator()(const T &a, const T &b)`
    *
-   * @param[in] d_temp_storage 
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no
    *   work is done.
    *
    * @param[in,out] temp_storage_bytes
@@ -595,7 +602,7 @@ struct DeviceScan
    *   Total number of input items (i.e., the length of \p d_in)
    *
    * @param[in] stream
-   *   **[optional]** CUDA stream to launch kernels within. Default is 
+   *   **[optional]** CUDA stream to launch kernels within. Default is
    *   stream<sub>0</sub>.
    *
    * [decoupled look-back]: https://research.nvidia.com/publication/single-pass-parallel-prefix-scan-decoupled-look-back
@@ -648,8 +655,8 @@ struct DeviceScan
   }
 
   /**
-   * @brief Computes a device-wide exclusive prefix scan using the specified 
-   *        binary `scan_op` functor. The `init_value` value is provided as 
+   * @brief Computes a device-wide exclusive prefix scan using the specified
+   *        binary `scan_op` functor. The `init_value` value is provided as
    *        a future value.
    *
    * @par
@@ -658,13 +665,13 @@ struct DeviceScan
    *   addition of floating-point types). Results for pseudo-associative
    *   operators may vary from run to run. Additional details can be found in
    *   the [decoupled look-back] description.
-   * - When `d_in` and `d_out` are equal, the scan is performed in-place. The 
-   *   range `[d_in, d_in + num_items)` and `[d_out, d_out + num_items)` 
+   * - When `d_in` and `d_out` are equal, the scan is performed in-place. The
+   *   range `[d_in, d_in + num_items)` and `[d_out, d_out + num_items)`
    *   shall not overlap in any other way.
    * - @devicestorage
    *
    * @par Snippet
-   * The code snippet below illustrates the exclusive prefix min-scan of an 
+   * The code snippet below illustrates the exclusive prefix min-scan of an
    * `int` device vector
    * @par
    * @code
@@ -681,7 +688,7 @@ struct DeviceScan
    *     }
    * };
    *
-   * // Declare, allocate, and initialize device-accessible pointers for 
+   * // Declare, allocate, and initialize device-accessible pointers for
    * // input and output
    * int          num_items;      // e.g., 7
    * int          *d_in;          // e.g., [8, 6, 7, 5, 3, 0, 9]
@@ -689,17 +696,17 @@ struct DeviceScan
    * int          *d_init_iter;   // e.g., INT_MAX
    * CustomMin    min_op;
    *
-   * auto future_init_value = 
+   * auto future_init_value =
    *   cub::FutureValue<InitialValueT, IterT>(d_init_iter);
    *
    * ...
    *
-   * // Determine temporary device storage requirements for exclusive 
+   * // Determine temporary device storage requirements for exclusive
    * // prefix scan
    * void     *d_temp_storage = NULL;
    * size_t   temp_storage_bytes = 0;
    * cub::DeviceScan::ExclusiveScan(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_in, d_out, min_op, future_init_value, num_items);
    *
    * // Allocate temporary storage for exclusive prefix scan
@@ -707,54 +714,54 @@ struct DeviceScan
    *
    * // Run exclusive prefix min-scan
    * cub::DeviceScan::ExclusiveScan(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_in, d_out, min_op, future_init_value, num_items);
    *
    * // d_out <-- [2147483647, 8, 6, 6, 5, 3, 0]
    *
    * @endcode
    *
-   * @tparam InputIteratorT   
-   *   **[inferred]** Random-access input iterator type for reading scan 
+   * @tparam InputIteratorT
+   *   **[inferred]** Random-access input iterator type for reading scan
    *   inputs \iterator
    *
-   * @tparam OutputIteratorT  
-   *   **[inferred]** Random-access output iterator type for writing scan 
+   * @tparam OutputIteratorT
+   *   **[inferred]** Random-access output iterator type for writing scan
    *   outputs \iterator
    *
-   * @tparam ScanOp           
-   *   **[inferred]** Binary scan functor type having member 
+   * @tparam ScanOp
+   *   **[inferred]** Binary scan functor type having member
    *   `T operator()(const T &a, const T &b)`
-   * 
-   * @tparam InitValueT       
-   *  **[inferred]** Type of the `init_value` used Binary scan functor type 
+   *
+   * @tparam InitValueT
+   *  **[inferred]** Type of the `init_value` used Binary scan functor type
    *   having member `T operator()(const T &a, const T &b)`
    *
-   * @param[in] d_temp_storage 
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no work 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no work
    *   is done.
    *
-   * @param[in,out] temp_storage_bytes 
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of \p d_temp_storage allocation
    *
-   * @param[in] d_in 
+   * @param[in] d_in
    *   Pointer to the input sequence of data items
    *
-   * @param[out] d_out 
+   * @param[out] d_out
    *   Pointer to the output sequence of data items
    *
-   * @param[in] scan_op 
+   * @param[in] scan_op
    *   Binary scan functor
    *
-   * @param[in] init_value 
+   * @param[in] init_value
    *   Initial value to seed the exclusive scan (and is assigned to `*d_out`)
    *
-   * @param[in] num_items 
+   * @param[in] num_items
    *   Total number of input items (i.e., the length of `d_in`)
    *
-   * @param[in] stream 
-   *   **[optional]** CUDA stream to launch kernels within. 
+   * @param[in] stream
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    *
    * [decoupled look-back]: https://research.nvidia.com/publication/single-pass-parallel-prefix-scan-decoupled-look-back
@@ -826,8 +833,8 @@ struct DeviceScan
   }
 
   /**
-   * @brief Computes a device-wide exclusive prefix scan using the specified 
-   *        binary `scan_op` functor. The `init_value` value is provided as 
+   * @brief Computes a device-wide exclusive prefix scan using the specified
+   *        binary `scan_op` functor. The `init_value` value is provided as
    *        a future value.
    *
    * @par
@@ -839,7 +846,7 @@ struct DeviceScan
    * - @devicestorage
    *
    * @par Snippet
-   * The code snippet below illustrates the exclusive prefix min-scan of an 
+   * The code snippet below illustrates the exclusive prefix min-scan of an
    * `int` device vector
    * @par
    * @code
@@ -856,24 +863,24 @@ struct DeviceScan
    *     }
    * };
    *
-   * // Declare, allocate, and initialize device-accessible pointers for 
+   * // Declare, allocate, and initialize device-accessible pointers for
    * // input and output
    * int          num_items;      // e.g., 7
    * int          *d_data;        // e.g., [8, 6, 7, 5, 3, 0, 9]
    * int          *d_init_iter;   // e.g., INT_MAX
    * CustomMin    min_op;
    *
-   * auto future_init_value = 
+   * auto future_init_value =
    *   cub::FutureValue<InitialValueT, IterT>(d_init_iter);
    *
    * ...
    *
-   * // Determine temporary device storage requirements for exclusive 
+   * // Determine temporary device storage requirements for exclusive
    * // prefix scan
    * void     *d_temp_storage = NULL;
    * size_t   temp_storage_bytes = 0;
    * cub::DeviceScan::ExclusiveScan(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_data, min_op, future_init_value, num_items);
    *
    * // Allocate temporary storage for exclusive prefix scan
@@ -881,47 +888,47 @@ struct DeviceScan
    *
    * // Run exclusive prefix min-scan
    * cub::DeviceScan::ExclusiveScan(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_data, min_op, future_init_value, num_items);
    *
    * // d_data <-- [2147483647, 8, 6, 6, 5, 3, 0]
    *
    * @endcode
    *
-   * @tparam IteratorT   
-   *   **[inferred]** Random-access input iterator type for reading scan 
+   * @tparam IteratorT
+   *   **[inferred]** Random-access input iterator type for reading scan
    *   inputs and writing scan outputs
    *
-   * @tparam ScanOp           
-   *   **[inferred]** Binary scan functor type having member 
+   * @tparam ScanOp
+   *   **[inferred]** Binary scan functor type having member
    *   `T operator()(const T &a, const T &b)`
-   * 
-   * @tparam InitValueT       
-   *  **[inferred]** Type of the `init_value` used Binary scan functor type 
+   *
+   * @tparam InitValueT
+   *  **[inferred]** Type of the `init_value` used Binary scan functor type
    *   having member `T operator()(const T &a, const T &b)`
    *
-   * @param[in] d_temp_storage 
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no work 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no work
    *   is done.
    *
-   * @param[in,out] temp_storage_bytes 
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of \p d_temp_storage allocation
    *
    * @param[in,out] d_data
    *   Pointer to the sequence of data items
    *
-   * @param[in] scan_op 
+   * @param[in] scan_op
    *   Binary scan functor
    *
-   * @param[in] init_value 
+   * @param[in] init_value
    *   Initial value to seed the exclusive scan (and is assigned to `*d_out`)
    *
-   * @param[in] num_items 
+   * @param[in] num_items
    *   Total number of input items (i.e., the length of `d_in`)
    *
-   * @param[in] stream 
-   *   **[optional]** CUDA stream to launch kernels within. 
+   * @param[in] stream
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    *
    * [decoupled look-back]: https://research.nvidia.com/publication/single-pass-parallel-prefix-scan-decoupled-look-back
@@ -992,8 +999,8 @@ struct DeviceScan
    *   addition of floating-point types). Results for pseudo-associative
    *   operators may vary from run to run. Additional details can be found in
    *   the [decoupled look-back] description.
-   * - When `d_in` and `d_out` are equal, the scan is performed in-place. The 
-   *   range `[d_in, d_in + num_items)` and `[d_out, d_out + num_items)` 
+   * - When `d_in` and `d_out` are equal, the scan is performed in-place. The
+   *   range `[d_in, d_in + num_items)` and `[d_out, d_out + num_items)`
    *   shall not overlap in any other way.
    * - @devicestorage
    *
@@ -1005,19 +1012,19 @@ struct DeviceScan
    * @code
    * #include <cub/cub.cuh>   // or equivalently <cub/device/device_scan.cuh>
    *
-   * // Declare, allocate, and initialize device-accessible pointers for 
+   * // Declare, allocate, and initialize device-accessible pointers for
    * // input and output
    * int  num_items;      // e.g., 7
    * int  *d_in;          // e.g., [8, 6, 7, 5, 3, 0, 9]
    * int  *d_out;         // e.g., [ ,  ,  ,  ,  ,  ,  ]
    * ...
    *
-   * // Determine temporary device storage requirements for inclusive 
+   * // Determine temporary device storage requirements for inclusive
    * // prefix sum
    * void     *d_temp_storage = nullptr;
    * size_t   temp_storage_bytes = 0;
    * cub::DeviceScan::InclusiveSum(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_in, d_out, num_items);
    *
    * // Allocate temporary storage for inclusive prefix sum
@@ -1025,40 +1032,40 @@ struct DeviceScan
    *
    * // Run inclusive prefix sum
    * cub::DeviceScan::InclusiveSum(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_in, d_out, num_items);
    *
    * // d_out <-- [8, 14, 21, 26, 29, 29, 38]
    *
    * @endcode
    *
-   * @tparam InputIteratorT     
-   *   **[inferred]** Random-access input iterator type for reading scan 
+   * @tparam InputIteratorT
+   *   **[inferred]** Random-access input iterator type for reading scan
    *   inputs \iterator
    *
-   * @tparam OutputIteratorT    
-   *   **[inferred]** Random-access output iterator type for writing scan 
+   * @tparam OutputIteratorT
+   *   **[inferred]** Random-access output iterator type for writing scan
    *   outputs \iterator
    *
-   * @param[in] d_temp_storage 
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no
    *   work is done.
    *
-   * @param[in,out] temp_storage_bytes  
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of `d_temp_storage` allocation
    *
-   * @param[in] d_in  
+   * @param[in] d_in
    *   Random-access iterator to the input sequence of data items
    *
-   * @param[out] d_out  
+   * @param[out] d_out
    *   Random-access iterator to the output sequence of data items
    *
-   * @param[in] num_items  
+   * @param[in] num_items
    *   Total number of input items (i.e., the length of `d_in`)
    *
-   * @param[in] stream 
-   *   **[optional]** CUDA stream to launch kernels within. 
+   * @param[in] stream
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    *
    * [decoupled look-back]: https://research.nvidia.com/publication/single-pass-parallel-prefix-scan-decoupled-look-back
@@ -1129,18 +1136,18 @@ struct DeviceScan
    * @code
    * #include <cub/cub.cuh>   // or equivalently <cub/device/device_scan.cuh>
    *
-   * // Declare, allocate, and initialize device-accessible pointers for 
+   * // Declare, allocate, and initialize device-accessible pointers for
    * // input and output
    * int  num_items;      // e.g., 7
    * int  *d_data;        // e.g., [8, 6, 7, 5, 3, 0, 9]
    * ...
    *
-   * // Determine temporary device storage requirements for inclusive 
+   * // Determine temporary device storage requirements for inclusive
    * // prefix sum
    * void     *d_temp_storage = nullptr;
    * size_t   temp_storage_bytes = 0;
    * cub::DeviceScan::InclusiveSum(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_data, num_items);
    *
    * // Allocate temporary storage for inclusive prefix sum
@@ -1148,33 +1155,33 @@ struct DeviceScan
    *
    * // Run inclusive prefix sum
    * cub::DeviceScan::InclusiveSum(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_data, num_items);
    *
    * // d_data <-- [8, 14, 21, 26, 29, 29, 38]
    *
    * @endcode
    *
-   * @tparam IteratorT     
-   *   **[inferred]** Random-access input iterator type for reading scan 
+   * @tparam IteratorT
+   *   **[inferred]** Random-access input iterator type for reading scan
    *   inputs and writing scan outputs
    *
-   * @param[in] d_temp_storage 
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no
    *   work is done.
    *
-   * @param[in,out] temp_storage_bytes  
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of `d_temp_storage` allocation
    *
    * @param[in,out] d_data
    *   Random-access iterator to the sequence of data items
    *
-   * @param[in] num_items  
+   * @param[in] num_items
    *   Total number of input items (i.e., the length of `d_in`)
    *
-   * @param[in] stream 
-   *   **[optional]** CUDA stream to launch kernels within. 
+   * @param[in] stream
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    *
    * [decoupled look-back]: https://research.nvidia.com/publication/single-pass-parallel-prefix-scan-decoupled-look-back
@@ -1215,7 +1222,7 @@ struct DeviceScan
   }
 
   /**
-   * @brief Computes a device-wide inclusive prefix scan using the specified 
+   * @brief Computes a device-wide inclusive prefix scan using the specified
    *        binary `scan_op` functor.
    *
    * @par
@@ -1224,13 +1231,13 @@ struct DeviceScan
    *   addition of floating-point types). Results for pseudo-associative
    *   operators may vary from run to run. Additional details can be found in
    *   the [decoupled look-back] description.
-   * - When `d_in` and `d_out` are equal, the scan is performed in-place. The 
-   *   range `[d_in, d_in + num_items)` and `[d_out, d_out + num_items)` 
+   * - When `d_in` and `d_out` are equal, the scan is performed in-place. The
+   *   range `[d_in, d_in + num_items)` and `[d_out, d_out + num_items)`
    *   shall not overlap in any other way.
    * - @devicestorage
    *
    * @par Snippet
-   * The code snippet below illustrates the inclusive prefix min-scan of an 
+   * The code snippet below illustrates the inclusive prefix min-scan of an
    * `int` device vector.
    *
    * @par
@@ -1248,7 +1255,7 @@ struct DeviceScan
    *     }
    * };
    *
-   * // Declare, allocate, and initialize device-accessible pointers for 
+   * // Declare, allocate, and initialize device-accessible pointers for
    * // input and output
    * int          num_items;      // e.g., 7
    * int          *d_in;          // e.g., [8, 6, 7, 5, 3, 0, 9]
@@ -1256,12 +1263,12 @@ struct DeviceScan
    * CustomMin    min_op;
    * ...
    *
-   * // Determine temporary device storage requirements for inclusive 
+   * // Determine temporary device storage requirements for inclusive
    * // prefix scan
    * void *d_temp_storage = nullptr;
    * size_t temp_storage_bytes = 0;
    * cub::DeviceScan::InclusiveScan(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_in, d_out, min_op, num_items);
    *
    * // Allocate temporary storage for inclusive prefix scan
@@ -1269,28 +1276,28 @@ struct DeviceScan
    *
    * // Run inclusive prefix min-scan
    * cub::DeviceScan::InclusiveScan(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_in, d_out, min_op, num_items);
    *
    * // d_out <-- [8, 6, 6, 5, 3, 0, 0]
    *
    * @endcode
    *
-   * @tparam InputIteratorT   
-   *   **[inferred]** Random-access input iterator type for reading scan 
+   * @tparam InputIteratorT
+   *   **[inferred]** Random-access input iterator type for reading scan
    *   inputs \iterator
    *
-   * @tparam OutputIteratorT  
-   *   **[inferred]** Random-access output iterator type for writing scan 
+   * @tparam OutputIteratorT
+   *   **[inferred]** Random-access output iterator type for writing scan
    *   outputs \iterator
    *
-   * @tparam ScanOp           
-   *   **[inferred]** Binary scan functor type having member 
+   * @tparam ScanOp
+   *   **[inferred]** Binary scan functor type having member
    *   `T operator()(const T &a, const T &b)`
    *
-   * @param[in]  
-   *   d_temp_storage Device-accessible allocation of temporary storage. 
-   *   When `nullptr`, the required allocation size is written to 
+   * @param[in]
+   *   d_temp_storage Device-accessible allocation of temporary storage.
+   *   When `nullptr`, the required allocation size is written to
    *   `temp_storage_bytes` and no work is done.
    *
    * @param[in,out] temp_storage_bytes
@@ -1309,7 +1316,7 @@ struct DeviceScan
    *   Total number of input items (i.e., the length of `d_in`)
    *
    * @param[in] stream
-   *   **[optional]** CUDA stream to launch kernels within. 
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    *
    * [decoupled look-back]: https://research.nvidia.com/publication/single-pass-parallel-prefix-scan-decoupled-look-back
@@ -1366,7 +1373,7 @@ struct DeviceScan
   }
 
   /**
-   * @brief Computes a device-wide inclusive prefix scan using the specified 
+   * @brief Computes a device-wide inclusive prefix scan using the specified
    *        binary `scan_op` functor.
    *
    * @par
@@ -1378,7 +1385,7 @@ struct DeviceScan
    * - @devicestorage
    *
    * @par Snippet
-   * The code snippet below illustrates the inclusive prefix min-scan of an 
+   * The code snippet below illustrates the inclusive prefix min-scan of an
    * `int` device vector.
    *
    * @par
@@ -1396,19 +1403,19 @@ struct DeviceScan
    *     }
    * };
    *
-   * // Declare, allocate, and initialize device-accessible pointers for 
+   * // Declare, allocate, and initialize device-accessible pointers for
    * // input and output
    * int          num_items;      // e.g., 7
    * int          *d_data;        // e.g., [8, 6, 7, 5, 3, 0, 9]
    * CustomMin    min_op;
    * ...
    *
-   * // Determine temporary device storage requirements for inclusive 
+   * // Determine temporary device storage requirements for inclusive
    * // prefix scan
    * void *d_temp_storage = nullptr;
    * size_t temp_storage_bytes = 0;
    * cub::DeviceScan::InclusiveScan(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_data, min_op, num_items);
    *
    * // Allocate temporary storage for inclusive prefix scan
@@ -1416,24 +1423,24 @@ struct DeviceScan
    *
    * // Run inclusive prefix min-scan
    * cub::DeviceScan::InclusiveScan(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_in, d_out, min_op, num_items);
    *
    * // d_data <-- [8, 6, 6, 5, 3, 0, 0]
    *
    * @endcode
    *
-   * @tparam IteratorT   
-   *   **[inferred]** Random-access input iterator type for reading scan 
+   * @tparam IteratorT
+   *   **[inferred]** Random-access input iterator type for reading scan
    *   inputs and writing scan outputs
    *
-   * @tparam ScanOp           
-   *   **[inferred]** Binary scan functor type having member 
+   * @tparam ScanOp
+   *   **[inferred]** Binary scan functor type having member
    *   `T operator()(const T &a, const T &b)`
    *
-   * @param[in]  
-   *   d_temp_storage Device-accessible allocation of temporary storage. 
-   *   When `nullptr`, the required allocation size is written to 
+   * @param[in]
+   *   d_temp_storage Device-accessible allocation of temporary storage.
+   *   When `nullptr`, the required allocation size is written to
    *   `temp_storage_bytes` and no work is done.
    *
    * @param[in,out] temp_storage_bytes
@@ -1449,7 +1456,7 @@ struct DeviceScan
    *   Total number of input items (i.e., the length of `d_in`)
    *
    * @param[in] stream
-   *   **[optional]** CUDA stream to launch kernels within. 
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    *
    * [decoupled look-back]: https://research.nvidia.com/publication/single-pass-parallel-prefix-scan-decoupled-look-back
@@ -1495,8 +1502,8 @@ struct DeviceScan
 
   /**
    * @brief Computes a device-wide exclusive prefix sum-by-key with key equality
-   *        defined by `equality_op`. The value of `0` is applied as the initial 
-   *        value, and is assigned to the beginning of each segment in 
+   *        defined by `equality_op`. The value of `0` is applied as the initial
+   *        value, and is assigned to the beginning of each segment in
    *        `d_values_out`.
    *
    * @par
@@ -1505,22 +1512,22 @@ struct DeviceScan
    *   addition of floating-point types). Results for pseudo-associative
    *   operators may vary from run to run. Additional details can be found in
    *   the [decoupled look-back] description.
-   * - `d_keys_in` may equal `d_values_out` but the range 
-   *   `[d_keys_in, d_keys_in + num_items)` and the range 
+   * - `d_keys_in` may equal `d_values_out` but the range
+   *   `[d_keys_in, d_keys_in + num_items)` and the range
    *   `[d_values_out, d_values_out + num_items)` shall not overlap otherwise.
-   * - `d_values_in` may equal `d_values_out` but the range 
-   *   `[d_values_in, d_values_in + num_items)` and the range 
+   * - `d_values_in` may equal `d_values_out` but the range
+   *   `[d_values_in, d_values_in + num_items)` and the range
    *   `[d_values_out, d_values_out + num_items)` shall not overlap otherwise.
    * - @devicestorage
    *
    * @par Snippet
-   * The code snippet below illustrates the exclusive prefix sum-by-key of an 
+   * The code snippet below illustrates the exclusive prefix sum-by-key of an
    * `int` device vector.
    * @par
    * @code
    * #include <cub/cub.cuh>   // or equivalently <cub/device/device_scan.cuh>
    *
-   * // Declare, allocate, and initialize device-accessible pointers for 
+   * // Declare, allocate, and initialize device-accessible pointers for
    * // input and output
    * int num_items;      // e.g., 7
    * int *d_keys_in;     // e.g., [0, 0, 1, 1, 1, 2, 2]
@@ -1532,7 +1539,7 @@ struct DeviceScan
    * void     *d_temp_storage = nullptr;
    * size_t   temp_storage_bytes = 0;
    * cub::DeviceScan::ExclusiveSumByKey(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_keys_in, d_values_in, d_values_out, num_items);
    *
    * // Allocate temporary storage
@@ -1540,57 +1547,57 @@ struct DeviceScan
    *
    * // Run exclusive prefix sum
    * cub::DeviceScan::ExclusiveSumByKey(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_keys_in, d_values_in, d_values_out, num_items);
    *
    * // d_values_out <-- [0, 8, 0, 7, 12, 0, 0]
    *
    * @endcode
    *
-   * @tparam KeysInputIteratorT      
-   *   **[inferred]** Random-access input iterator type for reading scan keys 
+   * @tparam KeysInputIteratorT
+   *   **[inferred]** Random-access input iterator type for reading scan keys
    *   inputs \iterator
-   * 
-   * @tparam ValuesInputIteratorT    
-   *   **[inferred]** Random-access input iterator type for reading scan 
+   *
+   * @tparam ValuesInputIteratorT
+   *   **[inferred]** Random-access input iterator type for reading scan
    *   values inputs \iterator
    *
-   * @tparam ValuesOutputIteratorT   
-   *   **[inferred]** Random-access output iterator type for writing scan 
+   * @tparam ValuesOutputIteratorT
+   *   **[inferred]** Random-access output iterator type for writing scan
    *   values outputs \iterator
    *
-   * @tparam EqualityOpT             
-   *   **[inferred]** Functor type having member 
-   *   `T operator()(const T &a, const T &b)` for binary operations that 
+   * @tparam EqualityOpT
+   *   **[inferred]** Functor type having member
+   *   `T operator()(const T &a, const T &b)` for binary operations that
    *   defines the equality of keys
    *
-   * @param[in] d_temp_storage 
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no
    *   work is done.
    *
-   * @param[in,out] temp_storage_bytes 
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of `d_temp_storage` allocation
    *
-   * @param[in] d_keys_in 
+   * @param[in] d_keys_in
    *   Random-access input iterator to the input sequence of key items
    *
-   * @param[in] d_values_in 
+   * @param[in] d_values_in
    *   Random-access input iterator to the input sequence of value items
    *
-   * @param[out] d_values_out 
+   * @param[out] d_values_out
    *   Random-access output iterator to the output sequence of value items
    *
-   * @param[in] num_items 
-   *   Total number of input items (i.e., the length of `d_keys_in` and 
+   * @param[in] num_items
+   *   Total number of input items (i.e., the length of `d_keys_in` and
    *   `d_values_in`)
    *
-   * @param[in] equality_op 
-   *   Binary functor that defines the equality of keys. 
+   * @param[in] equality_op
+   *   Binary functor that defines the equality of keys.
    *   Default is cub::Equality().
    *
-   * @param[in] stream 
-   *   **[optional]** CUDA stream to launch kernels within.  
+   * @param[in] stream
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    *
    * [decoupled look-back]: https://research.nvidia.com/publication/single-pass-parallel-prefix-scan-decoupled-look-back
@@ -1614,7 +1621,7 @@ struct DeviceScan
     using InitT = cub::detail::value_t<ValuesInputIteratorT>;
 
     // Initial value
-    InitT init_value{}; 
+    InitT init_value{};
 
     return DispatchScanByKey<KeysInputIteratorT,
                              ValuesInputIteratorT,
@@ -1666,10 +1673,10 @@ struct DeviceScan
   }
 
   /**
-   * @brief Computes a device-wide exclusive prefix scan-by-key using the 
-   *        specified binary `scan_op` functor. The key equality is defined by 
-   *        `equality_op`.  The `init_value` value is applied as the initial 
-   *        value, and is assigned to the beginning of each segment in 
+   * @brief Computes a device-wide exclusive prefix scan-by-key using the
+   *        specified binary `scan_op` functor. The key equality is defined by
+   *        `equality_op`.  The `init_value` value is applied as the initial
+   *        value, and is assigned to the beginning of each segment in
    *        `d_values_out`.
    *
    * @par
@@ -1678,16 +1685,16 @@ struct DeviceScan
    *   addition of floating-point types). Results for pseudo-associative
    *   operators may vary from run to run. Additional details can be found in
    *   the [decoupled look-back] description.
-   * - `d_keys_in` may equal `d_values_out` but the range 
-   *   `[d_keys_in, d_keys_in + num_items)` and the range 
+   * - `d_keys_in` may equal `d_values_out` but the range
+   *   `[d_keys_in, d_keys_in + num_items)` and the range
    *   `[d_values_out, d_values_out + num_items)` shall not overlap otherwise.
-   * - `d_values_in` may equal `d_values_out` but the range 
-   *   `[d_values_in, d_values_in + num_items)` and the range 
+   * - `d_values_in` may equal `d_values_out` but the range
+   *   `[d_values_in, d_values_in + num_items)` and the range
    *   `[d_values_out, d_values_out + num_items)` shall not overlap otherwise.
    * - @devicestorage
    *
    * @par Snippet
-   * The code snippet below illustrates the exclusive prefix min-scan-by-key of 
+   * The code snippet below illustrates the exclusive prefix min-scan-by-key of
    * an `int` device vector
    * @par
    * @code
@@ -1714,7 +1721,7 @@ struct DeviceScan
    *     }
    * };
    *
-   * // Declare, allocate, and initialize device-accessible pointers for 
+   * // Declare, allocate, and initialize device-accessible pointers for
    * // input and output
    * int          num_items;      // e.g., 7
    * int          *d_keys_in;     // e.g., [0, 0, 1, 1, 1, 2, 2]
@@ -1724,13 +1731,13 @@ struct DeviceScan
    * CustomEqual  equality_op;
    * ...
    *
-   * // Determine temporary device storage requirements for exclusive 
+   * // Determine temporary device storage requirements for exclusive
    * // prefix scan
    * void     *d_temp_storage = NULL;
    * size_t   temp_storage_bytes = 0;
    * cub::DeviceScan::ExclusiveScanByKey(
-   *   d_temp_storage, temp_storage_bytes, 
-   *   d_keys_in, d_values_in, d_values_out, min_op, 
+   *   d_temp_storage, temp_storage_bytes,
+   *   d_keys_in, d_values_in, d_values_out, min_op,
    *   (int) INT_MAX, num_items, equality_op);
    *
    * // Allocate temporary storage for exclusive prefix scan
@@ -1738,73 +1745,73 @@ struct DeviceScan
    *
    * // Run exclusive prefix min-scan
    * cub::DeviceScan::ExclusiveScanByKey(
-   *   d_temp_storage, temp_storage_bytes, 
-   *   d_keys_in, d_values_in, d_values_out, min_op, 
+   *   d_temp_storage, temp_storage_bytes,
+   *   d_keys_in, d_values_in, d_values_out, min_op,
    *   (int) INT_MAX, num_items, equality_op);
    *
    * // d_values_out <-- [2147483647, 8, 2147483647, 7, 5, 2147483647, 0]
    *
    * @endcode
    *
-   * @tparam KeysInputIteratorT      
-   *   **[inferred]** Random-access input iterator type for reading scan keys 
+   * @tparam KeysInputIteratorT
+   *   **[inferred]** Random-access input iterator type for reading scan keys
    *   inputs \iterator
    *
-   * @tparam ValuesInputIteratorT    
-   *   **[inferred]** Random-access input iterator type for reading scan values 
+   * @tparam ValuesInputIteratorT
+   *   **[inferred]** Random-access input iterator type for reading scan values
    *   inputs \iterator
    *
-   * @tparam ValuesOutputIteratorT   
-   *   **[inferred]** Random-access output iterator type for writing scan values 
+   * @tparam ValuesOutputIteratorT
+   *   **[inferred]** Random-access output iterator type for writing scan values
    *   outputs \iterator
    *
-   * @tparam ScanOp                  
-   *   **[inferred]** Binary scan functor type having member 
+   * @tparam ScanOp
+   *   **[inferred]** Binary scan functor type having member
    *   `T operator()(const T &a, const T &b)`
    *
-   * @tparam InitValueT              
-   *   **[inferred]** Type of the `init_value` value used in Binary scan 
+   * @tparam InitValueT
+   *   **[inferred]** Type of the `init_value` value used in Binary scan
    *   functor type having member `T operator()(const T &a, const T &b)`
    *
-   * @tparam EqualityOpT             
-   *   **[inferred]** Functor type having member 
-   *   `T operator()(const T &a, const T &b)` for binary operations that 
+   * @tparam EqualityOpT
+   *   **[inferred]** Functor type having member
+   *   `T operator()(const T &a, const T &b)` for binary operations that
    *   defines the equality of keys
    *
-   *  @param[in] d_temp_storage 
-   *    Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *    required allocation size is written to `temp_storage_bytes` and no work 
+   *  @param[in] d_temp_storage
+   *    Device-accessible allocation of temporary storage. When `nullptr`, the
+   *    required allocation size is written to `temp_storage_bytes` and no work
    *    is done.
    *
-   *  @param[in,out] temp_storage_bytes 
+   *  @param[in,out] temp_storage_bytes
    *    Reference to size in bytes of `d_temp_storage` allocation
    *
-   *  @param[in] d_keys_in 
+   *  @param[in] d_keys_in
    *    Random-access input iterator to the input sequence of key items
    *
-   *  @param[in] d_values_in 
+   *  @param[in] d_values_in
    *    Random-access input iterator to the input sequence of value items
    *
-   *  @param[out] d_values_out 
+   *  @param[out] d_values_out
    *    Random-access output iterator to the output sequence of value items
    *
-   *  @param[in] scan_op 
+   *  @param[in] scan_op
    *    Binary scan functor
    *
-   *  @param[in] init_value 
-   *    Initial value to seed the exclusive scan (and is assigned to the 
+   *  @param[in] init_value
+   *    Initial value to seed the exclusive scan (and is assigned to the
    *    beginning of each segment in `d_values_out`)
    *
-   *  @param[in] num_items 
-   *    Total number of input items (i.e., the length of `d_keys_in` and 
+   *  @param[in] num_items
+   *    Total number of input items (i.e., the length of `d_keys_in` and
    *    `d_values_in`)
    *
-   *  @param[in] equality_op 
-   *    Binary functor that defines the equality of keys. 
+   *  @param[in] equality_op
+   *    Binary functor that defines the equality of keys.
    *    Default is cub::Equality().
    *
-   *  @param[in] stream   
-   *    **[optional]** CUDA stream to launch kernels within.  
+   *  @param[in] stream
+   *    **[optional]** CUDA stream to launch kernels within.
    *    Default is stream<sub>0</sub>.
    *
    * [decoupled look-back]: https://research.nvidia.com/publication/single-pass-parallel-prefix-scan-decoupled-look-back
@@ -1888,7 +1895,7 @@ struct DeviceScan
   }
 
   /**
-   * @brief Computes a device-wide inclusive prefix sum-by-key with key 
+   * @brief Computes a device-wide inclusive prefix sum-by-key with key
    *        equality defined by `equality_op`.
    *
    * @par
@@ -1897,22 +1904,22 @@ struct DeviceScan
    *   addition of floating-point types). Results for pseudo-associative
    *   operators may vary from run to run. Additional details can be found in
    *   the [decoupled look-back] description.
-   * - `d_keys_in` may equal `d_values_out` but the range 
-   *   `[d_keys_in, d_keys_in + num_items)` and the range 
+   * - `d_keys_in` may equal `d_values_out` but the range
+   *   `[d_keys_in, d_keys_in + num_items)` and the range
    *   `[d_values_out, d_values_out + num_items)` shall not overlap otherwise.
-   * - `d_values_in` may equal `d_values_out` but the range 
-   *   `[d_values_in, d_values_in + num_items)` and the range 
+   * - `d_values_in` may equal `d_values_out` but the range
+   *   `[d_values_in, d_values_in + num_items)` and the range
    *   `[d_values_out, d_values_out + num_items)` shall not overlap otherwise.
    * - @devicestorage
    *
    * @par Snippet
-   * The code snippet below illustrates the inclusive prefix sum-by-key of an 
+   * The code snippet below illustrates the inclusive prefix sum-by-key of an
    * `int` device vector.
    * @par
    * @code
    * #include <cub/cub.cuh>   // or equivalently <cub/device/device_scan.cuh>
    *
-   * // Declare, allocate, and initialize device-accessible pointers for 
+   * // Declare, allocate, and initialize device-accessible pointers for
    * // input and output
    * int num_items;      // e.g., 7
    * int *d_keys_in;     // e.g., [0, 0, 1, 1, 1, 2, 2]
@@ -1924,7 +1931,7 @@ struct DeviceScan
    * void     *d_temp_storage = NULL;
    * size_t   temp_storage_bytes = 0;
    * cub::DeviceScan::InclusiveSumByKey(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_keys_in, d_values_in, d_values_out, num_items);
    *
    * // Allocate temporary storage for inclusive prefix sum
@@ -1932,59 +1939,59 @@ struct DeviceScan
    *
    * // Run inclusive prefix sum
    * cub::DeviceScan::InclusiveSumByKey(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_keys_in, d_values_in, d_values_out, num_items);
    *
    * // d_out <-- [8, 14, 7, 12, 15, 0, 9]
    *
    * @endcode
    *
-   * @tparam KeysInputIteratorT      
-   *   **[inferred]** Random-access input iterator type for reading scan 
+   * @tparam KeysInputIteratorT
+   *   **[inferred]** Random-access input iterator type for reading scan
    *   keys inputs \iterator
-   * 
-   * @tparam ValuesInputIteratorT    
-   *   **[inferred]** Random-access input iterator type for reading scan 
+   *
+   * @tparam ValuesInputIteratorT
+   *   **[inferred]** Random-access input iterator type for reading scan
    *   values inputs \iterator
-   * 
-   * @tparam ValuesOutputIteratorT   
-   *   **[inferred]** Random-access output iterator type for writing scan 
+   *
+   * @tparam ValuesOutputIteratorT
+   *   **[inferred]** Random-access output iterator type for writing scan
    *   values outputs \iterator
-   * 
-   * @tparam EqualityOpT             
-   *   **[inferred]** Functor type having member 
-   *   `T operator()(const T &a, const T &b)` for binary operations that 
+   *
+   * @tparam EqualityOpT
+   *   **[inferred]** Functor type having member
+   *   `T operator()(const T &a, const T &b)` for binary operations that
    *   defines the equality of keys
    *
-   *  @param[in] d_temp_storage 
-   *    Device-accessible allocation of temporary storage.  
-   *    When `nullptr`, the required allocation size is written to 
+   *  @param[in] d_temp_storage
+   *    Device-accessible allocation of temporary storage.
+   *    When `nullptr`, the required allocation size is written to
    *    `temp_storage_bytes` and no work is done.
-   * 
-   *  @param[in,out] temp_storage_bytes 
+   *
+   *  @param[in,out] temp_storage_bytes
    *    Reference to size in bytes of `d_temp_storage` allocation
-   * 
-   *  @param[in] d_keys_in 
+   *
+   *  @param[in] d_keys_in
    *    Random-access input iterator to the input sequence of key items
-   * 
-   *  @param[in] d_values_in 
+   *
+   *  @param[in] d_values_in
    *    Random-access input iterator to the input sequence of value items
-   * 
-   *  @param[out] d_values_out 
+   *
+   *  @param[out] d_values_out
    *    Random-access output iterator to the output sequence of value items
-   * 
-   *  @param[in] num_items 
-   *    Total number of input items (i.e., the length of `d_keys_in` and 
+   *
+   *  @param[in] num_items
+   *    Total number of input items (i.e., the length of `d_keys_in` and
    *    `d_values_in`)
-   * 
-   *  @param[in] equality_op 
-   *    Binary functor that defines the equality of keys. 
+   *
+   *  @param[in] equality_op
+   *    Binary functor that defines the equality of keys.
    *    Default is cub::Equality().
-   * 
-   *  @param[in] stream 
-   *    **[optional]** CUDA stream to launch kernels within.  
+   *
+   *  @param[in] stream
+   *    **[optional]** CUDA stream to launch kernels within.
    *    Default is stream<sub>0</sub>.
-   * 
+   *
    * [decoupled look-back]: https://research.nvidia.com/publication/single-pass-parallel-prefix-scan-decoupled-look-back
    */
   template <typename KeysInputIteratorT,
@@ -2054,8 +2061,8 @@ struct DeviceScan
   }
 
   /**
-   * @brief Computes a device-wide inclusive prefix scan-by-key using the 
-   *        specified binary `scan_op` functor. The key equality is defined 
+   * @brief Computes a device-wide inclusive prefix scan-by-key using the
+   *        specified binary `scan_op` functor. The key equality is defined
    *        by `equality_op`.
    *
    * @par
@@ -2064,16 +2071,16 @@ struct DeviceScan
    *   addition of floating-point types). Results for pseudo-associative
    *   operators may vary from run to run. Additional details can be found in
    *   the [decoupled look-back] description.
-   * - `d_keys_in` may equal `d_values_out` but the range 
-   *   `[d_keys_in, d_keys_in + num_items)` and the range 
+   * - `d_keys_in` may equal `d_values_out` but the range
+   *   `[d_keys_in, d_keys_in + num_items)` and the range
    *   `[d_values_out, d_values_out + num_items)` shall not overlap otherwise.
-   * - `d_values_in` may equal `d_values_out` but the range 
-   *   `[d_values_in, d_values_in + num_items)` and the range 
+   * - `d_values_in` may equal `d_values_out` but the range
+   *   `[d_values_in, d_values_in + num_items)` and the range
    *   `[d_values_out, d_values_out + num_items)` shall not overlap otherwise.
    * - @devicestorage
    *
    * @par Snippet
-   * The code snippet below illustrates the inclusive prefix min-scan-by-key 
+   * The code snippet below illustrates the inclusive prefix min-scan-by-key
    * of an `int` device vector.
    * @par
    * @code
@@ -2100,7 +2107,7 @@ struct DeviceScan
    *     }
    * };
    *
-   * // Declare, allocate, and initialize device-accessible pointers for 
+   * // Declare, allocate, and initialize device-accessible pointers for
    * // input and output
    * int          num_items;      // e.g., 7
    * int          *d_keys_in;     // e.g., [0, 0, 1, 1, 1, 2, 2]
@@ -2114,7 +2121,7 @@ struct DeviceScan
    * void *d_temp_storage = NULL;
    * size_t temp_storage_bytes = 0;
    * cub::DeviceScan::InclusiveScanByKey(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_keys_in, d_values_in, d_values_out, min_op, num_items, equality_op);
    *
    * // Allocate temporary storage for inclusive prefix scan
@@ -2122,66 +2129,66 @@ struct DeviceScan
    *
    * // Run inclusive prefix min-scan
    * cub::DeviceScan::InclusiveScanByKey(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_keys_in, d_values_in, d_values_out, min_op, num_items, equality_op);
    *
    * // d_out <-- [8, 6, 7, 5, 3, 0, 0]
    *
    * @endcode
    *
-   * @tparam KeysInputIteratorT      
-   *   **[inferred]** Random-access input iterator type for reading scan keys 
+   * @tparam KeysInputIteratorT
+   *   **[inferred]** Random-access input iterator type for reading scan keys
    *   inputs \iterator
    *
-   * @tparam ValuesInputIteratorT    
-   *   **[inferred]** Random-access input iterator type for reading scan 
+   * @tparam ValuesInputIteratorT
+   *   **[inferred]** Random-access input iterator type for reading scan
    *   values inputs \iterator
    *
-   * @tparam ValuesOutputIteratorT   
-   *   **[inferred]** Random-access output iterator type for writing scan 
+   * @tparam ValuesOutputIteratorT
+   *   **[inferred]** Random-access output iterator type for writing scan
    *   values outputs \iterator
    *
-   * @tparam ScanOp                  
-   *   **[inferred]** Binary scan functor type having member 
+   * @tparam ScanOp
+   *   **[inferred]** Binary scan functor type having member
    *   `T operator()(const T &a, const T &b)`
    *
-   * @tparam EqualityOpT             
-   *   **[inferred]** Functor type having member 
-   *   `T operator()(const T &a, const T &b)` for binary operations that 
+   * @tparam EqualityOpT
+   *   **[inferred]** Functor type having member
+   *   `T operator()(const T &a, const T &b)` for binary operations that
    *   defines the equality of keys
    *
-   *  @param[in] d_temp_storage 
-   *    Device-accessible allocation of temporary storage.  
-   *    When `nullptr`, the required allocation size is written to 
+   *  @param[in] d_temp_storage
+   *    Device-accessible allocation of temporary storage.
+   *    When `nullptr`, the required allocation size is written to
    *    `temp_storage_bytes` and no work is done.
-   * 
-   *  @param[in,out] temp_storage_bytes 
+   *
+   *  @param[in,out] temp_storage_bytes
    *    Reference to size in bytes of `d_temp_storage` allocation
-   * 
-   *  @param[in] d_keys_in 
+   *
+   *  @param[in] d_keys_in
    *    Random-access input iterator to the input sequence of key items
-   * 
-   *  @param[in] d_values_in 
+   *
+   *  @param[in] d_values_in
    *    Random-access input iterator to the input sequence of value items
-   * 
-   *  @param[out] d_values_out 
+   *
+   *  @param[out] d_values_out
    *    Random-access output iterator to the output sequence of value items
-   * 
-   *  @param[in] scan_op 
+   *
+   *  @param[in] scan_op
    *    Binary scan functor
-   * 
-   *  @param[in] num_items 
-   *    Total number of input items (i.e., the length of `d_keys_in` and 
+   *
+   *  @param[in] num_items
+   *    Total number of input items (i.e., the length of `d_keys_in` and
    *    `d_values_in`)
-   * 
-   *  @param[in] equality_op 
-   *    Binary functor that defines the equality of keys. 
+   *
+   *  @param[in] equality_op
+   *    Binary functor that defines the equality of keys.
    *    Default is cub::Equality().
-   * 
-   *  @param[in] stream 
-   *    **[optional]** CUDA stream to launch kernels within.  
+   *
+   *  @param[in] stream
+   *    **[optional]** CUDA stream to launch kernels within.
    *    Default is stream<sub>0</sub>.
-   * 
+   *
    * [decoupled look-back]: https://research.nvidia.com/publication/single-pass-parallel-prefix-scan-decoupled-look-back
    */
   template <typename KeysInputIteratorT,
diff --git a/cub/cub/device/device_segmented_radix_sort.cuh b/cub/cub/device/device_segmented_radix_sort.cuh
index c8f69223e94..dbec93997f0 100644
--- a/cub/cub/device/device_segmented_radix_sort.cuh
+++ b/cub/cub/device/device_segmented_radix_sort.cuh
@@ -13,9 +13,9 @@
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
  *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 
- * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 
- * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  * ARE DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE FOR ANY
  * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
@@ -27,18 +27,25 @@
  ******************************************************************************/
 
 /**
- * @file cub::DeviceSegmentedRadixSort provides device-wide, parallel 
- *       operations for computing a batched radix sort across multiple, 
- *       non-overlapping sequences of data items residing within 
+ * @file cub::DeviceSegmentedRadixSort provides device-wide, parallel
+ *       operations for computing a batched radix sort across multiple,
+ *       non-overlapping sequences of data items residing within
  *       device-accessible memory.
  */
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <stdio.h>
 #include <iterator>
 
-#include <cub/config.cuh>
 #include <cub/device/dispatch/dispatch_radix_sort.cuh>
 #include <cub/util_deprecated.cuh>
 
@@ -46,27 +53,27 @@ CUB_NAMESPACE_BEGIN
 
 
 /**
- * @brief DeviceSegmentedRadixSort provides device-wide, parallel operations 
- *        for computing a batched radix sort across multiple, non-overlapping 
- *        sequences of data items residing within device-accessible memory. 
+ * @brief DeviceSegmentedRadixSort provides device-wide, parallel operations
+ *        for computing a batched radix sort across multiple, non-overlapping
+ *        sequences of data items residing within device-accessible memory.
  *        ![](segmented_sorting_logo.png)
  * @ingroup SegmentedModule
  *
  * @par Overview
- * The [*radix sorting method*](http://en.wikipedia.org/wiki/Radix_sort) 
- * arranges items into ascending (or descending) order. The algorithm relies 
- * upon a positional representation for keys, i.e., each key is comprised of an 
- * ordered sequence of symbols (e.g., digits, characters, etc.) specified from 
- * least-significant to most-significant.  For a given input sequence of keys 
- * and a set of rules specifying a total ordering of the symbolic alphabet, the 
+ * The [*radix sorting method*](http://en.wikipedia.org/wiki/Radix_sort)
+ * arranges items into ascending (or descending) order. The algorithm relies
+ * upon a positional representation for keys, i.e., each key is comprised of an
+ * ordered sequence of symbols (e.g., digits, characters, etc.) specified from
+ * least-significant to most-significant.  For a given input sequence of keys
+ * and a set of rules specifying a total ordering of the symbolic alphabet, the
  * radix sorting method produces a lexicographic ordering of those keys.
  *
  * @par See Also
  * DeviceSegmentedRadixSort shares its implementation with DeviceRadixSort. See
  * that algorithm's documentation for more information.
  *
- * @par Segments are not required to be contiguous. Any element of input(s) or 
- * output(s) outside the specified segments will not be accessed nor modified.  
+ * @par Segments are not required to be contiguous. Any element of input(s) or
+ * output(s) outside the specified segments will not be accessed nor modified.
  *
  * @par Usage Considerations
  * @cdp_class{DeviceSegmentedRadixSort}
@@ -80,7 +87,7 @@ struct DeviceSegmentedRadixSort
   //@{
 
   /**
-   * @brief Sorts segments of key-value pairs into ascending order. 
+   * @brief Sorts segments of key-value pairs into ascending order.
    *        (`~2N` auxiliary storage required)
    *
    * @par
@@ -89,31 +96,31 @@ struct DeviceSegmentedRadixSort
    *   `segment_offsets` (of length `num_segments + 1`) can be aliased
    *   for both the `d_begin_offsets` and `d_end_offsets` parameters (where
    *   the latter is specified as `segment_offsets + 1`).
-   * - An optional bit subrange `[begin_bit, end_bit)` of differentiating key 
-   *   bits can be specified. This can reduce overall sorting overhead and 
+   * - An optional bit subrange `[begin_bit, end_bit)` of differentiating key
+   *   bits can be specified. This can reduce overall sorting overhead and
    *   yield a corresponding performance improvement.
    * - Let `in` be one of `{d_keys_in, d_values_in}` and `out` be any of
-   *   `{d_keys_out, d_values_out}`. The range `[out, out + num_items)` shall 
-   *   not overlap `[in, in + num_items)`, 
+   *   `{d_keys_out, d_values_out}`. The range `[out, out + num_items)` shall
+   *   not overlap `[in, in + num_items)`,
    *   `[d_begin_offsets, d_begin_offsets + num_segments)` nor
    *   `[d_end_offsets, d_end_offsets + num_segments)` in any way.
-   * - @devicestorageNP For sorting using only `O(P)` temporary storage, see 
+   * - @devicestorageNP For sorting using only `O(P)` temporary storage, see
    *   the sorting interface using DoubleBuffer wrappers below.
-   * - Segments are not required to be contiguous. For all index values `i` 
-   *   outside the specified segments `d_keys_in[i]`, `d_values_in[i]`, 
-   *   `d_keys_out[i]`, `d_values_out[i]` will not be accessed nor modified.   
+   * - Segments are not required to be contiguous. For all index values `i`
+   *   outside the specified segments `d_keys_in[i]`, `d_values_in[i]`,
+   *   `d_keys_out[i]`, `d_values_out[i]` will not be accessed nor modified.
    * - @devicestorage
    *
    * @par Snippet
-   * The code snippet below illustrates the batched sorting of three segments 
-   * (with one zero-length segment) of `int` keys with associated vector of 
+   * The code snippet below illustrates the batched sorting of three segments
+   * (with one zero-length segment) of `int` keys with associated vector of
    * `int` values.
    * @par
    * @code
-   * #include <cub/cub.cuh>  
+   * #include <cub/cub.cuh>
    * // or equivalently <cub/device/device_segmented_radix_sort.cuh>
    *
-   * // Declare, allocate, and initialize device-accessible pointers 
+   * // Declare, allocate, and initialize device-accessible pointers
    * // for sorting data
    * int  num_items;          // e.g., 7
    * int  num_segments;       // e.g., 3
@@ -145,72 +152,72 @@ struct DeviceSegmentedRadixSort
    * // d_values_out          <-- [1, 2, 0, 5, 4, 3, 6]
    * @endcode
    *
-   * @tparam KeyT                  
+   * @tparam KeyT
    *   **[inferred]** Key type
    *
-   * @tparam ValueT                
+   * @tparam ValueT
    *   **[inferred]** Value type
    *
-   * @tparam BeginOffsetIteratorT  
-   *   **[inferred]** Random-access input iterator type for reading segment 
+   * @tparam BeginOffsetIteratorT
+   *   **[inferred]** Random-access input iterator type for reading segment
    *   beginning offsets \iterator
    *
-   * @tparam EndOffsetIteratorT    
-   *   **[inferred]** Random-access input iterator type for reading segment 
+   * @tparam EndOffsetIteratorT
+   *   **[inferred]** Random-access input iterator type for reading segment
    *   ending offsets \iterator
    *
-   * @param[in] d_temp_storage 
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no work 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no work
    *   is done.
    *
-   * @param[in,out] temp_storage_bytes 
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of `d_temp_storage` allocation
    *
-   * @param[in] d_keys_in 
+   * @param[in] d_keys_in
    *   Device-accessible pointer to the input data of key data to sort
    *
-   * @param[out] d_keys_out 
+   * @param[out] d_keys_out
    *   Device-accessible pointer to the sorted output sequence of key data
    *
-   * @param[in] d_values_in 
-   *   Device-accessible pointer to the corresponding input sequence of 
+   * @param[in] d_values_in
+   *   Device-accessible pointer to the corresponding input sequence of
    *   associated value items
    *
-   * @param[out] d_values_out 
-   *   Device-accessible pointer to the correspondingly-reordered output 
+   * @param[out] d_values_out
+   *   Device-accessible pointer to the correspondingly-reordered output
    *   sequence of associated value items
    *
-   * @param[in] num_items 
+   * @param[in] num_items
    *   The total number of items within the segmented array, including items not
    *   covered by segments. `num_items` should match the largest element within
    *   the range `[d_end_offsets, d_end_offsets + num_segments)`.
    *
-   * @param[in] num_segments 
+   * @param[in] num_segments
    *   The number of segments that comprise the sorting data
    *
-   * @param[in] d_begin_offsets 
-   *   Random-access input iterator to the sequence of beginning offsets of 
-   *   length `num_segments`, such that `d_begin_offsets[i]` is the first 
-   *   element of the *i*<sup>th</sup> data segment in `d_keys_*` and 
+   * @param[in] d_begin_offsets
+   *   Random-access input iterator to the sequence of beginning offsets of
+   *   length `num_segments`, such that `d_begin_offsets[i]` is the first
+   *   element of the *i*<sup>th</sup> data segment in `d_keys_*` and
    *   `d_values_*`
    *
-   * @param[in] d_end_offsets 
-   *   Random-access input iterator to the sequence of ending offsets of length 
-   *   `num_segments`, such that `d_end_offsets[i] - 1` is the last element of 
-   *   the *i*<sup>th</sup> data segment in `d_keys_*` and `d_values_*`. If 
-   *   `d_end_offsets[i] - 1 <= d_begin_offsets[i]`, the *i*<sup>th</sup> is 
+   * @param[in] d_end_offsets
+   *   Random-access input iterator to the sequence of ending offsets of length
+   *   `num_segments`, such that `d_end_offsets[i] - 1` is the last element of
+   *   the *i*<sup>th</sup> data segment in `d_keys_*` and `d_values_*`. If
+   *   `d_end_offsets[i] - 1 <= d_begin_offsets[i]`, the *i*<sup>th</sup> is
    *   considered empty.
    *
-   * @param[in] begin_bit 
-   *   **[optional]** The least-significant bit index (inclusive) needed for 
+   * @param[in] begin_bit
+   *   **[optional]** The least-significant bit index (inclusive) needed for
    *   key comparison
    *
-   * @param[in] end_bit 
-   *   **[optional]** The most-significant bit index (exclusive) needed for key 
+   * @param[in] end_bit
+   *   **[optional]** The most-significant bit index (exclusive) needed for key
    *   comparison (e.g., `sizeof(unsigned int) * 8`)
    *
-   * @param[in] stream 
+   * @param[in] stream
    *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    */
@@ -299,7 +306,7 @@ struct DeviceSegmentedRadixSort
   }
 
   /**
-   * @brief Sorts segments of key-value pairs into ascending order. 
+   * @brief Sorts segments of key-value pairs into ascending order.
    *        (`~N` auxiliary storage required)
    *
    * @par
@@ -307,42 +314,42 @@ struct DeviceSegmentedRadixSort
    *   pair of associated value buffers.  Each pair is managed by a DoubleBuffer
    *   structure that indicates which of the two buffers is "current" (and thus
    *   contains the input data to be sorted).
-   * - The contents of both buffers within each pair may be altered by the 
+   * - The contents of both buffers within each pair may be altered by the
    *   sorting operation.
-   * - Upon completion, the sorting operation will update the "current" 
-   *   indicator within each DoubleBuffer wrapper to reference which of the two 
-   *   buffers now contains the sorted output sequence (a function of the number 
+   * - Upon completion, the sorting operation will update the "current"
+   *   indicator within each DoubleBuffer wrapper to reference which of the two
+   *   buffers now contains the sorted output sequence (a function of the number
    *   of key bits specified and the targeted device architecture).
    * - When input a contiguous sequence of segments, a single sequence
-   *   `segment_offsets` (of length `num_segments + 1`) can be aliased for both 
-   *   the `d_begin_offsets` and `d_end_offsets` parameters (where the latter is 
+   *   `segment_offsets` (of length `num_segments + 1`) can be aliased for both
+   *   the `d_begin_offsets` and `d_end_offsets` parameters (where the latter is
    *   specified as `segment_offsets + 1`).
-   * - An optional bit subrange `[begin_bit, end_bit)` of differentiating key 
-   *   bits can be specified. This can reduce overall sorting overhead and yield 
+   * - An optional bit subrange `[begin_bit, end_bit)` of differentiating key
+   *   bits can be specified. This can reduce overall sorting overhead and yield
    *   a corresponding performance improvement.
-   * - Let `cur` be one of `{d_keys.Current(), d_values.Current()}` and `alt` 
-   *   be any of `{d_keys.Alternate(), d_values.Alternate()}`. The range 
-   *   `[cur, cur + num_items)` shall not overlap 
+   * - Let `cur` be one of `{d_keys.Current(), d_values.Current()}` and `alt`
+   *   be any of `{d_keys.Alternate(), d_values.Alternate()}`. The range
+   *   `[cur, cur + num_items)` shall not overlap
    *   `[alt, alt + num_items)`. Both ranges shall not overlap
    *   `[d_begin_offsets, d_begin_offsets + num_segments)` nor
    *   `[d_end_offsets, d_end_offsets + num_segments)` in any way.
-   * - Segments are not required to be contiguous. For all index values `i` 
-   *   outside the specified segments `d_keys.Current()[i]`, 
-   *   `d_values.Current()[i]`, `d_keys.Alternate()[i]`, 
-   *   `d_values.Alternate()[i]` will not be accessed nor modified.   
+   * - Segments are not required to be contiguous. For all index values `i`
+   *   outside the specified segments `d_keys.Current()[i]`,
+   *   `d_values.Current()[i]`, `d_keys.Alternate()[i]`,
+   *   `d_values.Alternate()[i]` will not be accessed nor modified.
    * - @devicestorageP
    * - @devicestorage
    *
    * @par Snippet
-   * The code snippet below illustrates the batched sorting of three segments 
-   * (with one zero-length segment) of `int` keys with associated vector of 
+   * The code snippet below illustrates the batched sorting of three segments
+   * (with one zero-length segment) of `int` keys with associated vector of
    * `int` values.
    * @par
    * @code
-   * #include <cub/cub.cuh>   
+   * #include <cub/cub.cuh>
    * // or equivalently <cub/device/device_segmented_radix_sort.cuh>
    *
-   * // Declare, allocate, and initialize device-accessible pointers 
+   * // Declare, allocate, and initialize device-accessible pointers
    * // for sorting data
    * int  num_items;          // e.g., 7
    * int  num_segments;       // e.g., 3
@@ -377,69 +384,69 @@ struct DeviceSegmentedRadixSort
    *
    * @endcode
    *
-   * @tparam KeyT             
+   * @tparam KeyT
    *   **[inferred]** Key type
    *
-   * @tparam ValueT           
+   * @tparam ValueT
    *   **[inferred]** Value type
    *
-   * @tparam BeginOffsetIteratorT  
-   *   **[inferred]** Random-access input iterator type for reading segment 
+   * @tparam BeginOffsetIteratorT
+   *   **[inferred]** Random-access input iterator type for reading segment
    *   beginning offsets \iterator
    *
-   * @tparam EndOffsetIteratorT    
-   *   **[inferred]** Random-access input iterator type for reading segment 
+   * @tparam EndOffsetIteratorT
+   *   **[inferred]** Random-access input iterator type for reading segment
    *   ending offsets \iterator
    *
-   * @param[in] d_temp_storage 
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no work 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no work
    *   is done.
    *
-   * @param[in,out] temp_storage_bytes 
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of `d_temp_storage` allocation
    *
-   * @param[in,out] d_keys 
-   *   Reference to the double-buffer of keys whose "current" device-accessible 
-   *   buffer contains the unsorted input keys and, upon return, is updated to 
+   * @param[in,out] d_keys
+   *   Reference to the double-buffer of keys whose "current" device-accessible
+   *   buffer contains the unsorted input keys and, upon return, is updated to
    *   point to the sorted output keys
    *
-   * @param[in,out] d_values 
-   *   Double-buffer of values whose "current" device-accessible buffer 
-   *   contains the unsorted input values and, upon return, is updated to point 
+   * @param[in,out] d_values
+   *   Double-buffer of values whose "current" device-accessible buffer
+   *   contains the unsorted input values and, upon return, is updated to point
    *   to the sorted output values
    *
-   * @param[in] num_items 
+   * @param[in] num_items
    *   The total number of items within the segmented array, including items not
    *   covered by segments. `num_items` should match the largest element within
    *   the range `[d_end_offsets, d_end_offsets + num_segments)`.
    *
-   * @param[in] num_segments 
+   * @param[in] num_segments
    *   The number of segments that comprise the sorting data
    *
-   * @param[in] d_begin_offsets 
-   *   Random-access input iterator to the sequence of beginning offsets of 
-   *   length `num_segments`, such that `d_begin_offsets[i]` is the first 
-   *   element of the *i*<sup>th</sup> data segment in `d_keys_*` and 
+   * @param[in] d_begin_offsets
+   *   Random-access input iterator to the sequence of beginning offsets of
+   *   length `num_segments`, such that `d_begin_offsets[i]` is the first
+   *   element of the *i*<sup>th</sup> data segment in `d_keys_*` and
    *   `d_values_*`
    *
-   * @param[in] d_end_offsets 
-   *   Random-access input iterator to the sequence of ending offsets of length 
-   *   `num_segments`, such that `d_end_offsets[i] - 1` is the last element of 
-   *   the *i*<sup>th</sup> data segment in `d_keys_*` and `d_values_*`. 
-   *   If `d_end_offsets[i] - 1 <= d_begin_offsets[i]`, the *i*<sup>th</sup> is 
+   * @param[in] d_end_offsets
+   *   Random-access input iterator to the sequence of ending offsets of length
+   *   `num_segments`, such that `d_end_offsets[i] - 1` is the last element of
+   *   the *i*<sup>th</sup> data segment in `d_keys_*` and `d_values_*`.
+   *   If `d_end_offsets[i] - 1 <= d_begin_offsets[i]`, the *i*<sup>th</sup> is
    *   considered empty.
    *
-   * @param[in] begin_bit 
-   *   **[optional]** The least-significant bit index (inclusive) needed for 
+   * @param[in] begin_bit
+   *   **[optional]** The least-significant bit index (inclusive) needed for
    *   key comparison
    *
-   * @param[in] end_bit 
-   *   **[optional]** The most-significant bit index (exclusive) needed for key 
+   * @param[in] end_bit
+   *   **[optional]** The most-significant bit index (exclusive) needed for key
    *   comparison (e.g., `sizeof(unsigned int) * 8`)
    *
-   * @param[in] stream 
-   *   **[optional]** CUDA stream to launch kernels within.  
+   * @param[in] stream
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    */
   template <typename KeyT,
@@ -517,40 +524,40 @@ struct DeviceSegmentedRadixSort
   }
 
   /**
-   * @brief Sorts segments of key-value pairs into descending order. 
+   * @brief Sorts segments of key-value pairs into descending order.
    *        (`~2N` auxiliary storage required).
    *
    * @par
    * - The contents of the input data are not altered by the sorting operation
    * - When input a contiguous sequence of segments, a single sequence
-   *   `segment_offsets` (of length `num_segments + 1`) can be aliased for both 
-   *   the `d_begin_offsets` and `d_end_offsets` parameters (where the latter is 
+   *   `segment_offsets` (of length `num_segments + 1`) can be aliased for both
+   *   the `d_begin_offsets` and `d_end_offsets` parameters (where the latter is
    *   specified as `segment_offsets + 1`).
-   * - An optional bit subrange `[begin_bit, end_bit)` of differentiating key 
-   *   bits can be specified. This can reduce overall sorting overhead and 
+   * - An optional bit subrange `[begin_bit, end_bit)` of differentiating key
+   *   bits can be specified. This can reduce overall sorting overhead and
    *   yield a corresponding performance improvement.
    * - Let `in` be one of `{d_keys_in, d_values_in}` and `out` be any of
-   *   `{d_keys_out, d_values_out}`. The range `[out, out + num_items)` shall 
-   *   not overlap `[in, in + num_items)`, 
+   *   `{d_keys_out, d_values_out}`. The range `[out, out + num_items)` shall
+   *   not overlap `[in, in + num_items)`,
    *   `[d_begin_offsets, d_begin_offsets + num_segments)` nor
    *   `[d_end_offsets, d_end_offsets + num_segments)` in any way.
-   * - @devicestorageNP For sorting using only `O(P)` temporary storage, see 
+   * - @devicestorageNP For sorting using only `O(P)` temporary storage, see
    *   the sorting interface using DoubleBuffer wrappers below.
-   * - Segments are not required to be contiguous. For all index values `i` 
-   *   outside the specified segments `d_keys_in[i]`, `d_values_in[i]`, 
-   *   `d_keys_out[i]`, `d_values_out[i]` will not be accessed nor modified.   
+   * - Segments are not required to be contiguous. For all index values `i`
+   *   outside the specified segments `d_keys_in[i]`, `d_values_in[i]`,
+   *   `d_keys_out[i]`, `d_values_out[i]` will not be accessed nor modified.
    * - @devicestorage
    *
    * @par Snippet
-   * The code snippet below illustrates the batched sorting of three segments 
-   * (with one zero-length segment) of `int` keys with associated vector of 
+   * The code snippet below illustrates the batched sorting of three segments
+   * (with one zero-length segment) of `int` keys with associated vector of
    * `int` values.
    * @par
    * @code
-   * #include <cub/cub.cuh>   
+   * #include <cub/cub.cuh>
    * // or equivalently <cub/device/device_segmented_radix_sort.cuh>
    *
-   * // Declare, allocate, and initialize device-accessible pointers 
+   * // Declare, allocate, and initialize device-accessible pointers
    * // for sorting data
    * int  num_items;          // e.g., 7
    * int  num_segments;       // e.g., 3
@@ -582,73 +589,73 @@ struct DeviceSegmentedRadixSort
    * // d_values_out          <-- [0, 2, 1, 6, 3, 4, 5]
    * @endcode
    *
-   * @tparam KeyT             
+   * @tparam KeyT
    *   **[inferred]** Key type
    *
-   * @tparam ValueT           
+   * @tparam ValueT
    *   **[inferred]** Value type
    *
-   * @tparam BeginOffsetIteratorT  
-   *   **[inferred]** Random-access input iterator type for reading segment 
+   * @tparam BeginOffsetIteratorT
+   *   **[inferred]** Random-access input iterator type for reading segment
    *   beginning offsets \iterator
    *
-   * @tparam EndOffsetIteratorT    
-   *   **[inferred]** Random-access input iterator type for reading segment 
+   * @tparam EndOffsetIteratorT
+   *   **[inferred]** Random-access input iterator type for reading segment
    *   ending offsets \iterator
    *
-   * @param[in] d_temp_storage 
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no work 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no work
    *   is done.
    *
-   * @param[in,out] temp_storage_bytes 
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of `d_temp_storage` allocation
    *
-   * @param[in] d_keys_in 
+   * @param[in] d_keys_in
    *   Device-accessible pointer to the input data of key data to sort
    *
-   * @param[out] d_keys_out 
+   * @param[out] d_keys_out
    *   Device-accessible pointer to the sorted output sequence of key data
    *
-   * @param[in] d_values_in 
-   *   Device-accessible pointer to the corresponding input sequence of 
+   * @param[in] d_values_in
+   *   Device-accessible pointer to the corresponding input sequence of
    *   associated value items
    *
-   * @param[out] d_values_out 
-   *   Device-accessible pointer to the correspondingly-reordered output 
+   * @param[out] d_values_out
+   *   Device-accessible pointer to the correspondingly-reordered output
    *   sequence of associated value items
    *
-   * @param[in] num_items 
+   * @param[in] num_items
    *   The total number of items within the segmented array, including items not
    *   covered by segments. `num_items` should match the largest element within
    *   the range `[d_end_offsets, d_end_offsets + num_segments)`.
    *
-   * @param[in] num_segments 
+   * @param[in] num_segments
    *   The number of segments that comprise the sorting data
    *
-   * @param[in] d_begin_offsets 
-   *   Random-access input iterator to the sequence of beginning offsets of 
-   *   length `num_segments`, such that `d_begin_offsets[i]` is the first 
-   *   element of the *i*<sup>th</sup> data segment in `d_keys_*` and 
+   * @param[in] d_begin_offsets
+   *   Random-access input iterator to the sequence of beginning offsets of
+   *   length `num_segments`, such that `d_begin_offsets[i]` is the first
+   *   element of the *i*<sup>th</sup> data segment in `d_keys_*` and
    *   `d_values_*`
    *
-   * @param[in] d_end_offsets 
-   *   Random-access input iterator to the sequence of ending offsets of length 
-   *   `num_segments`, such that `d_end_offsets[i] - 1` is the last element of 
-   *   the *i*<sup>th</sup> data segment in `d_keys_*` and `d_values_*`. 
-   *   If `d_end_offsets[i] - 1 <= d_begin_offsets[i]`, the *i*<sup>th</sup> 
+   * @param[in] d_end_offsets
+   *   Random-access input iterator to the sequence of ending offsets of length
+   *   `num_segments`, such that `d_end_offsets[i] - 1` is the last element of
+   *   the *i*<sup>th</sup> data segment in `d_keys_*` and `d_values_*`.
+   *   If `d_end_offsets[i] - 1 <= d_begin_offsets[i]`, the *i*<sup>th</sup>
    *   is considered empty.
    *
-   * @param[in] begin_bit 
-   *   **[optional]** The least-significant bit index (inclusive) needed for 
+   * @param[in] begin_bit
+   *   **[optional]** The least-significant bit index (inclusive) needed for
    *   key comparison
    *
-   * @param[in] end_bit 
-   *   **[optional]** The most-significant bit index (exclusive) needed for key 
+   * @param[in] end_bit
+   *   **[optional]** The most-significant bit index (exclusive) needed for key
    *   comparison (e.g., `sizeof(unsigned int) * 8`)
    *
-   * @param[in] stream 
-   *   **[optional]** CUDA stream to launch kernels within.  
+   * @param[in] stream
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    */
   template <typename KeyT,
@@ -738,7 +745,7 @@ struct DeviceSegmentedRadixSort
   }
 
   /**
-   * @brief Sorts segments of key-value pairs into descending order. 
+   * @brief Sorts segments of key-value pairs into descending order.
    *        (`~N` auxiliary storage required).
    *
    * @par
@@ -746,43 +753,43 @@ struct DeviceSegmentedRadixSort
    *   pair of associated value buffers.  Each pair is managed by a DoubleBuffer
    *   structure that indicates which of the two buffers is "current" (and thus
    *   contains the input data to be sorted).
-   * - The contents of both buffers within each pair may be altered by the 
+   * - The contents of both buffers within each pair may be altered by the
    *   sorting operation.
-   * - Upon completion, the sorting operation will update the "current" 
-   *   indicator within each DoubleBuffer wrapper to reference which of the two 
-   *   buffers now contains the sorted output sequence (a function of the number 
+   * - Upon completion, the sorting operation will update the "current"
+   *   indicator within each DoubleBuffer wrapper to reference which of the two
+   *   buffers now contains the sorted output sequence (a function of the number
    *   of key bits specified and the targeted device architecture).
    * - When input a contiguous sequence of segments, a single sequence
-   *   `segment_offsets` (of length `num_segments + 1`) can be aliased for both 
-   *   the `d_begin_offsets` and `d_end_offsets` parameters (where the latter is 
+   *   `segment_offsets` (of length `num_segments + 1`) can be aliased for both
+   *   the `d_begin_offsets` and `d_end_offsets` parameters (where the latter is
    *   specified as `segment_offsets + 1`).
-   * - An optional bit subrange `[begin_bit, end_bit)` of differentiating key 
-   *   bits can be specified. This can reduce overall sorting overhead and 
+   * - An optional bit subrange `[begin_bit, end_bit)` of differentiating key
+   *   bits can be specified. This can reduce overall sorting overhead and
    *   yield a corresponding performance improvement.
-   * - Let `cur` be one of `{d_keys.Current(), d_values.Current()}` and `alt` 
-   *   be any of `{d_keys.Alternate(), d_values.Alternate()}`. The range 
-   *   `[cur, cur + num_items)` shall not overlap 
+   * - Let `cur` be one of `{d_keys.Current(), d_values.Current()}` and `alt`
+   *   be any of `{d_keys.Alternate(), d_values.Alternate()}`. The range
+   *   `[cur, cur + num_items)` shall not overlap
    *   `[alt, alt + num_items)`. Both ranges shall not overlap
    *   `[d_begin_offsets, d_begin_offsets + num_segments)` nor
    *   `[d_end_offsets, d_end_offsets + num_segments)` in any way.
-   * - Segments are not required to be contiguous. For all index values `i` 
-   *   outside the specified segments `d_keys.Current()[i]`, 
-   *   `d_values.Current()[i]`, `d_keys.Alternate()[i]`, 
-   *   `d_values.Alternate()[i]` will not be accessed nor modified.   
-   *   not to be modified. 
+   * - Segments are not required to be contiguous. For all index values `i`
+   *   outside the specified segments `d_keys.Current()[i]`,
+   *   `d_values.Current()[i]`, `d_keys.Alternate()[i]`,
+   *   `d_values.Alternate()[i]` will not be accessed nor modified.
+   *   not to be modified.
    * - @devicestorageP
    * - @devicestorage
    *
    * @par Snippet
-   * The code snippet below illustrates the batched sorting of three segments 
-   * (with one zero-length segment) of `int` keys with associated vector of 
+   * The code snippet below illustrates the batched sorting of three segments
+   * (with one zero-length segment) of `int` keys with associated vector of
    * `int` values.
    * @par
    * @code
-   * #include <cub/cub.cuh>   
+   * #include <cub/cub.cuh>
    * // or equivalently <cub/device/device_segmented_radix_sort.cuh>
    *
-   * // Declare, allocate, and initialize device-accessible pointers 
+   * // Declare, allocate, and initialize device-accessible pointers
    * // for sorting data
    * int  num_items;          // e.g., 7
    * int  num_segments;       // e.g., 3
@@ -816,69 +823,69 @@ struct DeviceSegmentedRadixSort
    * // d_values.Current()    <-- [0, 2, 1, 6, 3, 4, 5]
    * @endcode
    *
-   * @tparam KeyT             
+   * @tparam KeyT
    *   **[inferred]** Key type
    *
-   * @tparam ValueT           
+   * @tparam ValueT
    *   **[inferred]** Value type
    *
-   * @tparam BeginOffsetIteratorT  
-   *   **[inferred]** Random-access input iterator type for reading segment 
+   * @tparam BeginOffsetIteratorT
+   *   **[inferred]** Random-access input iterator type for reading segment
    *   beginning offsets \iterator
    *
-   * @tparam EndOffsetIteratorT    
-   *   **[inferred]** Random-access input iterator type for reading segment 
+   * @tparam EndOffsetIteratorT
+   *   **[inferred]** Random-access input iterator type for reading segment
    *   ending offsets \iterator
    *
-   * @param[in] d_temp_storage 
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no work 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no work
    *   is done.
    *
-   * @param[in,out] temp_storage_bytes 
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of `d_temp_storage` allocation
    *
-   * @param[in,out] d_keys 
-   *   Reference to the double-buffer of keys whose "current" device-accessible 
-   *   buffer contains the unsorted input keys and, upon return, is updated to 
+   * @param[in,out] d_keys
+   *   Reference to the double-buffer of keys whose "current" device-accessible
+   *   buffer contains the unsorted input keys and, upon return, is updated to
    *   point to the sorted output keys
    *
-   * @param[in,out] d_values 
-   *   Double-buffer of values whose "current" device-accessible buffer 
-   *   contains the unsorted input values and, upon return, is updated to point 
+   * @param[in,out] d_values
+   *   Double-buffer of values whose "current" device-accessible buffer
+   *   contains the unsorted input values and, upon return, is updated to point
    *   to the sorted output values
    *
-   * @param[in] num_items 
+   * @param[in] num_items
    *   The total number of items within the segmented array, including items not
    *   covered by segments. `num_items` should match the largest element within
    *   the range `[d_end_offsets, d_end_offsets + num_segments)`.
    *
-   * @param[in] num_segments 
+   * @param[in] num_segments
    *   The number of segments that comprise the sorting data
    *
-   * @param[in] d_begin_offsets 
-   *   Random-access input iterator to the sequence of beginning offsets of 
-   *   length `num_segments`, such that `d_begin_offsets[i]` is the first 
-   *   element of the *i*<sup>th</sup> data segment in `d_keys_*` and 
+   * @param[in] d_begin_offsets
+   *   Random-access input iterator to the sequence of beginning offsets of
+   *   length `num_segments`, such that `d_begin_offsets[i]` is the first
+   *   element of the *i*<sup>th</sup> data segment in `d_keys_*` and
    *   `d_values_*`
    *
-   * @param[in] d_end_offsets 
-   *   Random-access input iterator to the sequence of ending offsets of length 
-   *   `num_segments`, such that `d_end_offsets[i] - 1` is the last element of 
-   *   the *i*<sup>th</sup> data segment in `d_keys_*` and `d_values_*`.  
-   *   If `d_end_offsets[i] - 1 <= d_begin_offsets[i]`, the *i*<sup>th</sup> 
+   * @param[in] d_end_offsets
+   *   Random-access input iterator to the sequence of ending offsets of length
+   *   `num_segments`, such that `d_end_offsets[i] - 1` is the last element of
+   *   the *i*<sup>th</sup> data segment in `d_keys_*` and `d_values_*`.
+   *   If `d_end_offsets[i] - 1 <= d_begin_offsets[i]`, the *i*<sup>th</sup>
    *   is considered empty.
    *
-   * @param[in] begin_bit 
-   *   **[optional]** The least-significant bit index (inclusive) needed for 
+   * @param[in] begin_bit
+   *   **[optional]** The least-significant bit index (inclusive) needed for
    *   key comparison
    *
-   * @param[in] end_bit 
-   *   **[optional]** The most-significant bit index (exclusive) needed for key 
+   * @param[in] end_bit
+   *   **[optional]** The most-significant bit index (exclusive) needed for key
    *   comparison (e.g., `sizeof(unsigned int) * 8`)
    *
-   * @param[in] stream 
-   *   **[optional]** CUDA stream to launch kernels within. 
+   * @param[in] stream
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    */
   template <typename KeyT,
@@ -965,38 +972,38 @@ struct DeviceSegmentedRadixSort
 
 
   /**
-   * @brief Sorts segments of keys into ascending order. 
+   * @brief Sorts segments of keys into ascending order.
    *        (`~2N` auxiliary storage required)
    *
    * @par
    * - The contents of the input data are not altered by the sorting operation
-   * - An optional bit subrange `[begin_bit, end_bit)` of differentiating key 
-   *   bits can be specified. This can reduce overall sorting overhead and 
+   * - An optional bit subrange `[begin_bit, end_bit)` of differentiating key
+   *   bits can be specified. This can reduce overall sorting overhead and
    *   yield a corresponding performance improvement.
    * - When input a contiguous sequence of segments, a single sequence
-   *   `segment_offsets` (of length `num_segments + 1`) can be aliased for both 
-   *   the `d_begin_offsets` and `d_end_offsets` parameters (where the latter 
+   *   `segment_offsets` (of length `num_segments + 1`) can be aliased for both
+   *   the `d_begin_offsets` and `d_end_offsets` parameters (where the latter
    *   is specified as `segment_offsets + 1`).
    * - The range `[d_keys_out, d_keys_out + num_items)` shall not overlap
-   *   `[d_keys_in, d_keys_in + num_items)`, 
+   *   `[d_keys_in, d_keys_in + num_items)`,
    *   `[d_begin_offsets, d_begin_offsets + num_segments)` nor
    *   `[d_end_offsets, d_end_offsets + num_segments)` in any way.
-   * - @devicestorageNP For sorting using only `O(P)` temporary storage, see 
+   * - @devicestorageNP For sorting using only `O(P)` temporary storage, see
    *   the sorting interface using DoubleBuffer wrappers below.
-   * - Segments are not required to be contiguous. For all index values `i` 
-   *   outside the specified segments `d_keys_in[i]`, `d_keys_out[i]` will not 
-   *   be accessed nor modified.   
+   * - Segments are not required to be contiguous. For all index values `i`
+   *   outside the specified segments `d_keys_in[i]`, `d_keys_out[i]` will not
+   *   be accessed nor modified.
    * - @devicestorage
    *
    * @par Snippet
-   * The code snippet below illustrates the batched sorting of three segments 
+   * The code snippet below illustrates the batched sorting of three segments
    * (with one zero-length segment) of `int` keys.
    * @par
    * @code
-   * #include <cub/cub.cuh>   
+   * #include <cub/cub.cuh>
    * // or equivalently <cub/device/device_segmented_radix_sort.cuh>
    *
-   * // Declare, allocate, and initialize device-accessible pointers 
+   * // Declare, allocate, and initialize device-accessible pointers
    * // for sorting data
    * int  num_items;          // e.g., 7
    * int  num_segments;       // e.g., 3
@@ -1008,7 +1015,7 @@ struct DeviceSegmentedRadixSort
    * // Determine temporary device storage requirements
    * void     *d_temp_storage = NULL;
    * size_t   temp_storage_bytes = 0;
-   * cub::DeviceSegmentedRadixSort::SortKeys( 
+   * cub::DeviceSegmentedRadixSort::SortKeys(
    *     d_temp_storage, temp_storage_bytes, d_keys_in, d_keys_out,
    *     num_items, num_segments, d_offsets, d_offsets + 1);
    *
@@ -1016,7 +1023,7 @@ struct DeviceSegmentedRadixSort
    * cudaMalloc(&d_temp_storage, temp_storage_bytes);
    *
    * // Run sorting operation
-   * cub::DeviceSegmentedRadixSort::SortKeys( 
+   * cub::DeviceSegmentedRadixSort::SortKeys(
    *     d_temp_storage, temp_storage_bytes, d_keys_in, d_keys_out,
    *     num_items, num_segments, d_offsets, d_offsets + 1);
    *
@@ -1024,60 +1031,60 @@ struct DeviceSegmentedRadixSort
    *
    * @endcode
    *
-   * @tparam KeyT             
+   * @tparam KeyT
    *   **[inferred]** Key type
    *
-   * @tparam BeginOffsetIteratorT  
-   *   **[inferred]** Random-access input iterator type for reading segment 
+   * @tparam BeginOffsetIteratorT
+   *   **[inferred]** Random-access input iterator type for reading segment
    *   beginning offsets \iterator
    *
-   * @tparam EndOffsetIteratorT    
-   *   **[inferred]** Random-access input iterator type for reading segment 
+   * @tparam EndOffsetIteratorT
+   *   **[inferred]** Random-access input iterator type for reading segment
    *   ending offsets \iterator
    *
-   * @param[in] d_temp_storage 
+   * @param[in] d_temp_storage
    *   Device-accessible allocation of temporary storage.  When NULL, the required allocation size is written to \p temp_storage_bytes and no work is done.
    *
-   * @param[in,out] temp_storage_bytes 
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of \p d_temp_storage allocation
    *
-   * @param[in] d_keys_in  
+   * @param[in] d_keys_in
    *   Device-accessible pointer to the input data of key data to sort
    *
-   * @param[out] d_keys_out  
+   * @param[out] d_keys_out
    *   Device-accessible pointer to the sorted output sequence of key data
    *
-   * @param[in] num_items  
+   * @param[in] num_items
    *   The total number of items within the segmented array, including items not
    *   covered by segments. `num_items` should match the largest element within
    *   the range `[d_end_offsets, d_end_offsets + num_segments)`.
    *
-   * @param[in] num_segments  
+   * @param[in] num_segments
    *   The number of segments that comprise the sorting data
    *
-   * @param[in] d_begin_offsets  
-   *   Random-access input iterator to the sequence of beginning offsets of 
-   *   length `num_segments`, such that `d_begin_offsets[i]` is the first 
-   *   element of the *i*<sup>th</sup> data segment in `d_keys_*` and 
+   * @param[in] d_begin_offsets
+   *   Random-access input iterator to the sequence of beginning offsets of
+   *   length `num_segments`, such that `d_begin_offsets[i]` is the first
+   *   element of the *i*<sup>th</sup> data segment in `d_keys_*` and
    *   `d_values_*`
    *
-   * @param[in] d_end_offsets  
-   *   Random-access input iterator to the sequence of ending offsets of length 
-   *   `num_segments`, such that `d_end_offsets[i] - 1` is the last element of 
-   *   the *i*<sup>th</sup> data segment in `d_keys_*` and `d_values_*`.  
-   *   If `d_end_offsets[i] - 1 <= d_begin_offsets[i]`, the *i*<sup>th</sup> is 
+   * @param[in] d_end_offsets
+   *   Random-access input iterator to the sequence of ending offsets of length
+   *   `num_segments`, such that `d_end_offsets[i] - 1` is the last element of
+   *   the *i*<sup>th</sup> data segment in `d_keys_*` and `d_values_*`.
+   *   If `d_end_offsets[i] - 1 <= d_begin_offsets[i]`, the *i*<sup>th</sup> is
    *   considered empty.
    *
-   * @param[in] begin_bit  
-   *   **[optional]** The least-significant bit index (inclusive) needed for 
+   * @param[in] begin_bit
+   *   **[optional]** The least-significant bit index (inclusive) needed for
    *   key comparison
    *
-   * @param[in] end_bit  
-   *   **[optional]** The most-significant bit index (exclusive) needed for key 
+   * @param[in] end_bit
+   *   **[optional]** The most-significant bit index (exclusive) needed for key
    *   comparison (e.g., `sizeof(unsigned int) * 8`)
    *
-   * @param[in] stream  
-   *   **[optional]** CUDA stream to launch kernels within.  
+   * @param[in] stream
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    */
   template <typename KeyT,
@@ -1160,41 +1167,41 @@ struct DeviceSegmentedRadixSort
    * @brief Sorts segments of keys into ascending order. (~<em>N </em>auxiliary storage required).
    *
    * @par
-   * - The sorting operation is given a pair of key buffers managed by a 
+   * - The sorting operation is given a pair of key buffers managed by a
    *   DoubleBuffer structure that indicates which of the two buffers is
    *   "current" (and thus contains the input data to be sorted).
    * - The contents of both buffers may be altered by the sorting operation.
-   * - Upon completion, the sorting operation will update the "current" 
-   *   indicator within the DoubleBuffer wrapper to reference which of the two 
-   *   buffers now contains the sorted output sequence (a function of the 
+   * - Upon completion, the sorting operation will update the "current"
+   *   indicator within the DoubleBuffer wrapper to reference which of the two
+   *   buffers now contains the sorted output sequence (a function of the
    *   number of key bits specified and the targeted device architecture).
    * - When input a contiguous sequence of segments, a single sequence
-   *   `segment_offsets` (of length `num_segments + 1`) can be aliased for both 
-   *   the `d_begin_offsets` and `d_end_offsets` parameters (where the latter 
+   *   `segment_offsets` (of length `num_segments + 1`) can be aliased for both
+   *   the `d_begin_offsets` and `d_end_offsets` parameters (where the latter
    *   is specified as `segment_offsets + 1`).
-   * - An optional bit subrange `[begin_bit, end_bit)` of differentiating key 
-   *   bits can be specified. This can reduce overall sorting overhead and 
+   * - An optional bit subrange `[begin_bit, end_bit)` of differentiating key
+   *   bits can be specified. This can reduce overall sorting overhead and
    *   yield a corresponding performance improvement.
    * - Let `cur = d_keys.Current()` and `alt = d_keys.Alternate()`.
-   *   The range `[cur, cur + num_items)` shall not overlap 
+   *   The range `[cur, cur + num_items)` shall not overlap
    *   `[alt, alt + num_items)`. Both ranges shall not overlap
    *   `[d_begin_offsets, d_begin_offsets + num_segments)` nor
    *   `[d_end_offsets, d_end_offsets + num_segments)` in any way.
-   * - Segments are not required to be contiguous. For all index values `i` 
-   *   outside the specified segments `d_keys.Current()[i]`, 
-   *   `d_keys[i].Alternate()[i]` will not be accessed nor modified.   
+   * - Segments are not required to be contiguous. For all index values `i`
+   *   outside the specified segments `d_keys.Current()[i]`,
+   *   `d_keys[i].Alternate()[i]` will not be accessed nor modified.
    * - @devicestorageP
    * - @devicestorage
    *
    * @par Snippet
-   * The code snippet below illustrates the batched sorting of three segments 
+   * The code snippet below illustrates the batched sorting of three segments
    * (with one zero-length segment) of `int` keys.
    * @par
    * @code
-   * #include <cub/cub.cuh>   
+   * #include <cub/cub.cuh>
    * // or equivalently <cub/device/device_segmented_radix_sort.cuh>
    *
-   * // Declare, allocate, and initialize device-accessible pointers for 
+   * // Declare, allocate, and initialize device-accessible pointers for
    * // sorting data
    * int  num_items;          // e.g., 7
    * int  num_segments;       // e.g., 3
@@ -1225,61 +1232,61 @@ struct DeviceSegmentedRadixSort
    *
    * @endcode
    *
-   * @tparam KeyT             
+   * @tparam KeyT
    *   **[inferred]** Key type
    *
-   * @tparam BeginOffsetIteratorT  
-   *   **[inferred]** Random-access input iterator type for reading segment 
+   * @tparam BeginOffsetIteratorT
+   *   **[inferred]** Random-access input iterator type for reading segment
    *   beginning offsets \iterator
    *
-   * @tparam EndOffsetIteratorT    
-   *   **[inferred]** Random-access input iterator type for reading segment 
+   * @tparam EndOffsetIteratorT
+   *   **[inferred]** Random-access input iterator type for reading segment
    *   ending offsets \iterator
    *
-   * @param[in] d_temp_storage  
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no work 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no work
    *   is done.
    *
-   * @param[in,out] temp_storage_bytes  
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of `d_temp_storage` allocation
    *
-   * @param[in,out] d_keys  
-   *   Reference to the double-buffer of keys whose "current" device-accessible 
-   *   buffer contains the unsorted input keys and, upon return, is updated to 
+   * @param[in,out] d_keys
+   *   Reference to the double-buffer of keys whose "current" device-accessible
+   *   buffer contains the unsorted input keys and, upon return, is updated to
    *   point to the sorted output keys
    *
-   * @param[in] num_items  
+   * @param[in] num_items
    *   The total number of items within the segmented array, including items not
    *   covered by segments. `num_items` should match the largest element within
    *   the range `[d_end_offsets, d_end_offsets + num_segments)`.
    *
-   * @param[in] num_segments  
+   * @param[in] num_segments
    *   The number of segments that comprise the sorting data
    *
-   * @param[in] d_begin_offsets  
-   *   Random-access input iterator to the sequence of beginning offsets of 
-   *   length `num_segments`, such that `d_begin_offsets[i]` is the first 
-   *   element of the *i*<sup>th</sup> data segment in `d_keys_*` and 
+   * @param[in] d_begin_offsets
+   *   Random-access input iterator to the sequence of beginning offsets of
+   *   length `num_segments`, such that `d_begin_offsets[i]` is the first
+   *   element of the *i*<sup>th</sup> data segment in `d_keys_*` and
    *   `d_values_*`
    *
-   * @param[in] d_end_offsets  
-   *   Random-access input iterator to the sequence of ending offsets of length 
-   *   `num_segments`, such that `d_end_offsets[i] - 1` is the last element of 
-   *   the *i*<sup>th</sup> data segment in `d_keys_*` and `d_values_*`. 
+   * @param[in] d_end_offsets
+   *   Random-access input iterator to the sequence of ending offsets of length
+   *   `num_segments`, such that `d_end_offsets[i] - 1` is the last element of
+   *   the *i*<sup>th</sup> data segment in `d_keys_*` and `d_values_*`.
    *   If `d_end_offsets[i] - 1` <= d_begin_offsets[i]`, the *i*<sup>th</sup>
    *   is considered empty.
    *
-   * @param[in] begin_bit  
-   *   **[optional]** The least-significant bit index (inclusive)  
+   * @param[in] begin_bit
+   *   **[optional]** The least-significant bit index (inclusive)
    *   needed for key comparison
    *
-   * @param[in] end_bit  
-   *   **[optional]** The most-significant bit index (exclusive) needed for key 
+   * @param[in] end_bit
+   *   **[optional]** The most-significant bit index (exclusive) needed for key
    *   comparison (e.g., `sizeof(unsigned int) * 8`)
    *
-   * @param[in] stream  
-   *   **[optional]** CUDA stream to launch kernels within.  
+   * @param[in] stream
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    */
   template <typename KeyT,
@@ -1355,38 +1362,38 @@ struct DeviceSegmentedRadixSort
   }
 
   /**
-   * @brief Sorts segments of keys into descending order. 
+   * @brief Sorts segments of keys into descending order.
    * (`~2N` auxiliary storage required).
    *
    * @par
    * - The contents of the input data are not altered by the sorting operation
    * - When input a contiguous sequence of segments, a single sequence
-   *   `segment_offsets` (of length `num_segments + 1`) can be aliased for both 
-   *   the `d_begin_offsets` and `d_end_offsets` parameters (where the latter 
+   *   `segment_offsets` (of length `num_segments + 1`) can be aliased for both
+   *   the `d_begin_offsets` and `d_end_offsets` parameters (where the latter
    *   is specified as `segment_offsets + 1`).
-   * - An optional bit subrange `[begin_bit, end_bit)` of differentiating key 
-   *   bits can be specified. This can reduce overall sorting overhead and 
+   * - An optional bit subrange `[begin_bit, end_bit)` of differentiating key
+   *   bits can be specified. This can reduce overall sorting overhead and
    *   yield a corresponding performance improvement.
    * - The range `[d_keys_out, d_keys_out + num_items)` shall not overlap
-   *   `[d_keys_in, d_keys_in + num_items)`, 
+   *   `[d_keys_in, d_keys_in + num_items)`,
    *   `[d_begin_offsets, d_begin_offsets + num_segments)` nor
    *   `[d_end_offsets, d_end_offsets + num_segments)` in any way.
-   * - @devicestorageNP For sorting using only `O(P)` temporary storage, see 
+   * - @devicestorageNP For sorting using only `O(P)` temporary storage, see
    *   the sorting interface using DoubleBuffer wrappers below.
-   * - Segments are not required to be contiguous. For all index values `i` 
-   *   outside the specified segments `d_keys_in[i]`, `d_keys_out[i]` will not 
-   *   be accessed nor modified.   
+   * - Segments are not required to be contiguous. For all index values `i`
+   *   outside the specified segments `d_keys_in[i]`, `d_keys_out[i]` will not
+   *   be accessed nor modified.
    * - @devicestorage
    *
    * @par Snippet
-   * The code snippet below illustrates the batched sorting of three segments 
+   * The code snippet below illustrates the batched sorting of three segments
    * (with one zero-length segment) of `int` keys.
    * @par
    * @code
-   * #include <cub/cub.cuh>   
+   * #include <cub/cub.cuh>
    * // or equivalently <cub/device/device_segmented_radix_sort.cuh>
    *
-   * // Declare, allocate, and initialize device-accessible pointers 
+   * // Declare, allocate, and initialize device-accessible pointers
    * // for sorting data
    * int  num_items;          // e.g., 7
    * int  num_segments;       // e.g., 3
@@ -1417,62 +1424,62 @@ struct DeviceSegmentedRadixSort
    *
    * @endcode
    *
-   * @tparam KeyT             
+   * @tparam KeyT
    *   **[inferred]** Key type
    *
-   * @tparam BeginOffsetIteratorT  
-   *   **[inferred]** Random-access input iterator type for reading segment 
+   * @tparam BeginOffsetIteratorT
+   *   **[inferred]** Random-access input iterator type for reading segment
    *   beginning offsets \iterator
    *
-   * @tparam EndOffsetIteratorT    
-   *   **[inferred]** Random-access input iterator type for reading segment 
+   * @tparam EndOffsetIteratorT
+   *   **[inferred]** Random-access input iterator type for reading segment
    *   ending offsets \iterator
    *
-   * @param[in] d_temp_storage  
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no work 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no work
    *   is done.
    *
-   * @param[in,out] temp_storage_bytes  
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of `d_temp_storage` allocation
    *
-   * @param[in] d_keys_in  
+   * @param[in] d_keys_in
    *   Device-accessible pointer to the input data of key data to sort
    *
-   * @param[out] d_keys_out  
+   * @param[out] d_keys_out
    *   Device-accessible pointer to the sorted output sequence of key data
    *
-   * @param[in] num_items  
+   * @param[in] num_items
    *   The total number of items within the segmented array, including items not
    *   covered by segments. `num_items` should match the largest element within
    *   the range `[d_end_offsets, d_end_offsets + num_segments)`.
    *
-   * @param[in] num_segments  
+   * @param[in] num_segments
    *   The number of segments that comprise the sorting data
    *
-   * @param[in] d_begin_offsets  
-   *   Random-access input iterator to the sequence of beginning offsets of 
-   *   length `num_segments`, such that `d_begin_offsets[i]` is the first 
-   *   element of the *i*<sup>th</sup> data segment in `d_keys_*` and 
+   * @param[in] d_begin_offsets
+   *   Random-access input iterator to the sequence of beginning offsets of
+   *   length `num_segments`, such that `d_begin_offsets[i]` is the first
+   *   element of the *i*<sup>th</sup> data segment in `d_keys_*` and
    *   `d_values_*`
    *
-   * @param[in] d_end_offsets  
-   *   Random-access input iterator to the sequence of ending offsets of length 
-   *   `num_segments`, such that `d_end_offsets[i] - 1` is the last element of 
-   *   the *i*<sup>th</sup> data segment in `d_keys_*` and `d_values_*`. 
-   *   If `d_end_offsets[i] - 1 <= d_begin_offsets[i]`, the *i*<sup>th</sup> is 
+   * @param[in] d_end_offsets
+   *   Random-access input iterator to the sequence of ending offsets of length
+   *   `num_segments`, such that `d_end_offsets[i] - 1` is the last element of
+   *   the *i*<sup>th</sup> data segment in `d_keys_*` and `d_values_*`.
+   *   If `d_end_offsets[i] - 1 <= d_begin_offsets[i]`, the *i*<sup>th</sup> is
    *   considered empty.
    *
-   * @param[in] begin_bit  
-   *   **[optional]** The least-significant bit index (inclusive) needed for 
+   * @param[in] begin_bit
+   *   **[optional]** The least-significant bit index (inclusive) needed for
    *   key comparison
    *
-   * @param[in] end_bit  
-   *   **[optional]** The most-significant bit index (exclusive) needed for key 
+   * @param[in] end_bit
+   *   **[optional]** The most-significant bit index (exclusive) needed for key
    *   comparison (e.g., sizeof(unsigned int) * 8)
    *
-   * @param[in] stream  
-   *   **[optional]** CUDA stream to launch kernels within.  
+   * @param[in] stream
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    */
   template <typename KeyT,
@@ -1551,7 +1558,7 @@ struct DeviceSegmentedRadixSort
   }
 
   /**
-   * @brief Sorts segments of keys into descending order. 
+   * @brief Sorts segments of keys into descending order.
    * (`~N` auxiliary storage required).
    *
    * @par
@@ -1559,37 +1566,37 @@ struct DeviceSegmentedRadixSort
    *   DoubleBuffer structure that indicates which of the two buffers is
    *   "current" (and thus contains the input data to be sorted).
    * - The contents of both buffers may be altered by the sorting operation.
-   * - Upon completion, the sorting operation will update the "current" 
-   *   indicator within the DoubleBuffer wrapper to reference which of the two 
-   *   buffers now contains the sorted output sequence (a function of the 
+   * - Upon completion, the sorting operation will update the "current"
+   *   indicator within the DoubleBuffer wrapper to reference which of the two
+   *   buffers now contains the sorted output sequence (a function of the
    *   number of key bits specified and the targeted device architecture).
    * - When input a contiguous sequence of segments, a single sequence
    *   `segment_offsets` (of length `num_segments + 1`) can be aliased
    *   for both the `d_begin_offsets` and `d_end_offsets` parameters (where
    *   the latter is specified as `segment_offsets + 1`).
-   * - An optional bit subrange `[begin_bit, end_bit)` of differentiating key 
-   *   bits can be specified. This can reduce overall sorting overhead and 
+   * - An optional bit subrange `[begin_bit, end_bit)` of differentiating key
+   *   bits can be specified. This can reduce overall sorting overhead and
    *   yield a corresponding performance improvement.
    * - Let `cur = d_keys.Current()` and `alt = d_keys.Alternate()`.
-   *   The range `[cur, cur + num_items)` shall not overlap 
+   *   The range `[cur, cur + num_items)` shall not overlap
    *   `[alt, alt + num_items)`. Both ranges shall not overlap
    *   `[d_begin_offsets, d_begin_offsets + num_segments)` nor
    *   `[d_end_offsets, d_end_offsets + num_segments)` in any way.
-   * - Segments are not required to be contiguous. For all index values `i` 
-   *   outside the specified segments `d_keys.Current()[i]`, 
-   *   `d_keys[i].Alternate()[i]` will not be accessed nor modified.   
+   * - Segments are not required to be contiguous. For all index values `i`
+   *   outside the specified segments `d_keys.Current()[i]`,
+   *   `d_keys[i].Alternate()[i]` will not be accessed nor modified.
    * - @devicestorageP
    * - @devicestorage
    *
    * @par Snippet
-   * The code snippet below illustrates the batched sorting of three segments 
+   * The code snippet below illustrates the batched sorting of three segments
    * (with one zero-length segment) of `int` keys.
    * @par
    * @code
-   * #include <cub/cub.cuh>   
+   * #include <cub/cub.cuh>
    * // or equivalently <cub/device/device_segmented_radix_sort.cuh>
    *
-   * // Declare, allocate, and initialize device-accessible pointers 
+   * // Declare, allocate, and initialize device-accessible pointers
    * // for sorting data
    * int  num_items;          // e.g., 7
    * int  num_segments;       // e.g., 3
@@ -1619,61 +1626,61 @@ struct DeviceSegmentedRadixSort
    * // d_keys.Current()      <-- [8, 7, 6, 9, 5, 3, 0]
    * @endcode
    *
-   * @tparam KeyT             
+   * @tparam KeyT
    *   **[inferred]** Key type
    *
-   * @tparam BeginOffsetIteratorT  
-   *   **[inferred]** Random-access input iterator type for reading segment 
+   * @tparam BeginOffsetIteratorT
+   *   **[inferred]** Random-access input iterator type for reading segment
    *   beginning offsets \iterator
    *
-   * @tparam EndOffsetIteratorT    
-   *   **[inferred]** Random-access input iterator type for reading segment 
+   * @tparam EndOffsetIteratorT
+   *   **[inferred]** Random-access input iterator type for reading segment
    *   ending offsets \iterator
    *
-   * @param[in] d_temp_storage  
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no work 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no work
    *   is done.
    *
-   * @param[in,out] temp_storage_bytes  
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of `d_temp_storage` allocation
    *
-   * @param[in,out] d_keys  
-   *   Reference to the double-buffer of keys whose "current" device-accessible 
-   *   buffer contains the unsorted input keys and, upon return, is updated to 
+   * @param[in,out] d_keys
+   *   Reference to the double-buffer of keys whose "current" device-accessible
+   *   buffer contains the unsorted input keys and, upon return, is updated to
    *   point to the sorted output keys
    *
-   * @param[in] num_items  
+   * @param[in] num_items
    *   The total number of items within the segmented array, including items not
    *   covered by segments. `num_items` should match the largest element within
    *   the range `[d_end_offsets, d_end_offsets + num_segments)`.
    *
-   * @param[in] num_segments  
+   * @param[in] num_segments
    *   The number of segments that comprise the sorting data
    *
-   * @param[in] d_begin_offsets  
-   *   Random-access input iterator to the sequence of beginning offsets of 
-   *   length `num_segments`, such that `d_begin_offsets[i]` is the first 
-   *   element of the *i*<sup>th</sup> data segment in `d_keys_*` and 
+   * @param[in] d_begin_offsets
+   *   Random-access input iterator to the sequence of beginning offsets of
+   *   length `num_segments`, such that `d_begin_offsets[i]` is the first
+   *   element of the *i*<sup>th</sup> data segment in `d_keys_*` and
    *   `d_values_*`
    *
-   * @param[in] d_end_offsets  
-   *   Random-access input iterator to the sequence of ending offsets of length 
-   *   `num_segments`, such that `d_end_offsets[i] - 1` is the last element of 
-   *   the *i*<sup>th</sup> data segment in `d_keys_*` and `d_values_*`.  
-   *   If `d_end_offsets[i] - 1 <= d_begin_offsets[i], the *i*<sup>th</sup> is 
+   * @param[in] d_end_offsets
+   *   Random-access input iterator to the sequence of ending offsets of length
+   *   `num_segments`, such that `d_end_offsets[i] - 1` is the last element of
+   *   the *i*<sup>th</sup> data segment in `d_keys_*` and `d_values_*`.
+   *   If `d_end_offsets[i] - 1 <= d_begin_offsets[i], the *i*<sup>th</sup> is
    *   considered empty.
    *
-   * @param[in] begin_bit  
-   *   **[optional]** The least-significant bit index (inclusive) needed for 
+   * @param[in] begin_bit
+   *   **[optional]** The least-significant bit index (inclusive) needed for
    *   key comparison
    *
-   * @param[in] end_bit  
-   *   **[optional]** The most-significant bit index (exclusive) needed for key 
+   * @param[in] end_bit
+   *   **[optional]** The most-significant bit index (exclusive) needed for key
    *   comparison (e.g., `sizeof(unsigned int) * 8`)
    *
-   * @param[in] stream  
-   *   **[optional]** CUDA stream to launch kernels within.  
+   * @param[in] stream
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    */
   template <typename KeyT,
diff --git a/cub/cub/device/device_segmented_reduce.cuh b/cub/cub/device/device_segmented_reduce.cuh
index a776ac0f9e7..c0bc292a2ec 100644
--- a/cub/cub/device/device_segmented_reduce.cuh
+++ b/cub/cub/device/device_segmented_reduce.cuh
@@ -34,7 +34,14 @@
 
 #pragma once
 
-#include <cub/config.cuh>
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/detail/choose_offset.cuh>
 #include <cub/device/dispatch/dispatch_reduce.cuh>
 #include <cub/device/dispatch/dispatch_reduce_by_key.cuh>
diff --git a/cub/cub/device/device_segmented_sort.cuh b/cub/cub/device/device_segmented_sort.cuh
index 2b86da95078..823df17e249 100644
--- a/cub/cub/device/device_segmented_sort.cuh
+++ b/cub/cub/device/device_segmented_sort.cuh
@@ -34,7 +34,14 @@
 
 #pragma once
 
-#include <cub/config.cuh>
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/device/dispatch/dispatch_segmented_sort.cuh>
 #include <cub/util_deprecated.cuh>
 #include <cub/util_namespace.cuh>
@@ -72,8 +79,8 @@ CUB_NAMESPACE_BEGIN
  * (`unsigned char`, `int`, `double`, etc.) as well as CUDA's `__half` and
  * `__nv_bfloat16` 16-bit floating-point types.
  *
- * @par Segments are not required to be contiguous. Any element of input(s) or 
- * output(s) outside the specified segments will not be accessed nor modified.  
+ * @par Segments are not required to be contiguous. Any element of input(s) or
+ * output(s) outside the specified segments will not be accessed nor modified.
  *
  * @par A simple example
  * @code
@@ -135,12 +142,12 @@ struct DeviceSegmentedSort
    *   guaranteed that the relative order of these two elements will be
    *   preserved by sort.
    * - The range `[d_keys_out, d_keys_out + num_items)` shall not overlap
-   *   `[d_keys_in, d_keys_in + num_items)`, 
+   *   `[d_keys_in, d_keys_in + num_items)`,
    *   `[d_begin_offsets, d_begin_offsets + num_segments)` nor
    *   `[d_end_offsets, d_end_offsets + num_segments)` in any way.
-   * - Segments are not required to be contiguous. For all index values `i` 
-   *   outside the specified segments `d_keys_in[i]`, `d_keys_out[i]` will not 
-   *   be accessed nor modified.   
+   * - Segments are not required to be contiguous. For all index values `i`
+   *   outside the specified segments `d_keys_in[i]`, `d_keys_out[i]` will not
+   *   be accessed nor modified.
    *
    * @par Snippet
    * The code snippet below illustrates the batched sorting of three segments
@@ -309,12 +316,12 @@ struct DeviceSegmentedSort
    *   not guaranteed that the relative order of these two elements will be
    *   preserved by sort.
    * - The range `[d_keys_out, d_keys_out + num_items)` shall not overlap
-   *   `[d_keys_in, d_keys_in + num_items)`, 
+   *   `[d_keys_in, d_keys_in + num_items)`,
    *   `[d_begin_offsets, d_begin_offsets + num_segments)` nor
    *   `[d_end_offsets, d_end_offsets + num_segments)` in any way.
-   * - Segments are not required to be contiguous. For all index values `i` 
-   *   outside the specified segments `d_keys_in[i]`, `d_keys_out[i]` will not 
-   *   be accessed nor modified.   
+   * - Segments are not required to be contiguous. For all index values `i`
+   *   outside the specified segments `d_keys_in[i]`, `d_keys_out[i]` will not
+   *   be accessed nor modified.
    *
    * @par Snippet
    * The code snippet below illustrates the batched sorting of three segments
@@ -490,13 +497,13 @@ struct DeviceSegmentedSort
    *   not guaranteed that the relative order of these two elements will be
    *   preserved by sort.
    * - Let `cur = d_keys.Current()` and `alt = d_keys.Alternate()`.
-   *   The range `[cur, cur + num_items)` shall not overlap 
+   *   The range `[cur, cur + num_items)` shall not overlap
    *   `[alt, alt + num_items)`. Both ranges shall not overlap
    *   `[d_begin_offsets, d_begin_offsets + num_segments)` nor
    *   `[d_end_offsets, d_end_offsets + num_segments)` in any way.
-   * - Segments are not required to be contiguous. For all index values `i` 
-   *   outside the specified segments `d_keys.Current()[i]`, 
-   *   `d_keys[i].Alternate()[i]` will not be accessed nor modified.   
+   * - Segments are not required to be contiguous. For all index values `i`
+   *   outside the specified segments `d_keys.Current()[i]`,
+   *   `d_keys[i].Alternate()[i]` will not be accessed nor modified.
    *
    * @par Snippet
    * The code snippet below illustrates the batched sorting of three segments
@@ -671,13 +678,13 @@ struct DeviceSegmentedSort
    *   not guaranteed that the relative order of these two elements will be
    *   preserved by sort.
    * - Let `cur = d_keys.Current()` and `alt = d_keys.Alternate()`.
-   *   The range `[cur, cur + num_items)` shall not overlap 
+   *   The range `[cur, cur + num_items)` shall not overlap
    *   `[alt, alt + num_items)`. Both ranges shall not overlap
    *   `[d_begin_offsets, d_begin_offsets + num_segments)` nor
    *   `[d_end_offsets, d_end_offsets + num_segments)` in any way.
-   * - Segments are not required to be contiguous. For all index values `i` 
-   *   outside the specified segments `d_keys.Current()[i]`, 
-   *   `d_keys[i].Alternate()[i]` will not be accessed nor modified.   
+   * - Segments are not required to be contiguous. For all index values `i`
+   *   outside the specified segments `d_keys.Current()[i]`,
+   *   `d_keys[i].Alternate()[i]` will not be accessed nor modified.
    *
    * @par Snippet
    * The code snippet below illustrates the batched sorting of three segments
@@ -846,12 +853,12 @@ struct DeviceSegmentedSort
    *   @p x < @p y nor @p y < @p x) then a postcondition of stable sort is that
    *   @p x still precedes @p y.
    * - The range `[d_keys_out, d_keys_out + num_items)` shall not overlap
-   *   `[d_keys_in, d_keys_in + num_items)`, 
+   *   `[d_keys_in, d_keys_in + num_items)`,
    *   `[d_begin_offsets, d_begin_offsets + num_segments)` nor
    *   `[d_end_offsets, d_end_offsets + num_segments)` in any way.
-   * - Segments are not required to be contiguous. For all index values `i` 
-   *   outside the specified segments `d_keys_in[i]`, `d_keys_out[i]` will not 
-   *   be accessed nor modified.   
+   * - Segments are not required to be contiguous. For all index values `i`
+   *   outside the specified segments `d_keys_in[i]`, `d_keys_out[i]` will not
+   *   be accessed nor modified.
    *
    * @par Snippet
    * The code snippet below illustrates the batched sorting of three segments
@@ -1009,12 +1016,12 @@ struct DeviceSegmentedSort
    *   @p x < @p y nor @p y < @p x) then a postcondition of stable sort is that
    *   @p x still precedes @p y.
    * - The range `[d_keys_out, d_keys_out + num_items)` shall not overlap
-   *   `[d_keys_in, d_keys_in + num_items)`, 
+   *   `[d_keys_in, d_keys_in + num_items)`,
    *   `[d_begin_offsets, d_begin_offsets + num_segments)` nor
    *   `[d_end_offsets, d_end_offsets + num_segments)` in any way.
-   * - Segments are not required to be contiguous. For all index values `i` 
-   *   outside the specified segments `d_keys_in[i]`, `d_keys_out[i]` will not 
-   *   be accessed nor modified.   
+   * - Segments are not required to be contiguous. For all index values `i`
+   *   outside the specified segments `d_keys_in[i]`, `d_keys_out[i]` will not
+   *   be accessed nor modified.
    *
    * @par Snippet
    * The code snippet below illustrates the batched sorting of three segments
@@ -1181,13 +1188,13 @@ struct DeviceSegmentedSort
    *   @p x < @p y nor @p y < @p x) then a postcondition of stable sort is that
    *   @p x still precedes @p y.
    * - Let `cur = d_keys.Current()` and `alt = d_keys.Alternate()`.
-   *   The range `[cur, cur + num_items)` shall not overlap 
+   *   The range `[cur, cur + num_items)` shall not overlap
    *   `[alt, alt + num_items)`. Both ranges shall not overlap
    *   `[d_begin_offsets, d_begin_offsets + num_segments)` nor
    *   `[d_end_offsets, d_end_offsets + num_segments)` in any way.
-   * - Segments are not required to be contiguous. For all index values `i` 
-   *   outside the specified segments `d_keys.Current()[i]`, 
-   *   `d_keys[i].Alternate()[i]` will not be accessed nor modified.   
+   * - Segments are not required to be contiguous. For all index values `i`
+   *   outside the specified segments `d_keys.Current()[i]`,
+   *   `d_keys[i].Alternate()[i]` will not be accessed nor modified.
    *
    * @par Snippet
    * The code snippet below illustrates the batched sorting of three segments
@@ -1350,13 +1357,13 @@ struct DeviceSegmentedSort
    *   @p x < @p y nor @p y < @p x) then a postcondition of stable sort is that
    *   @p x still precedes @p y.
    * - Let `cur = d_keys.Current()` and `alt = d_keys.Alternate()`.
-   *   The range `[cur, cur + num_items)` shall not overlap 
+   *   The range `[cur, cur + num_items)` shall not overlap
    *   `[alt, alt + num_items)`. Both ranges shall not overlap
    *   `[d_begin_offsets, d_begin_offsets + num_segments)` nor
    *   `[d_end_offsets, d_end_offsets + num_segments)` in any way.
-   * - Segments are not required to be contiguous. For all index values `i` 
-   *   outside the specified segments `d_keys.Current()[i]`, 
-   *   `d_keys[i].Alternate()[i]` will not be accessed nor modified.   
+   * - Segments are not required to be contiguous. For all index values `i`
+   *   outside the specified segments `d_keys.Current()[i]`,
+   *   `d_keys[i].Alternate()[i]` will not be accessed nor modified.
    *
    * @par Snippet
    * The code snippet below illustrates the batched sorting of three segments
@@ -1520,13 +1527,13 @@ struct DeviceSegmentedSort
    *   guaranteed that the relative order of these two elements will be
    *   preserved by sort.
    * - Let `in` be one of `{d_keys_in, d_values_in}` and `out` be any of
-   *   `{d_keys_out, d_values_out}`. The range `[out, out + num_items)` shall 
-   *   not overlap `[in, in + num_items)`, 
+   *   `{d_keys_out, d_values_out}`. The range `[out, out + num_items)` shall
+   *   not overlap `[in, in + num_items)`,
    *   `[d_begin_offsets, d_begin_offsets + num_segments)` nor
    *   `[d_end_offsets, d_end_offsets + num_segments)` in any way.
-   * - Segments are not required to be contiguous. For all index values `i` 
-   *   outside the specified segments `d_keys_in[i]`, `d_values_in[i]`, 
-   *   `d_keys_out[i]`, `d_values_out[i]` will not be accessed nor modified.   
+   * - Segments are not required to be contiguous. For all index values `i`
+   *   outside the specified segments `d_keys_in[i]`, `d_values_in[i]`,
+   *   `d_keys_out[i]`, `d_values_out[i]` will not be accessed nor modified.
    *
    * @par Snippet
    * The code snippet below illustrates the batched sorting of three segments
@@ -1720,13 +1727,13 @@ struct DeviceSegmentedSort
    *   guaranteed that the relative order of these two elements will be
    *   preserved by sort.
    * - Let `in` be one of `{d_keys_in, d_values_in}` and `out` be any of
-   *   `{d_keys_out, d_values_out}`. The range `[out, out + num_items)` shall 
-   *   not overlap `[in, in + num_items)`, 
+   *   `{d_keys_out, d_values_out}`. The range `[out, out + num_items)` shall
+   *   not overlap `[in, in + num_items)`,
    *   `[d_begin_offsets, d_begin_offsets + num_segments)` nor
    *   `[d_end_offsets, d_end_offsets + num_segments)` in any way.
-   * - Segments are not required to be contiguous. For all index values `i` 
-   *   outside the specified segments `d_keys_in[i]`, `d_values_in[i]`, 
-   *   `d_keys_out[i]`, `d_values_out[i]` will not be accessed nor modified.   
+   * - Segments are not required to be contiguous. For all index values `i`
+   *   outside the specified segments `d_keys_in[i]`, `d_values_in[i]`,
+   *   `d_keys_out[i]`, `d_values_out[i]` will not be accessed nor modified.
    *
    * @par Snippet
    * The code snippet below illustrates the batched sorting of three segments
@@ -1930,16 +1937,16 @@ struct DeviceSegmentedSort
    *   @p j are equivalent: neither one is less than the other. It is not
    *   guaranteed that the relative order of these two elements will be
    *   preserved by sort.
-   * - Let `cur` be one of `{d_keys.Current(), d_values.Current()}` and `alt` 
-   *   be any of `{d_keys.Alternate(), d_values.Alternate()}`. The range 
-   *   `[cur, cur + num_items)` shall not overlap 
+   * - Let `cur` be one of `{d_keys.Current(), d_values.Current()}` and `alt`
+   *   be any of `{d_keys.Alternate(), d_values.Alternate()}`. The range
+   *   `[cur, cur + num_items)` shall not overlap
    *   `[alt, alt + num_items)`. Both ranges shall not overlap
    *   `[d_begin_offsets, d_begin_offsets + num_segments)` nor
    *   `[d_end_offsets, d_end_offsets + num_segments)` in any way.
-   * - Segments are not required to be contiguous. For all index values `i` 
-   *   outside the specified segments `d_keys.Current()[i]`, 
-   *   `d_values.Current()[i]`, `d_keys.Alternate()[i]`, 
-   *   `d_values.Alternate()[i]` will not be accessed nor modified.   
+   * - Segments are not required to be contiguous. For all index values `i`
+   *   outside the specified segments `d_keys.Current()[i]`,
+   *   `d_values.Current()[i]`, `d_keys.Alternate()[i]`,
+   *   `d_values.Alternate()[i]` will not be accessed nor modified.
    *
    * @par Snippet
    * The code snippet below illustrates the batched sorting of three segments
@@ -2131,16 +2138,16 @@ struct DeviceSegmentedSort
    *   @p i and @p j are equivalent: neither one is less than the other. It is
    *   not guaranteed that the relative order of these two elements will be
    *   preserved by sort.
-   * - Let `cur` be one of `{d_keys.Current(), d_values.Current()}` and `alt` 
-   *   be any of `{d_keys.Alternate(), d_values.Alternate()}`. The range 
-   *   `[cur, cur + num_items)` shall not overlap 
+   * - Let `cur` be one of `{d_keys.Current(), d_values.Current()}` and `alt`
+   *   be any of `{d_keys.Alternate(), d_values.Alternate()}`. The range
+   *   `[cur, cur + num_items)` shall not overlap
    *   `[alt, alt + num_items)`. Both ranges shall not overlap
    *   `[d_begin_offsets, d_begin_offsets + num_segments)` nor
    *   `[d_end_offsets, d_end_offsets + num_segments)` in any way.
-   * - Segments are not required to be contiguous. For all index values `i` 
-   *   outside the specified segments `d_keys.Current()[i]`, 
-   *   `d_values.Current()[i]`, `d_keys.Alternate()[i]`, 
-   *   `d_values.Alternate()[i]` will not be accessed nor modified.   
+   * - Segments are not required to be contiguous. For all index values `i`
+   *   outside the specified segments `d_keys.Current()[i]`,
+   *   `d_values.Current()[i]`, `d_keys.Alternate()[i]`,
+   *   `d_values.Alternate()[i]` will not be accessed nor modified.
    *
    * @par Snippet
    * The code snippet below illustrates the batched sorting of three segments
@@ -2327,13 +2334,13 @@ struct DeviceSegmentedSort
    *   @p x < @p y nor @p y < @p x) then a postcondition of stable sort is that
    *   @p x still precedes @p y.
    * - Let `in` be one of `{d_keys_in, d_values_in}` and `out` be any of
-   *   `{d_keys_out, d_values_out}`. The range `[out, out + num_items)` shall 
-   *   not overlap `[in, in + num_items)`, 
+   *   `{d_keys_out, d_values_out}`. The range `[out, out + num_items)` shall
+   *   not overlap `[in, in + num_items)`,
    *   `[d_begin_offsets, d_begin_offsets + num_segments)` nor
    *   `[d_end_offsets, d_end_offsets + num_segments)` in any way.
-   * - Segments are not required to be contiguous. For all index values `i` 
-   *   outside the specified segments `d_keys_in[i]`, `d_values_in[i]`, 
-   *   `d_keys_out[i]`, `d_values_out[i]` will not be accessed nor modified.   
+   * - Segments are not required to be contiguous. For all index values `i`
+   *   outside the specified segments `d_keys_in[i]`, `d_values_in[i]`,
+   *   `d_keys_out[i]`, `d_values_out[i]` will not be accessed nor modified.
    *
    * @par Snippet
    * The code snippet below illustrates the batched sorting of three segments
@@ -2520,13 +2527,13 @@ struct DeviceSegmentedSort
    *   @p x < @p y nor @p y < @p x) then a postcondition of stable sort is that
    *   @p x still precedes @p y.
    * - Let `in` be one of `{d_keys_in, d_values_in}` and `out` be any of
-   *   `{d_keys_out, d_values_out}`. The range `[out, out + num_items)` shall 
-   *   not overlap `[in, in + num_items)`, 
+   *   `{d_keys_out, d_values_out}`. The range `[out, out + num_items)` shall
+   *   not overlap `[in, in + num_items)`,
    *   `[d_begin_offsets, d_begin_offsets + num_segments)` nor
    *   `[d_end_offsets, d_end_offsets + num_segments)` in any way.
-   * - Segments are not required to be contiguous. For all index values `i` 
-   *   outside the specified segments `d_keys_in[i]`, `d_values_in[i]`, 
-   *   `d_keys_out[i]`, `d_values_out[i]` will not be accessed nor modified.   
+   * - Segments are not required to be contiguous. For all index values `i`
+   *   outside the specified segments `d_keys_in[i]`, `d_values_in[i]`,
+   *   `d_keys_out[i]`, `d_values_out[i]` will not be accessed nor modified.
    *
    * @par Snippet
    * The code snippet below illustrates the batched sorting of three segments
@@ -2723,16 +2730,16 @@ struct DeviceSegmentedSort
    *   @p x precedes @p y, and if the two elements are equivalent (neither
    *   @p x < @p y nor @p y < @p x) then a postcondition of stable sort is that
    *   @p x still precedes @p y.
-   * - Let `cur` be one of `{d_keys.Current(), d_values.Current()}` and `alt` 
-   *   be any of `{d_keys.Alternate(), d_values.Alternate()}`. The range 
-   *   `[cur, cur + num_items)` shall not overlap 
+   * - Let `cur` be one of `{d_keys.Current(), d_values.Current()}` and `alt`
+   *   be any of `{d_keys.Alternate(), d_values.Alternate()}`. The range
+   *   `[cur, cur + num_items)` shall not overlap
    *   `[alt, alt + num_items)`. Both ranges shall not overlap
    *   `[d_begin_offsets, d_begin_offsets + num_segments)` nor
    *   `[d_end_offsets, d_end_offsets + num_segments)` in any way.
-   * - Segments are not required to be contiguous. For all index values `i` 
-   *   outside the specified segments `d_keys.Current()[i]`, 
-   *   `d_values.Current()[i]`, `d_keys.Alternate()[i]`, 
-   *   `d_values.Alternate()[i]` will not be accessed nor modified.   
+   * - Segments are not required to be contiguous. For all index values `i`
+   *   outside the specified segments `d_keys.Current()[i]`,
+   *   `d_values.Current()[i]`, `d_keys.Alternate()[i]`,
+   *   `d_values.Alternate()[i]` will not be accessed nor modified.
    *
    * @par Snippet
    * The code snippet below illustrates the batched sorting of three segments
@@ -2918,16 +2925,16 @@ struct DeviceSegmentedSort
    *   @p x precedes @p y, and if the two elements are equivalent (neither
    *   @p x < @p y nor @p y < @p x) then a postcondition of stable sort is that
    *   @p x still precedes @p y.
-   * - Let `cur` be one of `{d_keys.Current(), d_values.Current()}` and `alt` 
-   *   be any of `{d_keys.Alternate(), d_values.Alternate()}`. The range 
-   *   `[cur, cur + num_items)` shall not overlap 
+   * - Let `cur` be one of `{d_keys.Current(), d_values.Current()}` and `alt`
+   *   be any of `{d_keys.Alternate(), d_values.Alternate()}`. The range
+   *   `[cur, cur + num_items)` shall not overlap
    *   `[alt, alt + num_items)`. Both ranges shall not overlap
    *   `[d_begin_offsets, d_begin_offsets + num_segments)` nor
    *   `[d_end_offsets, d_end_offsets + num_segments)` in any way.
-   * - Segments are not required to be contiguous. For all index values `i` 
-   *   outside the specified segments `d_keys.Current()[i]`, 
-   *   `d_values.Current()[i]`, `d_keys.Alternate()[i]`, 
-   *   `d_values.Alternate()[i]` will not be accessed nor modified.   
+   * - Segments are not required to be contiguous. For all index values `i`
+   *   outside the specified segments `d_keys.Current()[i]`,
+   *   `d_values.Current()[i]`, `d_keys.Alternate()[i]`,
+   *   `d_values.Alternate()[i]` will not be accessed nor modified.
    *
    * @par Snippet
    * The code snippet below illustrates the batched sorting of three segments
diff --git a/cub/cub/device/device_select.cuh b/cub/cub/device/device_select.cuh
index f21431391ab..22599ea82b1 100644
--- a/cub/cub/device/device_select.cuh
+++ b/cub/cub/device/device_select.cuh
@@ -13,9 +13,9 @@
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
  *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 
- * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 
- * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  * ARE DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE FOR ANY
  * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
@@ -27,17 +27,24 @@
  ******************************************************************************/
 
 /**
- * @file cub::DeviceSelect provides device-wide, parallel operations for 
- *       compacting selected items from sequences of data items residing within 
+ * @file cub::DeviceSelect provides device-wide, parallel operations for
+ *       compacting selected items from sequences of data items residing within
  *       device-accessible memory.
  */
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <iterator>
 #include <stdio.h>
 
-#include <cub/config.cuh>
 #include <cub/device/dispatch/dispatch_select_if.cuh>
 #include <cub/device/dispatch/dispatch_unique_by_key.cuh>
 #include <cub/util_deprecated.cuh>
@@ -46,8 +53,8 @@ CUB_NAMESPACE_BEGIN
 
 
 /**
- * @brief DeviceSelect provides device-wide, parallel operations for compacting 
- *        selected items from sequences of data items residing within 
+ * @brief DeviceSelect provides device-wide, parallel operations for compacting
+ *        selected items from sequences of data items residing within
  *        device-accessible memory. ![](select_logo.png)
  * @ingroup SingleModule
  *
@@ -62,15 +69,15 @@ CUB_NAMESPACE_BEGIN
  * @linear_performance{select-flagged, select-if, and select-unique}
  *
  * @par
- * The following chart illustrates DeviceSelect::If performance across 
- * different CUDA architectures for `int32` items, where 50% of the items are 
+ * The following chart illustrates DeviceSelect::If performance across
+ * different CUDA architectures for `int32` items, where 50% of the items are
  * randomly selected.
  *
  * @image html select_if_int32_50_percent.png
  *
  * @par
- * The following chart illustrates DeviceSelect::Unique performance across 
- * different CUDA architectures for `int32` items where segments have lengths 
+ * The following chart illustrates DeviceSelect::Unique performance across
+ * different CUDA architectures for `int32` items where segments have lengths
  * uniformly sampled from `[1, 1000]`.
  *
  * @image html select_unique_int32_len_500.png
@@ -82,28 +89,28 @@ CUB_NAMESPACE_BEGIN
 struct DeviceSelect
 {
   /**
-   * @brief Uses the `d_flags` sequence to selectively copy the corresponding 
-   *        items from `d_in` into `d_out`. The total number of items selected 
+   * @brief Uses the `d_flags` sequence to selectively copy the corresponding
+   *        items from `d_in` into `d_out`. The total number of items selected
    *        is written to `d_num_selected_out`. ![](select_flags_logo.png)
    *
    * @par
-   * - The value type of `d_flags` must be castable to `bool` (e.g., `bool`, 
+   * - The value type of `d_flags` must be castable to `bool` (e.g., `bool`,
    *   `char`, `int`, etc.).
-   * - Copies of the selected items are compacted into `d_out` and maintain 
+   * - Copies of the selected items are compacted into `d_out` and maintain
    *   their original relative ordering.
-   * - The range `[d_out, d_out + *d_num_selected_out)` shall not overlap 
+   * - The range `[d_out, d_out + *d_num_selected_out)` shall not overlap
    *   `[d_in, d_in + num_items)`, `[d_flags, d_flags + num_items)` nor
    *   `d_num_selected_out` in any way.
    * - @devicestorage
    *
    * @par Snippet
-   * The code snippet below illustrates the compaction of items selected from 
+   * The code snippet below illustrates the compaction of items selected from
    * an `int` device vector.
    * @par
    * @code
    * #include <cub/cub.cuh>  // or equivalently <cub/device/device_select.cuh>
    *
-   * // Declare, allocate, and initialize device-accessible pointers for input, 
+   * // Declare, allocate, and initialize device-accessible pointers for input,
    * // flags, and output
    * int  num_items;              // e.g., 8
    * int  *d_in;                  // e.g., [1, 2, 3, 4, 5, 6, 7, 8]
@@ -116,7 +123,7 @@ struct DeviceSelect
    * void     *d_temp_storage = NULL;
    * size_t   temp_storage_bytes = 0;
    * cub::DeviceSelect::Flagged(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_in, d_flags, d_out, d_num_selected_out, num_items);
    *
    * // Allocate temporary storage
@@ -124,7 +131,7 @@ struct DeviceSelect
    *
    * // Run selection
    * cub::DeviceSelect::Flagged(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_in, d_flags, d_out, d_num_selected_out, num_items);
    *
    * // d_out                 <-- [1, 4, 6, 7]
@@ -132,48 +139,48 @@ struct DeviceSelect
    *
    * @endcode
    *
-   * @tparam InputIteratorT       
-   *   **[inferred]** Random-access input iterator type for reading input 
+   * @tparam InputIteratorT
+   *   **[inferred]** Random-access input iterator type for reading input
    *   items \iterator
    *
-   * @tparam FlagIterator         
-   *   **[inferred]** Random-access input iterator type for reading selection 
+   * @tparam FlagIterator
+   *   **[inferred]** Random-access input iterator type for reading selection
    *   flags \iterator
    *
-   * @tparam OutputIteratorT      
-   *   **[inferred]** Random-access output iterator type for writing selected 
+   * @tparam OutputIteratorT
+   *   **[inferred]** Random-access output iterator type for writing selected
    *   items \iterator
    *
-   * @tparam NumSelectedIteratorT  
-   *   **[inferred]** Output iterator type for recording the number of items 
+   * @tparam NumSelectedIteratorT
+   *   **[inferred]** Output iterator type for recording the number of items
    *   selected \iterator
    *
-   * @param[in] d_temp_storage  
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no work 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no work
    *   is done.
    *
-   * @param[in,out] temp_storage_bytes  
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of `d_temp_storage` allocation
    *
-   * @param[in] d_in  
+   * @param[in] d_in
    *   Pointer to the input sequence of data items
    *
-   * @param[in] d_flags  
+   * @param[in] d_flags
    *   Pointer to the input sequence of selection flags
    *
-   * @param[out] d_out  
+   * @param[out] d_out
    *   Pointer to the output sequence of selected data items
    *
-   * @param[out] d_num_selected_out  
-   *   Pointer to the output total number of items selected 
+   * @param[out] d_num_selected_out
+   *   Pointer to the output total number of items selected
    *   (i.e., length of `d_out`)
    *
-   * @param[in] num_items  
+   * @param[in] num_items
    *   Total number of input items (i.e., length of `d_in`)
    *
-   * @param[in] stream  
-   *   **[optional]** CUDA stream to launch kernels within.  
+   * @param[in] stream
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    */
   template <typename InputIteratorT,
@@ -243,30 +250,30 @@ struct DeviceSelect
                                          num_items,
                                          stream);
   }
-   
+
   /**
-   * @brief Uses the `d_flags` sequence to selectively compact the items in 
-   *        `d_data`. The total number of items selected is written to 
+   * @brief Uses the `d_flags` sequence to selectively compact the items in
+   *        `d_data`. The total number of items selected is written to
    *        `d_num_selected_out`. ![](select_flags_logo.png)
    *
    * @par
-   * - The value type of `d_flags` must be castable to `bool` (e.g., `bool`, 
+   * - The value type of `d_flags` must be castable to `bool` (e.g., `bool`,
    *   `char`, `int`, etc.).
-   * - Copies of the selected items are compacted in-place and maintain 
+   * - Copies of the selected items are compacted in-place and maintain
    *   their original relative ordering.
-   * - The `d_data` may equal `d_flags`. The range 
-   *  `[d_data, d_data + num_items)` shall not overlap 
+   * - The `d_data` may equal `d_flags`. The range
+   *  `[d_data, d_data + num_items)` shall not overlap
    *  `[d_flags, d_flags + num_items)` in any other way.
    * - @devicestorage
    *
    * @par Snippet
-   * The code snippet below illustrates the compaction of items selected from 
+   * The code snippet below illustrates the compaction of items selected from
    * an `int` device vector.
    * @par
    * @code
    * #include <cub/cub.cuh>  // or equivalently <cub/device/device_select.cuh>
    *
-   * // Declare, allocate, and initialize device-accessible pointers for input, 
+   * // Declare, allocate, and initialize device-accessible pointers for input,
    * // flags, and output
    * int  num_items;              // e.g., 8
    * int  *d_data;                // e.g., [1, 2, 3, 4, 5, 6, 7, 8]
@@ -278,7 +285,7 @@ struct DeviceSelect
    * void     *d_temp_storage = NULL;
    * size_t   temp_storage_bytes = 0;
    * cub::DeviceSelect::Flagged(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_in, d_flags, d_num_selected_out, num_items);
    *
    * // Allocate temporary storage
@@ -286,7 +293,7 @@ struct DeviceSelect
    *
    * // Run selection
    * cub::DeviceSelect::Flagged(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_in, d_flags, d_num_selected_out, num_items);
    *
    * // d_data                <-- [1, 4, 6, 7]
@@ -294,40 +301,40 @@ struct DeviceSelect
    *
    * @endcode
    *
-   * @tparam IteratorT       
-   *   **[inferred]** Random-access iterator type for reading and writing 
+   * @tparam IteratorT
+   *   **[inferred]** Random-access iterator type for reading and writing
    *   selected items \iterator
    *
-   * @tparam FlagIterator         
-   *   **[inferred]** Random-access input iterator type for reading selection 
+   * @tparam FlagIterator
+   *   **[inferred]** Random-access input iterator type for reading selection
    *   flags \iterator
    *
-   * @tparam NumSelectedIteratorT  
-   *   **[inferred]** Output iterator type for recording the number of items 
+   * @tparam NumSelectedIteratorT
+   *   **[inferred]** Output iterator type for recording the number of items
    *   selected \iterator
    *
-   * @param[in] d_temp_storage  
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no work 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no work
    *   is done.
    *
-   * @param[in,out] temp_storage_bytes  
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of `d_temp_storage` allocation
    *
    * @param[in,out] d_data
    *   Pointer to the sequence of data items
    *
-   * @param[in] d_flags  
+   * @param[in] d_flags
    *   Pointer to the input sequence of selection flags
    *
-   * @param[out] d_num_selected_out  
-   *   Pointer to the output total number of items selected 
+   * @param[out] d_num_selected_out
+   *   Pointer to the output total number of items selected
    *
-   * @param[in] num_items  
+   * @param[in] num_items
    *   Total number of input items (i.e., length of `d_data`)
    *
-   * @param[in] stream  
-   *   **[optional]** CUDA stream to launch kernels within.  
+   * @param[in] stream
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    */
   template <typename IteratorT,
@@ -355,7 +362,7 @@ struct DeviceSelect
                             SelectOp,
                             EqualityOp,
                             OffsetT,
-                            false, 
+                            false,
                             may_alias>::Dispatch(d_temp_storage,
                                                  temp_storage_bytes,
                                                  d_data, // in
@@ -395,20 +402,20 @@ struct DeviceSelect
   }
 
   /**
-   * @brief Uses the `select_op` functor to selectively copy items from `d_in` 
-   *        into `d_out`. The total number of items selected is written to 
+   * @brief Uses the `select_op` functor to selectively copy items from `d_in`
+   *        into `d_out`. The total number of items selected is written to
    *        `d_num_selected_out`. ![](select_logo.png)
    *
    * @par
-   * - Copies of the selected items are compacted into `d_out` and maintain 
+   * - Copies of the selected items are compacted into `d_out` and maintain
    *   their original relative ordering.
-   * - The range `[d_out, d_out + *d_num_selected_out)` shall not overlap 
+   * - The range `[d_out, d_out + *d_num_selected_out)` shall not overlap
    *   `[d_in, d_in + num_items)` nor `d_num_selected_out` in any way.
    * - @devicestorage
    *
    * @par Performance
-   * The following charts illustrate saturated select-if performance across 
-   * different CUDA architectures for `int32` and `int64` items, respectively. 
+   * The following charts illustrate saturated select-if performance across
+   * different CUDA architectures for `int32` and `int64` items, respectively.
    * Items are selected with 50% probability.
    *
    * @image html select_if_int32_50_percent.png
@@ -421,7 +428,7 @@ struct DeviceSelect
    * @image html select_if_int64_5_percent.png
    *
    * @par Snippet
-   * The code snippet below illustrates the compaction of items selected from 
+   * The code snippet below illustrates the compaction of items selected from
    * an `int` device vector.
    * @par
    * @code
@@ -441,7 +448,7 @@ struct DeviceSelect
    *     }
    * };
    *
-   * // Declare, allocate, and initialize device-accessible pointers 
+   * // Declare, allocate, and initialize device-accessible pointers
    * // for input and output
    * int      num_items;              // e.g., 8
    * int      *d_in;                  // e.g., [0, 2, 3, 9, 5, 2, 81, 8]
@@ -454,7 +461,7 @@ struct DeviceSelect
    * void     *d_temp_storage = NULL;
    * size_t   temp_storage_bytes = 0;
    * cub::DeviceSelect::If(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_in, d_out, d_num_selected_out, num_items, select_op);
    *
    * // Allocate temporary storage
@@ -462,55 +469,55 @@ struct DeviceSelect
    *
    * // Run selection
    * cub::DeviceSelect::If(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_in, d_out, d_num_selected_out, num_items, select_op);
    *
    * // d_out                 <-- [0, 2, 3, 5, 2]
    * // d_num_selected_out    <-- [5]
    * @endcode
    *
-   * @tparam InputIteratorT       
-   *   **[inferred]** Random-access input iterator type for reading input 
+   * @tparam InputIteratorT
+   *   **[inferred]** Random-access input iterator type for reading input
    *   items \iterator
    *
-   * @tparam OutputIteratorT      
-   *   **[inferred]** Random-access output iterator type for writing selected 
+   * @tparam OutputIteratorT
+   *   **[inferred]** Random-access output iterator type for writing selected
    *   items \iterator
    *
-   * @tparam NumSelectedIteratorT  
-   *   **[inferred]** Output iterator type for recording the number of items 
+   * @tparam NumSelectedIteratorT
+   *   **[inferred]** Output iterator type for recording the number of items
    *   selected \iterator
    *
-   * @tparam SelectOp             
-   *   **[inferred]** Selection operator type having member 
+   * @tparam SelectOp
+   *   **[inferred]** Selection operator type having member
    *   `bool operator()(const T &a)`
    *
-   * @param[in] d_temp_storage  
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no work 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no work
    *   is done.
    *
-   * @param[in,out] temp_storage_bytes  
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of `d_temp_storage` allocation
    *
-   * @param[in] d_in  
+   * @param[in] d_in
    *   Pointer to the input sequence of data items
    *
-   * @param[out] d_out  
+   * @param[out] d_out
    *   Pointer to the output sequence of selected data items
    *
-   * @param[out] d_num_selected_out  
-   *   Pointer to the output total number of items selected 
+   * @param[out] d_num_selected_out
+   *   Pointer to the output total number of items selected
    *   (i.e., length of `d_out`)
    *
-   * @param[in] num_items  
+   * @param[in] num_items
    *   Total number of input items (i.e., length of `d_in`)
    *
-   * @param[in] select_op  
+   * @param[in] select_op
    *   Unary selection operator
    *
-   * @param[in] stream  
-   *   **[optional]** CUDA stream to launch kernels within.  
+   * @param[in] stream
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    */
   template <typename InputIteratorT,
@@ -580,17 +587,17 @@ struct DeviceSelect
   }
 
   /**
-   * @brief Uses the `select_op` functor to selectively compact items in 
-   *        `d_data`. The total number of items selected is written to 
+   * @brief Uses the `select_op` functor to selectively compact items in
+   *        `d_data`. The total number of items selected is written to
    *        `d_num_selected_out`. ![](select_logo.png)
    *
    * @par
-   * - Copies of the selected items are compacted in `d_data` and maintain 
+   * - Copies of the selected items are compacted in `d_data` and maintain
    *   their original relative ordering.
    * - @devicestorage
    *
    * @par Snippet
-   * The code snippet below illustrates the compaction of items selected from 
+   * The code snippet below illustrates the compaction of items selected from
    * an `int` device vector.
    * @par
    * @code
@@ -610,7 +617,7 @@ struct DeviceSelect
    *     }
    * };
    *
-   * // Declare, allocate, and initialize device-accessible pointers 
+   * // Declare, allocate, and initialize device-accessible pointers
    * // for input and output
    * int      num_items;              // e.g., 8
    * int      *d_data;                // e.g., [0, 2, 3, 9, 5, 2, 81, 8]
@@ -622,7 +629,7 @@ struct DeviceSelect
    * void     *d_temp_storage = NULL;
    * size_t   temp_storage_bytes = 0;
    * cub::DeviceSelect::If(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_data, d_num_selected_out, num_items, select_op);
    *
    * // Allocate temporary storage
@@ -630,47 +637,47 @@ struct DeviceSelect
    *
    * // Run selection
    * cub::DeviceSelect::If(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_data, d_num_selected_out, num_items, select_op);
    *
    * // d_data                <-- [0, 2, 3, 5, 2]
    * // d_num_selected_out    <-- [5]
    * @endcode
    *
-   * @tparam IteratorT       
-   *   **[inferred]** Random-access input iterator type for reading and 
+   * @tparam IteratorT
+   *   **[inferred]** Random-access input iterator type for reading and
    *   writing items \iterator
    *
-   * @tparam NumSelectedIteratorT  
-   *   **[inferred]** Output iterator type for recording the number of items 
+   * @tparam NumSelectedIteratorT
+   *   **[inferred]** Output iterator type for recording the number of items
    *   selected \iterator
    *
-   * @tparam SelectOp             
-   *   **[inferred]** Selection operator type having member 
+   * @tparam SelectOp
+   *   **[inferred]** Selection operator type having member
    *   `bool operator()(const T &a)`
    *
-   * @param[in] d_temp_storage  
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no work 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no work
    *   is done.
    *
-   * @param[in,out] temp_storage_bytes  
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of `d_temp_storage` allocation
    *
    * @param[in,out] d_data
    *   Pointer to the sequence of data items
    *
-   * @param[out] d_num_selected_out  
-   *   Pointer to the output total number of items selected 
+   * @param[out] d_num_selected_out
+   *   Pointer to the output total number of items selected
    *
-   * @param[in] num_items  
+   * @param[in] num_items
    *   Total number of input items (i.e., length of `d_data`)
    *
-   * @param[in] select_op  
+   * @param[in] select_op
    *   Unary selection operator
    *
-   * @param[in] stream  
-   *   **[optional]** CUDA stream to launch kernels within.  
+   * @param[in] stream
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    */
   template <typename IteratorT,
@@ -737,43 +744,43 @@ struct DeviceSelect
   }
 
   /**
-   * @brief Given an input sequence `d_in` having runs of consecutive 
-   *        equal-valued keys, only the first key from each run is selectively 
-   *        copied to `d_out`. The total number of items selected is written to 
+   * @brief Given an input sequence `d_in` having runs of consecutive
+   *        equal-valued keys, only the first key from each run is selectively
+   *        copied to `d_out`. The total number of items selected is written to
    *        `d_num_selected_out`. ![](unique_logo.png)
    *
    * @par
-   * - The `==` equality operator is used to determine whether keys are 
+   * - The `==` equality operator is used to determine whether keys are
    *   equivalent
-   * - Copies of the selected items are compacted into `d_out` and maintain 
+   * - Copies of the selected items are compacted into `d_out` and maintain
    *   their original relative ordering.
-   * - The range `[d_out, d_out + *d_num_selected_out)` shall not overlap 
+   * - The range `[d_out, d_out + *d_num_selected_out)` shall not overlap
    *   `[d_in, d_in + num_items)` nor `d_num_selected_out` in any way.
    * - @devicestorage
    *
    * @par Performance
    * The following charts illustrate saturated select-unique performance across different
-   * CUDA architectures for `int32` and `int64` items, respectively. Segments 
+   * CUDA architectures for `int32` and `int64` items, respectively. Segments
    * have lengths uniformly sampled from `[1, 1000]`.
    *
    * @image html select_unique_int32_len_500.png
    * @image html select_unique_int64_len_500.png
    *
    * @par
-   * The following charts are similar, but with segment lengths uniformly 
+   * The following charts are similar, but with segment lengths uniformly
    * sampled from `[1, 10]`:
    *
    * @image html select_unique_int32_len_5.png
    * @image html select_unique_int64_len_5.png
    *
    * @par Snippet
-   * The code snippet below illustrates the compaction of items selected from 
+   * The code snippet below illustrates the compaction of items selected from
    * an `int` device vector.
    * @par
    * @code
    * #include <cub/cub.cuh>   // or equivalently <cub/device/device_select.cuh>
    *
-   * // Declare, allocate, and initialize device-accessible pointers 
+   * // Declare, allocate, and initialize device-accessible pointers
    * // for input and output
    * int  num_items;              // e.g., 8
    * int  *d_in;                  // e.g., [0, 2, 2, 9, 5, 5, 5, 8]
@@ -785,7 +792,7 @@ struct DeviceSelect
    * void     *d_temp_storage = NULL;
    * size_t   temp_storage_bytes = 0;
    * cub::DeviceSelect::Unique(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_in, d_out, d_num_selected_out, num_items);
    *
    * // Allocate temporary storage
@@ -793,48 +800,48 @@ struct DeviceSelect
    *
    * // Run selection
    * cub::DeviceSelect::Unique(
-   *   d_temp_storage, temp_storage_bytes, 
+   *   d_temp_storage, temp_storage_bytes,
    *   d_in, d_out, d_num_selected_out, num_items);
    *
    * // d_out                 <-- [0, 2, 9, 5, 8]
    * // d_num_selected_out    <-- [5]
    * @endcode
    *
-   * @tparam InputIteratorT       
-   *   **[inferred]** Random-access input iterator type for reading input 
+   * @tparam InputIteratorT
+   *   **[inferred]** Random-access input iterator type for reading input
    *   items \iterator
    *
-   * @tparam OutputIteratorT      
-   *   **[inferred]** Random-access output iterator type for writing selected 
+   * @tparam OutputIteratorT
+   *   **[inferred]** Random-access output iterator type for writing selected
    *   items \iterator
    *
-   * @tparam NumSelectedIteratorT  
-   *   **[inferred]** Output iterator type for recording the number of items 
+   * @tparam NumSelectedIteratorT
+   *   **[inferred]** Output iterator type for recording the number of items
    *   selected \iterator
    *
-   * @param[in] d_temp_storage  
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no work 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no work
    *   is done.
    *
-   * @param[in,out] temp_storage_bytes  
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of `d_temp_storage` allocation
    *
-   * @param[in] d_in  
+   * @param[in] d_in
    *   Pointer to the input sequence of data items
    *
-   * @param[out] d_out  
+   * @param[out] d_out
    *   Pointer to the output sequence of selected data items
    *
-   * @param[out] d_num_selected_out  
-   *   Pointer to the output total number of items selected 
+   * @param[out] d_num_selected_out
+   *   Pointer to the output total number of items selected
    *   (i.e., length of `d_out`)
    *
-   * @param[in] num_items  
+   * @param[in] num_items
    *   Total number of input items (i.e., length of `d_in`)
    *
-   * @param[in] stream  
-   *   **[optional]** CUDA stream to launch kernels within.  
+   * @param[in] stream
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    */
   template <typename InputIteratorT,
@@ -900,16 +907,16 @@ struct DeviceSelect
   }
 
   /**
-   * @brief Given an input sequence `d_keys_in` and `d_values_in` with runs of 
-   *        key-value pairs with consecutive equal-valued keys, only the first 
-   *        key and its value from each run is selectively copied to 
-   *        `d_keys_out` and `d_values_out`. The total number of items selected 
+   * @brief Given an input sequence `d_keys_in` and `d_values_in` with runs of
+   *        key-value pairs with consecutive equal-valued keys, only the first
+   *        key and its value from each run is selectively copied to
+   *        `d_keys_out` and `d_values_out`. The total number of items selected
    *        is written to `d_num_selected_out`. ![](unique_logo.png)
    *
    * @par
-   * - The `==` equality operator is used to determine whether keys are 
+   * - The `==` equality operator is used to determine whether keys are
    *   equivalent
-   * - Copies of the selected items are compacted into `d_out` and maintain 
+   * - Copies of the selected items are compacted into `d_out` and maintain
    *   their original relative ordering.
    * - In-place operations are not supported. There must be no overlap between
    *   any of the provided ranges:
@@ -921,13 +928,13 @@ struct DeviceSelect
    * - @devicestorage
    *
    * @par Snippet
-   * The code snippet below illustrates the compaction of items selected from 
+   * The code snippet below illustrates the compaction of items selected from
    * an `int` device vector.
    * @par
    * @code
    * #include <cub/cub.cuh>   // or equivalently <cub/device/device_select.cuh>
    *
-   * // Declare, allocate, and initialize device-accessible pointers 
+   * // Declare, allocate, and initialize device-accessible pointers
    * // for input and output
    * int  num_items;              // e.g., 8
    * int  *d_keys_in;             // e.g., [0, 2, 2, 9, 5, 5, 5, 8]
@@ -941,8 +948,8 @@ struct DeviceSelect
    * void     *d_temp_storage = NULL;
    * size_t   temp_storage_bytes = 0;
    * cub::DeviceSelect::UniqueByKey(
-   *   d_temp_storage, temp_storage_bytes, 
-   *   d_keys_in, d_values_in, 
+   *   d_temp_storage, temp_storage_bytes,
+   *   d_keys_in, d_values_in,
    *   d_keys_out, d_values_out, d_num_selected_out, num_items);
    *
    * // Allocate temporary storage
@@ -950,8 +957,8 @@ struct DeviceSelect
    *
    * // Run selection
    * cub::DeviceSelect::UniqueByKey(
-   *   d_temp_storage, temp_storage_bytes, 
-   *   d_keys_in, d_values_in, 
+   *   d_temp_storage, temp_storage_bytes,
+   *   d_keys_in, d_values_in,
    *   d_keys_out, d_values_out, d_num_selected_out, num_items);
    *
    * // d_keys_out            <-- [0, 2, 9, 5, 8]
@@ -959,56 +966,56 @@ struct DeviceSelect
    * // d_num_selected_out    <-- [5]
    * @endcode
    *
-   * @tparam KeyInputIteratorT       
-   *   **[inferred]** Random-access input iterator type for reading input 
+   * @tparam KeyInputIteratorT
+   *   **[inferred]** Random-access input iterator type for reading input
    *   keys \iterator
    *
-   * @tparam ValueInputIteratorT     
-   *   **[inferred]** Random-access input iterator type for reading input 
+   * @tparam ValueInputIteratorT
+   *   **[inferred]** Random-access input iterator type for reading input
    *   values \iterator
    *
-   * @tparam KeyOutputIteratorT      
-   *   **[inferred]** Random-access output iterator type for writing selected 
+   * @tparam KeyOutputIteratorT
+   *   **[inferred]** Random-access output iterator type for writing selected
    *   keys \iterator
    *
-   * @tparam ValueOutputIteratorT    
-   *   **[inferred]** Random-access output iterator type for writing selected 
+   * @tparam ValueOutputIteratorT
+   *   **[inferred]** Random-access output iterator type for writing selected
    *   values \iterator
    *
-   * @tparam NumSelectedIteratorT    
-   *   **[inferred]** Output iterator type for recording the number of items 
+   * @tparam NumSelectedIteratorT
+   *   **[inferred]** Output iterator type for recording the number of items
    *   selected \iterator
    *
-   * @param[in] d_temp_storage  
-   *   Device-accessible allocation of temporary storage. When `nullptr`, the 
-   *   required allocation size is written to `temp_storage_bytes` and no work 
+   * @param[in] d_temp_storage
+   *   Device-accessible allocation of temporary storage. When `nullptr`, the
+   *   required allocation size is written to `temp_storage_bytes` and no work
    *   is done.
    *
-   * @param[in,out] temp_storage_bytes  
+   * @param[in,out] temp_storage_bytes
    *   Reference to size in bytes of `d_temp_storage` allocation
    *
-   * @param[in] d_keys_in  
+   * @param[in] d_keys_in
    *   Pointer to the input sequence of keys
    *
-   * @param[in] d_values_in  
+   * @param[in] d_values_in
    *   Pointer to the input sequence of values
    *
-   * @param[out] d_keys_out  
+   * @param[out] d_keys_out
    *   Pointer to the output sequence of selected keys
    *
-   * @param[out] d_values_out  
+   * @param[out] d_values_out
    *   Pointer to the output sequence of selected values
    *
-   * @param[out] d_num_selected_out  
-   *   Pointer to the total number of items selected (i.e., length of 
+   * @param[out] d_num_selected_out
+   *   Pointer to the total number of items selected (i.e., length of
    *   `d_keys_out` or `d_values_out`)
    *
-   * @param[in] num_items  
-   *   Total number of input items (i.e., length of `d_keys_in` or 
+   * @param[in] num_items
+   *   Total number of input items (i.e., length of `d_keys_in` or
    *   `d_values_in`)
    *
-   * @param[in] stream  
-   *   **[optional]** CUDA stream to launch kernels within. 
+   * @param[in] stream
+   *   **[optional]** CUDA stream to launch kernels within.
    *   Default is stream<sub>0</sub>.
    */
   template <typename KeyInputIteratorT,
diff --git a/cub/cub/device/device_spmv.cuh b/cub/cub/device/device_spmv.cuh
index 22dcbe5fcf5..ac6fd123142 100644
--- a/cub/cub/device/device_spmv.cuh
+++ b/cub/cub/device/device_spmv.cuh
@@ -34,6 +34,14 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <stdio.h>
 #include <iterator>
 #include <limits>
diff --git a/cub/cub/device/dispatch/dispatch_adjacent_difference.cuh b/cub/cub/device/dispatch/dispatch_adjacent_difference.cuh
index 6328e7d787b..a7fcc713ea7 100644
--- a/cub/cub/device/dispatch/dispatch_adjacent_difference.cuh
+++ b/cub/cub/device/dispatch/dispatch_adjacent_difference.cuh
@@ -27,8 +27,15 @@
 
 #pragma once
 
+#include "../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/agent/agent_adjacent_difference.cuh>
-#include <cub/config.cuh>
 #include <cub/detail/type_traits.cuh>
 #include <cub/util_debug.cuh>
 #include <cub/util_deprecated.cuh>
@@ -67,10 +74,10 @@ DeviceAdjacentDifferenceDifferenceKernel(InputIteratorT input,
                                          DifferenceOpT difference_op,
                                          OffsetT num_items)
 {
-  using ActivePolicyT = 
+  using ActivePolicyT =
     typename ChainedPolicyT::ActivePolicy::AdjacentDifferencePolicy;
 
-  // It is OK to introspect the return type or parameter types of the 
+  // It is OK to introspect the return type or parameter types of the
   // `operator()` function of `__device__` extended lambda within device code.
   using OutputT = detail::invoke_result_t<DifferenceOpT, InputT, InputT>;
 
@@ -94,7 +101,7 @@ DeviceAdjacentDifferenceDifferenceKernel(InputIteratorT input,
               num_items);
 
   int tile_idx = static_cast<int>(blockIdx.x);
-  OffsetT tile_base  = static_cast<OffsetT>(tile_idx) 
+  OffsetT tile_base  = static_cast<OffsetT>(tile_idx)
                      * ActivePolicyT::ITEMS_PER_TILE;
 
   agent.Process(tile_idx, tile_base);
@@ -313,7 +320,7 @@ struct DispatchAdjacentDifference : public SelectedPolicy
               num_items);
 
       error = CubDebug(detail::DebugSyncStream(stream));
-      
+
       if (cudaSuccess != error)
       {
         break;
diff --git a/cub/cub/device/dispatch/dispatch_batch_memcpy.cuh b/cub/cub/device/dispatch/dispatch_batch_memcpy.cuh
index 04384ae0451..ae9026df3a7 100644
--- a/cub/cub/device/dispatch/dispatch_batch_memcpy.cuh
+++ b/cub/cub/device/dispatch/dispatch_batch_memcpy.cuh
@@ -33,9 +33,16 @@
 
 #pragma once
 
+#include "../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/agent/agent_batch_memcpy.cuh>
 #include <cub/agent/single_pass_scan_operators.cuh>
-#include <cub/config.cuh>
 #include <cub/detail/temporary_storage.cuh>
 #include <cub/thread/thread_search.cuh>
 #include <cub/util_debug.cuh>
diff --git a/cub/cub/device/dispatch/dispatch_histogram.cuh b/cub/cub/device/dispatch/dispatch_histogram.cuh
index 5c9aab47a85..aa6bc4b23e6 100644
--- a/cub/cub/device/dispatch/dispatch_histogram.cuh
+++ b/cub/cub/device/dispatch/dispatch_histogram.cuh
@@ -35,8 +35,15 @@
 
 #pragma once
 
+#include "../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/agent/agent_histogram.cuh>
-#include <cub/config.cuh>
 #include <cub/detail/cpp_compatibility.cuh>
 #include <cub/util_device.cuh>
 #include <cub/util_temporary_storage.cuh>
@@ -567,8 +574,8 @@ struct dispatch_histogram
  * @tparam OffsetT
  *   Signed integer type for global offsets
  *
- * @tparam SelectedPolicy 
- *   Implementation detail, do not specify directly, requirements on the 
+ * @tparam SelectedPolicy
+ *   Implementation detail, do not specify directly, requirements on the
  *   content of this type are subject to breaking change.
  */
 template <int NUM_CHANNELS,
diff --git a/cub/cub/device/dispatch/dispatch_merge_sort.cuh b/cub/cub/device/dispatch/dispatch_merge_sort.cuh
index 79bdfa86775..c3b4f87d711 100644
--- a/cub/cub/device/dispatch/dispatch_merge_sort.cuh
+++ b/cub/cub/device/dispatch/dispatch_merge_sort.cuh
@@ -27,6 +27,14 @@
 
 #pragma once
 
+#include "../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/agent/agent_merge_sort.cuh>
 #include <cub/util_deprecated.cuh>
 #include <cub/util_device.cuh>
diff --git a/cub/cub/device/dispatch/dispatch_radix_sort.cuh b/cub/cub/device/dispatch/dispatch_radix_sort.cuh
index b57ccaace88..aef8df7a8da 100644
--- a/cub/cub/device/dispatch/dispatch_radix_sort.cuh
+++ b/cub/cub/device/dispatch/dispatch_radix_sort.cuh
@@ -33,13 +33,20 @@
 
 #pragma once
 
+#include "../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/agent/agent_radix_sort_downsweep.cuh>
 #include <cub/agent/agent_radix_sort_histogram.cuh>
 #include <cub/agent/agent_radix_sort_onesweep.cuh>
 #include <cub/agent/agent_radix_sort_upsweep.cuh>
 #include <cub/agent/agent_scan.cuh>
 #include <cub/block/block_radix_sort.cuh>
-#include <cub/config.cuh>
 #include <cub/grid/grid_even_share.cuh>
 #include <cub/util_debug.cuh>
 #include <cub/util_deprecated.cuh>
@@ -75,7 +82,7 @@ template <
     bool     IS_DESCENDING,                  ///< Whether or not the sorted-order is high-to-low
     typename KeyT,                           ///< Key type
     typename OffsetT,                        ///< Signed integer type for global offsets
-    typename DecomposerT = detail::identity_decomposer_t>                        
+    typename DecomposerT = detail::identity_decomposer_t>
 __launch_bounds__ (int((ALT_DIGIT_BITS) ?
     int(ChainedPolicyT::ActivePolicy::AltUpsweepPolicy::BLOCK_THREADS) :
     int(ChainedPolicyT::ActivePolicy::UpsweepPolicy::BLOCK_THREADS)))
@@ -316,7 +323,7 @@ CUB_DETAIL_KERNEL_ATTRIBUTES void DeviceRadixSortSingleTileKernel(
     ValueT          values[ITEMS_PER_THREAD];
 
     // Get default (min/max) value for out-of-bounds keys
-    bit_ordered_type default_key_bits = IS_DESCENDING 
+    bit_ordered_type default_key_bits = IS_DESCENDING
                                       ? traits::min_raw_binary_key(decomposer)
                                       : traits::max_raw_binary_key(decomposer);
 
@@ -540,7 +547,7 @@ CUB_DETAIL_KERNEL_ATTRIBUTES void DeviceSegmentedRadixSortKernel(
  * Onesweep kernels
  ******************************************************************************/
 
-/** 
+/**
  * Kernel for computing multiple histograms
  */
 
@@ -552,7 +559,7 @@ template <typename ChainedPolicyT,
           typename KeyT,
           typename OffsetT,
           typename DecomposerT = detail::identity_decomposer_t>
-CUB_DETAIL_KERNEL_ATTRIBUTES __launch_bounds__(ChainedPolicyT::ActivePolicy::HistogramPolicy::BLOCK_THREADS) 
+CUB_DETAIL_KERNEL_ATTRIBUTES __launch_bounds__(ChainedPolicyT::ActivePolicy::HistogramPolicy::BLOCK_THREADS)
 void DeviceRadixSortHistogramKernel(OffsetT *d_bins_out,
                                     const KeyT *d_keys_in,
                                     OffsetT num_items,
@@ -594,7 +601,7 @@ DeviceRadixSortOnesweepKernel
 }
 
 
-/** 
+/**
  * Exclusive sum kernel
  */
 template <
@@ -722,10 +729,10 @@ struct DeviceRadixSortPolicy
 
         // Histogram policy
         typedef AgentRadixSortHistogramPolicy <256, 8, 1, KeyT, ONESWEEP_RADIX_BITS> HistogramPolicy;
-        
+
         // Exclusive sum policy
         typedef AgentRadixSortExclusiveSumPolicy <256, ONESWEEP_RADIX_BITS> ExclusiveSumPolicy;
-        
+
         // Onesweep policy
         typedef AgentRadixSortOnesweepPolicy <256, 21, DominantT, 1,
             RADIX_RANK_MATCH_EARLY_COUNTS_ANY, BLOCK_SCAN_WARP_SCANS, RADIX_SORT_STORE_DIRECT,
@@ -778,10 +785,10 @@ struct DeviceRadixSortPolicy
 
         // Histogram policy
         typedef AgentRadixSortHistogramPolicy <256, 8, 1, KeyT, ONESWEEP_RADIX_BITS> HistogramPolicy;
-        
+
         // Exclusive sum policy
         typedef AgentRadixSortExclusiveSumPolicy <256, ONESWEEP_RADIX_BITS> ExclusiveSumPolicy;
-        
+
         // Onesweep policy
         typedef AgentRadixSortOnesweepPolicy <256, 21, DominantT, 1,
             RADIX_RANK_MATCH_EARLY_COUNTS_ANY, BLOCK_SCAN_WARP_SCANS, RADIX_SORT_STORE_DIRECT,
@@ -821,10 +828,10 @@ struct DeviceRadixSortPolicy
 
         // Histogram policy
         typedef AgentRadixSortHistogramPolicy <256, 8, 8, KeyT, ONESWEEP_RADIX_BITS> HistogramPolicy;
-        
+
         // Exclusive sum policy
         typedef AgentRadixSortExclusiveSumPolicy <256, ONESWEEP_RADIX_BITS> ExclusiveSumPolicy;
-        
+
         // Onesweep policy
         typedef AgentRadixSortOnesweepPolicy <256, OFFSET_64BIT ? 29 : 30, DominantT, 2,
             RADIX_RANK_MATCH_EARLY_COUNTS_ANY, BLOCK_SCAN_WARP_SCANS,
@@ -864,10 +871,10 @@ struct DeviceRadixSortPolicy
 
         // Histogram policy
         typedef AgentRadixSortHistogramPolicy <256, 8, 8, KeyT, ONESWEEP_RADIX_BITS> HistogramPolicy;
-        
+
         // Exclusive sum policy
         typedef AgentRadixSortExclusiveSumPolicy <256, ONESWEEP_RADIX_BITS> ExclusiveSumPolicy;
-        
+
         // Onesweep policy
         typedef AgentRadixSortOnesweepPolicy <256, 30, DominantT, 2,
             RADIX_RANK_MATCH_EARLY_COUNTS_ANY, BLOCK_SCAN_WARP_SCANS,
@@ -905,10 +912,10 @@ struct DeviceRadixSortPolicy
 
         // Histogram policy
         typedef AgentRadixSortHistogramPolicy <256, 8, 8, KeyT, ONESWEEP_RADIX_BITS> HistogramPolicy;
-        
+
         // Exclusive sum policy
         typedef AgentRadixSortExclusiveSumPolicy <256, ONESWEEP_RADIX_BITS> ExclusiveSumPolicy;
-        
+
         // Onesweep policy
         typedef AgentRadixSortOnesweepPolicy <256, 30, DominantT, 2,
             RADIX_RANK_MATCH_EARLY_COUNTS_ANY, BLOCK_SCAN_WARP_SCANS,
@@ -948,10 +955,10 @@ struct DeviceRadixSortPolicy
 
         // Histogram policy
         typedef AgentRadixSortHistogramPolicy <256, 8, 8, KeyT, ONESWEEP_RADIX_BITS> HistogramPolicy;
-        
+
         // Exclusive sum policy
         typedef AgentRadixSortExclusiveSumPolicy <256, ONESWEEP_RADIX_BITS> ExclusiveSumPolicy;
-        
+
         // Onesweep policy
         typedef AgentRadixSortOnesweepPolicy <256,
             sizeof(KeyT) == 4 && sizeof(ValueT) == 4 ? 46 : 23, DominantT, 4,
@@ -993,10 +1000,10 @@ struct DeviceRadixSortPolicy
 
         // Histogram policy
         typedef AgentRadixSortHistogramPolicy <128, 16, 1, KeyT, ONESWEEP_RADIX_BITS> HistogramPolicy;
-        
+
         // Exclusive sum policy
         typedef AgentRadixSortExclusiveSumPolicy <256, ONESWEEP_RADIX_BITS> ExclusiveSumPolicy;
-        
+
         // Onesweep policy
         typedef AgentRadixSortOnesweepPolicy <384,
             OFFSET_64BIT && sizeof(KeyT) == 4 && !KEYS_ONLY ? 17 : 21, DominantT, 1,
@@ -1162,8 +1169,8 @@ struct DeviceRadixSortPolicy
  * @tparam OffsetT
  *   Signed integer type for global offsets
  *
- * @tparam DecomposerT 
- *   Implementation detail, do not specify directly, requirements on the 
+ * @tparam DecomposerT
+ *   Implementation detail, do not specify directly, requirements on the
  *   content of this type are subject to breaking change.
  */
 template <bool IS_DESCENDING,
@@ -1503,19 +1510,19 @@ struct DispatchRadixSort : SelectedPolicy
                 radix_digits            = 1 << radix_bits;
 
                 error = CubDebug(upsweep_config.Init<UpsweepPolicyT>(upsweep_kernel));
-                if (cudaSuccess != error) 
+                if (cudaSuccess != error)
                 {
                     break;
                 }
 
                 error = CubDebug(scan_config.Init<ScanPolicyT>(scan_kernel));
-                if (cudaSuccess != error) 
+                if (cudaSuccess != error)
                 {
                     break;
                 }
 
                 error = CubDebug(downsweep_config.Init<DownsweepPolicyT>(downsweep_kernel));
-                if (cudaSuccess != error) 
+                if (cudaSuccess != error)
                 {
                     break;
                 }
@@ -1629,7 +1636,7 @@ struct DispatchRadixSort : SelectedPolicy
                                                                            histogram_kernel,
                                                                            HISTO_BLOCK_THREADS,
                                                                            0));
-            if (cudaSuccess != error) 
+            if (cudaSuccess != error)
             {
                 break;
             }
@@ -1713,7 +1720,7 @@ struct DispatchRadixSort : SelectedPolicy
                                                0,
                                                num_blocks * RADIX_DIGITS * sizeof(AtomicOffsetT),
                                                stream));
-                    if (cudaSuccess != error) 
+                    if (cudaSuccess != error)
                     {
                         break;
                     }
@@ -1760,7 +1767,7 @@ struct DispatchRadixSort : SelectedPolicy
                 {
                     break;
                 }
-                
+
                 // use the temporary buffers if no overwrite is allowed
                 if (!is_overwrite_okay && pass == 0)
                 {
@@ -1775,7 +1782,7 @@ struct DispatchRadixSort : SelectedPolicy
                 d_values.selector ^= 1;
             }
         } while (0);
-        
+
         return error;
     }
 
@@ -1799,7 +1806,7 @@ struct DispatchRadixSort : SelectedPolicy
             // Get device ordinal
             int device_ordinal;
             error = CubDebug(cudaGetDevice(&device_ordinal));
-            if (cudaSuccess != error) 
+            if (cudaSuccess != error)
             {
                 break;
             }
@@ -1808,7 +1815,7 @@ struct DispatchRadixSort : SelectedPolicy
             int sm_count;
             error = CubDebug(
               cudaDeviceGetAttribute(&sm_count, cudaDevAttrMultiProcessorCount, device_ordinal));
-            if (cudaSuccess != error) 
+            if (cudaSuccess != error)
             {
                 break;
             }
@@ -1899,7 +1906,7 @@ struct DispatchRadixSort : SelectedPolicy
                                   spine_length,
                                   current_bit,
                                   (current_bit < alt_end_bit) ? alt_pass_config : pass_config));
-            if (cudaSuccess != error) 
+            if (cudaSuccess != error)
             {
                 break;
             }
@@ -1917,7 +1924,7 @@ struct DispatchRadixSort : SelectedPolicy
                   current_bit,
                   (current_bit < alt_end_bit) ? alt_pass_config : pass_config));
 
-                if (cudaSuccess != error) 
+                if (cudaSuccess != error)
                 {
                     break;
                 }
@@ -1956,7 +1963,7 @@ struct DispatchRadixSort : SelectedPolicy
             DeviceRadixSortUpsweepKernel<   MaxPolicyT, true,    IS_DESCENDING, KeyT, OffsetT, DecomposerT>,
             RadixSortScanBinsKernel<        MaxPolicyT, OffsetT>,
             DeviceRadixSortDownsweepKernel< MaxPolicyT, false,   IS_DESCENDING, KeyT, ValueT, OffsetT, DecomposerT>,
-            DeviceRadixSortDownsweepKernel< MaxPolicyT, true,    IS_DESCENDING, KeyT, ValueT, OffsetT, DecomposerT>);        
+            DeviceRadixSortDownsweepKernel< MaxPolicyT, true,    IS_DESCENDING, KeyT, ValueT, OffsetT, DecomposerT>);
     }
 
     template <typename ActivePolicyT>
@@ -1977,7 +1984,7 @@ struct DispatchRadixSort : SelectedPolicy
             temp_storage_bytes = 1;
             return cudaSuccess;
         }
-        
+
         // Copy keys
         #ifdef CUB_DETAIL_DEBUG_ENABLE_LOG
         _CubLog("Invoking async copy of %lld keys on stream %lld\n", (long long)num_items,
@@ -2103,7 +2110,7 @@ struct DispatchRadixSort : SelectedPolicy
             int ptx_version = 0;
 
             error = CubDebug(PtxVersion(ptx_version));
-            if (cudaSuccess != error) 
+            if (cudaSuccess != error)
             {
                 break;
             }
@@ -2397,7 +2404,7 @@ struct DispatchSegmentedRadixSort : SelectedPolicy
 
             // Alias the temporary allocations from the single storage blob (or compute the necessary size of the blob)
             error = CubDebug(AliasTemporaries(d_temp_storage, temp_storage_bytes, allocations, allocation_sizes));
-            if (cudaSuccess != error) 
+            if (cudaSuccess != error)
             {
                 break;
             }
@@ -2437,7 +2444,7 @@ struct DispatchSegmentedRadixSort : SelectedPolicy
                                   d_values_remaining_passes.Current(),
                                   current_bit,
                                   (current_bit < alt_end_bit) ? alt_pass_config : pass_config));
-            if (cudaSuccess != error) 
+            if (cudaSuccess != error)
             {
                 break;
             }
@@ -2452,7 +2459,7 @@ struct DispatchSegmentedRadixSort : SelectedPolicy
                   d_values_remaining_passes.d_buffers[d_keys_remaining_passes.selector ^ 1],
                   current_bit,
                   (current_bit < alt_end_bit) ? alt_pass_config : pass_config));
-                if (cudaSuccess != error) 
+                if (cudaSuccess != error)
                 {
                     break;
                 }
@@ -2533,7 +2540,7 @@ struct DispatchSegmentedRadixSort : SelectedPolicy
             int ptx_version = 0;
 
             error = CubDebug(PtxVersion(ptx_version));
-            if (cudaSuccess != error) 
+            if (cudaSuccess != error)
             {
                 break;
             }
@@ -2548,7 +2555,7 @@ struct DispatchSegmentedRadixSort : SelectedPolicy
 
             // Dispatch to chained policy
             error = CubDebug(MaxPolicyT::Invoke(ptx_version, dispatch));
-            if (cudaSuccess != error) 
+            if (cudaSuccess != error)
             {
                 break;
             }
diff --git a/cub/cub/device/dispatch/dispatch_reduce.cuh b/cub/cub/device/dispatch/dispatch_reduce.cuh
index 39a777c8e3d..ca4f1ee5b05 100644
--- a/cub/cub/device/dispatch/dispatch_reduce.cuh
+++ b/cub/cub/device/dispatch/dispatch_reduce.cuh
@@ -34,8 +34,15 @@
 
 #pragma once
 
+#include "../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/agent/agent_reduce.cuh>
-#include <cub/config.cuh>
 #include <cub/grid/grid_even_share.cuh>
 #include <cub/iterator/arg_index_input_iterator.cuh>
 #include <cub/thread/thread_operators.cuh>
diff --git a/cub/cub/device/dispatch/dispatch_reduce_by_key.cuh b/cub/cub/device/dispatch/dispatch_reduce_by_key.cuh
index a758fa60992..3ff8b936821 100644
--- a/cub/cub/device/dispatch/dispatch_reduce_by_key.cuh
+++ b/cub/cub/device/dispatch/dispatch_reduce_by_key.cuh
@@ -33,8 +33,15 @@
 
 #pragma once
 
+#include "../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/agent/agent_reduce_by_key.cuh>
-#include <cub/config.cuh>
 #include <cub/device/dispatch/dispatch_scan.cuh>
 #include <cub/device/dispatch/tuning/tuning_reduce_by_key.cuh>
 #include <cub/grid/grid_queue.cuh>
@@ -205,8 +212,8 @@ __launch_bounds__(int(ChainedPolicyT::ActivePolicy::ReduceByKeyPolicyT::BLOCK_TH
  * @tparam OffsetT
  *   Signed integer type for global offsets
  *
- * @tparam SelectedPolicy 
- *   Implementation detail, do not specify directly, requirements on the 
+ * @tparam SelectedPolicy
+ *   Implementation detail, do not specify directly, requirements on the
  *   content of this type are subject to breaking change.
  */
 template <typename KeysInputIteratorT,
diff --git a/cub/cub/device/dispatch/dispatch_rle.cuh b/cub/cub/device/dispatch/dispatch_rle.cuh
index 18309b44c7d..ca7b7242b80 100644
--- a/cub/cub/device/dispatch/dispatch_rle.cuh
+++ b/cub/cub/device/dispatch/dispatch_rle.cuh
@@ -34,8 +34,15 @@
 
 #pragma once
 
+#include "../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/agent/agent_rle.cuh>
-#include <cub/config.cuh>
 #include <cub/device/dispatch/dispatch_scan.cuh>
 #include <cub/device/dispatch/tuning/tuning_run_length_encode.cuh>
 #include <cub/grid/grid_queue.cuh>
@@ -172,8 +179,8 @@ __launch_bounds__(int(ChainedPolicyT::ActivePolicy::RleSweepPolicyT::BLOCK_THREA
  * @tparam OffsetT
  *   Signed integer type for global offsets
  *
- * @tparam SelectedPolicy 
- *   Implementation detail, do not specify directly, requirements on the 
+ * @tparam SelectedPolicy
+ *   Implementation detail, do not specify directly, requirements on the
  *   content of this type are subject to breaking change.
  */
 template <typename InputIteratorT,
diff --git a/cub/cub/device/dispatch/dispatch_scan.cuh b/cub/cub/device/dispatch/dispatch_scan.cuh
index 6893f24e1dc..7b76c02ba80 100644
--- a/cub/cub/device/dispatch/dispatch_scan.cuh
+++ b/cub/cub/device/dispatch/dispatch_scan.cuh
@@ -14,9 +14,9 @@
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
  *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 
- * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 
- * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  * ARE DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE FOR ANY
  * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
@@ -35,8 +35,15 @@
 
 #pragma once
 
+#include "../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/agent/agent_scan.cuh>
-#include <cub/config.cuh>
 #include <cub/device/dispatch/tuning/tuning_scan.cuh>
 #include <cub/grid/grid_queue.cuh>
 #include <cub/thread/thread_operators.cuh>
@@ -211,7 +218,7 @@ __launch_bounds__(int(ChainedPolicyT::ActivePolicy::ScanPolicyT::BLOCK_THREADS))
  *   Random-access output iterator type for writing scan outputs \iterator
  *
  * @tparam ScanOpT
- *   Binary scan functor type having member 
+ *   Binary scan functor type having member
  *   `auto operator()(const T &a, const U &b)`
  *
  * @tparam InitValueT
@@ -226,9 +233,9 @@ template <typename InputIteratorT,
           typename ScanOpT,
           typename InitValueT,
           typename OffsetT,
-          typename AccumT = 
+          typename AccumT =
             detail::accumulator_t<
-              ScanOpT, 
+              ScanOpT,
               cub::detail::conditional_t<
                 std::is_same<InitValueT, NullType>::value,
                 cub::detail::value_t<InputIteratorT>,
diff --git a/cub/cub/device/dispatch/dispatch_scan_by_key.cuh b/cub/cub/device/dispatch/dispatch_scan_by_key.cuh
index 62df5c6b913..75cb2829bc4 100644
--- a/cub/cub/device/dispatch/dispatch_scan_by_key.cuh
+++ b/cub/cub/device/dispatch/dispatch_scan_by_key.cuh
@@ -12,9 +12,9 @@
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
  *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 
- * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 
- * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  * ARE DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE FOR ANY
  * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
@@ -26,15 +26,22 @@
  ******************************************************************************/
 
 /**
- * @file DeviceScan provides device-wide, parallel operations for computing a 
- *       prefix scan across a sequence of data items residing within 
+ * @file DeviceScan provides device-wide, parallel operations for computing a
+ *       prefix scan across a sequence of data items residing within
  *       device-accessible memory.
  */
 
 #pragma once
 
+#include "../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/agent/agent_scan_by_key.cuh>
-#include <cub/config.cuh>
 #include <cub/device/dispatch/dispatch_scan.cuh>
 #include <cub/device/dispatch/tuning/tuning_scan_by_key.cuh>
 #include <cub/thread/thread_operators.cuh>
@@ -465,7 +472,7 @@ struct DispatchScanByKey : SelectedPolicy
       }
 
       // Sync the stream if specified to flush runtime errors
-      
+
       error = CubDebug(detail::DebugSyncStream(stream));
       if (cudaSuccess != error)
       {
diff --git a/cub/cub/device/dispatch/dispatch_segmented_sort.cuh b/cub/cub/device/dispatch/dispatch_segmented_sort.cuh
index e2d5da09669..e31fcebe321 100644
--- a/cub/cub/device/dispatch/dispatch_segmented_sort.cuh
+++ b/cub/cub/device/dispatch/dispatch_segmented_sort.cuh
@@ -27,6 +27,14 @@
 
 #pragma once
 
+#include "../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/agent/agent_segmented_radix_sort.cuh>
 #include <cub/agent/agent_sub_warp_merge_sort.cuh>
 #include <cub/block/block_load.cuh>
diff --git a/cub/cub/device/dispatch/dispatch_select_if.cuh b/cub/cub/device/dispatch/dispatch_select_if.cuh
index bb9fe685768..32fe0a153a2 100644
--- a/cub/cub/device/dispatch/dispatch_select_if.cuh
+++ b/cub/cub/device/dispatch/dispatch_select_if.cuh
@@ -28,14 +28,21 @@
 
 /**
  * @file
- *   cub::DeviceSelect provides device-wide, parallel operations for selecting items from sequences 
+ *   cub::DeviceSelect provides device-wide, parallel operations for selecting items from sequences
  *   of data items residing within device-accessible memory.
  */
 
 #pragma once
 
+#include "../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/agent/agent_select_if.cuh>
-#include <cub/config.cuh>
 #include <cub/device/dispatch/dispatch_scan.cuh>
 #include <cub/device/dispatch/tuning/tuning_select_if.cuh>
 #include <cub/grid/grid_queue.cuh>
@@ -64,39 +71,39 @@ CUB_NAMESPACE_BEGIN
  * Otherwise performs flag-based selection if FlagsInputIterator's value type != NullType
  * Otherwise performs discontinuity selection (keep unique)
  *
- * @tparam InputIteratorT 
+ * @tparam InputIteratorT
  *   Random-access input iterator type for reading input items
  *
- * @tparam FlagsInputIteratorT 
- *   Random-access input iterator type for reading selection flags (NullType* if a selection functor 
+ * @tparam FlagsInputIteratorT
+ *   Random-access input iterator type for reading selection flags (NullType* if a selection functor
  *   or discontinuity flagging is to be used for selection)
  *
- * @tparam SelectedOutputIteratorT 
+ * @tparam SelectedOutputIteratorT
  *   Random-access output iterator type for writing selected items
  *
- * @tparam NumSelectedIteratorT 
+ * @tparam NumSelectedIteratorT
  *   Output iterator type for recording the number of items selected
  *
- * @tparam ScanTileStateT 
+ * @tparam ScanTileStateT
  *   Tile status interface type
  *
- * @tparam SelectOpT 
- *   Selection operator type (NullType if selection flags or discontinuity flagging is 
+ * @tparam SelectOpT
+ *   Selection operator type (NullType if selection flags or discontinuity flagging is
  *   to be used for selection)
  *
- * @tparam EqualityOpT 
- *   Equality operator type (NullType if selection functor or selection flags is 
+ * @tparam EqualityOpT
+ *   Equality operator type (NullType if selection functor or selection flags is
  *   to be used for selection)
  *
- * @tparam OffsetT 
+ * @tparam OffsetT
  *   Signed integer type for global offsets
  *
- * @tparam KEEP_REJECTS 
+ * @tparam KEEP_REJECTS
  *   Whether or not we push rejected items to the back of the output
  *
  * @param[in] d_in
  *   Pointer to the input sequence of data items
- * 
+ *
  * @param[in] d_flags
  *   Pointer to the input sequence of selection flags (if applicable)
  *
@@ -106,18 +113,18 @@ CUB_NAMESPACE_BEGIN
  * @param[out] d_num_selected_out
  *   Pointer to the total number of items selected (i.e., length of \p d_selected_out)
  *
- * @param[in] tile_status 
+ * @param[in] tile_status
  *   Tile status interface
  *
  * @param[in] select_op
  *   Selection operator
- * 
+ *
  * @param[in] equality_op
  *   Equality operator
- * 
+ *
  * @param[in] num_items
  *   Total number of input items (i.e., length of \p d_in)
- * 
+ *
  * @param[in] num_tiles
  *   Total number of tiles for the entire problem
  */
@@ -176,7 +183,7 @@ __launch_bounds__(int(ChainedPolicyT::ActivePolicy::SelectIfPolicyT::BLOCK_THREA
  *   Random-access input iterator type for reading input items
  *
  * @tparam FlagsInputIteratorT
- *   Random-access input iterator type for reading selection flags 
+ *   Random-access input iterator type for reading selection flags
  *   (NullType* if a selection functor or discontinuity flagging is to be used for selection)
  *
  * @tparam SelectedOutputIteratorT
@@ -186,11 +193,11 @@ __launch_bounds__(int(ChainedPolicyT::ActivePolicy::SelectIfPolicyT::BLOCK_THREA
  *   Output iterator type for recording the number of items selected
  *
  * @tparam SelectOpT
- *   Selection operator type (NullType if selection flags or discontinuity flagging is 
+ *   Selection operator type (NullType if selection flags or discontinuity flagging is
  *   to be used for selection)
  *
  * @tparam EqualityOpT
- *   Equality operator type (NullType if selection functor or selection flags is to 
+ *   Equality operator type (NullType if selection functor or selection flags is to
  *   be used for selection)
  *
  * @tparam OffsetT
@@ -225,13 +232,13 @@ struct DispatchSelectIf : SelectedPolicy
 
     static constexpr int INIT_KERNEL_THREADS = 128;
 
-    /// Device-accessible allocation of temporary storage. 
-    /// When `nullptr`, the required allocation size is written to `temp_storage_bytes` 
+    /// Device-accessible allocation of temporary storage.
+    /// When `nullptr`, the required allocation size is written to `temp_storage_bytes`
     /// and no work is done.
     void* d_temp_storage;
 
     /// Reference to size in bytes of `d_temp_storage` allocation
-    size_t& temp_storage_bytes; 
+    size_t& temp_storage_bytes;
 
     /// Pointer to the input sequence of data items
     InputIteratorT d_in;
@@ -261,11 +268,11 @@ struct DispatchSelectIf : SelectedPolicy
 
     /**
      * @param d_temp_storage
-     *   Device-accessible allocation of temporary storage. 
-     *   When `nullptr`, the required allocation size is written to `temp_storage_bytes` 
+     *   Device-accessible allocation of temporary storage.
+     *   When `nullptr`, the required allocation size is written to `temp_storage_bytes`
      *   and no work is done.
-     * 
-     * @param temp_storage_bytes 
+     *
+     * @param temp_storage_bytes
      *   Reference to size in bytes of `d_temp_storage` allocation
      *
      * @param d_in
@@ -339,7 +346,7 @@ struct DispatchSelectIf : SelectedPolicy
             // Get device ordinal
             int device_ordinal;
             error = CubDebug(cudaGetDevice(&device_ordinal));
-            if (cudaSuccess != error) 
+            if (cudaSuccess != error)
             {
                 break;
             }
@@ -352,7 +359,7 @@ struct DispatchSelectIf : SelectedPolicy
 
             // bytes needed for tile status descriptors
             error = CubDebug(ScanTileStateT::AllocationSize(num_tiles, allocation_sizes[0]));
-            if (cudaSuccess != error) 
+            if (cudaSuccess != error)
             {
                 break;
             }
@@ -361,7 +368,7 @@ struct DispatchSelectIf : SelectedPolicy
             void* allocations[1] = {};
 
             error = CubDebug(AliasTemporaries(d_temp_storage, temp_storage_bytes, allocations, allocation_sizes));
-            if (cudaSuccess != error) 
+            if (cudaSuccess != error)
             {
                 break;
             }
@@ -375,7 +382,7 @@ struct DispatchSelectIf : SelectedPolicy
             // Construct the tile status interface
             ScanTileStateT tile_status;
             error = CubDebug(tile_status.Init(num_tiles, allocations[0], allocation_sizes[0]));
-            if (cudaSuccess != error) 
+            if (cudaSuccess != error)
             {
                 break;
             }
@@ -418,7 +425,7 @@ struct DispatchSelectIf : SelectedPolicy
             // Get max x-dimension of grid
             int max_dim_x;
             error = CubDebug(cudaDeviceGetAttribute(&max_dim_x, cudaDevAttrMaxGridDimX, device_ordinal));
-            if (cudaSuccess != error) 
+            if (cudaSuccess != error)
             {
                 break;
             }
@@ -509,11 +516,11 @@ struct DispatchSelectIf : SelectedPolicy
      * Internal dispatch routine
      *
      * @param d_temp_storage
-     *   Device-accessible allocation of temporary storage. 
-     *   When `nullptr`, the required allocation size is written to `temp_storage_bytes` 
+     *   Device-accessible allocation of temporary storage.
+     *   When `nullptr`, the required allocation size is written to `temp_storage_bytes`
      *   and no work is done.
-     * 
-     * @param temp_storage_bytes 
+     *
+     * @param temp_storage_bytes
      *   Reference to size in bytes of `d_temp_storage` allocation
      *
      * @param d_in
@@ -555,7 +562,7 @@ struct DispatchSelectIf : SelectedPolicy
         using MaxPolicyT = typename SelectedPolicy::MaxPolicy;
 
         int ptx_version = 0;
-        if (cudaError_t error = CubDebug(PtxVersion(ptx_version))) 
+        if (cudaError_t error = CubDebug(PtxVersion(ptx_version)))
         {
             return error;
         }
@@ -571,23 +578,23 @@ struct DispatchSelectIf : SelectedPolicy
                                     num_items,
                                     stream,
                                     ptx_version);
-        
+
         return CubDebug(MaxPolicyT::Invoke(ptx_version, dispatch));
     }
 
     CUB_DETAIL_RUNTIME_DEBUG_SYNC_IS_NOT_SUPPORTED
     CUB_RUNTIME_FUNCTION __forceinline__
     static cudaError_t Dispatch(
-        void*                       d_temp_storage,          
-        size_t&                     temp_storage_bytes,       
-        InputIteratorT              d_in,                      
-        FlagsInputIteratorT         d_flags,                    
-        SelectedOutputIteratorT     d_selected_out,              
-        NumSelectedIteratorT        d_num_selected_out,           
-        SelectOpT                   select_op,                     
-        EqualityOpT                 equality_op,                    
-        OffsetT                     num_items,            
-        cudaStream_t                stream,                
+        void*                       d_temp_storage,
+        size_t&                     temp_storage_bytes,
+        InputIteratorT              d_in,
+        FlagsInputIteratorT         d_flags,
+        SelectedOutputIteratorT     d_selected_out,
+        NumSelectedIteratorT        d_num_selected_out,
+        SelectOpT                   select_op,
+        EqualityOpT                 equality_op,
+        OffsetT                     num_items,
+        cudaStream_t                stream,
         bool                        debug_synchronous)
     {
       CUB_DETAIL_RUNTIME_DEBUG_SYNC_USAGE_LOG
diff --git a/cub/cub/device/dispatch/dispatch_spmv_orig.cuh b/cub/cub/device/dispatch/dispatch_spmv_orig.cuh
index 8f2a68f967c..fda0d518d58 100644
--- a/cub/cub/device/dispatch/dispatch_spmv_orig.cuh
+++ b/cub/cub/device/dispatch/dispatch_spmv_orig.cuh
@@ -34,10 +34,17 @@
 
 #pragma once
 
+#include "../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/agent/agent_segment_fixup.cuh>
 #include <cub/agent/agent_spmv_orig.cuh>
 #include <cub/agent/single_pass_scan_operators.cuh>
-#include <cub/config.cuh>
 #include <cub/detail/cpp_compatibility.cuh>
 #include <cub/grid/grid_queue.cuh>
 #include <cub/thread/thread_search.cuh>
@@ -196,9 +203,9 @@ DeviceSpmvEmptyMatrixKernel(SpmvParams<ValueT, OffsetT> spmv_params)
     {
         ValueT result = 0.0;
 
-        CUB_IF_CONSTEXPR(HAS_BETA) 
+        CUB_IF_CONSTEXPR(HAS_BETA)
         {
-            result += spmv_params.beta * spmv_params.d_vector_y[row]; 
+            result += spmv_params.beta * spmv_params.d_vector_y[row];
         }
 
         spmv_params.d_vector_y[row] = result;
@@ -838,7 +845,7 @@ struct DispatchSpmv
             constexpr bool has_beta = false;
 
             if (CubDebug(error = Dispatch(
-                d_temp_storage, temp_storage_bytes, spmv_params, stream, 
+                d_temp_storage, temp_storage_bytes, spmv_params, stream,
                 DeviceSpmv1ColKernel<PtxSpmvPolicyT, ValueT, OffsetT>,
                 DeviceSpmvSearchKernel<PtxSpmvPolicyT, OffsetT, CoordinateT, SpmvParamsT>,
                 DeviceSpmvKernel<PtxSpmvPolicyT, ScanTileStateT, ValueT, OffsetT, CoordinateT, has_alpha, has_beta>,
diff --git a/cub/cub/device/dispatch/dispatch_three_way_partition.cuh b/cub/cub/device/dispatch/dispatch_three_way_partition.cuh
index 8b1849fa24c..083c84f6274 100644
--- a/cub/cub/device/dispatch/dispatch_three_way_partition.cuh
+++ b/cub/cub/device/dispatch/dispatch_three_way_partition.cuh
@@ -27,8 +27,15 @@
 
 #pragma once
 
+#include "../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/agent/agent_three_way_partition.cuh>
-#include <cub/config.cuh>
 #include <cub/device/dispatch/dispatch_scan.cuh>
 #include <cub/device/dispatch/tuning/tuning_three_way_partition.cuh>
 #include <cub/thread/thread_operators.cuh>
diff --git a/cub/cub/device/dispatch/dispatch_unique_by_key.cuh b/cub/cub/device/dispatch/dispatch_unique_by_key.cuh
index c924e71ef7e..36ab55be61c 100644
--- a/cub/cub/device/dispatch/dispatch_unique_by_key.cuh
+++ b/cub/cub/device/dispatch/dispatch_unique_by_key.cuh
@@ -34,8 +34,8 @@
 #include <cub/agent/agent_unique_by_key.cuh>
 #include <cub/device/dispatch/dispatch_scan.cuh>
 #include <cub/device/dispatch/tuning/tuning_unique_by_key.cuh>
+#include <cub/config.cuh>
 #include <cub/util_deprecated.cuh>
-#include <cub/util_macro.cuh>
 #include <cub/util_math.cuh>
 
 #include <iterator>
diff --git a/cub/cub/device/dispatch/tuning/tuning_histogram.cuh b/cub/cub/device/dispatch/tuning/tuning_histogram.cuh
index 2b595e91e79..eca09e9b4cf 100644
--- a/cub/cub/device/dispatch/tuning/tuning_histogram.cuh
+++ b/cub/cub/device/dispatch/tuning/tuning_histogram.cuh
@@ -27,9 +27,16 @@
 
 #pragma once
 
+#include "../../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/agent/agent_histogram.cuh>
 #include <cub/block/block_load.cuh>
-#include <cub/config.cuh>
 #include <cub/util_device.cuh>
 #include <cub/util_type.cuh>
 
diff --git a/cub/cub/device/dispatch/tuning/tuning_reduce_by_key.cuh b/cub/cub/device/dispatch/tuning/tuning_reduce_by_key.cuh
index e06ddb5019e..38900bdee60 100644
--- a/cub/cub/device/dispatch/tuning/tuning_reduce_by_key.cuh
+++ b/cub/cub/device/dispatch/tuning/tuning_reduce_by_key.cuh
@@ -27,12 +27,19 @@
 
 #pragma once
 
+#include "../../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/agent/agent_reduce_by_key.cuh>
 #include <cub/agent/single_pass_scan_operators.cuh>
 #include <cub/block/block_load.cuh>
 #include <cub/block/block_scan.cuh>
 #include <cub/block/block_store.cuh>
-#include <cub/config.cuh>
 #include <cub/util_device.cuh>
 #include <cub/util_type.cuh>
 
diff --git a/cub/cub/device/dispatch/tuning/tuning_run_length_encode.cuh b/cub/cub/device/dispatch/tuning/tuning_run_length_encode.cuh
index d7ad21c808f..c057d31c60c 100644
--- a/cub/cub/device/dispatch/tuning/tuning_run_length_encode.cuh
+++ b/cub/cub/device/dispatch/tuning/tuning_run_length_encode.cuh
@@ -27,13 +27,20 @@
 
 #pragma once
 
+#include "../../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/agent/agent_reduce_by_key.cuh>
 #include <cub/agent/agent_rle.cuh>
 #include <cub/agent/single_pass_scan_operators.cuh>
 #include <cub/block/block_load.cuh>
 #include <cub/block/block_scan.cuh>
 #include <cub/block/block_store.cuh>
-#include <cub/config.cuh>
 #include <cub/util_device.cuh>
 #include <cub/util_type.cuh>
 
@@ -157,7 +164,7 @@ struct sm90_tuning<LengthT, KeyT, primitive_length::yes, primitive_key::yes, len
   using delay_constructor = detail::no_delay_constructor_t<515>;
 };
 
-#if CUB_IS_INT128_ENABLED 
+#if CUB_IS_INT128_ENABLED
 template <class LengthT>
 struct sm90_tuning<LengthT, __int128_t, primitive_length::yes, primitive_key::no, length_size::_4, key_size::_16>
 {
@@ -259,7 +266,7 @@ struct sm80_tuning<LengthT, KeyT, primitive_length::yes, primitive_key::yes, len
   using delay_constructor = detail::no_delay_constructor_t<1075>;
 };
 
-#if CUB_IS_INT128_ENABLED 
+#if CUB_IS_INT128_ENABLED
 template <class LengthT>
 struct sm80_tuning<LengthT, __int128_t, primitive_length::yes, primitive_key::no, length_size::_4, key_size::_16>
 {
@@ -287,7 +294,7 @@ struct sm80_tuning<LengthT, __uint128_t, primitive_length::yes, primitive_key::n
 
 } // namespace encode
 
-namespace non_trivial_runs 
+namespace non_trivial_runs
 {
 
 template <class LengthT,
@@ -304,7 +311,7 @@ struct sm90_tuning
 
   static constexpr int items = CUB_MIN(nominal_4b_items_per_thread,
                                  CUB_MAX(1, (nominal_4b_items_per_thread * 4 / sizeof(KeyT))));
-  
+
   static constexpr BlockLoadAlgorithm load_algorithm = BLOCK_LOAD_WARP_TRANSPOSE;
 
   static constexpr bool store_with_time_slicing = true;
@@ -368,7 +375,7 @@ struct sm90_tuning<LengthT, KeyT, primitive_length::yes, primitive_key::yes, len
   using delay_constructor = detail::no_delay_constructor_t<840>;
 };
 
-#if CUB_IS_INT128_ENABLED 
+#if CUB_IS_INT128_ENABLED
 template <class LengthT>
 struct sm90_tuning<LengthT, __int128_t, primitive_length::yes, primitive_key::no, length_size::_4, key_size::_16>
 {
@@ -412,7 +419,7 @@ struct sm80_tuning
 
   static constexpr int items = CUB_MIN(nominal_4b_items_per_thread,
                                  CUB_MAX(1, (nominal_4b_items_per_thread * 4 / sizeof(KeyT))));
-  
+
   static constexpr BlockLoadAlgorithm load_algorithm = BLOCK_LOAD_WARP_TRANSPOSE;
 
   static constexpr bool store_with_time_slicing = true;
@@ -476,7 +483,7 @@ struct sm80_tuning<LengthT, KeyT, primitive_length::yes, primitive_key::yes, len
   using delay_constructor = detail::no_delay_constructor_t<1065>;
 };
 
-#if CUB_IS_INT128_ENABLED 
+#if CUB_IS_INT128_ENABLED
 template <class LengthT>
 struct sm80_tuning<LengthT, __int128_t, primitive_length::yes, primitive_key::no, length_size::_4, key_size::_16>
 {
diff --git a/cub/cub/device/dispatch/tuning/tuning_scan.cuh b/cub/cub/device/dispatch/tuning/tuning_scan.cuh
index eff4f29fcd2..dae9ce4c9c2 100644
--- a/cub/cub/device/dispatch/tuning/tuning_scan.cuh
+++ b/cub/cub/device/dispatch/tuning/tuning_scan.cuh
@@ -27,12 +27,19 @@
 
 #pragma once
 
+#include "../../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/agent/agent_scan.cuh>
 #include <cub/agent/single_pass_scan_operators.cuh>
 #include <cub/block/block_load.cuh>
 #include <cub/block/block_scan.cuh>
 #include <cub/block/block_store.cuh>
-#include <cub/config.cuh>
 #include <cub/util_device.cuh>
 #include <cub/util_type.cuh>
 
@@ -102,7 +109,7 @@ template <class T> struct sm90_tuning<T, primitive_op::yes, primitive_accum::yes
 template <> struct sm90_tuning<float,  primitive_op::yes, primitive_accum::yes, accum_size::_4> : tuning<128, 24, 688, 1140> {};
 template <> struct sm90_tuning<double, primitive_op::yes, primitive_accum::yes, accum_size::_8> : tuning<224, 24, 576, 1215> {};
 
-#if CUB_IS_INT128_ENABLED 
+#if CUB_IS_INT128_ENABLED
 template <> struct sm90_tuning< __int128_t, primitive_op::yes, primitive_accum::no, accum_size::_16> : tuning<576, 21, 860, 630> {};
 template <> struct sm90_tuning<__uint128_t, primitive_op::yes, primitive_accum::no, accum_size::_16> : tuning<576, 21, 860, 630> {};
 #endif
@@ -229,7 +236,7 @@ struct sm80_tuning<__uint128_t, primitive_op::yes, primitive_accum::no, accum_si
 } // namespace detail
 
 
-template <typename AccumT, typename ScanOpT = Sum> 
+template <typename AccumT, typename ScanOpT = Sum>
 struct DeviceScanPolicy
 {
   // For large values, use timesliced loads/stores to fit shared memory.
diff --git a/cub/cub/device/dispatch/tuning/tuning_scan_by_key.cuh b/cub/cub/device/dispatch/tuning/tuning_scan_by_key.cuh
index 92e4931fe81..b6756c92a3d 100644
--- a/cub/cub/device/dispatch/tuning/tuning_scan_by_key.cuh
+++ b/cub/cub/device/dispatch/tuning/tuning_scan_by_key.cuh
@@ -27,12 +27,19 @@
 
 #pragma once
 
+#include "../../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/agent/agent_scan_by_key.cuh>
 #include <cub/agent/single_pass_scan_operators.cuh>
 #include <cub/block/block_load.cuh>
 #include <cub/block/block_scan.cuh>
 #include <cub/block/block_store.cuh>
-#include <cub/config.cuh>
 #include <cub/thread/thread_operators.cuh>
 #include <cub/util_device.cuh>
 #include <cub/util_math.cuh>
@@ -169,7 +176,7 @@ struct sm90_tuning<KeyT, ValueT, primitive_op::yes, key_size::_1, val_size::_8,
   using delay_constructor = detail::fixed_delay_constructor_t<488, 1070>;
 };
 
-#if CUB_IS_INT128_ENABLED 
+#if CUB_IS_INT128_ENABLED
 template <class KeyT>
 struct sm90_tuning<KeyT, __int128_t, primitive_op::yes, key_size::_1, val_size::_16, primitive_accum::no>
 {
@@ -255,7 +262,7 @@ struct sm90_tuning<KeyT, ValueT, primitive_op::yes, key_size::_2, val_size::_8,
   using delay_constructor = detail::fixed_delay_constructor_t<352, 1170>;
 };
 
-#if CUB_IS_INT128_ENABLED 
+#if CUB_IS_INT128_ENABLED
 template <class KeyT>
 struct sm90_tuning<KeyT, __int128_t, primitive_op::yes, key_size::_2, val_size::_16, primitive_accum::no>
 {
@@ -341,7 +348,7 @@ struct sm90_tuning<KeyT, ValueT, primitive_op::yes, key_size::_4, val_size::_8,
   using delay_constructor = detail::fixed_delay_constructor_t<556, 1195>;
 };
 
-#if CUB_IS_INT128_ENABLED 
+#if CUB_IS_INT128_ENABLED
 template <class KeyT>
 struct sm90_tuning<KeyT, __int128_t, primitive_op::yes, key_size::_4, val_size::_16, primitive_accum::no>
 {
@@ -427,7 +434,7 @@ struct sm90_tuning<KeyT, ValueT, primitive_op::yes, key_size::_8, val_size::_8,
   using delay_constructor = detail::fixed_delay_constructor_t<600, 930>;
 };
 
-#if CUB_IS_INT128_ENABLED 
+#if CUB_IS_INT128_ENABLED
 template <class KeyT>
 struct sm90_tuning<KeyT, __int128_t, primitive_op::yes, key_size::_8, val_size::_16, primitive_accum::no>
 {
@@ -513,7 +520,7 @@ struct sm90_tuning<KeyT, ValueT, primitive_op::yes, key_size::_16, val_size::_8,
   using delay_constructor = detail::fixed_delay_constructor_t<320, 1200>;
 };
 
-#if CUB_IS_INT128_ENABLED 
+#if CUB_IS_INT128_ENABLED
 template <class KeyT>
 struct sm90_tuning<KeyT, __int128_t, primitive_op::yes, key_size::_16, val_size::_16, primitive_accum::no>
 {
@@ -627,7 +634,7 @@ struct sm80_tuning<KeyT, ValueT, primitive_op::yes, key_size::_1, val_size::_8,
   using delay_constructor = detail::fixed_delay_constructor_t<124, 1040>;
 };
 
-#if CUB_IS_INT128_ENABLED 
+#if CUB_IS_INT128_ENABLED
 template <class KeyT>
 struct sm80_tuning<KeyT, __int128_t, primitive_op::yes, key_size::_1, val_size::_16, primitive_accum::no>
 {
@@ -713,7 +720,7 @@ struct sm80_tuning<KeyT, ValueT, primitive_op::yes, key_size::_2, val_size::_8,
   using delay_constructor = detail::fixed_delay_constructor_t<160, 695>;
 };
 
-#if CUB_IS_INT128_ENABLED 
+#if CUB_IS_INT128_ENABLED
 template <class KeyT>
 struct sm80_tuning<KeyT, __int128_t, primitive_op::yes, key_size::_2, val_size::_16, primitive_accum::no>
 {
@@ -799,7 +806,7 @@ struct sm80_tuning<KeyT, ValueT, primitive_op::yes, key_size::_4, val_size::_8,
   using delay_constructor = detail::fixed_delay_constructor_t<888, 635>;
 };
 
-#if CUB_IS_INT128_ENABLED 
+#if CUB_IS_INT128_ENABLED
 template <class KeyT>
 struct sm80_tuning<KeyT, __int128_t, primitive_op::yes, key_size::_4, val_size::_16, primitive_accum::no>
 {
@@ -885,7 +892,7 @@ struct sm80_tuning<KeyT, ValueT, primitive_op::yes, key_size::_8, val_size::_8,
   using delay_constructor = detail::no_delay_constructor_t<1160>;
 };
 
-#if CUB_IS_INT128_ENABLED 
+#if CUB_IS_INT128_ENABLED
 template <class KeyT>
 struct sm80_tuning<KeyT, __int128_t, primitive_op::yes, key_size::_8, val_size::_16, primitive_accum::no>
 {
@@ -971,7 +978,7 @@ struct sm80_tuning<KeyT, ValueT, primitive_op::yes, key_size::_16, val_size::_8,
   using delay_constructor = detail::no_delay_constructor_t<1030>;
 };
 
-#if CUB_IS_INT128_ENABLED 
+#if CUB_IS_INT128_ENABLED
 template <class KeyT>
 struct sm80_tuning<KeyT, __int128_t, primitive_op::yes, key_size::_16, val_size::_16, primitive_accum::no>
 {
diff --git a/cub/cub/device/dispatch/tuning/tuning_select_if.cuh b/cub/cub/device/dispatch/tuning/tuning_select_if.cuh
index 41e5b34f5ae..865c376b739 100644
--- a/cub/cub/device/dispatch/tuning/tuning_select_if.cuh
+++ b/cub/cub/device/dispatch/tuning/tuning_select_if.cuh
@@ -27,20 +27,27 @@
 
 #pragma once
 
+#include "../../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/agent/agent_select_if.cuh>
 #include <cub/agent/single_pass_scan_operators.cuh>
 #include <cub/block/block_load.cuh>
 #include <cub/block/block_scan.cuh>
-#include <cub/config.cuh>
 #include <cub/util_device.cuh>
 #include <cub/util_type.cuh>
 
 CUB_NAMESPACE_BEGIN
 
-namespace detail 
+namespace detail
 {
 
-namespace select 
+namespace select
 {
 
 enum class flagged { no, yes };
@@ -152,7 +159,7 @@ struct sm90_tuning<Input, flagged::no, keep_rejects::no, offset_size::_4, primit
     using delay_constructor = detail::fixed_delay_constructor_t<380, 1140>;
 };
 
-#if CUB_IS_INT128_ENABLED 
+#if CUB_IS_INT128_ENABLED
 template <>
 struct sm90_tuning<__int128_t, flagged::no, keep_rejects::no, offset_size::_4, primitive::no, input_size::_16>
 {
@@ -221,7 +228,7 @@ struct sm90_tuning<Input, flagged::yes, keep_rejects::no, offset_size::_4, primi
     using delay_constructor = detail::fixed_delay_constructor_t<360, 1170>;
 };
 
-#if CUB_IS_INT128_ENABLED 
+#if CUB_IS_INT128_ENABLED
 template <>
 struct sm90_tuning<__int128_t, flagged::yes, keep_rejects::no, offset_size::_4, primitive::no, input_size::_16>
 {
@@ -290,7 +297,7 @@ struct sm90_tuning<Input, flagged::no, keep_rejects::yes, offset_size::_4, primi
     using delay_constructor = detail::fixed_delay_constructor_t<512, 1075>;
 };
 
-#if CUB_IS_INT128_ENABLED 
+#if CUB_IS_INT128_ENABLED
 template <>
 struct sm90_tuning<__int128_t, flagged::no, keep_rejects::yes, offset_size::_4, primitive::no, input_size::_16>
 {
@@ -359,7 +366,7 @@ struct sm90_tuning<Input, flagged::yes, keep_rejects::yes, offset_size::_4, prim
     using delay_constructor = detail::fixed_delay_constructor_t<532, 1180>;
 };
 
-#if CUB_IS_INT128_ENABLED 
+#if CUB_IS_INT128_ENABLED
 template <>
 struct sm90_tuning<__int128_t, flagged::yes, keep_rejects::yes, offset_size::_4, primitive::no, input_size::_16>
 {
@@ -450,7 +457,7 @@ struct sm80_tuning<Input, flagged::no, keep_rejects::no, offset_size::_4, primit
     using delay_constructor = detail::fixed_delay_constructor_t<832, 1165>;
 };
 
-#if CUB_IS_INT128_ENABLED 
+#if CUB_IS_INT128_ENABLED
 template <>
 struct sm80_tuning<__int128_t, flagged::no, keep_rejects::no, offset_size::_4, primitive::no, input_size::_16>
 {
@@ -519,7 +526,7 @@ struct sm80_tuning<Input, flagged::yes, keep_rejects::no, offset_size::_4, primi
     using delay_constructor = detail::no_delay_constructor_t<1130>;
 };
 
-#if CUB_IS_INT128_ENABLED 
+#if CUB_IS_INT128_ENABLED
 template <>
 struct sm80_tuning<__int128_t, flagged::yes, keep_rejects::no, offset_size::_4, primitive::no, input_size::_16>
 {
@@ -588,7 +595,7 @@ struct sm80_tuning<Input, flagged::no, keep_rejects::yes, offset_size::_4, primi
     using delay_constructor = detail::fixed_delay_constructor_t<68, 1160>;
 };
 
-#if CUB_IS_INT128_ENABLED 
+#if CUB_IS_INT128_ENABLED
 template <>
 struct sm80_tuning<__int128_t, flagged::no, keep_rejects::yes, offset_size::_4, primitive::no, input_size::_16>
 {
@@ -657,7 +664,7 @@ struct sm80_tuning<Input, flagged::yes, keep_rejects::yes, offset_size::_4, prim
     using delay_constructor = detail::fixed_delay_constructor_t<884, 1130>;
 };
 
-#if CUB_IS_INT128_ENABLED 
+#if CUB_IS_INT128_ENABLED
 template <>
 struct sm80_tuning<__int128_t, flagged::yes, keep_rejects::yes, offset_size::_4, primitive::no, input_size::_16>
 {
diff --git a/cub/cub/device/dispatch/tuning/tuning_three_way_partition.cuh b/cub/cub/device/dispatch/tuning/tuning_three_way_partition.cuh
index c4b09d8495f..0144c9c8f9d 100644
--- a/cub/cub/device/dispatch/tuning/tuning_three_way_partition.cuh
+++ b/cub/cub/device/dispatch/tuning/tuning_three_way_partition.cuh
@@ -27,11 +27,18 @@
 
 #pragma once
 
+#include "../../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/agent/agent_three_way_partition.cuh>
 #include <cub/agent/single_pass_scan_operators.cuh>
 #include <cub/block/block_load.cuh>
 #include <cub/block/block_scan.cuh>
-#include <cub/config.cuh>
 #include <cub/util_device.cuh>
 #include <cub/util_math.cuh>
 #include <cub/util_type.cuh>
diff --git a/cub/cub/device/dispatch/tuning/tuning_unique_by_key.cuh b/cub/cub/device/dispatch/tuning/tuning_unique_by_key.cuh
index d16f770965b..32c9b05c3bd 100644
--- a/cub/cub/device/dispatch/tuning/tuning_unique_by_key.cuh
+++ b/cub/cub/device/dispatch/tuning/tuning_unique_by_key.cuh
@@ -27,11 +27,18 @@
 
 #pragma once
 
+#include "../../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/agent/agent_unique_by_key.cuh>
 #include <cub/agent/single_pass_scan_operators.cuh>
 #include <cub/block/block_load.cuh>
 #include <cub/block/block_scan.cuh>
-#include <cub/config.cuh>
 #include <cub/util_device.cuh>
 #include <cub/util_math.cuh>
 #include <cub/util_type.cuh>
@@ -716,7 +723,7 @@ struct DeviceUniqueByKeyPolicy
                                                       detail::default_delay_constructor_t<int>>;
   };
 
-  struct DefaultTuning 
+  struct DefaultTuning
   {
     static constexpr int INPUT_SIZE = sizeof(KeyT);
     enum
diff --git a/cub/cub/grid/grid_barrier.cuh b/cub/cub/grid/grid_barrier.cuh
index 063a2c395cd..516703cdbdb 100644
--- a/cub/cub/grid/grid_barrier.cuh
+++ b/cub/cub/grid/grid_barrier.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -33,8 +33,15 @@
 
 #pragma once
 
-#include "../util_debug.cuh"
 #include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
+#include "../util_debug.cuh"
 #include "../thread/thread_load.cuh"
 
 CUB_NAMESPACE_BEGIN
@@ -195,13 +202,13 @@ public:
 
                 // Allocate and initialize to zero
                 retval = CubDebug(cudaMalloc((void**) &d_sync, sync_bytes));
-                if (cudaSuccess != retval) 
+                if (cudaSuccess != retval)
                 {
                     break;
                 }
 
                 retval = CubDebug(cudaMemset(d_sync, 0, new_sync_bytes));
-                if (cudaSuccess != retval) 
+                if (cudaSuccess != retval)
                 {
                     break;
                 }
diff --git a/cub/cub/grid/grid_even_share.cuh b/cub/cub/grid/grid_even_share.cuh
index d2150511321..f86b885a5c8 100644
--- a/cub/cub/grid/grid_even_share.cuh
+++ b/cub/cub/grid/grid_even_share.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -35,8 +35,13 @@
 #pragma once
 
 #include "../config.cuh"
-#include "../util_namespace.cuh"
-#include "../util_macro.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include "../util_math.cuh"
 #include "../util_type.cuh"
 #include "grid_mapping.cuh"
diff --git a/cub/cub/grid/grid_mapping.cuh b/cub/cub/grid/grid_mapping.cuh
index b57f193deb8..3b4abf41bff 100644
--- a/cub/cub/grid/grid_mapping.cuh
+++ b/cub/cub/grid/grid_mapping.cuh
@@ -35,6 +35,13 @@
 
 #include "../config.cuh"
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
+
 CUB_NAMESPACE_BEGIN
 
 
diff --git a/cub/cub/grid/grid_queue.cuh b/cub/cub/grid/grid_queue.cuh
index e1933e3d381..7a8f2887cce 100644
--- a/cub/cub/grid/grid_queue.cuh
+++ b/cub/cub/grid/grid_queue.cuh
@@ -33,7 +33,14 @@
 
 #pragma once
 
-#include <cub/config.cuh>
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/util_debug.cuh>
 
 #include <nv/target>
diff --git a/cub/cub/host/mutex.cuh b/cub/cub/host/mutex.cuh
index 4ee40288452..22773be0014 100644
--- a/cub/cub/host/mutex.cuh
+++ b/cub/cub/host/mutex.cuh
@@ -33,9 +33,16 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <mutex>
 
-#include <cub/config.cuh>
 #include <cub/util_deprecated.cuh>
 
 
@@ -43,8 +50,8 @@ CUB_NAMESPACE_BEGIN
 
 
 /**
- * Wraps std::mutex 
- * @deprecated [Since CUB 2.1.0] The `cub::Mutex` is deprecated and will be removed 
+ * Wraps std::mutex
+ * @deprecated [Since CUB 2.1.0] The `cub::Mutex` is deprecated and will be removed
  *             in a future release. Use `std::mutex` instead.
  */
 struct CUB_DEPRECATED Mutex
diff --git a/cub/cub/iterator/arg_index_input_iterator.cuh b/cub/cub/iterator/arg_index_input_iterator.cuh
index 7ea860981e6..d895a53e72f 100644
--- a/cub/cub/iterator/arg_index_input_iterator.cuh
+++ b/cub/cub/iterator/arg_index_input_iterator.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -33,10 +33,17 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <iterator>
 #include <iostream>
 
-#include "../config.cuh"
 #include "../thread/thread_load.cuh"
 #include "../thread/thread_store.cuh"
 
diff --git a/cub/cub/iterator/cache_modified_input_iterator.cuh b/cub/cub/iterator/cache_modified_input_iterator.cuh
index 9a5936d5ac6..b42e5b3cb49 100644
--- a/cub/cub/iterator/cache_modified_input_iterator.cuh
+++ b/cub/cub/iterator/cache_modified_input_iterator.cuh
@@ -33,10 +33,17 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <iterator>
 #include <iostream>
 
-#include "../config.cuh"
 #include "../thread/thread_load.cuh"
 #include "../thread/thread_store.cuh"
 
diff --git a/cub/cub/iterator/cache_modified_output_iterator.cuh b/cub/cub/iterator/cache_modified_output_iterator.cuh
index 91d4fc91a7e..92024af04d8 100644
--- a/cub/cub/iterator/cache_modified_output_iterator.cuh
+++ b/cub/cub/iterator/cache_modified_output_iterator.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -33,12 +33,19 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <iterator>
 #include <iostream>
 
 #include "../thread/thread_load.cuh"
 #include "../thread/thread_store.cuh"
-#include "../config.cuh"
 
 #if (THRUST_VERSION >= 100700)
     // This iterator is compatible with Thrust API 1.7 and newer
diff --git a/cub/cub/iterator/constant_input_iterator.cuh b/cub/cub/iterator/constant_input_iterator.cuh
index 3de5123df34..38816f5cbe1 100644
--- a/cub/cub/iterator/constant_input_iterator.cuh
+++ b/cub/cub/iterator/constant_input_iterator.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -33,12 +33,19 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <iterator>
 #include <iostream>
 
 #include "../thread/thread_load.cuh"
 #include "../thread/thread_store.cuh"
-#include "../config.cuh"
 
 #if (THRUST_VERSION >= 100700)
     // This iterator is compatible with Thrust API 1.7 and newer
diff --git a/cub/cub/iterator/counting_input_iterator.cuh b/cub/cub/iterator/counting_input_iterator.cuh
index 700455f420c..dc0108ac6f8 100644
--- a/cub/cub/iterator/counting_input_iterator.cuh
+++ b/cub/cub/iterator/counting_input_iterator.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -33,12 +33,19 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <iterator>
 #include <iostream>
 
 #include "../thread/thread_load.cuh"
 #include "../thread/thread_store.cuh"
-#include "../config.cuh"
 
 #if (THRUST_VERSION >= 100700)
     // This iterator is compatible with Thrust API 1.7 and newer
diff --git a/cub/cub/iterator/discard_output_iterator.cuh b/cub/cub/iterator/discard_output_iterator.cuh
index ac47a3ff344..66e764412b5 100644
--- a/cub/cub/iterator/discard_output_iterator.cuh
+++ b/cub/cub/iterator/discard_output_iterator.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -33,11 +33,17 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <iterator>
 #include <iostream>
 
-#include "../config.cuh"
-
 #if (THRUST_VERSION >= 100700)
     // This iterator is compatible with Thrust API 1.7 and newer
     #include <thrust/iterator/iterator_facade.h>
diff --git a/cub/cub/iterator/tex_obj_input_iterator.cuh b/cub/cub/iterator/tex_obj_input_iterator.cuh
index cd3d015aab8..24cdd165a34 100644
--- a/cub/cub/iterator/tex_obj_input_iterator.cuh
+++ b/cub/cub/iterator/tex_obj_input_iterator.cuh
@@ -33,7 +33,14 @@
 
 #pragma once
 
-#include <cub/config.cuh>
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/thread/thread_load.cuh>
 #include <cub/thread/thread_store.cuh>
 #include <cub/util_debug.cuh>
diff --git a/cub/cub/iterator/tex_ref_input_iterator.cuh b/cub/cub/iterator/tex_ref_input_iterator.cuh
index 0d877e1db49..67217a0f803 100644
--- a/cub/cub/iterator/tex_ref_input_iterator.cuh
+++ b/cub/cub/iterator/tex_ref_input_iterator.cuh
@@ -33,7 +33,14 @@
 
 #pragma once
 
-#include <cub/config.cuh>
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/iterator/tex_obj_input_iterator.cuh>
 
 #include <cstddef>
diff --git a/cub/cub/iterator/transform_input_iterator.cuh b/cub/cub/iterator/transform_input_iterator.cuh
index 0b3350e88a4..de69a64ae76 100644
--- a/cub/cub/iterator/transform_input_iterator.cuh
+++ b/cub/cub/iterator/transform_input_iterator.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -33,12 +33,19 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <iterator>
 #include <iostream>
 
 #include "../thread/thread_load.cuh"
 #include "../thread/thread_store.cuh"
-#include "../config.cuh"
 
 #if (THRUST_VERSION >= 100700)
     // This iterator is compatible with Thrust API 1.7 and newer
diff --git a/cub/cub/thread/thread_load.cuh b/cub/cub/thread/thread_load.cuh
index 0a8456898ec..fc83bbe8c3a 100644
--- a/cub/cub/thread/thread_load.cuh
+++ b/cub/cub/thread/thread_load.cuh
@@ -33,9 +33,16 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <iterator>
 
-#include "../config.cuh"
 #include "../util_ptx.cuh"
 #include "../util_type.cuh"
 
diff --git a/cub/cub/thread/thread_operators.cuh b/cub/cub/thread/thread_operators.cuh
index c6b097ea354..af1fe845b71 100644
--- a/cub/cub/thread/thread_operators.cuh
+++ b/cub/cub/thread/thread_operators.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2022, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,10 +12,10 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 
- * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 
- * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  * ARE DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE FOR ANY
  * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
@@ -37,11 +37,20 @@
 
 #pragma once
 
-#include <cub/config.cuh>
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/util_cpp_dialect.cuh>
 #include <cub/util_type.cuh>
 
+_CCCL_DIAG_SUPPRESS_DEPRECATED_PUSH
 #include <cuda/std/functional>
+_CCCL_DIAG_SUPPRESS_DEPRECATED_POP
 #include <cuda/std/type_traits>
 #include <cuda/std/utility>
 
@@ -164,8 +173,8 @@ struct ArgMax
              const KeyValuePair<OffsetT, T> &b) const
   {
     // Mooch BUG (device reduce argmax gk110 3.2 million random fp32)
-    // return ((b.value > a.value) || 
-    //         ((a.value == b.value) && (b.key < a.key))) 
+    // return ((b.value > a.value) ||
+    //         ((a.value == b.value) && (b.key < a.key)))
     //      ? b : a;
 
     if ((b.value > a.value) || ((a.value == b.value) && (b.key < a.key)))
@@ -282,14 +291,14 @@ public:
 /**
  * @brief Reduce-by-segment functor.
  *
- * Given two cub::KeyValuePair inputs `a` and `b` and a binary associative 
- * combining operator `f(const T &x, const T &y)`, an instance of this functor 
- * returns a cub::KeyValuePair whose `key` field is `a.key + b.key`, and whose 
- * `value` field is either `b.value` if `b.key` is non-zero, or 
+ * Given two cub::KeyValuePair inputs `a` and `b` and a binary associative
+ * combining operator `f(const T &x, const T &y)`, an instance of this functor
+ * returns a cub::KeyValuePair whose `key` field is `a.key + b.key`, and whose
+ * `value` field is either `b.value` if `b.key` is non-zero, or
  * `f(a.value, b.value)` otherwise.
  *
- * ReduceBySegmentOp is an associative, non-commutative binary combining 
- * operator for input sequences of cub::KeyValuePair pairings. Such sequences 
+ * ReduceBySegmentOp is an associative, non-commutative binary combining
+ * operator for input sequences of cub::KeyValuePair pairings. Such sequences
  * are typically used to represent a segmented set of values to be reduced
  * and a corresponding set of {0,1}-valued integer "head flags" demarcating the
  * first value of each segment.
@@ -348,7 +357,7 @@ struct ReduceBySegmentOp
     // else {
     //   The second partial reduction does not span a reset, so accumulate both
     //   into the running aggregate
-    // } 
+    // }
     retval.value = (second.key) ? second.value : op(first.value, second.value);
 #endif
     return retval;
diff --git a/cub/cub/thread/thread_reduce.cuh b/cub/cub/thread/thread_reduce.cuh
index 82042a17667..8384b0cc401 100644
--- a/cub/cub/thread/thread_reduce.cuh
+++ b/cub/cub/thread/thread_reduce.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -33,9 +33,16 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include "../thread/thread_operators.cuh"
 #include "../detail/type_traits.cuh"
-#include "../config.cuh"
 
 CUB_NAMESPACE_BEGIN
 
@@ -50,7 +57,7 @@ template <
     typename    T,
     typename    ReductionOp,
     typename    PrefixT,
-    typename    AccumT = detail::accumulator_t<ReductionOp, PrefixT, T>> 
+    typename    AccumT = detail::accumulator_t<ReductionOp, PrefixT, T>>
 __device__ __forceinline__ AccumT ThreadReduce(
     T*                  input,                  ///< [in] Input array
     ReductionOp         reduction_op,           ///< [in] Binary reduction operator
@@ -79,7 +86,7 @@ template <
     typename    T,
     typename    ReductionOp,
     typename    PrefixT,
-    typename    AccumT = detail::accumulator_t<ReductionOp, PrefixT, T>> 
+    typename    AccumT = detail::accumulator_t<ReductionOp, PrefixT, T>>
 __device__ __forceinline__ AccumT ThreadReduce(
     T*          input,                  ///< [in] Input array
     ReductionOp reduction_op,           ///< [in] Binary reduction operator
@@ -121,7 +128,7 @@ template <
     typename    T,
     typename    ReductionOp,
     typename    PrefixT,
-    typename    AccumT = detail::accumulator_t<ReductionOp, PrefixT, T>> 
+    typename    AccumT = detail::accumulator_t<ReductionOp, PrefixT, T>>
 __device__ __forceinline__ AccumT ThreadReduce(
     T           (&input)[LENGTH],       ///< [in] Input array
     ReductionOp reduction_op,           ///< [in] Binary reduction operator
diff --git a/cub/cub/thread/thread_scan.cuh b/cub/cub/thread/thread_scan.cuh
index b5e42710fcd..bc3840ec9c8 100644
--- a/cub/cub/thread/thread_scan.cuh
+++ b/cub/cub/thread/thread_scan.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -34,6 +34,13 @@
 #pragma once
 
 #include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include "../thread/thread_operators.cuh"
 
 CUB_NAMESPACE_BEGIN
diff --git a/cub/cub/thread/thread_search.cuh b/cub/cub/thread/thread_search.cuh
index 62b3cdb4e5d..e18caf09a09 100644
--- a/cub/cub/thread/thread_search.cuh
+++ b/cub/cub/thread/thread_search.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -33,10 +33,17 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <iterator>
 #include <cub/util_namespace.cuh>
 #include <cub/util_type.cuh>
-#include <cub/config.cuh>
 
 #include <nv/target>
 
diff --git a/cub/cub/thread/thread_sort.cuh b/cub/cub/thread/thread_sort.cuh
index 5d486789684..caeb22e7801 100644
--- a/cub/cub/thread/thread_sort.cuh
+++ b/cub/cub/thread/thread_sort.cuh
@@ -28,6 +28,13 @@
 #pragma once
 
 #include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include "../util_ptx.cuh"
 #include "../util_type.cuh"
 
diff --git a/cub/cub/thread/thread_store.cuh b/cub/cub/thread/thread_store.cuh
index d1f055df158..9d24aa54ad9 100644
--- a/cub/cub/thread/thread_store.cuh
+++ b/cub/cub/thread/thread_store.cuh
@@ -33,7 +33,14 @@
 
 #pragma once
 
-#include <cub/config.cuh>
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/util_ptx.cuh>
 #include <cub/util_type.cuh>
 
diff --git a/cub/cub/util_allocator.cuh b/cub/cub/util_allocator.cuh
index 7dc12fb920e..8172dd919f8 100644
--- a/cub/cub/util_allocator.cuh
+++ b/cub/cub/util_allocator.cuh
@@ -33,6 +33,14 @@
 
 #pragma once
 
+#include "config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include "util_namespace.cuh"
 #include "util_debug.cuh"
 
@@ -453,13 +461,13 @@ struct CachingDeviceAllocator
             if (device != entrypoint_device)
             {
                 error = CubDebug(cudaGetDevice(&entrypoint_device));
-                if (cudaSuccess != error) 
+                if (cudaSuccess != error)
                 {
                     return error;
                 }
 
                 error = CubDebug(cudaSetDevice(device));
-                if (cudaSuccess != error) 
+                if (cudaSuccess != error)
                 {
                     return error;
                 }
@@ -491,13 +499,13 @@ struct CachingDeviceAllocator
 
                     // Free device memory and destroy stream event.
                     error = CubDebug(cudaFree(block_itr->d_ptr));
-                    if (cudaSuccess != error) 
+                    if (cudaSuccess != error)
                     {
                         break;
                     }
 
                     error = CubDebug(cudaEventDestroy(block_itr->ready_event));
-                    if (cudaSuccess != error) 
+                    if (cudaSuccess != error)
                     {
                         break;
                     }
@@ -519,7 +527,7 @@ struct CachingDeviceAllocator
 
                 // Try to allocate again
                 error = CubDebug(cudaMalloc(&search_key.d_ptr, search_key.bytes));
-                if (cudaSuccess != error) 
+                if (cudaSuccess != error)
                 {
                     return error;
                 }
@@ -547,7 +555,7 @@ struct CachingDeviceAllocator
             if ((entrypoint_device != INVALID_DEVICE_ORDINAL) && (entrypoint_device != device))
             {
                 error = CubDebug(cudaSetDevice(entrypoint_device));
-                if (cudaSuccess != error) 
+                if (cudaSuccess != error)
                 {
                     return error;
                 }
@@ -639,13 +647,13 @@ struct CachingDeviceAllocator
         if (device != entrypoint_device)
         {
             error = CubDebug(cudaGetDevice(&entrypoint_device));
-            if (cudaSuccess != error) 
+            if (cudaSuccess != error)
             {
                 return error;
             }
 
             error = CubDebug(cudaSetDevice(device));
-            if (cudaSuccess != error) 
+            if (cudaSuccess != error)
             {
                 return error;
             }
@@ -655,7 +663,7 @@ struct CachingDeviceAllocator
         {
             // Insert the ready event in the associated stream (must have current device set properly)
             error = CubDebug(cudaEventRecord(search_key.ready_event, search_key.associated_stream));
-            if (cudaSuccess != error) 
+            if (cudaSuccess != error)
             {
                 return error;
             }
@@ -665,13 +673,13 @@ struct CachingDeviceAllocator
         {
             // Free the allocation from the runtime and cleanup the event.
             error = CubDebug(cudaFree(d_ptr));
-            if (cudaSuccess != error) 
+            if (cudaSuccess != error)
             {
                 return error;
             }
 
             error = CubDebug(cudaEventDestroy(search_key.ready_event));
-            if (cudaSuccess != error) 
+            if (cudaSuccess != error)
             {
                 return error;
             }
@@ -684,7 +692,7 @@ struct CachingDeviceAllocator
         if ((entrypoint_device != INVALID_DEVICE_ORDINAL) && (entrypoint_device != device))
         {
             error = CubDebug(cudaSetDevice(entrypoint_device));
-            if (cudaSuccess != error) 
+            if (cudaSuccess != error)
             {
                 return error;
             }
@@ -728,7 +736,7 @@ struct CachingDeviceAllocator
             if (entrypoint_device == INVALID_DEVICE_ORDINAL)
             {
                 error = CubDebug(cudaGetDevice(&entrypoint_device));
-                if (cudaSuccess != error) 
+                if (cudaSuccess != error)
                 {
                     break;
                 }
@@ -738,7 +746,7 @@ struct CachingDeviceAllocator
             if (begin->device != current_device)
             {
                 error = CubDebug(cudaSetDevice(begin->device));
-                if (cudaSuccess != error) 
+                if (cudaSuccess != error)
                 {
                     break;
                 }
@@ -747,13 +755,13 @@ struct CachingDeviceAllocator
 
             // Free device memory
             error = CubDebug(cudaFree(begin->d_ptr));
-            if (cudaSuccess != error) 
+            if (cudaSuccess != error)
             {
                 break;
             }
 
             error = CubDebug(cudaEventDestroy(begin->ready_event));
-            if (cudaSuccess != error) 
+            if (cudaSuccess != error)
             {
                 break;
             }
diff --git a/cub/cub/util_arch.cuh b/cub/cub/util_arch.cuh
index d2506e93cfc..2a09635ac34 100644
--- a/cub/cub/util_arch.cuh
+++ b/cub/cub/util_arch.cuh
@@ -33,9 +33,17 @@
 
 #pragma once
 
-#include <cub/util_cpp_dialect.cuh>
-#include <cub/util_namespace.cuh>
-#include <cub/util_macro.cuh>
+#include "config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
+#include "util_cpp_dialect.cuh"
+#include "util_namespace.cuh"
+#include "util_macro.cuh"
 
 // Legacy include; this functionality used to be defined in here.
 #include <cub/detail/detect_cuda_runtime.cuh>
@@ -44,7 +52,7 @@ CUB_NAMESPACE_BEGIN
 
 #ifndef DOXYGEN_SHOULD_SKIP_THIS    // Do not document
 
-// \deprecated [Since 2.1.0] 
+// \deprecated [Since 2.1.0]
 #define CUB_USE_COOPERATIVE_GROUPS
 
 /// In device code, CUB_PTX_ARCH expands to the PTX version for which we are
diff --git a/cub/cub/util_compiler.cuh b/cub/cub/util_compiler.cuh
index 7cda3c44012..3286c07db43 100644
--- a/cub/cub/util_compiler.cuh
+++ b/cub/cub/util_compiler.cuh
@@ -32,6 +32,15 @@
 
 #pragma once
 
+// For _CCCL_IMPLICIT_SYSTEM_HEADER
+#include <cuda/__cccl_config>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // enumerate host compilers we know about
 #define CUB_HOST_COMPILER_UNKNOWN 0
 #define CUB_HOST_COMPILER_MSVC 1
diff --git a/cub/cub/util_cpp_dialect.cuh b/cub/cub/util_cpp_dialect.cuh
index 693177974a2..81c01203fe0 100644
--- a/cub/cub/util_cpp_dialect.cuh
+++ b/cub/cub/util_cpp_dialect.cuh
@@ -31,6 +31,14 @@
 
 #pragma once
 
+#include <cuda/__cccl_config>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include "util_compiler.cuh"
 
 // Deprecation warnings may be silenced by defining the following macros. These
diff --git a/cub/cub/util_debug.cuh b/cub/cub/util_debug.cuh
index 5023524dd6e..4f7bdd86205 100644
--- a/cub/cub/util_debug.cuh
+++ b/cub/cub/util_debug.cuh
@@ -36,8 +36,13 @@
 
 #pragma once
 
-#include <cub/util_namespace.cuh>
-#include <cub/util_arch.cuh>
+#include "config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #include <nv/target>
 
@@ -58,8 +63,8 @@ CUB_NAMESPACE_BEGIN
 /**
  * @def CUB_DEBUG_SYNC
  *
- * Causes synchronization of the stream after every kernel launch to check 
- * for errors. Also causes kernel launch configurations to be printed to the 
+ * Causes synchronization of the stream after every kernel launch to check
+ * for errors. Also causes kernel launch configurations to be printed to the
  * console.
  */
 #define CUB_DEBUG_SYNC
@@ -67,7 +72,7 @@ CUB_NAMESPACE_BEGIN
 /**
  * @def CUB_DEBUG_HOST_ASSERTIONS
  *
- * Extends `CUB_DEBUG_SYNC` effects by checking host-side precondition 
+ * Extends `CUB_DEBUG_SYNC` effects by checking host-side precondition
  * assertions.
  */
 #define CUB_DEBUG_HOST_ASSERTIONS
@@ -75,7 +80,7 @@ CUB_NAMESPACE_BEGIN
 /**
  * @def CUB_DEBUG_DEVICE_ASSERTIONS
  *
- * Extends `CUB_DEBUG_HOST_ASSERTIONS` effects by checking device-side 
+ * Extends `CUB_DEBUG_HOST_ASSERTIONS` effects by checking device-side
  * precondition assertions.
  */
 #define CUB_DEBUG_DEVICE_ASSERTIONS
@@ -83,14 +88,14 @@ CUB_NAMESPACE_BEGIN
 /**
  * @def CUB_DEBUG_ALL
  *
- * Causes host and device-side precondition assertions to be checked. Apart 
- * from that, causes synchronization of the stream after every kernel launch to 
- * check for errors. Also causes kernel launch configurations to be printed to 
+ * Causes host and device-side precondition assertions to be checked. Apart
+ * from that, causes synchronization of the stream after every kernel launch to
+ * check for errors. Also causes kernel launch configurations to be printed to
  * the console.
  */
 #define CUB_DEBUG_ALL
 
-#endif // DOXYGEN_SHOULD_SKIP_THIS 
+#endif // DOXYGEN_SHOULD_SKIP_THIS
 
 /**
  * \addtogroup UtilMgmt
@@ -132,7 +137,7 @@ CUB_NAMESPACE_BEGIN
 
 // All
 #ifdef CUB_DEBUG_ALL
-#define CUB_DETAIL_DEBUG_LEVEL CUB_DETAIL_DEBUG_LEVEL_ALL 
+#define CUB_DETAIL_DEBUG_LEVEL CUB_DETAIL_DEBUG_LEVEL_ALL
 #endif
 
 // Default case, no extra debugging:
@@ -196,11 +201,11 @@ cudaError_t Debug(cudaError_t error, const char *filename, int line)
   cudaError_t last_error = cudaSuccess;
 
   NV_IF_TARGET(
-    NV_IS_HOST, 
+    NV_IS_HOST,
     (last_error = cudaGetLastError();),
     (CUB_TEMP_DEVICE_CODE;)
   );
-  
+
   #undef CUB_TEMP_DEVICE_CODE
   // clang-format on
 
diff --git a/cub/cub/util_deprecated.cuh b/cub/cub/util_deprecated.cuh
index a988c9fca90..7b8834cc929 100644
--- a/cub/cub/util_deprecated.cuh
+++ b/cub/cub/util_deprecated.cuh
@@ -32,13 +32,17 @@
 
 #pragma once
 
+#include "config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #include <cub/detail/type_traits.cuh>
-#include <cub/util_compiler.cuh>
-#include <cub/util_cpp_dialect.cuh>
 #include <cub/util_debug.cuh>
 
-
 #if defined(THRUST_IGNORE_DEPRECATED_API) && !defined(CUB_IGNORE_DEPRECATED_API)
 #  define CUB_IGNORE_DEPRECATED_API
 #endif
diff --git a/cub/cub/util_device.cuh b/cub/cub/util_device.cuh
index c7e15cafe06..d8bb40cdcc4 100644
--- a/cub/cub/util_device.cuh
+++ b/cub/cub/util_device.cuh
@@ -37,14 +37,18 @@
 
 #pragma once
 
+#include "config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cuda/std/utility>
 
 #include <cub/detail/device_synchronize.cuh>
-#include <cub/util_arch.cuh>
-#include <cub/util_cpp_dialect.cuh>
 #include <cub/util_debug.cuh>
-#include <cub/util_macro.cuh>
-#include <cub/util_namespace.cuh>
 #include <cub/util_type.cuh>
 // for backward compatibility
 #include <cub/util_temporary_storage.cuh>
@@ -412,7 +416,7 @@ CUB_RUNTIME_FUNCTION inline cudaError_t SmVersionUncached(int& sm_version, int d
     {
         int major = 0, minor = 0;
         error = CubDebug(cudaDeviceGetAttribute(&major, cudaDevAttrComputeCapabilityMajor, device));
-        if (cudaSuccess != error) 
+        if (cudaSuccess != error)
         {
             break;
         }
@@ -544,7 +548,7 @@ CUB_RUNTIME_FUNCTION inline cudaError_t HasUVA(bool& has_uva)
     cudaError_t error = cudaSuccess;
     int device = -1;
     error = CubDebug(cudaGetDevice(&device));
-    if (cudaSuccess != error) 
+    if (cudaSuccess != error)
     {
         return error;
     }
diff --git a/cub/cub/util_macro.cuh b/cub/cub/util_macro.cuh
index 73a97c44df4..99bd0165fe7 100644
--- a/cub/cub/util_macro.cuh
+++ b/cub/cub/util_macro.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -32,15 +32,21 @@
 
 #pragma once
 
+#include "version.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/detail/detect_cuda_runtime.cuh>
 #include <cub/util_namespace.cuh>
 
 #include <cuda/std/utility>
-#include <cuda/std/version> // _LIBCUDACXX_HIDDEN, _LIBCUDACXX_{CLANG,GCC}_DIAGNOSTIC_IGNORED
 
 CUB_NAMESPACE_BEGIN
 
-
 /**
  * \addtogroup UtilModule
  * @{
@@ -116,7 +122,7 @@ constexpr __host__ __device__ auto max CUB_PREVENT_MACRO_SUBSTITUTION(T &&t,
 #endif
 
 #ifndef CUB_DETAIL_KERNEL_ATTRIBUTES
-#define CUB_DETAIL_KERNEL_ATTRIBUTES __global__ _LIBCUDACXX_HIDDEN
+#define CUB_DETAIL_KERNEL_ATTRIBUTES CCCL_DETAIL_KERNEL_ATTRIBUTES
 #endif
 
 /**
@@ -124,12 +130,12 @@ constexpr __host__ __device__ auto max CUB_PREVENT_MACRO_SUBSTITUTION(T &&t,
  * If defined, the default suppression of kernel visibility attribute warning is disabled.
  */
 #if !defined(CUB_DISABLE_KERNEL_VISIBILITY_WARNING_SUPPRESSION)
-_LIBCUDACXX_GCC_DIAGNOSTIC_IGNORED("-Wattributes")
-_LIBCUDACXX_CLANG_DIAGNOSTIC_IGNORED("-Wattributes")                      
-#if !defined(_LIBCUDACXX_COMPILER_NVHPC_CUDA)
-_LIBCUDACXX_NVHPC_DIAGNOSTIC_IGNORED(attribute_requires_external_linkage)
-#endif
-#endif
+_CCCL_DIAG_SUPPRESS_GCC("-Wattributes")
+_CCCL_DIAG_SUPPRESS_CLANG("-Wattributes")
+#if !defined(_CCCL_CUDA_COMPILER_NVHPC)
+_CCCL_DIAG_SUPPRESS_NVHPC(attribute_requires_external_linkage)
+#endif // !_CCCL_CUDA_COMPILER_NVHPC
+#endif // !CUB_DISABLE_KERNEL_VISIBILITY_WARNING_SUPPRESSION
 
 /** @} */       // end group UtilModule
 
diff --git a/cub/cub/util_math.cuh b/cub/cub/util_math.cuh
index d69fc2ee2d5..018c2a615d5 100644
--- a/cub/cub/util_math.cuh
+++ b/cub/cub/util_math.cuh
@@ -32,10 +32,15 @@
 
 #pragma once
 
-#include <type_traits>
+#include "config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
-#include "util_namespace.cuh"
-#include "util_macro.cuh"
+#include <type_traits>
 
 CUB_NAMESPACE_BEGIN
 
diff --git a/cub/cub/util_namespace.cuh b/cub/cub/util_namespace.cuh
index 27ff12dbbaa..5139c8a571b 100644
--- a/cub/cub/util_namespace.cuh
+++ b/cub/cub/util_namespace.cuh
@@ -38,7 +38,14 @@
 // This is not used by this file; this is a hack so that we can detect the
 // CUB version from Thrust on older versions of CUB that did not have
 // version.cuh.
-#include <cub/version.cuh>
+#include "version.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/detail/detect_cuda_runtime.cuh>
 
 // Prior to 1.13.1, only the PREFIX/POSTFIX macros were used. Notify users
diff --git a/cub/cub/util_ptx.cuh b/cub/cub/util_ptx.cuh
index ff6fdb07f50..4c5d01e5e54 100644
--- a/cub/cub/util_ptx.cuh
+++ b/cub/cub/util_ptx.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -34,14 +34,18 @@
 
 #pragma once
 
-#include "util_type.cuh"
-#include "util_arch.cuh"
-#include "util_namespace.cuh"
-#include "util_debug.cuh"
+#include "config.cuh"
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
-CUB_NAMESPACE_BEGIN
+#include "util_debug.cuh"
+#include "util_type.cuh"
 
+CUB_NAMESPACE_BEGIN
 
 /**
  * \addtogroup UtilPtx
@@ -163,7 +167,7 @@ __device__ __forceinline__ unsigned int BFE(
     return (source >> bit_start) & MASK;
 }
 
-#if CUB_IS_INT128_ENABLED 
+#if CUB_IS_INT128_ENABLED
 /**
  * Bitfield-extract for 128-bit types.
  */
@@ -328,7 +332,7 @@ __device__  __forceinline__ int WARP_BALLOT(int predicate, unsigned int member_m
 /**
  * Warp synchronous shfl_up
  */
-__device__ __forceinline__ 
+__device__ __forceinline__
 unsigned int SHFL_UP_SYNC(unsigned int word, int src_offset, int flags, unsigned int member_mask)
 {
     asm volatile("shfl.sync.up.b32 %0, %1, %2, %3, %4;"
@@ -339,7 +343,7 @@ unsigned int SHFL_UP_SYNC(unsigned int word, int src_offset, int flags, unsigned
 /**
  * Warp synchronous shfl_down
  */
-__device__ __forceinline__ 
+__device__ __forceinline__
 unsigned int SHFL_DOWN_SYNC(unsigned int word, int src_offset, int flags, unsigned int member_mask)
 {
     asm volatile("shfl.sync.down.b32 %0, %1, %2, %3, %4;"
@@ -350,7 +354,7 @@ unsigned int SHFL_DOWN_SYNC(unsigned int word, int src_offset, int flags, unsign
 /**
  * Warp synchronous shfl_idx
  */
-__device__ __forceinline__ 
+__device__ __forceinline__
 unsigned int SHFL_IDX_SYNC(unsigned int word, int src_lane, int flags, unsigned int member_mask)
 {
     asm volatile("shfl.sync.idx.b32 %0, %1, %2, %3, %4;"
@@ -361,7 +365,7 @@ unsigned int SHFL_IDX_SYNC(unsigned int word, int src_lane, int flags, unsigned
 /**
  * Warp synchronous shfl_idx
  */
-__device__ __forceinline__ 
+__device__ __forceinline__
 unsigned int SHFL_IDX_SYNC(unsigned int word, int src_lane, unsigned int member_mask)
 {
     return __shfl_sync(member_mask, word, src_lane);
@@ -395,7 +399,7 @@ __device__ __forceinline__ float FFMA_RZ(float a, float b, float c)
  */
 __device__ __forceinline__ void ThreadExit() {
     asm volatile("exit;");
-}    
+}
 
 
 /**
@@ -561,7 +565,7 @@ __device__ __forceinline__ T ShuffleUp(
     typedef typename UnitWord<T>::ShuffleWord ShuffleWord;
 
     constexpr int   WORDS           = (sizeof(T) + sizeof(ShuffleWord) - 1) / sizeof(ShuffleWord);
- 
+
     T               output;
     ShuffleWord     *output_alias   = reinterpret_cast<ShuffleWord *>(&output);
     ShuffleWord     *input_alias    = reinterpret_cast<ShuffleWord *>(&input);
@@ -728,26 +732,26 @@ __device__ __forceinline__ T ShuffleIndex(
 
 
 #ifndef DOXYGEN_SHOULD_SKIP_THIS    // Do not document
-namespace detail 
+namespace detail
 {
 
-/** 
- * Implementation detail for `MatchAny`. It provides specializations for full and partial warps. 
- * For partial warps, inactive threads must be masked out. This is done in the partial warp 
- * specialization below. 
+/**
+ * Implementation detail for `MatchAny`. It provides specializations for full and partial warps.
+ * For partial warps, inactive threads must be masked out. This is done in the partial warp
+ * specialization below.
  * Usage:
  * ```
- * // returns a mask of threads with the same 4 least-significant bits of `label` 
+ * // returns a mask of threads with the same 4 least-significant bits of `label`
  * // in a warp with 16 active threads
- * warp_matcher_t<4, 16>::match_any(label); 
+ * warp_matcher_t<4, 16>::match_any(label);
  *
- * // returns a mask of threads with the same 4 least-significant bits of `label` 
+ * // returns a mask of threads with the same 4 least-significant bits of `label`
  * // in a warp with 32 active threads (no extra work is done)
- * warp_matcher_t<4, 32>::match_any(label); 
+ * warp_matcher_t<4, 32>::match_any(label);
  * ```
  */
 template <int LABEL_BITS, int WARP_ACTIVE_THREADS>
-struct warp_matcher_t 
+struct warp_matcher_t
 {
 
   static __device__ unsigned int match_any(unsigned int label)
@@ -758,7 +762,7 @@ struct warp_matcher_t
 };
 
 template <int LABEL_BITS>
-struct warp_matcher_t<LABEL_BITS, CUB_PTX_WARP_THREADS> 
+struct warp_matcher_t<LABEL_BITS, CUB_PTX_WARP_THREADS>
 {
 
   // match.any.sync.b32 is slower when matching a few bits
diff --git a/cub/cub/util_temporary_storage.cuh b/cub/cub/util_temporary_storage.cuh
index 588c554a32f..17548600a47 100644
--- a/cub/cub/util_temporary_storage.cuh
+++ b/cub/cub/util_temporary_storage.cuh
@@ -33,6 +33,14 @@
 
 #pragma once
 
+#include "config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/util_debug.cuh>
 #include <cub/util_namespace.cuh>
 
diff --git a/cub/cub/util_type.cuh b/cub/cub/util_type.cuh
index 2beecbc892c..6fbe3413760 100644
--- a/cub/cub/util_type.cuh
+++ b/cub/cub/util_type.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -33,6 +33,14 @@
 
 #pragma once
 
+#include "config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cfloat>
 #include <iostream>
 #include <iterator>
@@ -48,11 +56,6 @@
 #endif
 
 #include <cub/detail/uninitialized_copy.cuh>
-#include <cub/util_arch.cuh>
-#include <cub/util_compiler.cuh>
-#include <cub/util_deprecated.cuh>
-#include <cub/util_macro.cuh>
-#include <cub/util_namespace.cuh>
 
 #include <cuda/std/type_traits>
 
@@ -1220,7 +1223,7 @@ template <> struct NumericTraits<unsigned long> :       BaseTraits<UNSIGNED_INTE
 template <> struct NumericTraits<unsigned long long> :  BaseTraits<UNSIGNED_INTEGER, true, false, unsigned long long, unsigned long long> {};
 
 
-#if CUB_IS_INT128_ENABLED 
+#if CUB_IS_INT128_ENABLED
 template <>
 struct NumericTraits<__uint128_t>
 {
diff --git a/cub/cub/version.cuh b/cub/cub/version.cuh
index d51023912fa..9934b92eff9 100644
--- a/cub/cub/version.cuh
+++ b/cub/cub/version.cuh
@@ -35,6 +35,15 @@
 
 #pragma once
 
+// For _CCCL_IMPLICIT_SYSTEM_HEADER
+#include <cuda/__cccl_config>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 /*! \def CUB_VERSION
  *  \brief The preprocessor macro \p CUB_VERSION encodes the version
  *         number of the CUB library.
diff --git a/cub/cub/warp/specializations/warp_exchange_shfl.cuh b/cub/cub/warp/specializations/warp_exchange_shfl.cuh
index fa73509b319..ae1327ee6c1 100644
--- a/cub/cub/warp/specializations/warp_exchange_shfl.cuh
+++ b/cub/cub/warp/specializations/warp_exchange_shfl.cuh
@@ -27,7 +27,14 @@
 
 #pragma once
 
-#include <cub/config.cuh>
+#include "../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/util_ptx.cuh>
 #include <cub/util_type.cuh>
 
@@ -51,7 +58,7 @@ class WarpExchangeShfl
   static constexpr bool IS_ARCH_WARP = LOGICAL_WARP_THREADS == CUB_WARP_THREADS(0);
 
   // concrete recursion class
-  template <typename OutputT, int IDX, int SIZE> 
+  template <typename OutputT, int IDX, int SIZE>
   class CompileTimeArray : protected CompileTimeArray<OutputT, IDX + 1, SIZE>
   {
   protected:
@@ -239,7 +246,7 @@ class WarpExchangeShfl
   };
 
   // terminating partial specialization
-  template <typename OutputT, int SIZE> 
+  template <typename OutputT, int SIZE>
   class CompileTimeArray<OutputT, SIZE, SIZE>
   {
   protected:
diff --git a/cub/cub/warp/specializations/warp_exchange_smem.cuh b/cub/cub/warp/specializations/warp_exchange_smem.cuh
index f1c2edce60f..fc9fcbc5939 100644
--- a/cub/cub/warp/specializations/warp_exchange_smem.cuh
+++ b/cub/cub/warp/specializations/warp_exchange_smem.cuh
@@ -33,7 +33,14 @@
 
 #pragma once
 
-#include <cub/config.cuh>
+#include "../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/util_ptx.cuh>
 #include <cub/util_type.cuh>
 
diff --git a/cub/cub/warp/specializations/warp_reduce_shfl.cuh b/cub/cub/warp/specializations/warp_reduce_shfl.cuh
index ad6d2512f92..dd0e82bf0b2 100644
--- a/cub/cub/warp/specializations/warp_reduce_shfl.cuh
+++ b/cub/cub/warp/specializations/warp_reduce_shfl.cuh
@@ -34,6 +34,13 @@
 #pragma once
 
 #include "../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include "../../thread/thread_operators.cuh"
 #include "../../util_ptx.cuh"
 #include "../../util_type.cuh"
@@ -46,31 +53,31 @@
 CUB_NAMESPACE_BEGIN
 
 
-namespace detail 
+namespace detail
 {
 
 template <class A = int, class = A>
-struct reduce_add_exists : ::cuda::std::false_type 
+struct reduce_add_exists : ::cuda::std::false_type
 {};
 
 template <class T>
-struct reduce_add_exists<T, decltype(__reduce_add_sync(0xFFFFFFFF, T{}))> : ::cuda::std::true_type 
+struct reduce_add_exists<T, decltype(__reduce_add_sync(0xFFFFFFFF, T{}))> : ::cuda::std::true_type
 {};
 
 template <class T = int, class = T>
-struct reduce_min_exists : ::cuda::std::false_type 
+struct reduce_min_exists : ::cuda::std::false_type
 {};
 
 template <class T>
-struct reduce_min_exists<T, decltype(__reduce_min_sync(0xFFFFFFFF, T{}))> : ::cuda::std::true_type 
+struct reduce_min_exists<T, decltype(__reduce_min_sync(0xFFFFFFFF, T{}))> : ::cuda::std::true_type
 {};
 
 template <class T = int, class = T>
-struct reduce_max_exists : ::cuda::std::false_type 
+struct reduce_max_exists : ::cuda::std::false_type
 {};
 
 template <class T>
-struct reduce_max_exists<T, decltype(__reduce_max_sync(0xFFFFFFFF, T{}))> : ::cuda::std::true_type 
+struct reduce_max_exists<T, decltype(__reduce_max_sync(0xFFFFFFFF, T{}))> : ::cuda::std::true_type
 {};
 
 }
@@ -419,7 +426,7 @@ struct WarpReduceShfl
     //---------------------------------------------------------------------
     template <typename ReductionOp>
     __device__ __forceinline__ T ReduceImpl(
-        Int2Type<0>     /* all_lanes_valid */, 
+        Int2Type<0>     /* all_lanes_valid */,
         T               input,                  ///< [in] Calling thread's input
         int             valid_items,            ///< [in] Total number of valid items across the logical warp
         ReductionOp     reduction_op)           ///< [in] Binary reduction operator
@@ -436,7 +443,7 @@ struct WarpReduceShfl
 
     template <typename ReductionOp>
     __device__ __forceinline__ T ReduceImpl(
-        Int2Type<1>     /* all_lanes_valid */, 
+        Int2Type<1>     /* all_lanes_valid */,
         T               input,                  ///< [in] Calling thread's input
         int             /* valid_items */,      ///< [in] Total number of valid items across the logical warp
         ReductionOp     reduction_op)           ///< [in] Binary reduction operator
@@ -452,7 +459,7 @@ struct WarpReduceShfl
     }
 
     template <class U = T>
-    __device__ __forceinline__ 
+    __device__ __forceinline__
     typename std::enable_if<
                (std::is_same<int, U>::value || std::is_same<unsigned int, U>::value)
             && detail::reduce_add_exists<>::value, T>::type
@@ -474,7 +481,7 @@ struct WarpReduceShfl
     }
 
     template <class U = T>
-    __device__ __forceinline__ 
+    __device__ __forceinline__
     typename std::enable_if<
                (std::is_same<int, U>::value || std::is_same<unsigned int, U>::value)
             && detail::reduce_min_exists<>::value, T>::type
@@ -496,7 +503,7 @@ struct WarpReduceShfl
     }
 
     template <class U = T>
-    __device__ __forceinline__ 
+    __device__ __forceinline__
     typename std::enable_if<
                (std::is_same<int, U>::value || std::is_same<unsigned int, U>::value)
             && detail::reduce_max_exists<>::value, T>::type
diff --git a/cub/cub/warp/specializations/warp_reduce_smem.cuh b/cub/cub/warp/specializations/warp_reduce_smem.cuh
index 7ffb73a3c27..a2077940c7b 100644
--- a/cub/cub/warp/specializations/warp_reduce_smem.cuh
+++ b/cub/cub/warp/specializations/warp_reduce_smem.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -34,6 +34,13 @@
 #pragma once
 
 #include "../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include "../../thread/thread_operators.cuh"
 #include "../../thread/thread_load.cuh"
 #include "../../thread/thread_store.cuh"
diff --git a/cub/cub/warp/specializations/warp_scan_shfl.cuh b/cub/cub/warp/specializations/warp_scan_shfl.cuh
index c2988711c8d..81db566c184 100644
--- a/cub/cub/warp/specializations/warp_scan_shfl.cuh
+++ b/cub/cub/warp/specializations/warp_scan_shfl.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -34,6 +34,13 @@
 #pragma once
 
 #include "../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include "../../thread/thread_operators.cuh"
 #include "../../util_type.cuh"
 #include "../../util_ptx.cuh"
diff --git a/cub/cub/warp/specializations/warp_scan_smem.cuh b/cub/cub/warp/specializations/warp_scan_smem.cuh
index f5290e979a4..8f76b3c6253 100644
--- a/cub/cub/warp/specializations/warp_scan_smem.cuh
+++ b/cub/cub/warp/specializations/warp_scan_smem.cuh
@@ -1,7 +1,7 @@
 /******************************************************************************
  * Copyright (c) 2011, Duane Merrill.  All rights reserved.
  * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  *     * Redistributions of source code must retain the above copyright
@@ -12,7 +12,7 @@
  *     * Neither the name of the NVIDIA CORPORATION nor the
  *       names of its contributors may be used to endorse or promote products
  *       derived from this software without specific prior written permission.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
@@ -34,6 +34,13 @@
 #pragma once
 
 #include "../../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include "../../thread/thread_operators.cuh"
 #include "../../thread/thread_load.cuh"
 #include "../../thread/thread_store.cuh"
diff --git a/cub/cub/warp/warp_exchange.cuh b/cub/cub/warp/warp_exchange.cuh
index e863f67b103..9bb36bdfe6f 100644
--- a/cub/cub/warp/warp_exchange.cuh
+++ b/cub/cub/warp/warp_exchange.cuh
@@ -33,7 +33,14 @@
 
 #pragma once
 
-#include <cub/config.cuh>
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/util_ptx.cuh>
 #include <cub/util_type.cuh>
 #include <cub/warp/specializations/warp_exchange_shfl.cuh>
diff --git a/cub/cub/warp/warp_load.cuh b/cub/cub/warp/warp_load.cuh
index 424145588c2..4af6223c533 100644
--- a/cub/cub/warp/warp_load.cuh
+++ b/cub/cub/warp/warp_load.cuh
@@ -29,8 +29,15 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/block/block_load.cuh>
-#include <cub/config.cuh>
 #include <cub/iterator/cache_modified_input_iterator.cuh>
 #include <cub/util_ptx.cuh>
 #include <cub/util_type.cuh>
diff --git a/cub/cub/warp/warp_merge_sort.cuh b/cub/cub/warp/warp_merge_sort.cuh
index 3ad5dccd9c4..4da00d162c7 100644
--- a/cub/cub/warp/warp_merge_sort.cuh
+++ b/cub/cub/warp/warp_merge_sort.cuh
@@ -27,8 +27,15 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/block/block_merge_sort.cuh>
-#include <cub/config.cuh>
 #include <cub/util_ptx.cuh>
 #include <cub/util_type.cuh>
 
diff --git a/cub/cub/warp/warp_reduce.cuh b/cub/cub/warp/warp_reduce.cuh
index 2901f56bbcc..f00f13a5b06 100644
--- a/cub/cub/warp/warp_reduce.cuh
+++ b/cub/cub/warp/warp_reduce.cuh
@@ -34,7 +34,14 @@
 
 #pragma once
 
-#include <cub/config.cuh>
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/thread/thread_operators.cuh>
 #include <cub/util_type.cuh>
 #include <cub/warp/specializations/warp_reduce_shfl.cuh>
diff --git a/cub/cub/warp/warp_scan.cuh b/cub/cub/warp/warp_scan.cuh
index feff7ffe69d..80df80a2356 100644
--- a/cub/cub/warp/warp_scan.cuh
+++ b/cub/cub/warp/warp_scan.cuh
@@ -34,7 +34,14 @@
 
 #pragma once
 
-#include <cub/config.cuh>
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/thread/thread_operators.cuh>
 #include <cub/util_type.cuh>
 #include <cub/warp/specializations/warp_scan_shfl.cuh>
diff --git a/cub/cub/warp/warp_store.cuh b/cub/cub/warp/warp_store.cuh
index d40e63e65ee..f3564e4315e 100644
--- a/cub/cub/warp/warp_store.cuh
+++ b/cub/cub/warp/warp_store.cuh
@@ -29,8 +29,15 @@
 
 #pragma once
 
+#include "../config.cuh"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/block/block_store.cuh>
-#include <cub/config.cuh>
 #include <cub/util_ptx.cuh>
 #include <cub/util_type.cuh>
 #include <cub/warp/warp_exchange.cuh>
diff --git a/libcudacxx/.upstream-tests/utils/generate_feature_test_macro_components.py b/libcudacxx/.upstream-tests/utils/generate_feature_test_macro_components.py
index 2f028627ac5..a0530a2de50 100755
--- a/libcudacxx/.upstream-tests/utils/generate_feature_test_macro_components.py
+++ b/libcudacxx/.upstream-tests/utils/generate_feature_test_macro_components.py
@@ -722,9 +722,11 @@ def produce_version_header():
 
 #include <__config>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #if _LIBCUDACXX_STD_VER > 11
 {cxx14_macros}
diff --git a/libcudacxx/include/cuda/__cccl_config b/libcudacxx/include/cuda/__cccl_config
new file mode 100644
index 00000000000..50c16356e10
--- /dev/null
+++ b/libcudacxx/include/cuda/__cccl_config
@@ -0,0 +1,16 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+// SPDX-FileCopyrightText: Copyright (c) 2023 NVIDIA CORPORATION & AFFILIATES.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _CUDA__CCCL_CONFIG
+#define _CUDA__CCCL_CONFIG
+
+#include "std/detail/libcxx/include/__cccl_config"
+
+#endif // _CUDA__CCCL_CONFIG
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/CMakeLists.txt b/libcudacxx/include/cuda/std/detail/libcxx/include/CMakeLists.txt
index 0cbf21f272a..bc2ba9881a2 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/CMakeLists.txt
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/CMakeLists.txt
@@ -6,6 +6,7 @@ set(files
   __bsd_locale_defaults.h
   __bsd_locale_fallbacks.h
   __errc
+  __cccl_config
   __concepts/__concept_macros.h
   __concepts/_One_of.h
   __concepts/arithmetic.h
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__algorithm/swap_ranges.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__algorithm/swap_ranges.h
index fc315ba6285..52524acc387 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__algorithm/swap_ranges.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__algorithm/swap_ranges.h
@@ -16,9 +16,11 @@
 
 #include "../__utility/swap.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__assert b/libcudacxx/include/cuda/std/detail/libcxx/include/__assert
index 1d3eea00d84..7d8d58f3eb2 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__assert
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__assert
@@ -17,9 +17,11 @@
 
 #include "__verbose_abort"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 // This is for backwards compatibility with code that might have been enabling
 // assertions through the Debug mode previously.
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__availability b/libcudacxx/include/cuda/std/detail/libcxx/include/__availability
index ef401c005c6..b90ebf1f278 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__availability
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__availability
@@ -15,9 +15,11 @@
 #include <__config>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 // Libc++ is shipped by various vendors. In particular, it is used as a system
 // library on macOS, iOS and other Apple platforms. In order for users to be
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__bit_reference b/libcudacxx/include/cuda/std/detail/libcxx/include/__bit_reference
index 6fe85cb956a..84411dfacda 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__bit_reference
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__bit_reference
@@ -14,9 +14,11 @@
 #include <bit>
 #include <algorithm>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_PUSH_MACROS
 #include <__undef_macros>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__bsd_locale_defaults.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__bsd_locale_defaults.h
index 609865abd16..71ed40cff60 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__bsd_locale_defaults.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__bsd_locale_defaults.h
@@ -14,9 +14,11 @@
 #ifndef _LIBCUDACXX_BSD_LOCALE_DEFAULTS_H
 #define _LIBCUDACXX_BSD_LOCALE_DEFAULTS_H
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #define __libcpp_mb_cur_max_l(loc)                          MB_CUR_MAX_L(loc)
 #define __libcpp_btowc_l(ch, loc)                           btowc_l(ch, loc)
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__bsd_locale_fallbacks.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__bsd_locale_fallbacks.h
index 58220977d76..128ced77337 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__bsd_locale_fallbacks.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__bsd_locale_fallbacks.h
@@ -17,9 +17,11 @@
 #include <stdarg.h>
 #include <memory>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__cccl/visibility.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__cccl/visibility.h
new file mode 100644
index 00000000000..b8e9d337414
--- /dev/null
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__cccl/visibility.h
@@ -0,0 +1,46 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+// SPDX-FileCopyrightText: Copyright (c) 2023 NVIDIA CORPORATION & AFFILIATES.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef __CCCL_VISIBILITY_H
+#define __CCCL_VISIBILITY_H
+
+#ifndef __CCCL_CONFIG
+#error "<__cccl/visibility.h> should only be included in from <cuda/__cccl_config>"
+#endif // __CCCL_CONFIG
+
+// We want to ensure that all warning emmiting from this header are supressed
+// We define cub and thrust kernels as hidden. However, this triggers errors about missing external linkage iff the
+// definition of the _CCCL_ATTRIBUTE_HIDDEN macro is not in a system header :shrug:
+#if defined(_CCCL_COMPILER_NVHPC)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_FORCE_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
+// For unknown reasons, nvc++ need to selectively disable this warning
+// We do not want to use our usual macro because that would have push / pop semantics
+#if defined(_CCCL_COMPILER_NVHPC)
+#pragma nv_diag_suppress 1407
+#endif // _CCCL_COMPILER_NVHPC
+
+// Enable us to hide kernels
+#if defined(_CCCL_COMPILER_MSVC)
+#  define _CCCL_ATTRIBUTE_HIDDEN
+#elif defined(_CCCL_COMPILER_NVRTC)
+#  define _CCCL_ATTRIBUTE_HIDDEN
+#else // ^^^ _CCCL_COMPILER_NVRTC ^^^ / vvv _CCCL_COMPILER_NVRTC vvv
+#  define _CCCL_ATTRIBUTE_HIDDEN __attribute__ ((__visibility__("hidden")))
+#endif // !_CCCL_COMPILER_NVRTC
+
+#if !defined(CCCL_DETAIL_KERNEL_ATTRIBUTES)
+#  define CCCL_DETAIL_KERNEL_ATTRIBUTES __global__ _CCCL_ATTRIBUTE_HIDDEN
+#endif // !CCCL_DETAIL_KERNEL_ATTRIBUTES
+
+#endif // __CCCL_VISIBILITY_H
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__cccl_config b/libcudacxx/include/cuda/std/detail/libcxx/include/__cccl_config
new file mode 100644
index 00000000000..49cafa2db2e
--- /dev/null
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__cccl_config
@@ -0,0 +1,166 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+// SPDX-FileCopyrightText: Copyright (c) 2023 NVIDIA CORPORATION & AFFILIATES.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef __CCCL_CONFIG
+#define __CCCL_CONFIG
+
+// Determine the host compiler
+#if defined(__NVCOMPILER)
+#  define _CCCL_COMPILER_NVHPC
+#elif defined(__clang__)
+#  define _CCCL_COMPILER_CLANG
+#elif defined(__GNUC__)
+#  define _CCCL_COMPILER_GCC
+#elif defined(_MSC_VER)
+#  define _CCCL_COMPILER_MSVC
+#elif defined(__IBMCPP__)
+#  define _CCCL_COMPILER_IBM
+#elif defined(__CUDACC_RTC__)
+#  define _CCCL_COMPILER_NVRTC
+#endif
+
+// Convenient shortcut to determine which version of MSVC we are dealing with
+#if defined(_CCCL_COMPILER_MSVC)
+#  if _MSC_VER < 1917
+#    define _CCCL_COMPILER_MSVC_2017
+#  elif _MSC_VER < 1930
+#    define _CCCL_COMPILER_MSVC_2019
+#  else // _MSC_VER < 1940
+#    define _CCCL_COMPILER_MSVC_2022
+#  endif // _MSC_VER < 1940
+#endif // _CCCL_COMPILER_MSVC
+
+// Determine the cuda compiler
+#if defined(__NVCC__)
+#  define _CCCL_CUDA_COMPILER_NVCC
+#elif defined(_NVHPC_CUDA)
+#  define _CCCL_CUDA_COMPILER_NVHPC
+#elif defined(__CUDA__) && defined(_CCCL_COMPILER_CLANG)
+#  define _CCCL_CUDA_COMPILER_CLANG
+#endif
+
+// Enforce that cccl headers are treated as system headers
+#if defined(_CCCL_COMPILER_GCC)
+#  define _CCCL_FORCE_SYSTEM_HEADER _Pragma("GCC system_header")
+#elif defined(_CCCL_COMPILER_CLANG)
+#  define _CCCL_FORCE_SYSTEM_HEADER _Pragma("clang system_header")
+#elif defined(_CCCL_COMPILER_MSVC)
+#  define _CCCL_FORCE_SYSTEM_HEADER __pragma(system_header)
+#elif defined(_CCCL_COMPILER_NVHPC)
+#  define _CCCL_FORCE_SYSTEM_HEADER _Pragma("GCC system_header")
+#else
+#  define _CCCL_FORCE_SYSTEM_HEADER
+#endif
+
+// Potentially enable that cccl headers are treated as system headers
+#if defined(_CCCL_NO_SYSTEM_HEADER)
+#  define _CCCL_IMPLICIT_SYSTEM_HEADER
+#elif defined(_CCCL_COMPILER_MSVC) \
+  && defined(_LIBCUDACXX_DISABLE_PRAGMA_MSVC_WARNING)
+#  define _CCCL_IMPLICIT_SYSTEM_HEADER
+#elif defined(_CCCL_COMPILER_NVRTC)
+#  define _CCCL_IMPLICIT_SYSTEM_HEADER
+#elif defined(_LIBCUDACXX_DISABLE_PRAGMA_GCC_SYSTEM_HEADER)
+#  define _CCCL_IMPLICIT_SYSTEM_HEADER
+#else
+#  define _CCCL_USE_IMPLICIT_SYSTEM_DEADER
+#  define _CCCL_IMPLICIT_SYSTEM_HEADER _CCCL_FORCE_SYSTEM_HEADER
+#endif
+
+// We want to ensure that all warning emmiting from this header are supressed
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
+// Enable us to selectively silence host compiler warnings
+#define _CCCL_TOSTRING2(_STR) #_STR
+#define _CCCL_TOSTRING(_STR) _CCCL_TOSTRING2(_STR)
+#ifdef _CCCL_COMPILER_CLANG
+#  define _CCCL_DIAG_PUSH _Pragma("clang diagnostic push")
+#  define _CCCL_DIAG_POP  _Pragma("clang diagnostic pop")
+#  define _CCCL_DIAG_SUPPRESS_CLANG(str) _Pragma(_CCCL_TOSTRING(clang diagnostic ignored str))
+#  define _CCCL_DIAG_SUPPRESS_GCC(str)
+#  define _CCCL_DIAG_SUPPRESS_NVHPC(str)
+#elif defined(_CCCL_COMPILER_GCC)
+#  define _CCCL_DIAG_PUSH _Pragma("GCC diagnostic push")
+#  define _CCCL_DIAG_POP  _Pragma("GCC diagnostic pop")
+#  define _CCCL_DIAG_SUPPRESS_CLANG(str)
+#  define _CCCL_DIAG_SUPPRESS_GCC(str) _Pragma(_CCCL_TOSTRING(GCC diagnostic ignored str))
+#  define _CCCL_DIAG_SUPPRESS_NVHPC(str)
+#elif defined(_CCCL_COMPILER_NVHPC)
+#  define _CCCL_DIAG_PUSH _Pragma("diagnostic push")
+#  define _CCCL_DIAG_POP _Pragma("diagnostic pop")
+#  define _CCCL_DIAG_SUPPRESS_CLANG(str)
+#  define _CCCL_DIAG_SUPPRESS_GCC(str)
+#  define _CCCL_DIAG_SUPPRESS_NVHPC(str) _Pragma(_LIBCUDACXX_TOSTRING(diag_suppress str))
+#else
+#  define _CCCL_DIAG_PUSH
+#  define _CCCL_DIAG_POP
+#  define _CCCL_DIAG_SUPPRESS_CLANG(str)
+#  define _CCCL_DIAG_SUPPRESS_GCC(str)
+#  define _CCCL_DIAG_SUPPRESS_NVHPC(str)
+#endif
+
+// Convenient shortcuts to silence common warnings
+#if defined(_CCCL_COMPILER_CLANG)
+#  define _CCCL_DIAG_SUPPRESS_DEPRECATED_PUSH _CCCL_DIAG_PUSH                                       \
+                                              _CCCL_DIAG_SUPPRESS_CLANG("-Wdeprecated")             \
+                                              _CCCL_DIAG_SUPPRESS_CLANG("-Wdeprecated-declarations")
+#  define _CCCL_DIAG_SUPPRESS_DEPRECATED_POP  _CCCL_DIAG_POP
+#elif defined(_CCCL_COMPILER_GCC)
+#  define _CCCL_DIAG_SUPPRESS_DEPRECATED_PUSH _CCCL_DIAG_PUSH                                     \
+                                              _CCCL_DIAG_SUPPRESS_GCC("-Wdeprecated")             \
+                                              _CCCL_DIAG_SUPPRESS_GCC("-Wdeprecated-declarations")
+#  define _CCCL_DIAG_SUPPRESS_DEPRECATED_POP  _CCCL_DIAG_POP
+#else // !_CCCL_COMPILER_CLANG && !_CCCL_COMPILER_GCC
+#  define _CCCL_DIAG_SUPPRESS_DEPRECATED_PUSH
+#  define _CCCL_DIAG_SUPPRESS_DEPRECATED_POP
+#endif // !_CCCL_COMPILER_CLANG && !_CCCL_COMPILER_GCC
+
+// Enable us to selectively silence cuda compiler warnings
+#if defined(__CUDACC__)
+#  if !defined(_CCCL_CUDA_COMPILER_CLANG)
+#    define _CCCL_NV_DIAG_SUPPRESS(_WARNING)
+#    define _CCCL_NV_DIAG_DEFAULT(_WARNING)
+#  elif defined(__NVCC_DIAG_PRAGMA_SUPPORT__)
+#    if defined(_CCCL_COMPILER_MSVC)
+#      define _CCCL_NV_DIAG_SUPPRESS(_WARNING) __pragma(_CCCL_TOSTRING(nv_diag_suppress _WARNING))
+#      define _CCCL_NV_DIAG_DEFAULT(_WARNING)  __pragma(_CCCL_TOSTRING(nv_diag_default _WARNING))
+#    else // ^^^ _CCCL_COMPILER_MSVC ^^^ / vvv !_CCCL_COMPILER_MSVC vvv
+#      define _CCCL_NV_DIAG_SUPPRESS(_WARNING) _Pragma(_CCCL_TOSTRING(nv_diagnostic push)) \
+                                               _Pragma(_CCCL_TOSTRING(nv_diag_suppress _WARNING))
+#      define _CCCL_NV_DIAG_DEFAULT(_WARNING)  _Pragma(_CCCL_TOSTRING(nv_diagnostic pop))
+#    endif // !_CCCL_COMPILER_MSVC
+#  elif defined(_CCCL_COMPILER_NVHPC)
+#    define _CCCL_NV_DIAG_SUPPRESS(_WARNING) _Pragma(_CCCL_TOSTRING(diagnostic push)) \
+                                             _Pragma(_CCCL_TOSTRING(diag_suppress _WARNING))
+#    define _CCCL_NV_DIAG_DEFAULT(_WARNING)  _Pragma(_CCCL_TOSTRING(diagnostic pop))
+#  else // ^^^ __NVCC_DIAG_PRAGMA_SUPPORT__ ^^^ / vvv !__NVCC_DIAG_PRAGMA_SUPPORT__ vvv
+#    if defined(_CCCL_COMPILER_MSVC_2017) // MSVC 2017 has issues with restoring the warning
+#      define _CCCL_NV_DIAG_SUPPRESS(_WARNING) __pragma(_CCCL_TOSTRING(diag_suppress _WARNING))
+#      define _CCCL_NV_DIAG_DEFAULT(_WARNING)
+#    elif defined(_CCCL_COMPILER_MSVC)
+#      define _CCCL_NV_DIAG_SUPPRESS(_WARNING) __pragma(_CCCL_TOSTRING(diag_suppress _WARNING))
+#      define _CCCL_NV_DIAG_DEFAULT(_WARNING)  __pragma(_CCCL_TOSTRING(diag_default _WARNING))
+#    else // ^^^ _CCCL_COMPILER_MSVC ^^^ / vvv !_CCCL_COMPILER_MSVC vvv
+#      define _CCCL_NV_DIAG_SUPPRESS(_WARNING) _Pragma(_CCCL_TOSTRING(diag_suppress _WARNING))
+#      define _CCCL_NV_DIAG_DEFAULT(_WARNING)  _Pragma(_CCCL_TOSTRING(diag_default _WARNING))
+#    endif // !_CCCL_COMPILER_MSVC
+#  endif // !__NVCC_DIAG_PRAGMA_SUPPORT__
+#else // ^^^ __CUDACC__ ^^^ / vvv !__CUDACC__ vvv
+#  define _CCCL_NV_DIAG_SUPPRESS(_WARNING)
+#  define _CCCL_NV_DIAG_DEFAULT(_WARNING)
+#endif // other compilers
+
+#include "__cccl/visibility.h"
+
+#endif // __CCCL_CONFIG
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/_One_of.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/_One_of.h
index 3b9f9976017..aabfa399425 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/_One_of.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/_One_of.h
@@ -18,9 +18,11 @@
 #include "../__type_traits/disjunction.h"
 #include "../__type_traits/is_same.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/__concept_macros.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/__concept_macros.h
index fb95cbf4552..347e2ae9b5a 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/__concept_macros.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/__concept_macros.h
@@ -18,6 +18,12 @@
 #include <__config>
 #endif //__cuda_std__
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if _LIBCUDACXX_STD_VER > 11
 
 #define _LIBCUDACXX_PP_CAT_(_Xp, ...) _Xp##__VA_ARGS__
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/arithmetic.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/arithmetic.h
index eb0030bdf12..a05c05b0183 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/arithmetic.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/arithmetic.h
@@ -22,9 +22,11 @@
 #include "../__type_traits/is_signed.h"
 #include "../__type_traits/is_unsigned_integer.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/assignable.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/assignable.h
index b942a2bb0ef..bb5a0e3aadf 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/assignable.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/assignable.h
@@ -21,9 +21,11 @@
 #include "../__type_traits/make_const_lvalue_ref.h"
 #include "../__utility/forward.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/boolean_testable.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/boolean_testable.h
index d1fa4ec7861..6ed6b8246c3 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/boolean_testable.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/boolean_testable.h
@@ -18,9 +18,11 @@
 #include "../__concepts/convertible_to.h"
 #include "../__utility/forward.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/class_or_enum.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/class_or_enum.h
index 6e6b2a5cc49..026e99af01e 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/class_or_enum.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/class_or_enum.h
@@ -20,9 +20,11 @@
 #include "../__type_traits/is_union.h"
 #include "../__type_traits/remove_cvref.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/common_reference_with.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/common_reference_with.h
index d0f8ef6d127..777a50f12ca 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/common_reference_with.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/common_reference_with.h
@@ -21,9 +21,11 @@
 #include "../__type_traits/copy_cv.h"
 #include "../__type_traits/copy_cvref.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/common_with.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/common_with.h
index de51e0398dc..4ad58c4f20c 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/common_with.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/common_with.h
@@ -21,9 +21,11 @@
 #include "../__type_traits/common_type.h"
 #include "../__utility/declval.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/constructible.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/constructible.h
index a0f33cb27d0..0443ebab28f 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/constructible.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/constructible.h
@@ -20,9 +20,11 @@
 #include "../__type_traits/add_lvalue_reference.h"
 #include "../__type_traits/is_constructible.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/convertible_to.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/convertible_to.h
index b2a869dcd7c..85a1eea229d 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/convertible_to.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/convertible_to.h
@@ -18,9 +18,11 @@
 #include "../__type_traits/is_convertible.h"
 #include "../__utility/declval.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/copyable.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/copyable.h
index cd309f71e84..15ad7790030 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/copyable.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/copyable.h
@@ -19,9 +19,11 @@
 #include "../__concepts/constructible.h"
 #include "../__concepts/movable.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/derived_from.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/derived_from.h
index 1e92e5f2f13..571430b5341 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/derived_from.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/derived_from.h
@@ -19,9 +19,11 @@
 #include "../__type_traits/is_base_of.h"
 #include "../__type_traits/is_convertible.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/destructible.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/destructible.h
index 6a0ad1bff44..6e24a7bb899 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/destructible.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/destructible.h
@@ -22,9 +22,11 @@
 #include "../__type_traits/void_t.h"
 #include "../__utility/declval.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/different_from.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/different_from.h
index c4b146ded36..a99effd2e77 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/different_from.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/different_from.h
@@ -18,9 +18,11 @@
 #include "../__concepts/same_as.h"
 #include "../__type_traits/remove_cvref.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/equality_comparable.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/equality_comparable.h
index 821e6f4abcd..cfe374fce77 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/equality_comparable.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/equality_comparable.h
@@ -20,9 +20,11 @@
 #include "../__type_traits/common_reference.h"
 #include "../__type_traits/make_const_lvalue_ref.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/invocable.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/invocable.h
index 44be8f5164d..227a9c2694b 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/invocable.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/invocable.h
@@ -18,9 +18,11 @@
 #include "../__functional/invoke.h"
 #include "../__utility/forward.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/movable.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/movable.h
index c72143f0c85..28d01625e4a 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/movable.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/movable.h
@@ -20,9 +20,11 @@
 #include "../__concepts/swappable.h"
 #include "../__type_traits/is_object.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/predicate.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/predicate.h
index 946f7993d78..b88aab51641 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/predicate.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/predicate.h
@@ -19,9 +19,11 @@
 #include "../__concepts/invocable.h"
 #include "../__functional/invoke.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/regular.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/regular.h
index 118e88c3018..05979362388 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/regular.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/regular.h
@@ -18,9 +18,11 @@
 #include "../__concepts/equality_comparable.h"
 #include "../__concepts/semiregular.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/relation.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/relation.h
index 5e46686cdcb..c3ed2e90dc3 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/relation.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/relation.h
@@ -17,9 +17,11 @@
 #include "../__concepts/__concept_macros.h"
 #include "../__concepts/predicate.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/same_as.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/same_as.h
index 039b09e2dad..0b4668a86a1 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/same_as.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/same_as.h
@@ -17,9 +17,11 @@
 #include "../__concepts/__concept_macros.h"
 #include "../__type_traits/is_same.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/semiregular.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/semiregular.h
index 22663849435..46c3bca6d4d 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/semiregular.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/semiregular.h
@@ -18,9 +18,11 @@
 #include "../__concepts/constructible.h"
 #include "../__concepts/copyable.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/swappable.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/swappable.h
index aa34fd55380..b5c1f1b666b 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/swappable.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/swappable.h
@@ -31,9 +31,11 @@
 #include "../__utility/forward.h"
 #include "../__utility/move.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #if defined(_LIBCUDACXX_COMPILER_MSVC)
 _LIBCUDACXX_NV_DIAG_SUPPRESS(461) // nonstandard cast to array type ignored
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/totally_ordered.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/totally_ordered.h
index 73b5e9991a8..d1973c7c80d 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/totally_ordered.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__concepts/totally_ordered.h
@@ -20,9 +20,11 @@
 #include "../__type_traits/common_reference.h"
 #include "../__type_traits/make_const_lvalue_ref.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__config b/libcudacxx/include/cuda/std/detail/libcxx/include/__config
index af8b95be7be..f71aa623243 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__config
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__config
@@ -10,6 +10,14 @@
 #ifndef _LIBCUDACXX_CONFIG
 #define _LIBCUDACXX_CONFIG
 
+#include "__cccl_config"
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if defined(__NVCOMPILER)
 #  define _LIBCUDACXX_COMPILER_NVHPC
 #elif defined(__clang__)
@@ -76,20 +84,12 @@
 #define _LIBCUDACXX_CUDACC_BELOW_12_3
 #endif // defined(_LIBCUDACXX_CUDACC) && _LIBCUDACXX_CUDACC_VER < 1203000
 
-#if defined(_MSC_VER) && !defined(__clang__)
+#if defined(_LIBCUDACXX_COMPILER_MSVC)
 #  define _LIBCUDACXX_HAS_PRAGMA_MSVC_WARNING
 #  if !defined(_LIBCUDACXX_DISABLE_PRAGMA_MSVC_WARNING)
 #    define _LIBCUDACXX_USE_PRAGMA_MSVC_WARNING
-#  endif
-#else
-#  if !defined(_LIBCUDACXX_DISABLE_PRAGMA_GCC_SYSTEM_HEADER) && !defined(_LIBCUDACXX_COMPILER_NVRTC)
-#    define _LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER
-#  endif
-#endif
-
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
-#pragma GCC system_header
-#endif
+#  endif // !_LIBCUDACXX_DISABLE_PRAGMA_MSVC_WARNING
+#endif // _LIBCUDACXX_COMPILER_MSVC
 
 #ifdef __cplusplus
 
@@ -2245,7 +2245,7 @@ extern "C" _LIBCUDACXX_FUNC_VIS void __sanitizer_annotate_contiguous_container(
 #endif
 
 #if !defined(_LIBCUDACXX_DISABLE_EXEC_CHECK)
-#if defined(_LIBCUDACXX_CUDACC)                       \
+#if defined(_LIBCUDACXX_CUDACC)               \
  && !defined(_LIBCUDACXX_COMPILER_NVRTC)      \
  && !defined(_LIBCUDACXX_COMPILER_NVHPC_CUDA) \
  && !defined(_LIBCUDACXX_COMPILER_CLANG_CUDA)
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__cuda/atomic.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__cuda/atomic.h
index 20249e8ffcd..9e53acce011 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__cuda/atomic.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__cuda/atomic.h
@@ -15,9 +15,11 @@
 #error "<__cuda/atomic> should only be included in from <cuda/std/atomic>"
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_CUDA
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__cuda/barrier.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__cuda/barrier.h
index bc7d992902f..115e57a27d9 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__cuda/barrier.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__cuda/barrier.h
@@ -19,9 +19,11 @@
 #  error "CUDA synchronization primitives are only supported for sm_70 and up."
 #endif
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #include "../cstdlib"                // _LIBCUDACXX_UNREACHABLE
 #include "../__type_traits/void_t.h" // _CUDA_VSTD::__void_t
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__cuda/chrono.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__cuda/chrono.h
index 346bf146edd..225123ff80c 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__cuda/chrono.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__cuda/chrono.h
@@ -17,9 +17,11 @@
 
 #include <nv/target>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__cuda/semaphore.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__cuda/semaphore.h
index 25624f9ba74..0e7b9ae95c8 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__cuda/semaphore.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__cuda/semaphore.h
@@ -15,9 +15,11 @@
 #error "<__cuda/semaphore> should only be included in from <cuda/std/semaphore>"
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_CUDA
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__debug b/libcudacxx/include/cuda/std/detail/libcxx/include/__debug
index 450ca1b2bc6..70f25f4241f 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__debug
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__debug
@@ -19,9 +19,11 @@
 #include "__type_traits/is_constant_evaluated.h"
 #include "cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #if defined(_LIBCUDACXX_ENABLE_DEBUG_MODE) && !defined(_LIBCUDACXX_DEBUG_RANDOMIZE_UNSPECIFIED_STABILITY)
 # define _LIBCUDACXX_DEBUG_RANDOMIZE_UNSPECIFIED_STABILITY
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__errc b/libcudacxx/include/cuda/std/detail/libcxx/include/__errc
index 03382fa8650..1143fcf7460 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__errc
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__errc
@@ -103,9 +103,11 @@ enum class errc
 #include <__config>
 #include <cerrno>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__expected/bad_expected_access.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__expected/bad_expected_access.h
index fe5dc6e50b7..82880af230a 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__expected/bad_expected_access.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__expected/bad_expected_access.h
@@ -16,9 +16,11 @@
 #include "../__utility/move.h"
 #include "../exception"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #if _LIBCUDACXX_STD_VER > 11
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__expected/expected.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__expected/expected.h
index 07f11fc702b..39fce36d6da 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__expected/expected.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__expected/expected.h
@@ -60,9 +60,11 @@
 #include "../cstdlib"
 #include "../initializer_list"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #if _LIBCUDACXX_STD_VER > 11
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__expected/expected_base.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__expected/expected_base.h
index b1744a258f9..26e911667d5 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__expected/expected_base.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__expected/expected_base.h
@@ -45,9 +45,11 @@
 #include "../__utility/in_place.h"
 #include "../__utility/move.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #if _LIBCUDACXX_STD_VER > 11
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__expected/unexpect.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__expected/unexpect.h
index 7dd8c52ebbc..f4a55f5bdd2 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__expected/unexpect.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__expected/unexpect.h
@@ -13,9 +13,11 @@
 #include <__config>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #if _LIBCUDACXX_STD_VER > 11
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__expected/unexpected.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__expected/unexpected.h
index afefdb4ab51..d8529a04344 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__expected/unexpected.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__expected/unexpected.h
@@ -31,9 +31,11 @@
 
 #include "../initializer_list"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/binary_function.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/binary_function.h
index e476e68c076..25751f5be0b 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/binary_function.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/binary_function.h
@@ -14,9 +14,11 @@
 #include <__config>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
@@ -43,6 +45,7 @@ template <class _Arg1, class _Arg2, class _Result> struct __binary_function_keep
 #if _LIBCUDACXX_STD_VER <= 14 || defined(_LIBCUDACXX_ENABLE_CXX17_REMOVED_UNARY_BINARY_FUNCTION)
 _LIBCUDACXX_DIAGNOSTIC_PUSH
 _LIBCUDACXX_CLANG_DIAGNOSTIC_IGNORED("-Wdeprecated-declarations")
+_LIBCUDACXX_GCC_DIAGNOSTIC_IGNORED("-Wdeprecated-declarations")
 template <class _Arg1, class _Arg2, class _Result>
 using __binary_function = binary_function<_Arg1, _Arg2, _Result>;
 _LIBCUDACXX_DIAGNOSTIC_POP
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/binary_negate.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/binary_negate.h
index 3adaaa2dd01..96d72a5515d 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/binary_negate.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/binary_negate.h
@@ -17,9 +17,11 @@
 
 #include "../__functional/binary_function.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/bind.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/bind.h
index c90edb81439..1d7f9d38125 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/bind.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/bind.h
@@ -41,9 +41,11 @@
 #include "../cstddef"
 #include "../tuple"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/bind_back.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/bind_back.h
index a9a4a889664..9d7b7eddc1d 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/bind_back.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/bind_back.h
@@ -28,9 +28,11 @@
 
 #include "../tuple"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/bind_front.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/bind_front.h
index c73800b366e..0ee3921c118 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/bind_front.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/bind_front.h
@@ -25,9 +25,11 @@
 #include "../__type_traits/is_nothrow_constructible.h"
 #include "../__utility/forward.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/binder1st.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/binder1st.h
index 094b92cd972..702187d0197 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/binder1st.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/binder1st.h
@@ -17,9 +17,11 @@
 
 #include "../__functional/unary_function.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/binder2nd.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/binder2nd.h
index e02ce01a771..f8238309f7c 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/binder2nd.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/binder2nd.h
@@ -17,9 +17,11 @@
 
 #include "../__functional/unary_function.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/compose.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/compose.h
index b1ab0445856..6a61be4e4ee 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/compose.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/compose.h
@@ -20,9 +20,11 @@
 #include "../__type_traits/decay.h"
 #include "../__utility/forward.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/default_searcher.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/default_searcher.h
index 3b76674073f..1e7a5cd7b2b 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/default_searcher.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/default_searcher.h
@@ -21,9 +21,11 @@
 #include "../__iterator/iterator_traits.h"
 #include "../__utility/pair.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/function.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/function.h
index c031d7dd2e7..14707ceb605 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/function.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/function.h
@@ -44,9 +44,11 @@
 #include "../__utility/swap.h"
 #include "../tuple"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/hash.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/hash.h
index fc9e1d66fe0..4ed859a0379 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/hash.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/hash.h
@@ -31,9 +31,11 @@
 #include "../__utility/swap.h"
 #include "../cstdint"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #ifndef __cuda_std__
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/identity.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/identity.h
index 9e2a532bb25..b1a2d4116ce 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/identity.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/identity.h
@@ -17,9 +17,11 @@
 
 #include "../__utility/forward.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/invoke.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/invoke.h
index b4d327d924a..95ebdae81f6 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/invoke.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/invoke.h
@@ -33,9 +33,11 @@
 #include "../__utility/declval.h"
 #include "../__utility/forward.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 // TODO: Disentangle the type traits and _CUDA_VSTD::invoke properly
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/is_transparent.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/is_transparent.h
index 997718aa72f..42b512ef1ec 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/is_transparent.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/is_transparent.h
@@ -18,9 +18,11 @@
 #include "../__type_traits/integral_constant.h"
 #include "../__type_traits/void_t.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/mem_fn.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/mem_fn.h
index 6f51ca939f4..6f9674e5b1c 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/mem_fn.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/mem_fn.h
@@ -20,9 +20,11 @@
 #include "../__functional/weak_result_type.h"
 #include "../__utility/forward.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/mem_fun_ref.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/mem_fun_ref.h
index 105adbff6e4..ad2e6ec1146 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/mem_fun_ref.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/mem_fun_ref.h
@@ -18,9 +18,11 @@
 #include "../__functional/binary_function.h"
 #include "../__functional/unary_function.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/not_fn.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/not_fn.h
index d36a7d6cb73..31e48c544ec 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/not_fn.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/not_fn.h
@@ -23,9 +23,11 @@
 #include "../__type_traits/is_move_constructible.h"
 #include "../__utility/forward.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/operations.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/operations.h
index a4400bf43d1..5c35a84a307 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/operations.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/operations.h
@@ -19,9 +19,11 @@
 #include "../__functional/unary_function.h"
 #include "../__utility/forward.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/perfect_forward.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/perfect_forward.h
index 6e20b2f79f7..a84212a0cb2 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/perfect_forward.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/perfect_forward.h
@@ -27,9 +27,11 @@
 
 #include "../tuple"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/pointer_to_binary_function.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/pointer_to_binary_function.h
index d76f447e4b4..72c46a411a4 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/pointer_to_binary_function.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/pointer_to_binary_function.h
@@ -17,9 +17,11 @@
 
 #include "../__functional/binary_function.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/pointer_to_unary_function.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/pointer_to_unary_function.h
index eb87609f8b6..a528d8c5c2b 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/pointer_to_unary_function.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/pointer_to_unary_function.h
@@ -17,9 +17,11 @@
 
 #include "../__functional/unary_function.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/ranges_operations.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/ranges_operations.h
index 172522b891e..210ce402a3e 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/ranges_operations.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/ranges_operations.h
@@ -18,9 +18,11 @@
 #include "../__concepts/totally_ordered.h"
 #include "../__utility/forward.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #if _LIBCUDACXX_STD_VER > 14
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/reference_wrapper.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/reference_wrapper.h
index a9cb9a37da5..c8e3c0f7e7e 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/reference_wrapper.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/reference_wrapper.h
@@ -17,9 +17,11 @@
 #include "../__functional/weak_result_type.h"
 #include "../__memory/addressof.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/unary_function.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/unary_function.h
index f6b27e71a00..6545c451f33 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/unary_function.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/unary_function.h
@@ -13,9 +13,11 @@
 #include <__config>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
@@ -40,6 +42,7 @@ template <class _Arg, class _Result> struct __unary_function_keep_layout_base {
 #if _LIBCUDACXX_STD_VER <= 14 || defined(_LIBCUDACXX_ENABLE_CXX17_REMOVED_UNARY_BINARY_FUNCTION)
 _LIBCUDACXX_DIAGNOSTIC_PUSH
 _LIBCUDACXX_CLANG_DIAGNOSTIC_IGNORED("-Wdeprecated-declarations")
+_LIBCUDACXX_GCC_DIAGNOSTIC_IGNORED("-Wdeprecated-declarations")
 template <class _Arg, class _Result>
 using __unary_function = unary_function<_Arg, _Result>;
 _LIBCUDACXX_DIAGNOSTIC_POP
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/unary_negate.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/unary_negate.h
index 601c2728229..2a6adfd5e97 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/unary_negate.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/unary_negate.h
@@ -17,9 +17,11 @@
 
 #include "../__functional/unary_function.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/unwrap_ref.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/unwrap_ref.h
index 291de09f111..c0f522ede1a 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/unwrap_ref.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/unwrap_ref.h
@@ -14,9 +14,11 @@
 #include <__config>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/weak_result_type.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/weak_result_type.h
index 1b131530677..911263b8331 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/weak_result_type.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional/weak_result_type.h
@@ -21,9 +21,11 @@
 #include "../__type_traits/is_same.h"
 #include "../__utility/declval.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional_base b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional_base
index 1c35dd55dfa..dc28633036e 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__functional_base
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__functional_base
@@ -32,9 +32,11 @@
 #include <__pragma_push>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/array.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/array.h
index af3240d40bb..9161771b523 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/array.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/array.h
@@ -16,9 +16,11 @@
 
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/get.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/get.h
index 69b3a53fe13..2d68f304009 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/get.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/get.h
@@ -20,9 +20,11 @@
 #include "../__tuple_dir/tuple_element.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
-#  pragma GCC system_header
-#endif
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/hash.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/hash.h
index 67cef524e45..16df8ee9d6a 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/hash.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/hash.h
@@ -14,9 +14,11 @@
 #include <__config>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/memory_resource.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/memory_resource.h
index fe6b74d38ac..130da942a52 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/memory_resource.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/memory_resource.h
@@ -14,9 +14,11 @@
 #include <__config>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/pair.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/pair.h
index eeb3b0a9bc2..7d19a66520e 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/pair.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/pair.h
@@ -14,9 +14,11 @@
 #include <__config>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/span.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/span.h
index 10d19205499..06b3a667e98 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/span.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/span.h
@@ -17,9 +17,11 @@
 
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/string.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/string.h
index a24c849692e..5a408b7df5f 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/string.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/string.h
@@ -16,9 +16,11 @@
 
 #include "../__fwd/memory_resource.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/string_view.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/string_view.h
index e82026ddd74..670de6771c2 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/string_view.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/string_view.h
@@ -17,9 +17,11 @@
 
 #include "../__fwd/string.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/tuple.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/tuple.h
index 981325e8a38..7fbf47e9174 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/tuple.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__fwd/tuple.h
@@ -14,9 +14,11 @@
 #include <__config>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__hash_table b/libcudacxx/include/cuda/std/detail/libcxx/include/__hash_table
index ce7c46cf6f6..899800339c5 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__hash_table
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__hash_table
@@ -22,9 +22,11 @@
 #include "__assert" // all public C++ headers provide the assertion handler
 #include "__debug"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_PUSH_MACROS
 #include <__undef_macros>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/access.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/access.h
index 946b20e03e1..248ec11f575 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/access.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/access.h
@@ -17,9 +17,11 @@
 
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/advance.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/advance.h
index e8e52742cb0..a9257b833a9 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/advance.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/advance.h
@@ -19,9 +19,11 @@
 #include "../__iterator/iterator_traits.h"
 #include "../cstdlib"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/back_insert_iterator.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/back_insert_iterator.h
index 455a344305b..8d62a4e9b32 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/back_insert_iterator.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/back_insert_iterator.h
@@ -21,9 +21,11 @@
 #include "../__utility/move.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/bounded_iter.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/bounded_iter.h
index 924279e4c96..adb69566e8e 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/bounded_iter.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/bounded_iter.h
@@ -22,9 +22,11 @@
 #include "../__type_traits/is_convertible.h"
 #include "../__utility/move.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/concepts.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/concepts.h
index 2cf3bd4fdd6..7b0cee8406c 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/concepts.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/concepts.h
@@ -45,9 +45,11 @@
 #include "../__type_traits/remove_cvref.h"
 #include "../__type_traits/void_t.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/data.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/data.h
index 433a56478a8..4d2bda2bed7 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/data.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/data.h
@@ -18,9 +18,11 @@
 #include "../cstddef"
 #include "../initializer_list"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/default_sentinel.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/default_sentinel.h
index 0ee17a956c3..f1859e1a413 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/default_sentinel.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/default_sentinel.h
@@ -15,9 +15,11 @@
 #include <__config>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/distance.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/distance.h
index e0b0cad4a79..d4cd2de3128 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/distance.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/distance.h
@@ -17,9 +17,11 @@
 
 #include "../__iterator/iterator_traits.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/empty.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/empty.h
index 3aebeafafb5..7a2a655f80d 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/empty.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/empty.h
@@ -18,9 +18,11 @@
 #include "../cstddef"
 #include "../initializer_list"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/erase_if_container.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/erase_if_container.h
index ec26718362c..31160203a6c 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/erase_if_container.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/erase_if_container.h
@@ -15,9 +15,11 @@
 #include <__config>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/front_insert_iterator.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/front_insert_iterator.h
index 873ae3686aa..0fdde41a2db 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/front_insert_iterator.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/front_insert_iterator.h
@@ -21,9 +21,11 @@
 #include "../__utility/move.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/incrementable_traits.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/incrementable_traits.h
index f935531c907..be3c3488923 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/incrementable_traits.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/incrementable_traits.h
@@ -27,9 +27,11 @@
 #include "../__utility/declval.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/indirectly_comparable.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/indirectly_comparable.h
index ce0f4626a17..f787950ca67 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/indirectly_comparable.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/indirectly_comparable.h
@@ -19,9 +19,11 @@
 #include "../__iterator/concepts.h"
 #include "../__iterator/projected.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/insert_iterator.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/insert_iterator.h
index da3d7bac8e7..f0b3716a96f 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/insert_iterator.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/insert_iterator.h
@@ -21,9 +21,11 @@
 #include "../__utility/move.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/istream_iterator.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/istream_iterator.h
index 62e48d19c09..6e97e0aa71b 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/istream_iterator.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/istream_iterator.h
@@ -20,9 +20,11 @@
 #include "../cstddef"
 #include "../iosfwd"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/istreambuf_iterator.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/istreambuf_iterator.h
index 9a9695490ec..63a4845b9b3 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/istreambuf_iterator.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/istreambuf_iterator.h
@@ -19,9 +19,11 @@
 #include "../__iterator/iterator_traits.h"
 #include "../iosfwd"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/iter_move.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/iter_move.h
index dfed7fd63b6..6b326cf8c3e 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/iter_move.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/iter_move.h
@@ -24,9 +24,11 @@
 #include "../__utility/forward.h"
 #include "../__utility/move.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #if defined(__clang__)
 #pragma clang diagnostic push
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/iter_swap.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/iter_swap.h
index c4ca7624cc0..8bc0c8dbe05 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/iter_swap.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/iter_swap.h
@@ -24,9 +24,11 @@
 #include "../__utility/forward.h"
 #include "../__utility/move.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #if _LIBCUDACXX_STD_VER > 14
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/iterator.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/iterator.h
index a22b37161c8..f9289e79da6 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/iterator.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/iterator.h
@@ -17,9 +17,11 @@
 
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/iterator_traits.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/iterator_traits.h
index 64851ce5403..4b5340d0353 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/iterator_traits.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/iterator_traits.h
@@ -32,9 +32,11 @@
 #include "../__type_traits/void_t.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/mergeable.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/mergeable.h
index 069d3fd0f2d..1af14cf573a 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/mergeable.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/mergeable.h
@@ -20,9 +20,11 @@
 #include "../__iterator/concepts.h"
 #include "../__iterator/projected.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/move_iterator.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/move_iterator.h
index 989cba4bcb5..91813072c13 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/move_iterator.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/move_iterator.h
@@ -22,9 +22,11 @@
 #include "../__type_traits/remove_reference.h"
 #include "../__utility/move.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/move_sentinel.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/move_sentinel.h
index c8f3f16a521..56530cdce9b 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/move_sentinel.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/move_sentinel.h
@@ -19,9 +19,11 @@
 #include "../__concepts/semiregular.h"
 #include "../__utility/move.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #if _LIBCUDACXX_STD_VER > 14
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/next.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/next.h
index 3ea7022f82e..ef24e788a0f 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/next.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/next.h
@@ -20,9 +20,11 @@
 #include "../__iterator/iterator_traits.h"
 #include "../__type_traits/enable_if.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/ostream_iterator.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/ostream_iterator.h
index 4e2e3bf89e7..6bd59d3ce5e 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/ostream_iterator.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/ostream_iterator.h
@@ -21,9 +21,11 @@
 #include "../cstddef"
 #include "../iosfwd"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/ostreambuf_iterator.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/ostreambuf_iterator.h
index 3e284ad919a..99d2222de7a 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/ostreambuf_iterator.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/ostreambuf_iterator.h
@@ -20,9 +20,11 @@
 #include "../cstddef"
 #include "../iosfwd"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/permutable.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/permutable.h
index 2c751c6d0e3..3c5a0eb4ff5 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/permutable.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/permutable.h
@@ -18,9 +18,11 @@
 #include "../__iterator/concepts.h"
 #include "../__iterator/iter_swap.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/prev.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/prev.h
index 2a429366222..63deb9faa76 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/prev.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/prev.h
@@ -20,9 +20,11 @@
 #include "../__iterator/iterator_traits.h"
 #include "../__type_traits/enable_if.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/projected.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/projected.h
index bb872119bb9..c19a22405a4 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/projected.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/projected.h
@@ -19,9 +19,11 @@
 #include "../__type_traits/enable_if.h"
 #include "../__type_traits/remove_cvref.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/readable_traits.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/readable_traits.h
index d6679c41d86..fb540ace1c9 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/readable_traits.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/readable_traits.h
@@ -27,9 +27,11 @@
 #include "../__type_traits/remove_extent.h"
 #include "../__type_traits/void_t.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/reverse_access.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/reverse_access.h
index 6b9140a3609..d68adaed755 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/reverse_access.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/reverse_access.h
@@ -19,9 +19,11 @@
 #include "../cstddef"
 #include "../initializer_list"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/reverse_iterator.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/reverse_iterator.h
index dfcfaa36e0b..ba68afb5215 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/reverse_iterator.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/reverse_iterator.h
@@ -20,9 +20,11 @@
 #include "../__memory/addressof.h"
 #include "../__type_traits/void_t.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/size.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/size.h
index ed52da48aca..aca8f78ac95 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/size.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/size.h
@@ -19,9 +19,11 @@
 #include "../__type_traits/make_signed.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/sortable.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/sortable.h
index 42d4c1ac078..135954960f3 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/sortable.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/sortable.h
@@ -21,9 +21,11 @@
 #include "../__iterator/permutable.h"
 #include "../__iterator/projected.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/unreachable_sentinel.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/unreachable_sentinel.h
index a10ccb4719a..46cb67acd2c 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/unreachable_sentinel.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/unreachable_sentinel.h
@@ -18,9 +18,11 @@
 
 #include "../__iterator/concepts.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #if _LIBCUDACXX_STD_VER > 14
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/wrap_iter.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/wrap_iter.h
index 63dec48fa7e..5375661e814 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/wrap_iter.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__iterator/wrap_iter.h
@@ -23,9 +23,11 @@
 #include "../__type_traits/is_convertible.h"
 #include "../__type_traits/is_trivially_copy_assignable.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__locale b/libcudacxx/include/cuda/std/detail/libcxx/include/__locale
index 749a25ce19e..c8fc12ba5ec 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__locale
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__locale
@@ -42,9 +42,11 @@
 # include <support/musl/xlocale.h>
 #endif
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/compressed_pair.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/compressed_pair.h
index 07fe2bdd55e..6c7cb6fa976 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/compressed_pair.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/compressed_pair.h
@@ -55,9 +55,11 @@
 #include "../__type_traits/enable_if.h"
 #include "../__type_traits/is_empty.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/default_accessor.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/default_accessor.h
index 45f87823de1..52a3a88a93b 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/default_accessor.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/default_accessor.h
@@ -52,9 +52,11 @@
 #include "../__type_traits/is_convertible.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/dynamic_extent.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/dynamic_extent.h
index 0fb760c9f2d..bcfb9b24273 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/dynamic_extent.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/dynamic_extent.h
@@ -53,7 +53,7 @@
 #include "../cstddef"
 #include "../limits" // numeric_limits
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#ifdef _LIBCUDACXX_IMPLICIT_SYSTEM_HEADER
 #endif
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/extents.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/extents.h
index a827c49ce5a..707438e643c 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/extents.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/extents.h
@@ -64,9 +64,11 @@
 #include "../cstddef"
 #include "../span"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #if defined(_LIBCUDACXX_PUSH_MACROS)
   _LIBCUDACXX_PUSH_MACROS
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/full_extent_t.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/full_extent_t.h
index 294ba6c6af8..45d9b45e179 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/full_extent_t.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/full_extent_t.h
@@ -50,9 +50,11 @@
 
 #include "../__mdspan/macros.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/layout_left.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/layout_left.h
index febe3f86ec5..cc70c3d9634 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/layout_left.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/layout_left.h
@@ -58,9 +58,11 @@
 #include "../__utility/integer_sequence.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/layout_right.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/layout_right.h
index 74ad81f92fa..a7c037b317f 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/layout_right.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/layout_right.h
@@ -58,9 +58,11 @@
 #include "../__utility/integer_sequence.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/layout_stride.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/layout_stride.h
index b5ef97af4db..f7ab6b88546 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/layout_stride.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/layout_stride.h
@@ -70,9 +70,11 @@
 #include "../numeric"
 #include "../span"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/macros.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/macros.h
index 0a14f21debf..7bad432c40f 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/macros.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/macros.h
@@ -55,9 +55,11 @@
 #include "../__type_traits/remove_reference.h"
 #include "../__utility/declval.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #if _LIBCUDACXX_STD_VER > 11
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/maybe_static_value.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/maybe_static_value.h
index a6a41aca872..0070e063a35 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/maybe_static_value.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/maybe_static_value.h
@@ -54,9 +54,11 @@
 #include "../__mdspan/no_unique_address.h"
 #endif
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 // This is only needed for the non-standard-layout version of partially
 // static array.
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/mdspan.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/mdspan.h
index 37cc94c732c..19f84739cd7 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/mdspan.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/mdspan.h
@@ -69,9 +69,11 @@
 #include "../array"
 #include "../span"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/no_unique_address.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/no_unique_address.h
index c26892f031c..4d071531b89 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/no_unique_address.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/no_unique_address.h
@@ -54,9 +54,11 @@
 #include "../__type_traits/is_trivially_destructible.h"
 #include "../__utility/move.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/standard_layout_static_array.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/standard_layout_static_array.h
index 13d9c89f0ec..f58f2dfa394 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/standard_layout_static_array.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/standard_layout_static_array.h
@@ -60,9 +60,11 @@
 #include "../cstddef"
 #include "../span"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/static_array.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/static_array.h
index 84df90d64c1..ed707f69d12 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/static_array.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/static_array.h
@@ -58,9 +58,11 @@
 #include "../array"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #if !__MDSPAN_PRESERVE_STANDARD_LAYOUT
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/submdspan.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/submdspan.h
index b551fab61a8..60071147ba2 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/submdspan.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/submdspan.h
@@ -67,9 +67,11 @@
 #include "../__utility/pair.h"
 #include "../tuple"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/type_list.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/type_list.h
index 95fb3a3c6c1..10ca0c3bc73 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/type_list.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__mdspan/type_list.h
@@ -51,9 +51,11 @@
 #include "../__mdspan/macros.h"
 #include "../__utility/integer_sequence.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__memory/addressof.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__memory/addressof.h
index 758bf1fe090..93e6b0c8971 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__memory/addressof.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__memory/addressof.h
@@ -15,9 +15,11 @@
 #include <__config>
 #endif //__cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__memory/construct_at.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__memory/construct_at.h
index a0577fac227..9171ba2fffe 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__memory/construct_at.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__memory/construct_at.h
@@ -27,14 +27,16 @@
 #include "../__utility/forward.h"
 #include "../__utility/move.h"
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #ifdef _LIBCUDACXX_COMPILER_CLANG_CUDA
 #include <new>
 #endif // _LIBCUDACXX_COMPILER_CLANG_CUDA
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
-#pragma GCC system_header
-#endif
-
 #if defined(__cuda_std__) && _LIBCUDACXX_STD_VER > 17 // need to backfill ::std::construct_at
 #ifndef _LIBCUDACXX_COMPILER_NVRTC
 #include <memory>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__memory/pointer_traits.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__memory/pointer_traits.h
index eed321f7432..781c629ef71 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__memory/pointer_traits.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__memory/pointer_traits.h
@@ -27,9 +27,11 @@
 #include "../__utility/declval.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__memory/voidify.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__memory/voidify.h
index e487b3f3af7..9b7ffb135f8 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__memory/voidify.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__memory/voidify.h
@@ -17,9 +17,11 @@
 
 #include "../__memory/addressof.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__mutex_base b/libcudacxx/include/cuda/std/detail/libcxx/include/__mutex_base
index 9208ca23c0a..e09afba6848 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__mutex_base
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__mutex_base
@@ -17,9 +17,11 @@
 
 #include <time.h>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_PUSH_MACROS
 #include <__undef_macros>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__node_handle b/libcudacxx/include/cuda/std/detail/libcxx/include/__node_handle
index 63791cdafa3..d6c3c35ec0b 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__node_handle
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__node_handle
@@ -15,9 +15,11 @@
 #include <memory>
 #include <optional>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_PUSH_MACROS
 #include <__undef_macros>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__nullptr b/libcudacxx/include/cuda/std/detail/libcxx/include/__nullptr
index 153f88f1cf6..0fe93766dcc 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__nullptr
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__nullptr
@@ -12,9 +12,11 @@
 
 #include <__config>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #ifdef _LIBCUDACXX_HAS_NO_NULLPTR
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__pragma_push b/libcudacxx/include/cuda/std/detail/libcxx/include/__pragma_push
index ad34a38d936..9b4ee961937 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__pragma_push
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__pragma_push
@@ -7,7 +7,7 @@
 //
 //===----------------------------------------------------------------------===//
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#ifdef _LIBCUDACXX_IMPLICIT_SYSTEM_HEADER
   #pragma GCC system_header
 #endif
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__split_buffer b/libcudacxx/include/cuda/std/detail/libcxx/include/__split_buffer
index 3f5757ab82c..b4d861cd3b4 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__split_buffer
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__split_buffer
@@ -6,9 +6,11 @@
 #include <type_traits>
 #include <algorithm>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_PUSH_MACROS
 #include <__undef_macros>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__sso_allocator b/libcudacxx/include/cuda/std/detail/libcxx/include/__sso_allocator
index 94f774df415..05479a6807f 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__sso_allocator
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__sso_allocator
@@ -14,9 +14,11 @@
 #include <type_traits>
 #include <new>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__std_stream b/libcudacxx/include/cuda/std/detail/libcxx/include/__std_stream
index 996db038734..603051531f9 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__std_stream
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__std_stream
@@ -16,9 +16,11 @@
 #include <__locale>
 #include <cstdio>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_PUSH_MACROS
 #include <__undef_macros>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__string b/libcudacxx/include/cuda/std/detail/libcxx/include/__string
index 972355e8bd8..558dd690d80 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__string
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__string
@@ -60,9 +60,11 @@ template <> struct char_traits<char8_t>;  // c++20
 #include "__assert" // all public C++ headers provide the assertion handler
 #include <__debug>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_PUSH_MACROS
 #include <__undef_macros>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__threading_support b/libcudacxx/include/cuda/std/detail/libcxx/include/__threading_support
index c383adcfb4f..8a5b2c98af3 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__threading_support
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__threading_support
@@ -27,9 +27,11 @@
 #include <__pragma_push>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #if defined(_LIBCUDACXX_HAS_THREAD_API_EXTERNAL)
 # ifndef __cuda_std__
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__tree b/libcudacxx/include/cuda/std/detail/libcxx/include/__tree
index af65a06a04d..63b9963eeec 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__tree
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__tree
@@ -16,9 +16,11 @@
 #include <stdexcept>
 #include <algorithm>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_PUSH_MACROS
 #include <__undef_macros>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple_dir/apply_cv.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple_dir/apply_cv.h
index ff670542dda..48292d4bb27 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple_dir/apply_cv.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple_dir/apply_cv.h
@@ -19,9 +19,11 @@
 #include "../__type_traits/is_volatile.h"
 #include "../__type_traits/remove_reference.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
-#  pragma GCC system_header
-#endif
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple_dir/make_tuple_types.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple_dir/make_tuple_types.h
index 6c3a4d1482f..e119fab819f 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple_dir/make_tuple_types.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple_dir/make_tuple_types.h
@@ -25,9 +25,11 @@
 #include "../__type_traits/remove_reference.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
-#  pragma GCC system_header
-#endif
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple_dir/sfinae_helpers.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple_dir/sfinae_helpers.h
index 77993f05b57..c22f2065b55 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple_dir/sfinae_helpers.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple_dir/sfinae_helpers.h
@@ -30,9 +30,11 @@
 #include "../__type_traits/remove_reference.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
-#  pragma GCC system_header
-#endif
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple_dir/tuple_element.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple_dir/tuple_element.h
index d2b3bdd4f67..a0c64d9cd15 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple_dir/tuple_element.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple_dir/tuple_element.h
@@ -21,9 +21,11 @@
 #include "../__type_traits/add_volatile.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
-#  pragma GCC system_header
-#endif
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple_dir/tuple_indices.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple_dir/tuple_indices.h
index 67787321c90..350e8df0a65 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple_dir/tuple_indices.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple_dir/tuple_indices.h
@@ -17,9 +17,11 @@
 #include "../__utility/integer_sequence.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
-#  pragma GCC system_header
-#endif
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple_dir/tuple_like.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple_dir/tuple_like.h
index 8557bd56162..566d49ada3d 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple_dir/tuple_like.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple_dir/tuple_like.h
@@ -21,9 +21,11 @@
 #include "../__type_traits/integral_constant.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
-#  pragma GCC system_header
-#endif
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple_dir/tuple_size.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple_dir/tuple_size.h
index 74c8ebd1bcc..85bf46d1945 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple_dir/tuple_size.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple_dir/tuple_size.h
@@ -22,9 +22,11 @@
 #include "../__type_traits/is_volatile.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
-#  pragma GCC system_header
-#endif
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple_dir/tuple_types.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple_dir/tuple_types.h
index c98f7f23c35..389cf889126 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple_dir/tuple_types.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple_dir/tuple_types.h
@@ -14,9 +14,11 @@
 #  include <__config>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
-#  pragma GCC system_header
-#endif
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/add_const.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/add_const.h
index 0c97b01dad9..ab3dc037d62 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/add_const.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/add_const.h
@@ -14,9 +14,11 @@
 #include <__config>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/add_cv.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/add_cv.h
index 3836cd4985d..24cf172c531 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/add_cv.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/add_cv.h
@@ -14,9 +14,11 @@
 #include <__config>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/add_lvalue_reference.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/add_lvalue_reference.h
index 404fdba123b..10fc9c2878e 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/add_lvalue_reference.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/add_lvalue_reference.h
@@ -16,9 +16,11 @@
 
 #include "../__type_traits/is_referenceable.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/add_pointer.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/add_pointer.h
index dc5f7b1e7ab..fba8cbb8425 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/add_pointer.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/add_pointer.h
@@ -20,9 +20,11 @@
 #include "../__type_traits/remove_cv.h"
 #include "../__type_traits/remove_reference.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/add_rvalue_reference.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/add_rvalue_reference.h
index 18ee9487f78..d4273a04323 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/add_rvalue_reference.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/add_rvalue_reference.h
@@ -16,9 +16,11 @@
 
 #include "../__type_traits/is_referenceable.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/add_volatile.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/add_volatile.h
index 28c89838c72..438806a1af2 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/add_volatile.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/add_volatile.h
@@ -14,9 +14,11 @@
 #include <__config>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/aligned_storage.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/aligned_storage.h
index 07f874edc06..dceabc43ec5 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/aligned_storage.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/aligned_storage.h
@@ -19,9 +19,11 @@
 #include "../__type_traits/type_list.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/aligned_union.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/aligned_union.h
index d441ada18a4..f216765c95e 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/aligned_union.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/aligned_union.h
@@ -18,9 +18,11 @@
 #include "../__type_traits/integral_constant.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/alignment_of.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/alignment_of.h
index 7838428480e..16b8e36cbbc 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/alignment_of.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/alignment_of.h
@@ -17,9 +17,11 @@
 #include "../__type_traits/integral_constant.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/apply_cv.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/apply_cv.h
index ec25ca1841b..14ca61f6433 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/apply_cv.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/apply_cv.h
@@ -19,9 +19,11 @@
 #include "../__type_traits/remove_reference.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/can_extract_key.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/can_extract_key.h
index 1f8aab7858d..e0789f42a6b 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/can_extract_key.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/can_extract_key.h
@@ -21,9 +21,11 @@
 #include "../__type_traits/remove_const.h"
 #include "../__type_traits/remove_const_ref.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/common_reference.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/common_reference.h
index 42833313f07..812a21fbafb 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/common_reference.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/common_reference.h
@@ -31,9 +31,11 @@
 #include "../__type_traits/void_t.h"
 #include "../__utility/declval.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/common_type.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/common_type.h
index e52e77d09eb..270b411f9f4 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/common_type.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/common_type.h
@@ -20,9 +20,11 @@
 #include "../__type_traits/void_t.h"
 #include "../__utility/declval.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/conditional.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/conditional.h
index 8419631b483..aa776667203 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/conditional.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/conditional.h
@@ -14,9 +14,11 @@
 #include <__config>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/conjunction.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/conjunction.h
index 9bfb91f00a7..8c602450d37 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/conjunction.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/conjunction.h
@@ -18,9 +18,11 @@
 #include "../__type_traits/enable_if.h"
 #include "../__type_traits/integral_constant.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/copy_cv.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/copy_cv.h
index 4b6b4b97402..75eb8644fcb 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/copy_cv.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/copy_cv.h
@@ -18,9 +18,11 @@
 #include "../__type_traits/add_cv.h"
 #include "../__type_traits/add_volatile.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/copy_cvref.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/copy_cvref.h
index d75a0e9043d..8205dc3fe85 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/copy_cvref.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/copy_cvref.h
@@ -18,9 +18,11 @@
 #include "../__type_traits/add_rvalue_reference.h"
 #include "../__type_traits/copy_cv.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/decay.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/decay.h
index c5ef28403be..bb8d449a26d 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/decay.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/decay.h
@@ -23,9 +23,11 @@
 #include "../__type_traits/remove_extent.h"
 #include "../__type_traits/remove_reference.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/dependent_type.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/dependent_type.h
index a1b04a1c470..faca9d784f6 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/dependent_type.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/dependent_type.h
@@ -14,9 +14,11 @@
 #include <__config>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/disjunction.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/disjunction.h
index 931763c6f92..8cff53cf9a6 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/disjunction.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/disjunction.h
@@ -16,9 +16,11 @@
 
 #include "../__type_traits/integral_constant.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/enable_if.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/enable_if.h
index 3b41a8474da..e3fe02bb277 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/enable_if.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/enable_if.h
@@ -14,9 +14,11 @@
 #include <__config>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/extent.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/extent.h
index 1fc4d8b9727..7bbf6a9d42b 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/extent.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/extent.h
@@ -18,9 +18,11 @@
 #include "../__type_traits/integral_constant.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/has_unique_object_representation.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/has_unique_object_representation.h
index be8dd4c0d29..5f21ebd4b97 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/has_unique_object_representation.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/has_unique_object_representation.h
@@ -18,9 +18,11 @@
 #include "../__type_traits/remove_all_extents.h"
 #include "../__type_traits/remove_cv.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/has_virtual_destructor.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/has_virtual_destructor.h
index 776d3a4747d..9f62201263c 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/has_virtual_destructor.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/has_virtual_destructor.h
@@ -16,9 +16,11 @@
 
 #include "../__type_traits/integral_constant.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/integral_constant.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/integral_constant.h
index 79d27bbae78..a008ca9642e 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/integral_constant.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/integral_constant.h
@@ -14,9 +14,11 @@
 #include <__config>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_abstract.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_abstract.h
index 503f0366976..53ee5a68238 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_abstract.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_abstract.h
@@ -16,9 +16,11 @@
 
 #include "../__type_traits/integral_constant.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_aggregate.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_aggregate.h
index cfd9731c6cc..84b515e80b1 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_aggregate.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_aggregate.h
@@ -16,9 +16,11 @@
 
 #include "../__type_traits/integral_constant.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_allocator.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_allocator.h
index 130578d8eea..9e9cb2bf32b 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_allocator.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_allocator.h
@@ -19,9 +19,11 @@
 #include "../__utility/declval.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_arithmetic.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_arithmetic.h
index c30c89c9e17..9e84dab5c19 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_arithmetic.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_arithmetic.h
@@ -18,9 +18,11 @@
 #include "../__type_traits/is_floating_point.h"
 #include "../__type_traits/is_integral.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_array.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_array.h
index d290a77159d..82b24fe3c3a 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_array.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_array.h
@@ -17,9 +17,11 @@
 #include "../__type_traits/integral_constant.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_assignable.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_assignable.h
index d2486b38b90..ba577332a45 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_assignable.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_assignable.h
@@ -18,9 +18,11 @@
 #include "../__type_traits/is_void.h"
 #include "../__utility/declval.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_base_of.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_base_of.h
index 7726973421a..a6857826424 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_base_of.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_base_of.h
@@ -16,9 +16,11 @@
 
 #include "../__type_traits/integral_constant.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_bounded_array.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_bounded_array.h
index 45ac57dab12..694063647ef 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_bounded_array.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_bounded_array.h
@@ -17,9 +17,11 @@
 #include "../__type_traits/integral_constant.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_callable.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_callable.h
index 173d84b520e..6d03b6d2a98 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_callable.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_callable.h
@@ -17,9 +17,11 @@
 #include "../__type_traits/integral_constant.h"
 #include "../__utility/declval.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_char_like_type.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_char_like_type.h
index cacc3383046..d49b17a5d42 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_char_like_type.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_char_like_type.h
@@ -18,9 +18,11 @@
 #include "../__type_traits/is_standard_layout.h"
 #include "../__type_traits/is_trivial.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_class.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_class.h
index 1e08722cfa9..703fc9c3bb0 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_class.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_class.h
@@ -17,9 +17,11 @@
 #include "../__type_traits/integral_constant.h"
 #include "../__type_traits/is_union.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_compound.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_compound.h
index 44dd92282a9..d1e20ba3374 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_compound.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_compound.h
@@ -17,9 +17,11 @@
 #include "../__type_traits/integral_constant.h"
 #include "../__type_traits/is_fundamental.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_const.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_const.h
index fd695a83ca6..24f4687bd2e 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_const.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_const.h
@@ -16,9 +16,11 @@
 
 #include "../__type_traits/integral_constant.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_constant_evaluated.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_constant_evaluated.h
index 2744c851bf5..3b91dabd176 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_constant_evaluated.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_constant_evaluated.h
@@ -14,9 +14,11 @@
 #include <__config>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_constructible.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_constructible.h
index 599fd644609..0f3b41dac1b 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_constructible.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_constructible.h
@@ -26,9 +26,11 @@
 #include "../__type_traits/remove_cvref.h"
 #include "../__utility/declval.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_convertible.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_convertible.h
index 9506d86c4c8..080928beefc 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_convertible.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_convertible.h
@@ -23,9 +23,11 @@
 #include "../__utility/declval.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_copy_assignable.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_copy_assignable.h
index 09739c1f563..4acf903e415 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_copy_assignable.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_copy_assignable.h
@@ -18,9 +18,11 @@
 #include "../__type_traits/add_lvalue_reference.h"
 #include "../__type_traits/is_assignable.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_copy_constructible.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_copy_constructible.h
index 470c88e4ed1..37cb709f50d 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_copy_constructible.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_copy_constructible.h
@@ -18,9 +18,11 @@
 #include "../__type_traits/add_lvalue_reference.h"
 #include "../__type_traits/is_constructible.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_core_convertible.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_core_convertible.h
index 7f63d93dfb0..f4ab6b26c27 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_core_convertible.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_core_convertible.h
@@ -16,9 +16,11 @@
 
 #include "../__type_traits/integral_constant.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_default_constructible.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_default_constructible.h
index 4dd528d0df0..988b1f9b6ae 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_default_constructible.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_default_constructible.h
@@ -16,9 +16,11 @@
 
 #include "../__type_traits/is_constructible.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_destructible.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_destructible.h
index b5bdede7426..59af3f3ab6e 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_destructible.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_destructible.h
@@ -20,9 +20,11 @@
 #include "../__type_traits/remove_all_extents.h"
 #include "../__utility/declval.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_empty.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_empty.h
index ac334191a79..54f99c4f5f1 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_empty.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_empty.h
@@ -16,9 +16,11 @@
 
 #include "../__type_traits/integral_constant.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_enum.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_enum.h
index 979815395a2..9e7f1482311 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_enum.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_enum.h
@@ -26,9 +26,11 @@
 #include "../__type_traits/is_union.h"
 #include "../__type_traits/is_void.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_final.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_final.h
index 6d9ff1781d0..42056c53fff 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_final.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_final.h
@@ -16,9 +16,11 @@
 
 #include "../__type_traits/integral_constant.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_floating_point.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_floating_point.h
index 430c18326a2..fee4d040d69 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_floating_point.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_floating_point.h
@@ -17,9 +17,11 @@
 #include "../__type_traits/integral_constant.h"
 #include "../__type_traits/remove_cv.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_function.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_function.h
index b52d8cfabc3..17de151f89f 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_function.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_function.h
@@ -18,9 +18,11 @@
 #include "../__type_traits/is_const.h"
 #include "../__type_traits/is_reference.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_fundamental.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_fundamental.h
index 08006183b0f..c2ff42c8ab6 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_fundamental.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_fundamental.h
@@ -19,9 +19,11 @@
 #include "../__type_traits/is_null_pointer.h"
 #include "../__type_traits/is_void.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_implicitly_default_constructible.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_implicitly_default_constructible.h
index 7f314ab586c..0564e5b05e4 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_implicitly_default_constructible.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_implicitly_default_constructible.h
@@ -17,9 +17,11 @@
 #include "../__type_traits/integral_constant.h"
 #include "../__type_traits/is_default_constructible.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_integral.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_integral.h
index b03da44a648..0c5303be6f6 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_integral.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_integral.h
@@ -17,9 +17,11 @@
 #include "../__type_traits/integral_constant.h"
 #include "../__type_traits/remove_cv.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_literal_type.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_literal_type.h
index 3f74f164f20..25f1675d5b8 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_literal_type.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_literal_type.h
@@ -19,9 +19,11 @@
 #include "../__type_traits/is_scalar.h"
 #include "../__type_traits/remove_all_extents.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_member_function_pointer.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_member_function_pointer.h
index 95c7fb1253d..eee3802f843 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_member_function_pointer.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_member_function_pointer.h
@@ -19,9 +19,11 @@
 #include "../__type_traits/remove_cv.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_member_object_pointer.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_member_object_pointer.h
index 40055e9caf2..642df0fe977 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_member_object_pointer.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_member_object_pointer.h
@@ -18,9 +18,11 @@
 #include "../__type_traits/is_member_function_pointer.h"
 #include "../__type_traits/remove_cv.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_member_pointer.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_member_pointer.h
index 00d3c74b8c9..7023e637cb7 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_member_pointer.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_member_pointer.h
@@ -18,9 +18,11 @@
 #include "../__type_traits/is_member_function_pointer.h"
 #include "../__type_traits/remove_cv.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_move_assignable.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_move_assignable.h
index 5d7697c34f4..5b049e5c601 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_move_assignable.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_move_assignable.h
@@ -18,9 +18,11 @@
 #include "../__type_traits/add_rvalue_reference.h"
 #include "../__type_traits/is_assignable.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_move_constructible.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_move_constructible.h
index 4a976a6c951..6d58e0b82bf 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_move_constructible.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_move_constructible.h
@@ -18,9 +18,11 @@
 #include "../__type_traits/add_rvalue_reference.h"
 #include "../__type_traits/is_constructible.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_assignable.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_assignable.h
index e1d8e030f29..97a5f766a45 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_assignable.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_assignable.h
@@ -19,9 +19,11 @@
 #include "../__type_traits/is_scalar.h"
 #include "../__utility/declval.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_constructible.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_constructible.h
index 6a2e3b3c21b..8541a4bef91 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_constructible.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_constructible.h
@@ -19,9 +19,11 @@
 #include "../__type_traits/is_scalar.h"
 #include "../__utility/declval.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_convertible.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_convertible.h
index 660a73ac450..bbd398dad7b 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_convertible.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_convertible.h
@@ -22,9 +22,11 @@
 #include "../__type_traits/lazy.h"
 #include "../__utility/declval.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_copy_assignable.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_copy_assignable.h
index 36eacc318d8..68bf4cef285 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_copy_assignable.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_copy_assignable.h
@@ -18,9 +18,11 @@
 #include "../__type_traits/add_lvalue_reference.h"
 #include "../__type_traits/is_nothrow_assignable.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_copy_constructible.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_copy_constructible.h
index b0421b5533f..3f5302834fc 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_copy_constructible.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_copy_constructible.h
@@ -18,9 +18,11 @@
 #include "../__type_traits/add_lvalue_reference.h"
 #include "../__type_traits/is_nothrow_constructible.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_default_constructible.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_default_constructible.h
index 8543041c3ac..9d407025adf 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_default_constructible.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_default_constructible.h
@@ -16,9 +16,11 @@
 
 #include "../__type_traits/is_nothrow_constructible.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_destructible.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_destructible.h
index 760151584a5..e8f843181dd 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_destructible.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_destructible.h
@@ -22,9 +22,11 @@
 #include "../__utility/declval.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_move_assignable.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_move_assignable.h
index 834d3bf99bd..a2d3caaf934 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_move_assignable.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_move_assignable.h
@@ -18,9 +18,11 @@
 #include "../__type_traits/add_rvalue_reference.h"
 #include "../__type_traits/is_nothrow_assignable.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_move_constructible.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_move_constructible.h
index f8dc72d0ab4..57cf7fabcd6 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_move_constructible.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_nothrow_move_constructible.h
@@ -17,9 +17,11 @@
 #include "../__type_traits/add_rvalue_reference.h"
 #include "../__type_traits/is_nothrow_constructible.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_null_pointer.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_null_pointer.h
index 0782df40a61..fb9874052fc 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_null_pointer.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_null_pointer.h
@@ -18,9 +18,11 @@
 #include "../__type_traits/remove_cv.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_object.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_object.h
index 477cb111de5..7be9b3b7afd 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_object.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_object.h
@@ -20,9 +20,11 @@
 #include "../__type_traits/is_scalar.h"
 #include "../__type_traits/is_union.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_pod.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_pod.h
index 0758bee19f4..5ec7f7bd39d 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_pod.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_pod.h
@@ -21,9 +21,11 @@
 #include "../__type_traits/is_trivially_destructible.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_pointer.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_pointer.h
index 5e6f0478373..2eff37ba0fc 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_pointer.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_pointer.h
@@ -17,9 +17,11 @@
 #include "../__type_traits/integral_constant.h"
 #include "../__type_traits/remove_cv.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_polymorphic.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_polymorphic.h
index 67c65ca6022..7fa80ec8467 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_polymorphic.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_polymorphic.h
@@ -17,9 +17,11 @@
 #include "../__type_traits/enable_if.h"
 #include "../__type_traits/integral_constant.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_primary_template.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_primary_template.h
index 62489fd8d0b..bdff70f97da 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_primary_template.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_primary_template.h
@@ -19,9 +19,11 @@
 #include "../__type_traits/is_valid_expansion.h"
 #include "../__type_traits/void_t.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_reference.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_reference.h
index f0a9b712e60..3b79fb89c3e 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_reference.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_reference.h
@@ -16,9 +16,11 @@
 
 #include "../__type_traits/integral_constant.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_reference_wrapper.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_reference_wrapper.h
index 55859fae2d5..f2d73a359c9 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_reference_wrapper.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_reference_wrapper.h
@@ -17,9 +17,11 @@
 #include "../__type_traits/integral_constant.h"
 #include "../__type_traits/remove_cv.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_referenceable.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_referenceable.h
index 7f965e94022..8d61a50c59e 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_referenceable.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_referenceable.h
@@ -17,9 +17,11 @@
 #include "../__type_traits/integral_constant.h"
 #include "../__type_traits/is_same.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_same.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_same.h
index 1fc3a4de6e9..ec24fa50ef9 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_same.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_same.h
@@ -16,9 +16,11 @@
 
 #include "../__type_traits/integral_constant.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_scalar.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_scalar.h
index 3fef34b4452..ee43b1b3269 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_scalar.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_scalar.h
@@ -22,9 +22,11 @@
 #include "../__type_traits/is_pointer.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_scoped_enum.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_scoped_enum.h
index 65c2b3c14bc..94a530b11f5 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_scoped_enum.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_scoped_enum.h
@@ -19,9 +19,11 @@
 #include "../__type_traits/is_enum.h"
 #include "../__type_traits/underlying_type.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_signed.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_signed.h
index a74507c2994..e7f2c1543f2 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_signed.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_signed.h
@@ -17,9 +17,11 @@
 #include "../__type_traits/integral_constant.h"
 #include "../__type_traits/is_arithmetic.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_signed_integer.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_signed_integer.h
index 68b4e4c9485..bbaba0140b2 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_signed_integer.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_signed_integer.h
@@ -16,9 +16,11 @@
 
 #include "../__type_traits/integral_constant.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_standard_layout.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_standard_layout.h
index 11322fd91d1..bf367af2e6d 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_standard_layout.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_standard_layout.h
@@ -17,9 +17,11 @@
 
 #include "../__type_traits/integral_constant.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_swappable.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_swappable.h
index 2a0e9f9858b..79a875993d8 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_swappable.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_swappable.h
@@ -28,9 +28,11 @@
 #include "../__utility/declval.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivial.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivial.h
index 0b4b0752d89..0f79032f701 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivial.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivial.h
@@ -18,9 +18,11 @@
 #include "../__type_traits/is_trivially_copyable.h"
 #include "../__type_traits/is_trivially_default_constructible.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_assignable.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_assignable.h
index 8175c259a50..5adb13e4878 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_assignable.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_assignable.h
@@ -17,9 +17,11 @@
 #include "../__type_traits/integral_constant.h"
 #include "../__type_traits/is_scalar.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_constructible.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_constructible.h
index 6f1d3e6d74c..0e7ab17cdb4 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_constructible.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_constructible.h
@@ -17,9 +17,11 @@
 #include "../__type_traits/integral_constant.h"
 #include "../__type_traits/is_scalar.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_copy_assignable.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_copy_assignable.h
index 87edfd5a891..7af35798147 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_copy_assignable.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_copy_assignable.h
@@ -18,9 +18,11 @@
 #include "../__type_traits/add_lvalue_reference.h"
 #include "../__type_traits/is_trivially_assignable.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_copy_constructible.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_copy_constructible.h
index e9405bb9d74..c659dc65b78 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_copy_constructible.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_copy_constructible.h
@@ -18,9 +18,11 @@
 #include "../__type_traits/add_lvalue_reference.h"
 #include "../__type_traits/is_trivially_constructible.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_copyable.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_copyable.h
index f593aec0c4a..213eaeee173 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_copyable.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_copyable.h
@@ -18,9 +18,11 @@
 #include "../__type_traits/is_scalar.h"
 #include "../__type_traits/remove_all_extents.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_default_constructible.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_default_constructible.h
index c1332492f52..844b14b198d 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_default_constructible.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_default_constructible.h
@@ -16,9 +16,11 @@
 
 #include "../__type_traits/is_trivially_constructible.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_destructible.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_destructible.h
index daf3edbd943..45b718791bd 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_destructible.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_destructible.h
@@ -20,9 +20,11 @@
 #include "../__type_traits/is_scalar.h"
 #include "../__type_traits/remove_all_extents.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_move_assignable.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_move_assignable.h
index 65775913143..30d002e79e9 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_move_assignable.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_move_assignable.h
@@ -18,9 +18,11 @@
 #include "../__type_traits/add_rvalue_reference.h"
 #include "../__type_traits/is_trivially_assignable.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_move_constructible.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_move_constructible.h
index 814981fd71d..99ecf5f13d7 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_move_constructible.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_trivially_move_constructible.h
@@ -17,9 +17,11 @@
 #include "../__type_traits/add_rvalue_reference.h"
 #include "../__type_traits/is_trivially_constructible.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_unbounded_array.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_unbounded_array.h
index e272fc0719b..28563057dae 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_unbounded_array.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_unbounded_array.h
@@ -16,9 +16,11 @@
 
 #include "../__type_traits/integral_constant.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_union.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_union.h
index 71e76287847..655f4984d32 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_union.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_union.h
@@ -17,9 +17,11 @@
 #include "../__type_traits/integral_constant.h"
 #include "../__type_traits/remove_cv.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_unsigned.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_unsigned.h
index 47a10ea575c..551ba0b14d5 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_unsigned.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_unsigned.h
@@ -18,9 +18,11 @@
 #include "../__type_traits/is_arithmetic.h"
 #include "../__type_traits/is_integral.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_unsigned_integer.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_unsigned_integer.h
index 2a27c7c3f86..5cf645c9572 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_unsigned_integer.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_unsigned_integer.h
@@ -16,9 +16,11 @@
 
 #include "../__type_traits/integral_constant.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_valid_expansion.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_valid_expansion.h
index 985e25a6e29..f7ce4fb7f61 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_valid_expansion.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_valid_expansion.h
@@ -16,9 +16,11 @@
 
 #include "../__type_traits/integral_constant.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_void.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_void.h
index fb034c4345c..ff8987f5949 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_void.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_void.h
@@ -18,9 +18,11 @@
 #include "../__type_traits/is_same.h"
 #include "../__type_traits/remove_cvref.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_volatile.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_volatile.h
index 95fbed6326a..b0c8698089e 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_volatile.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/is_volatile.h
@@ -16,9 +16,11 @@
 
 #include "../__type_traits/integral_constant.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/lazy.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/lazy.h
index f54bf4dd77d..9982483380f 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/lazy.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/lazy.h
@@ -14,9 +14,11 @@
 #include <__config>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/make_32_64_or_128_bit.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/make_32_64_or_128_bit.h
index 0ac1cb676ef..90f0efe55ac 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/make_32_64_or_128_bit.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/make_32_64_or_128_bit.h
@@ -21,9 +21,11 @@
 #include "../__type_traits/make_unsigned.h"
 #include "../cstdint"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/make_const_lvalue_ref.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/make_const_lvalue_ref.h
index 4d8e6b6a4cb..30248365aa2 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/make_const_lvalue_ref.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/make_const_lvalue_ref.h
@@ -16,9 +16,11 @@
 
 #include "../__type_traits/remove_reference.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/make_signed.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/make_signed.h
index 14b5ab42c51..eee78f5a504 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/make_signed.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/make_signed.h
@@ -21,9 +21,11 @@
 #include "../__type_traits/remove_cv.h"
 #include "../__type_traits/type_list.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/make_unsigned.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/make_unsigned.h
index edb074ec646..e4a8648f8eb 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/make_unsigned.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/make_unsigned.h
@@ -24,9 +24,11 @@
 #include "../__type_traits/type_list.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/maybe_const.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/maybe_const.h
index 1a90b10ee77..654d28b8646 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/maybe_const.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/maybe_const.h
@@ -16,9 +16,11 @@
 
 #include "../__type_traits/conditional.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/nat.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/nat.h
index 93df885716b..fda3b1b17f8 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/nat.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/nat.h
@@ -14,9 +14,11 @@
 #include <__config>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/negation.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/negation.h
index abd5e32c322..dde1f3ec12a 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/negation.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/negation.h
@@ -16,9 +16,11 @@
 
 #include "../__type_traits/integral_constant.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/promote.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/promote.h
index 5a968c722f4..4dcde79c22b 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/promote.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/promote.h
@@ -19,9 +19,11 @@
 #include "../__utility/declval.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/rank.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/rank.h
index 990d20ad08c..41e60e8077a 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/rank.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/rank.h
@@ -17,9 +17,11 @@
 #include "../__type_traits/integral_constant.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_all_extents.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_all_extents.h
index a77c359df81..03b07e79e4b 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_all_extents.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_all_extents.h
@@ -16,9 +16,11 @@
 
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_const.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_const.h
index 4a270fa23c7..e72135dfc1d 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_const.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_const.h
@@ -14,9 +14,11 @@
 #include <__config>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_const_ref.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_const_ref.h
index 3001531798e..7b9cb321f3b 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_const_ref.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_const_ref.h
@@ -17,9 +17,11 @@
 #include "../__type_traits/remove_const.h"
 #include "../__type_traits/remove_reference.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_cv.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_cv.h
index 42ce4e987fc..d858bcc8b8a 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_cv.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_cv.h
@@ -17,9 +17,11 @@
 #include "../__type_traits/remove_const.h"
 #include "../__type_traits/remove_volatile.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_cvref.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_cvref.h
index 05ee4fbd0ba..4b946579590 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_cvref.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_cvref.h
@@ -18,9 +18,11 @@
 #include "../__type_traits/remove_cv.h"
 #include "../__type_traits/remove_reference.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_extent.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_extent.h
index 5b412c809ab..1cf6f3eb93b 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_extent.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_extent.h
@@ -16,9 +16,11 @@
 
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_pointer.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_pointer.h
index 4a15a66aafc..d152a362bc7 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_pointer.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_pointer.h
@@ -14,9 +14,11 @@
 #include <__config>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_reference.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_reference.h
index 4b88c477434..0e800ba4702 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_reference.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_reference.h
@@ -16,9 +16,11 @@
 
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_volatile.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_volatile.h
index 58e70a50e30..1ccb3ff08b5 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_volatile.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/remove_volatile.h
@@ -14,9 +14,11 @@
 #include <__config>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/result_of.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/result_of.h
index 0c8d772c325..58378ff2938 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/result_of.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/result_of.h
@@ -16,9 +16,11 @@
 
 #include "../__functional/invoke.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/type_identity.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/type_identity.h
index f76ed081d1a..125fff289c6 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/type_identity.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/type_identity.h
@@ -14,9 +14,11 @@
 #include <__config>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/type_list.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/type_list.h
index 328e88b811e..3e287075900 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/type_list.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/type_list.h
@@ -16,9 +16,11 @@
 
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/underlying_type.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/underlying_type.h
index 88f860ee0a2..ac083427537 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/underlying_type.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/underlying_type.h
@@ -16,9 +16,11 @@
 
 #include "../__type_traits/is_enum.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/void_t.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/void_t.h
index 7cf29a19c98..4aeb40f0b85 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/void_t.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__type_traits/void_t.h
@@ -14,9 +14,11 @@
 #include <__config>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/as_const.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/as_const.h
index 4d8e92a48ca..1fd69cfda50 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/as_const.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/as_const.h
@@ -16,9 +16,11 @@
 
 #include "../__type_traits/add_const.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/auto_cast.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/auto_cast.h
index dba7b3f2209..3712bd10d8c 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/auto_cast.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/auto_cast.h
@@ -17,9 +17,11 @@
 
 #include "../__type_traits/decay.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #define _LIBCUDACXX_AUTO_CAST(expr) static_cast<_CUDA_VSTD::decay_t<decltype((expr))>>(expr)
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/cmp.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/cmp.h
index ab4c7476fed..4f48dd02463 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/cmp.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/cmp.h
@@ -27,9 +27,11 @@
 #include <__pragma_push>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/convert_to_integral.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/convert_to_integral.h
index d6f189cfdd3..7f4974b71b5 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/convert_to_integral.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/convert_to_integral.h
@@ -19,9 +19,11 @@
 #include "../__type_traits/is_floating_point.h"
 #include "../__type_traits/underlying_type.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/declval.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/declval.h
index 750e57a1a3d..e7be76bb445 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/declval.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/declval.h
@@ -14,9 +14,11 @@
 #include <__config>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/exception_guard.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/exception_guard.h
index 52725013aae..94a84e71b08 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/exception_guard.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/exception_guard.h
@@ -19,9 +19,11 @@
 #include "../__utility/exchange.h"
 #include "../__utility/move.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/exchange.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/exchange.h
index d0ac808e86c..883b5330b4b 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/exchange.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/exchange.h
@@ -19,9 +19,11 @@
 #include "../__utility/forward.h"
 #include "../__utility/move.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/forward.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/forward.h
index 3bac753a1e2..0b093d94cd3 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/forward.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/forward.h
@@ -19,9 +19,11 @@
 #include "../__type_traits/remove_reference.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/forward_like.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/forward_like.h
index b4d740765fa..0a65b7e37d2 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/forward_like.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/forward_like.h
@@ -20,9 +20,11 @@
 #include "../__type_traits/is_reference.h"
 #include "../__type_traits/remove_reference.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/in_place.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/in_place.h
index bcab29b8fa7..e7648ec8dbe 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/in_place.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/in_place.h
@@ -18,9 +18,11 @@
 #include "../__type_traits/remove_reference.h"
 #include "../__type_traits/remove_cvref.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/integer_sequence.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/integer_sequence.h
index f0c7d7406e2..e9ee2d7e4bf 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/integer_sequence.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/integer_sequence.h
@@ -16,9 +16,11 @@
 
 #include "../__type_traits/is_integral.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/move.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/move.h
index 31dfcc4a1b8..6726e4aaaa6 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/move.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/move.h
@@ -20,9 +20,11 @@
 #include "../__type_traits/is_nothrow_move_constructible.h"
 #include "../__type_traits/remove_reference.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/pair.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/pair.h
index c4f66a62071..718aa29f887 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/pair.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/pair.h
@@ -59,9 +59,11 @@
 #include <utility>
 #endif // defined(__cuda_std__) && !defined(__CUDACC_RTC__)
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/piecewise_construct.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/piecewise_construct.h
index 30e86e19fa4..fc503f819b3 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/piecewise_construct.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/piecewise_construct.h
@@ -14,9 +14,11 @@
 #include <__config>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/priority_tag.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/priority_tag.h
index 09898c5362b..4c636309484 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/priority_tag.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/priority_tag.h
@@ -16,9 +16,11 @@
 
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/rel_ops.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/rel_ops.h
index 72f95609063..aed8fc43d43 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/rel_ops.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/rel_ops.h
@@ -17,9 +17,11 @@
 #include "../__utility/forward.h"
 #include "../__utility/move.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/swap.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/swap.h
index e68295e60df..cc5ed96de7b 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/swap.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/swap.h
@@ -23,9 +23,11 @@
 #include "../__utility/move.h"
 #include "../cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/to_underlying.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/to_underlying.h
index d9e54fccc38..bd91e060165 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/to_underlying.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/to_underlying.h
@@ -17,9 +17,11 @@
 
 #include "../__type_traits/underlying_type.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/unreachable.h b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/unreachable.h
index 3a136921b67..5ef2e996f00 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/unreachable.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__utility/unreachable.h
@@ -16,9 +16,11 @@
 
 #include "../cstdlib"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/__verbose_abort b/libcudacxx/include/cuda/std/detail/libcxx/include/__verbose_abort
index be1a16ff916..862ecd08789 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/__verbose_abort
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/__verbose_abort
@@ -17,9 +17,11 @@
 
 #include "__availability"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 // Provide a default implementation of __libcpp_verbose_abort if we know that neither the built
 // library nor the user is providing one. Otherwise, just declare it and use the one from the
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/algorithm b/libcudacxx/include/cuda/std/detail/libcxx/include/algorithm
index 310652e03fb..503abf21fa6 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/algorithm
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/algorithm
@@ -668,9 +668,11 @@ template <class BidirectionalIterator, class Compare>
 #include <__pragma_push>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/any b/libcudacxx/include/cuda/std/detail/libcxx/include/any
index 0031c5cda82..83ab42b7456 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/any
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/any
@@ -88,9 +88,11 @@ namespace std {
 #include <cstdlib>
 #include <version>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 namespace std {
 class _LIBCUDACXX_EXCEPTION_ABI _LIBCUDACXX_AVAILABILITY_BAD_ANY_CAST bad_any_cast : public bad_cast
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/array b/libcudacxx/include/cuda/std/detail/libcxx/include/array
index 73f2b2ca869..05839350cc5 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/array
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/array
@@ -149,9 +149,11 @@ template <size_t I, class T, size_t N> const T&& get(const array<T, N>&&) noexce
 #include "__tuple_dir/tuple_element.h"
 #include "__tuple_dir/tuple_size.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/atomic b/libcudacxx/include/cuda/std/detail/libcxx/include/atomic
index 45c2a5a558a..1d045e42b37 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/atomic
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/atomic
@@ -572,9 +572,11 @@ void atomic_signal_fence(memory_order m) noexcept;
 #include <__pragma_push>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #ifdef _LIBCUDACXX_HAS_NO_THREADS
 # error <atomic> is not supported on this single threaded system
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/barrier b/libcudacxx/include/cuda/std/detail/libcxx/include/barrier
index ab4c0fb851f..ad4298b3e8e 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/barrier
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/barrier
@@ -65,9 +65,11 @@ namespace std
 #include <__pragma_push>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #ifdef _LIBCUDACXX_HAS_NO_THREADS
 # error <barrier> is not supported on this single threaded system
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/bit b/libcudacxx/include/cuda/std/detail/libcxx/include/bit
index 32263c56b50..b3ea29e4e18 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/bit
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/bit
@@ -68,9 +68,11 @@ namespace std {
 #include <__pragma_push>
 #endif //__cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #if defined(_LIBCUDACXX_COMPILER_MSVC)
 #include <intrin.h>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/bitset b/libcudacxx/include/cuda/std/detail/libcxx/include/bitset
index a6b19b81ce9..cbf68020ea8 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/bitset
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/bitset
@@ -121,9 +121,11 @@ template <size_t N> struct hash<std::bitset<N>>;
 #include <iosfwd>
 #include <__functional_base>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_PUSH_MACROS
 #include <__undef_macros>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/ccomplex b/libcudacxx/include/cuda/std/detail/libcxx/include/ccomplex
index 728af56ffaf..3e727a933b5 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/ccomplex
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/ccomplex
@@ -19,9 +19,11 @@
 
 #include <complex>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 // hh 080623 Created
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/cctype b/libcudacxx/include/cuda/std/detail/libcxx/include/cctype
index 16173ec7674..61ae214c384 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/cctype
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/cctype
@@ -37,9 +37,11 @@ int toupper(int c);
 #include <__config>
 #include <ctype.h>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/cerrno b/libcudacxx/include/cuda/std/detail/libcxx/include/cerrno
index 717f0f481d3..8f8c5533a64 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/cerrno
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/cerrno
@@ -25,8 +25,10 @@ Macros:
 #include <__config>
 #include <errno.h>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #endif  // _LIBCUDACXX_CERRNO
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/cfenv b/libcudacxx/include/cuda/std/detail/libcxx/include/cfenv
index 65e776ff64c..06e311c7014 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/cfenv
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/cfenv
@@ -55,9 +55,11 @@ int feupdateenv(const fenv_t* envp);
 #include <__config>
 #include <fenv.h>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/charconv b/libcudacxx/include/cuda/std/detail/libcxx/include/charconv
index 29cf317e97e..e51090e8fd2 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/charconv
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/charconv
@@ -82,9 +82,11 @@ namespace std {
 
 #include <__debug>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_PUSH_MACROS
 #include <__undef_macros>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/chrono b/libcudacxx/include/cuda/std/detail/libcxx/include/chrono
index f00bdeedbcc..828aa923813 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/chrono
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/chrono
@@ -850,9 +850,11 @@ constexpr chrono::year                                  operator ""y(unsigned lo
 #include <__pragma_push>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 // Silence NVCC warnings `long double` arising from chrono floating pointer
 // user-defined literals which are defined in terms of `long double`.
@@ -3203,6 +3205,9 @@ constexpr hours make24(const hours& __h, bool __is_pm) noexcept
 #pragma GCC system_header
 #endif
 
+_LIBCUDACXX_DIAGNOSTIC_PUSH
+_LIBCUDACXX_GCC_DIAGNOSTIC_IGNORED("-Wliteral-suffix")
+_LIBCUDACXX_CLANG_DIAGNOSTIC_IGNORED("-Wuser-defined-literals")
 // Suffixes for duration literals [time.duration.literals]
 inline namespace literals
 {
@@ -3295,6 +3300,7 @@ inline namespace literals
     }
 #endif //_LIBCUDACXX_STD_VER > 17 && !defined(_LIBCUDACXX_HAS_NO_CXX20_CHRONO_LITERALS)
 }}
+_LIBCUDACXX_DIAGNOSTIC_POP
 
 namespace chrono { // hoist the literals into namespace std::chrono
    using namespace literals::chrono_literals;
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/cinttypes b/libcudacxx/include/cuda/std/detail/libcxx/include/cinttypes
index 0a995a61420..8029b59ebe2 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/cinttypes
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/cinttypes
@@ -239,9 +239,11 @@ uintmax_t wcstoumax(const wchar_t* restrict nptr, wchar_t** restrict endptr, int
 
 #include "cstdint"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/climits b/libcudacxx/include/cuda/std/detail/libcxx/include/climits
index 92cfad5a052..5574afe450a 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/climits
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/climits
@@ -48,9 +48,11 @@ Macros:
 #include <__pragma_push>
 #endif //__cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #if defined(_LIBCUDACXX_COMPILER_MSVC)
 #include "support/win32/limits_msvc_win32.h"
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/clocale b/libcudacxx/include/cuda/std/detail/libcxx/include/clocale
index 64d6fe37774..52c61aaa6a3 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/clocale
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/clocale
@@ -37,9 +37,11 @@ lconv* localeconv();
 #include <__config>
 #include <locale.h>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/cmath b/libcudacxx/include/cuda/std/detail/libcxx/include/cmath
index 2855220e72d..7638eaf5db1 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/cmath
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/cmath
@@ -320,9 +320,11 @@ long double    truncl(long double x);
 #include <__pragma_push>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #ifdef _LIBCUDACXX_COMPILER_NVRTC
 #include "climits"
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/codecvt b/libcudacxx/include/cuda/std/detail/libcxx/include/codecvt
index cbb09db11ab..4c02915ab10 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/codecvt
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/codecvt
@@ -57,9 +57,11 @@ class codecvt_utf8_utf16
 #include <__config>
 #include <__locale>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/compare b/libcudacxx/include/cuda/std/detail/libcxx/include/compare
index d3482f0e23c..63d2b3f20b1 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/compare
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/compare
@@ -50,9 +50,11 @@ namespace std {
 #include <type_traits>
 #include <array>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/complex b/libcudacxx/include/cuda/std/detail/libcxx/include/complex
index 1abd36617db..e064ade753b 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/complex
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/complex
@@ -267,9 +267,11 @@ template<class T> complex<T> tanh (const complex<T>&);
 #include <__pragma_push>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 # if _LIBCUDACXX_CUDA_ABI_VERSION > 3
 #  define _LIBCUDACXX_COMPLEX_ALIGNAS(V) _ALIGNAS(V)
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/complex.h b/libcudacxx/include/cuda/std/detail/libcxx/include/complex.h
index 505d11cef35..22ad5415020 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/complex.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/complex.h
@@ -19,9 +19,11 @@
 
 #include <__config>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #ifdef __cplusplus
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/concepts b/libcudacxx/include/cuda/std/detail/libcxx/include/concepts
index 8b99044509e..b68750dd601 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/concepts
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/concepts
@@ -162,8 +162,10 @@ namespace std {
 
 #include "version"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #endif // _LIBCUDACXX_CONCEPTS
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/condition_variable b/libcudacxx/include/cuda/std/detail/libcxx/include/condition_variable
index 0d11724365f..ded33441f72 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/condition_variable
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/condition_variable
@@ -110,9 +110,11 @@ public:
 #include <__mutex_base>
 #include <memory>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #ifndef _LIBCUDACXX_HAS_NO_THREADS
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/csetjmp b/libcudacxx/include/cuda/std/detail/libcxx/include/csetjmp
index d6b46eff984..a5faf91b9c2 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/csetjmp
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/csetjmp
@@ -33,9 +33,11 @@ void longjmp(jmp_buf env, int val);
 #include <__config>
 #include <setjmp.h>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/csignal b/libcudacxx/include/cuda/std/detail/libcxx/include/csignal
index 39c13590e02..39259812809 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/csignal
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/csignal
@@ -42,9 +42,11 @@ int raise(int sig);
 #include <__config>
 #include <signal.h>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/cstdarg b/libcudacxx/include/cuda/std/detail/libcxx/include/cstdarg
index a7b2fb9a1d0..aca5c592846 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/cstdarg
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/cstdarg
@@ -37,9 +37,11 @@ Types:
 #include <__pragma_push>
 #endif //__cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/cstdbool b/libcudacxx/include/cuda/std/detail/libcxx/include/cstdbool
index 026f9574fcd..23a2ef4504e 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/cstdbool
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/cstdbool
@@ -24,9 +24,11 @@ Macros:
 #include <__pragma_push>
 #endif //__cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #undef __bool_true_false_are_defined
 #define __bool_true_false_are_defined 1
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/cstddef b/libcudacxx/include/cuda/std/detail/libcxx/include/cstddef
index da6f5610413..83f0d8c2994 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/cstddef
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/cstddef
@@ -48,9 +48,11 @@ Types:
 #include <__pragma_push>
 #endif //__cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #ifndef __cuda_std__
 // Don't include our own <stddef.h>; we don't want to declare ::nullptr_t.
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/cstdint b/libcudacxx/include/cuda/std/detail/libcxx/include/cstdint
index 1f0337a6008..ef6ff78accf 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/cstdint
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/cstdint
@@ -157,9 +157,11 @@ Types:
 #include <__pragma_push>
 #endif //__cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/cstdio b/libcudacxx/include/cuda/std/detail/libcxx/include/cstdio
index e17d75386e0..c067f6177d7 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/cstdio
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/cstdio
@@ -98,9 +98,11 @@ void perror(const char* s);
 #include <__config>
 #include <stdio.h>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/cstdlib b/libcudacxx/include/cuda/std/detail/libcxx/include/cstdlib
index f1659f57553..306c319ce64 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/cstdlib
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/cstdlib
@@ -94,9 +94,11 @@ void *aligned_alloc(size_t alignment, size_t size);                       // C11
 #include <__pragma_push>
 #endif //__cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #if defined(_LIBCUDACXX_COMPILER_CLANG_CUDA)
 #  define _LIBCUDACXX_UNREACHABLE() __builtin_unreachable()
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/cstring b/libcudacxx/include/cuda/std/detail/libcxx/include/cstring
index 06e1c95f8b3..7e4a904ac25 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/cstring
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/cstring
@@ -62,9 +62,11 @@ size_t strlen(const char* s);
 #include <__pragma_push>
 #endif //__simet__ <- a tribute to ogiroux
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/ctgmath b/libcudacxx/include/cuda/std/detail/libcxx/include/ctgmath
index d43fbcce9d8..11c0746caee 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/ctgmath
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/ctgmath
@@ -21,8 +21,10 @@
 #include <ccomplex>
 #include "cmath"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #endif  // _LIBCUDACXX_CTGMATH
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/ctime b/libcudacxx/include/cuda/std/detail/libcxx/include/ctime
index 1b045f9f2d5..fa17dfa0f36 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/ctime
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/ctime
@@ -59,9 +59,11 @@ typedef long long int time_t;
 #include <__pragma_push>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/ctype.h b/libcudacxx/include/cuda/std/detail/libcxx/include/ctype.h
index 4957f6eb134..173c16e4273 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/ctype.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/ctype.h
@@ -31,9 +31,11 @@ int toupper(int c);
 
 #include <__config>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #include_next <ctype.h>
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/cwchar b/libcudacxx/include/cuda/std/detail/libcxx/include/cwchar
index 750b774c6a1..9d0062a94b8 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/cwchar
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/cwchar
@@ -106,9 +106,11 @@ size_t wcsrtombs(char* restrict dst, const wchar_t** restrict src, size_t len,
 #include <cwctype>
 #include <wchar.h>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/cwctype b/libcudacxx/include/cuda/std/detail/libcxx/include/cwctype
index ce8a7933c29..3a1ccc36a26 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/cwctype
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/cwctype
@@ -53,9 +53,11 @@ wctrans_t wctrans(const char* property);
 #include <cctype>
 #include <wctype.h>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/deque b/libcudacxx/include/cuda/std/detail/libcxx/include/deque
index d730129136f..271ba5e3720 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/deque
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/deque
@@ -167,9 +167,11 @@ template <class T, class Allocator, class Predicate>
 #include <stdexcept>
 #include <version>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_PUSH_MACROS
 #include <__undef_macros>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/errno.h b/libcudacxx/include/cuda/std/detail/libcxx/include/errno.h
index b35af2570a0..589d29d5092 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/errno.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/errno.h
@@ -24,9 +24,11 @@
 
 #include <__config>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #include_next <errno.h>
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/exception b/libcudacxx/include/cuda/std/detail/libcxx/include/exception
index 971154e1e3f..4efc1841828 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/exception
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/exception
@@ -89,9 +89,11 @@ template <class E> void rethrow_if_nested(const E& e);
 #include <vcruntime_exception.h>
 #endif
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD_NOVERSION  // purposefully not using versioning namespace
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/expected b/libcudacxx/include/cuda/std/detail/libcxx/include/expected
index 2f01dd9de16..68f61f6e1d5 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/expected
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/expected
@@ -32,8 +32,10 @@ namespace std {
 
 #include "version"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #endif // _LIBCUDACXX_EXPECTED
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/__config b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/__config
index 4571a089240..802708e3277 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/__config
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/__config
@@ -12,9 +12,11 @@
 
 #include <__config>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #define _LIBCUDACXX_BEGIN_NAMESPACE_EXPERIMENTAL namespace std { namespace experimental {
 #define _LIBCUDACXX_END_NAMESPACE_EXPERIMENTAL  } }
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/algorithm b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/algorithm
index 7fe305cf0b4..1da9a4a26f8 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/algorithm
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/algorithm
@@ -37,9 +37,11 @@ ForwardIterator search(ForwardIterator first, ForwardIterator last,
 
 #include <__debug>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_PUSH_MACROS
 #include <__undef_macros>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/coroutine b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/coroutine
index a5e60fe3ef5..bd94e5b6d02 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/coroutine
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/coroutine
@@ -53,9 +53,11 @@ template <class P> struct hash<coroutine_handle<P>>;
 #include <cstddef>
 #include <__debug>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #ifdef _LIBCUDACXX_HAS_NO_COROUTINES
 # if defined(_LIBCUDACXX_WARNING)
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/deque b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/deque
index 1f2291796db..06673b613bc 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/deque
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/deque
@@ -32,9 +32,11 @@ namespace pmr {
 #include <deque>
 #include <experimental/memory_resource>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_LFTS_PMR
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/filesystem b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/filesystem
index e0e6b9bbab0..736b2917302 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/filesystem
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/filesystem
@@ -232,9 +232,11 @@
 #include <experimental/__config>
 #include <filesystem>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_PUSH_MACROS
 #include <__undef_macros>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/forward_list b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/forward_list
index 585ca5a8d6f..2222319f75a 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/forward_list
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/forward_list
@@ -32,9 +32,11 @@ namespace pmr {
 #include <forward_list>
 #include <experimental/memory_resource>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_LFTS_PMR
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/functional b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/functional
index 719396c07c7..e45c071a8d6 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/functional
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/functional
@@ -96,9 +96,11 @@ inline namespace fundamentals_v1 {
 
 #include <__debug>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_PUSH_MACROS
 #include <__undef_macros>
@@ -112,7 +114,7 @@ template<class _ForwardIterator, class _BinaryPredicate = equal_to<>>
 class _LIBCUDACXX_TYPE_VIS default_searcher {
 public:
     _LIBCUDACXX_INLINE_VISIBILITY
-    default_searcher(_ForwardIterator __f, _ForwardIterator __l, 
+    default_searcher(_ForwardIterator __f, _ForwardIterator __l,
                        _BinaryPredicate __p = _BinaryPredicate())
         : __first_(__f), __last_(__l), __pred_(__p) {}
 
@@ -151,12 +153,12 @@ public: // TODO private:
 
     const _Value __default_value_;
     std::unordered_map<_Key, _Value, _Hash, _BinaryPredicate> __table;
-    
+
 public:
     _LIBCUDACXX_INLINE_VISIBILITY
     _BMSkipTable(std::size_t __sz, _Value __default, _Hash __hf, _BinaryPredicate __pred)
         : __default_value_(__default), __table(__sz, __hf, __pred) {}
-    
+
     _LIBCUDACXX_INLINE_VISIBILITY
     void insert(const key_type &__key, value_type __val)
     {
@@ -170,7 +172,7 @@ public:
         return __it == __table.end() ? __default_value_ : __it->second;
     }
 };
-    
+
 
 //  Special case small numeric values; use an array
 template<class _Key, typename _Value, class _Hash, class _BinaryPredicate>
@@ -189,7 +191,7 @@ public:
     {
         std::fill_n(__table.begin(), __table.size(), __default);
     }
-    
+
     _LIBCUDACXX_INLINE_VISIBILITY
     void insert(key_type __key, value_type __val)
     {
@@ -204,8 +206,8 @@ public:
 };
 
 
-template <class _RandomAccessIterator1, 
-          class _Hash = hash<typename iterator_traits<_RandomAccessIterator1>::value_type>, 
+template <class _RandomAccessIterator1,
+          class _Hash = hash<typename iterator_traits<_RandomAccessIterator1>::value_type>,
           class _BinaryPredicate = equal_to<>>
 class _LIBCUDACXX_TYPE_VIS boyer_moore_searcher {
 private:
@@ -217,9 +219,9 @@ private:
                     is_same<_Hash, hash<value_type>>::value &&
                     is_same<_BinaryPredicate, equal_to<>>::value
             > skip_table_type;
-    
+
 public:
-    boyer_moore_searcher(_RandomAccessIterator1 __f, _RandomAccessIterator1 __l, 
+    boyer_moore_searcher(_RandomAccessIterator1 __f, _RandomAccessIterator1 __l,
                 _Hash __hf = _Hash(), _BinaryPredicate __pred = _BinaryPredicate())
             : __first_(__f), __last_(__l), __pred_(__pred),
               __pattern_length_(_CUDA_VSTD::distance(__first_, __last_)),
@@ -232,13 +234,13 @@ public:
 
         this->__build_suffix_table ( __first_, __last_, __pred_ );
         }
-        
+
     template <typename _RandomAccessIterator2>
     pair<_RandomAccessIterator2, _RandomAccessIterator2>
     operator ()(_RandomAccessIterator2 __f, _RandomAccessIterator2 __l) const
     {
         static_assert ( std::is_same<
-                std::__remove_cvref_t<typename std::iterator_traits<_RandomAccessIterator1>::value_type>, 
+                std::__remove_cvref_t<typename std::iterator_traits<_RandomAccessIterator1>::value_type>,
                 std::__remove_cvref_t<typename std::iterator_traits<_RandomAccessIterator2>::value_type>
                     >::value,
                 "Corpus and Pattern iterators must point to the same type" );
@@ -247,13 +249,13 @@ public:
         if (__first_ == __last_) return make_pair(__f, __f); // empty pattern
 
     //  If the pattern is larger than the corpus, we can't find it!
-        if ( __pattern_length_ > _CUDA_VSTD::distance (__f, __l)) 
+        if ( __pattern_length_ > _CUDA_VSTD::distance (__f, __l))
             return make_pair(__l, __l);
 
-    //  Do the search 
+    //  Do the search
         return this->__search(__f, __l);
     }
-        
+
 public: // TODO private:
     _RandomAccessIterator1               __first_;
     _RandomAccessIterator1               __last_;
@@ -270,7 +272,7 @@ public: // TODO private:
         const _RandomAccessIterator2 __last = __l - __pattern_length_;
         const skip_table_type &         __skip   = *__skip_.get();
         const vector<difference_type> & __suffix = *__suffix_.get();
-        
+
         while (__cur <= __last)
         {
 
@@ -282,7 +284,7 @@ public: // TODO private:
                 if ( __j == 0 )
                     return make_pair(__cur, __cur + __pattern_length_);
                 }
-            
+
         //  Since we didn't match, figure out how far to skip forward
             difference_type __k = __skip[__cur [ __j - 1 ]];
             difference_type __m = __j - __k - 1;
@@ -291,7 +293,7 @@ public: // TODO private:
             else
                 __cur += __suffix[ __j ];
         }
-    
+
         return make_pair(__l, __l);     // We didn't find anything
     }
 
@@ -300,21 +302,21 @@ public: // TODO private:
     void __compute_bm_prefix ( _Iterator __f, _Iterator __l, _BinaryPredicate __pred, _Container &__prefix )
     {
         const std::size_t __count = _CUDA_VSTD::distance(__f, __l);
-                        
+
         __prefix[0] = 0;
         std::size_t __k = 0;
         for ( std::size_t __i = 1; __i < __count; ++__i )
         {
             while ( __k > 0 && !__pred ( __f[__k], __f[__i] ))
                 __k = __prefix [ __k - 1 ];
-                
+
             if ( __pred ( __f[__k], __f[__i] ))
                 __k++;
             __prefix [ __i ] = __k;
         }
     }
 
-    void __build_suffix_table(_RandomAccessIterator1 __f, _RandomAccessIterator1 __l, 
+    void __build_suffix_table(_RandomAccessIterator1 __f, _RandomAccessIterator1 __l,
                                                     _BinaryPredicate __pred)
     {
         const std::size_t __count = _CUDA_VSTD::distance(__f, __l);
@@ -322,19 +324,19 @@ public: // TODO private:
         if (__count > 0)
         {
             _CUDA_VSTD::vector<value_type> __scratch(__count);
-            
+
             __compute_bm_prefix(__f, __l, __pred, __scratch);
             for ( std::size_t __i = 0; __i <= __count; __i++ )
                 __suffix[__i] = __count - __scratch[__count-1];
-    
+
             typedef _CUDA_VSTD::reverse_iterator<_RandomAccessIterator1> _RevIter;
             __compute_bm_prefix(_RevIter(__l), _RevIter(__f), __pred, __scratch);
-     
+
             for ( std::size_t __i = 0; __i < __count; __i++ )
             {
                 const std::size_t     __j = __count - __scratch[__i];
                 const difference_type __k = __i     - __scratch[__i] + 1;
-     
+
                 if (__suffix[__j] > __k)
                     __suffix[__j] = __k;
             }
@@ -343,20 +345,20 @@ public: // TODO private:
 
 };
 
-template<class _RandomAccessIterator, 
-         class _Hash = hash<typename iterator_traits<_RandomAccessIterator>::value_type>, 
+template<class _RandomAccessIterator,
+         class _Hash = hash<typename iterator_traits<_RandomAccessIterator>::value_type>,
          class _BinaryPredicate = equal_to<>>
 _LIBCUDACXX_INLINE_VISIBILITY
 boyer_moore_searcher<_RandomAccessIterator, _Hash, _BinaryPredicate>
-make_boyer_moore_searcher( _RandomAccessIterator __f, _RandomAccessIterator __l, 
+make_boyer_moore_searcher( _RandomAccessIterator __f, _RandomAccessIterator __l,
                     _Hash __hf = _Hash(), _BinaryPredicate __p = _BinaryPredicate ())
 {
     return boyer_moore_searcher<_RandomAccessIterator, _Hash, _BinaryPredicate>(__f, __l, __hf, __p);
 }
 
 // boyer-moore-horspool
-template <class _RandomAccessIterator1, 
-          class _Hash = hash<typename iterator_traits<_RandomAccessIterator1>::value_type>, 
+template <class _RandomAccessIterator1,
+          class _Hash = hash<typename iterator_traits<_RandomAccessIterator1>::value_type>,
           class _BinaryPredicate = equal_to<>>
 class _LIBCUDACXX_TYPE_VIS boyer_moore_horspool_searcher {
 private:
@@ -370,7 +372,7 @@ private:
             > skip_table_type;
 
 public:
-    boyer_moore_horspool_searcher(_RandomAccessIterator1 __f, _RandomAccessIterator1 __l, 
+    boyer_moore_horspool_searcher(_RandomAccessIterator1 __f, _RandomAccessIterator1 __l,
                 _Hash __hf = _Hash(), _BinaryPredicate __pred = _BinaryPredicate())
             : __first_(__f), __last_(__l), __pred_(__pred),
               __pattern_length_(_CUDA_VSTD::distance(__first_, __last_)),
@@ -384,13 +386,13 @@ public:
                     __skip_->insert(*__f, __pattern_length_ - 1 - __i);
             }
         }
-            
+
     template <typename _RandomAccessIterator2>
     pair<_RandomAccessIterator2, _RandomAccessIterator2>
     operator ()(_RandomAccessIterator2 __f, _RandomAccessIterator2 __l) const
     {
         static_assert ( std::is_same<
-                std::__remove_cvref_t<typename std::iterator_traits<_RandomAccessIterator1>::value_type>, 
+                std::__remove_cvref_t<typename std::iterator_traits<_RandomAccessIterator1>::value_type>,
                 std::__remove_cvref_t<typename std::iterator_traits<_RandomAccessIterator2>::value_type>
                     >::value,
                 "Corpus and Pattern iterators must point to the same type" );
@@ -399,13 +401,13 @@ public:
         if (__first_ == __last_) return make_pair(__f, __f); // empty pattern
 
     //  If the pattern is larger than the corpus, we can't find it!
-        if ( __pattern_length_ > _CUDA_VSTD::distance (__f, __l)) 
+        if ( __pattern_length_ > _CUDA_VSTD::distance (__f, __l))
             return make_pair(__l, __l);
 
-    //  Do the search 
+    //  Do the search
         return this->__search(__f, __l);
     }
-        
+
 private:
     _RandomAccessIterator1      __first_;
     _RandomAccessIterator1      __last_;
@@ -433,17 +435,17 @@ private:
             }
             __cur += __skip[__cur[__pattern_length_-1]];
         }
-        
+
         return make_pair(__l, __l);
     }
 };
 
-template<class _RandomAccessIterator, 
-         class _Hash = hash<typename iterator_traits<_RandomAccessIterator>::value_type>, 
+template<class _RandomAccessIterator,
+         class _Hash = hash<typename iterator_traits<_RandomAccessIterator>::value_type>,
          class _BinaryPredicate = equal_to<>>
 _LIBCUDACXX_INLINE_VISIBILITY
 boyer_moore_horspool_searcher<_RandomAccessIterator, _Hash, _BinaryPredicate>
-make_boyer_moore_horspool_searcher( _RandomAccessIterator __f, _RandomAccessIterator __l, 
+make_boyer_moore_horspool_searcher( _RandomAccessIterator __f, _RandomAccessIterator __l,
                     _Hash __hf = _Hash(), _BinaryPredicate __p = _BinaryPredicate ())
 {
     return boyer_moore_horspool_searcher<_RandomAccessIterator, _Hash, _BinaryPredicate>(__f, __l, __hf, __p);
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/list b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/list
index 0600d27e65a..8f6fbced635 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/list
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/list
@@ -32,9 +32,11 @@ namespace pmr {
 #include <list>
 #include <experimental/memory_resource>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_LFTS_PMR
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/map b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/map
index 222836e1801..339b149f623 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/map
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/map
@@ -37,9 +37,11 @@ namespace pmr {
 #include <map>
 #include <experimental/memory_resource>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_LFTS_PMR
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/memory_resource b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/memory_resource
index 7b32dd558d9..93e62b656f2 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/memory_resource
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/memory_resource
@@ -77,9 +77,11 @@ namespace pmr {
 #include <cstdlib>
 #include <__debug>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_PUSH_MACROS
 #include <__undef_macros>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/propagate_const b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/propagate_const
index 476d307692b..70afdfa086b 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/propagate_const
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/propagate_const
@@ -106,11 +106,11 @@
 
 */
 
-#include <experimental/__config>
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#include <experimental/__config>#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
-
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #if _LIBCUDACXX_STD_VER > 11
 
 #include <type_traits>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/regex b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/regex
index 3a7cd32ccec..ef2dc3aea12 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/regex
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/regex
@@ -40,9 +40,11 @@ namespace pmr {
 #include <experimental/string>
 #include <experimental/memory_resource>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_LFTS_PMR
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/set b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/set
index 995d50a3d48..a267b7acd1c 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/set
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/set
@@ -37,9 +37,11 @@ namespace pmr {
 #include <set>
 #include <experimental/memory_resource>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_LFTS_PMR
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/simd b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/simd
index 99453767b5f..ed37b0f212b 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/simd
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/simd
@@ -655,9 +655,11 @@ public:
 #include <cstddef>
 #include <functional>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_EXPERIMENTAL_SIMD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/string b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/string
index 0e502e823b3..6509bd8b9cd 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/string
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/string
@@ -41,9 +41,11 @@ namespace pmr {
 #include <string>
 #include <experimental/memory_resource>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_LFTS_PMR
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/type_traits b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/type_traits
index 83859d47216..c73143bbb0c 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/type_traits
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/type_traits
@@ -60,7 +60,7 @@ inline namespace fundamentals_v1 {
      using is_detected_convertible = is_convertible<detected_t<Op, Args...>, To>;
   template <class To, template<class...> class Op, class... Args>
      constexpr bool is_detected_convertible_v
-       = is_detected_convertible<To, Op, Args...>::value;  
+       = is_detected_convertible<To, Op, Args...>::value;
 
 } // namespace fundamentals_v1
 } // namespace experimental
@@ -75,9 +75,11 @@ inline namespace fundamentals_v1 {
 #include <initializer_list>
 #include <type_traits>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_LFTS
 
@@ -122,7 +124,7 @@ struct _DETECTOR<_Default, void_t<_Op<_Args...>>, _Op, _Args...> {
    using value_t = true_type;
    using type = _Op<_Args...>;
    };
-     
+
 
 template <template<class...> class _Op, class... _Args>
   using is_detected = typename _DETECTOR<nonesuch, void, _Op, _Args...>::value_t;
@@ -144,7 +146,7 @@ template <class Expected, template<class...> class _Op, class... _Args>
 template <class To, template<class...> class _Op, class... _Args>
   using is_detected_convertible = is_convertible<detected_t<_Op, _Args...>, To>;
 template <class To, template<class...> class _Op, class... _Args>
-  constexpr bool is_detected_convertible_v = is_detected_convertible<To, _Op, _Args...>::value;  
+  constexpr bool is_detected_convertible_v = is_detected_convertible<To, _Op, _Args...>::value;
 
 
 _LIBCUDACXX_END_NAMESPACE_LFTS
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/unordered_map b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/unordered_map
index 221156a3e84..a110a918fe0 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/unordered_map
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/unordered_map
@@ -43,9 +43,11 @@ namespace pmr {
 #include <unordered_map>
 #include <experimental/memory_resource>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_LFTS_PMR
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/unordered_set b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/unordered_set
index 2c3243e06c7..4714c8421f5 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/unordered_set
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/unordered_set
@@ -37,9 +37,11 @@ namespace pmr {
 #include <unordered_set>
 #include <experimental/memory_resource>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_LFTS_PMR
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/utility b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/utility
index 8f7f7a149d9..6d2c15d5704 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/utility
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/utility
@@ -33,9 +33,11 @@ inline namespace fundamentals_v1 {
 #include <experimental/__config>
 #include <utility>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_LFTS
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/vector b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/vector
index d3c593467c5..31c06fca578 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/vector
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/vector
@@ -32,9 +32,11 @@ namespace pmr {
 #include <vector>
 #include <experimental/memory_resource>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_LFTS_PMR
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/ext/hash_map b/libcudacxx/include/cuda/std/detail/libcxx/include/ext/hash_map
index 292ee3156d7..4dbebc16f3b 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/ext/hash_map
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/ext/hash_map
@@ -216,9 +216,11 @@ template <class Key, class T, class Hash, class Pred, class Alloc>
 #endif
 #endif
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 namespace __gnu_cxx {
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/fenv.h b/libcudacxx/include/cuda/std/detail/libcxx/include/fenv.h
index 30b1eff084d..766517eea32 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/fenv.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/fenv.h
@@ -52,9 +52,11 @@ int feupdateenv(const fenv_t* envp);
 
 #include <__config>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #include_next <fenv.h>
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/filesystem b/libcudacxx/include/cuda/std/detail/libcxx/include/filesystem
index 73242d42cda..3ec35cdee0f 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/filesystem
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/filesystem
@@ -247,9 +247,11 @@
 
 #include <__debug>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_PUSH_MACROS
 #include <__undef_macros>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/float.h b/libcudacxx/include/cuda/std/detail/libcxx/include/float.h
index afd612c76e1..404eebabce1 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/float.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/float.h
@@ -74,9 +74,11 @@
 
 #include <__pragma_push>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #include_next <float.h>
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/forward_list b/libcudacxx/include/cuda/std/detail/libcxx/include/forward_list
index 633cc575527..25ae3efacf2 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/forward_list
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/forward_list
@@ -185,9 +185,11 @@ template <class T, class Allocator, class Predicate>
 #include <algorithm>
 #include <version>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_PUSH_MACROS
 #include <__undef_macros>
@@ -1498,7 +1500,7 @@ forward_list<_Tp, _Alloc>::remove(const value_type& __v)
         else
             ++__i;
     }
-    
+
     return (__remove_return_type) __count_removed;
 }
 
@@ -1526,7 +1528,7 @@ forward_list<_Tp, _Alloc>::remove_if(_Predicate __pred)
         else
             ++__i;
     }
-    
+
     return (__remove_return_type) __count_removed;
 }
 
@@ -1546,7 +1548,7 @@ forward_list<_Tp, _Alloc>::unique(_BinaryPredicate __binary_pred)
             __deleted_nodes.splice_after(__deleted_nodes.before_begin(), *this, __i, __j);
         __i = __j;
     }
-    
+
     return (__remove_return_type) __count_removed;
 }
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/fstream b/libcudacxx/include/cuda/std/detail/libcxx/include/fstream
index 772abc6b83a..e86f7427eae 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/fstream
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/fstream
@@ -187,9 +187,11 @@ typedef basic_fstream<wchar_t> wfstream;
 #include <cstdlib>
 #include <filesystem>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_PUSH_MACROS
 #include <__undef_macros>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/functional b/libcudacxx/include/cuda/std/detail/libcxx/include/functional
index d2c937ea655..1e13ca7a14c 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/functional
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/functional
@@ -541,8 +541,10 @@ POLICY:  For non-variadic implementations, the number of arguments is limited
 #include "type_traits"
 #include "version"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #endif  // _LIBCUDACXX_FUNCTIONAL
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/future b/libcudacxx/include/cuda/std/detail/libcxx/include/future
index c2c08623add..c911590b2aa 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/future
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/future
@@ -369,9 +369,11 @@ template <class R, class Alloc> struct uses_allocator<packaged_task<R>, Alloc>;
 #include <mutex>
 #include <thread>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #ifdef _LIBCUDACXX_HAS_NO_THREADS
 #error <future> is not supported on this single threaded system
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/initializer_list b/libcudacxx/include/cuda/std/detail/libcxx/include/initializer_list
index 0470bf4e2bd..303f8f343b7 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/initializer_list
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/initializer_list
@@ -47,9 +47,11 @@ template<class E> const E* end(initializer_list<E> il) noexcept; // constexpr in
 
 #include "cstddef"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 namespace std  // purposefully not versioned
 {
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/inttypes.h b/libcudacxx/include/cuda/std/detail/libcxx/include/inttypes.h
index c40180098d3..abc3e0896f7 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/inttypes.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/inttypes.h
@@ -237,9 +237,11 @@ uintmax_t wcstoumax(const wchar_t* restrict nptr, wchar_t** restrict endptr, int
 
 #include <__config>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 /* C99 stdlib (e.g. glibc < 2.18) does not provide format macros needed
    for C++11 unless __STDC_FORMAT_MACROS is defined
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/iomanip b/libcudacxx/include/cuda/std/detail/libcxx/include/iomanip
index 3684b01d69d..2f01d9199ed 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/iomanip
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/iomanip
@@ -47,9 +47,11 @@ template <class charT, class traits, class Allocator>
 #include <istream>
 #include <version>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/ios b/libcudacxx/include/cuda/std/detail/libcxx/include/ios
index f26d370c1e9..eb1c5607397 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/ios
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/ios
@@ -219,9 +219,11 @@ storage-class-specifier const error_category& iostream_category() noexcept;
 #include <atomic>     // for __xindex_
 #endif
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/iosfwd b/libcudacxx/include/cuda/std/detail/libcxx/include/iosfwd
index 3e7c0503804..4b705112043 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/iosfwd
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/iosfwd
@@ -100,9 +100,11 @@ typedef fpos<char_traits<wchar_t>::state_type> wstreampos;
 #include "__fwd/string.h"
 #include "version"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/iostream b/libcudacxx/include/cuda/std/detail/libcxx/include/iostream
index e6f0d86dfdf..1863a346f5b 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/iostream
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/iostream
@@ -39,9 +39,11 @@ extern wostream wclog;
 #include <istream>
 #include <ostream>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/istream b/libcudacxx/include/cuda/std/detail/libcxx/include/istream
index d09ad5c861e..2834f6dedd3 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/istream
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/istream
@@ -162,9 +162,11 @@ template <class charT, class traits, class T>
 #include <version>
 #include <ostream>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_PUSH_MACROS
 #include <__undef_macros>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/iterator b/libcudacxx/include/cuda/std/detail/libcxx/include/iterator
index 6fb9daf30ff..12b8f3a3646 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/iterator
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/iterator
@@ -729,9 +729,11 @@ template <class E> constexpr const E* data(initializer_list<E> il) noexcept;
 #include <__pragma_push>
 #endif //__cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/latch b/libcudacxx/include/cuda/std/detail/libcxx/include/latch
index c4367b65f7f..6ed62527f6b 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/latch
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/latch
@@ -50,9 +50,11 @@ namespace std
 #include <__pragma_push>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #ifdef _LIBCUDACXX_HAS_NO_THREADS
 # error <latch> is not supported on this single threaded system
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/limits b/libcudacxx/include/cuda/std/detail/libcxx/include/limits
index 93aef89a59c..942ef9e039e 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/limits
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/limits
@@ -118,9 +118,11 @@ template<> class numeric_limits<cv long double>;
 #include <__pragma_push>
 #endif //__cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #if defined(_LIBCUDACXX_COMPILER_MSVC)
 #include "support/win32/limits_msvc_win32.h"
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/limits.h b/libcudacxx/include/cuda/std/detail/libcxx/include/limits.h
index 6cb5334dee3..29a3601f7fb 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/limits.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/limits.h
@@ -41,9 +41,11 @@
 
 #include <__pragma_push>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #ifndef __GNUC__
 #include_next <limits.h>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/list b/libcudacxx/include/cuda/std/detail/libcxx/include/list
index 50e20059b2e..977062f9345 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/list
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/list
@@ -190,9 +190,11 @@ template <class T, class Allocator, class Predicate>
 
 #include <__debug>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_PUSH_MACROS
 #include <__undef_macros>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/locale b/libcudacxx/include/cuda/std/detail/libcxx/include/locale
index 08e36d539a2..5c1e61d39f8 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/locale
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/locale
@@ -207,9 +207,11 @@ template <class charT> class messages_byname;
 #include <__bsd_locale_fallbacks.h>
 #endif
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_PUSH_MACROS
 #include <__undef_macros>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/locale.h b/libcudacxx/include/cuda/std/detail/libcxx/include/locale.h
index d156f318608..29d1f4cc1ce 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/locale.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/locale.h
@@ -35,9 +35,11 @@
 
 #include <__config>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #include_next <locale.h>
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/map b/libcudacxx/include/cuda/std/detail/libcxx/include/map
index 527ca400d35..9d7c419ef3e 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/map
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/map
@@ -486,9 +486,11 @@ template <class Key, class T, class Compare, class Allocator, class Predicate>
 #include <type_traits>
 #include <version>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/math.h b/libcudacxx/include/cuda/std/detail/libcxx/include/math.h
index f6f4433a2a7..2215e722e82 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/math.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/math.h
@@ -293,9 +293,11 @@ long double    truncl(long double x);
 
 #include <__config>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #define _LIBCUDACXX_STDLIB_INCLUDE_NEXT
 #include <stdlib.h>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/memory b/libcudacxx/include/cuda/std/detail/libcxx/include/memory
index 33dc40e5d27..2dbae984671 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/memory
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/memory
@@ -704,9 +704,11 @@ void* align(size_t alignment, size_t size, void*& ptr, size_t& space);
 #include <__pragma_push>
 #endif //__cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/mutex b/libcudacxx/include/cuda/std/detail/libcxx/include/mutex
index 37e5045196b..a3979a102e1 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/mutex
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/mutex
@@ -195,9 +195,11 @@ template<class Callable, class ...Args>
 #include <version>
 #include <__threading_support>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_PUSH_MACROS
 #include <__undef_macros>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/new b/libcudacxx/include/cuda/std/detail/libcxx/include/new
index a2521caccbb..304fddcda94 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/new
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/new
@@ -106,9 +106,11 @@ void  operator delete[](void* ptr, void*) noexcept;
 #include <__pragma_push>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #if !defined(__cpp_sized_deallocation) || __cpp_sized_deallocation  < 201309L
 #define _LIBCUDACXX_HAS_NO_LANGUAGE_SIZED_DEALLOCATION
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/numeric b/libcudacxx/include/cuda/std/detail/libcxx/include/numeric
index b8ea76f905f..e240f28d99d 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/numeric
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/numeric
@@ -161,9 +161,11 @@ floating_point  midpoint(floating_point a, floating_point b);    // C++20
 #include <__pragma_push>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/optional b/libcudacxx/include/cuda/std/detail/libcxx/include/optional
index c483376f544..8112bd4c47e 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/optional
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/optional
@@ -205,9 +205,11 @@ template<class T>
 #include <__pragma_push>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #ifndef __cuda_std__
 namespace std  // purposefully not using versioning namespace
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/ostream b/libcudacxx/include/cuda/std/detail/libcxx/include/ostream
index 32125a62ed1..c196ba70cdc 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/ostream
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/ostream
@@ -142,9 +142,11 @@ template <class charT, class traits, class T>
 #include <bitset>
 #include <version>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/queue b/libcudacxx/include/cuda/std/detail/libcxx/include/queue
index 1b2a459309f..181e165c2a5 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/queue
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/queue
@@ -191,9 +191,11 @@ template <class T, class Container, class Compare>
 #include <functional>
 #include <algorithm>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/random b/libcudacxx/include/cuda/std/detail/libcxx/include/random
index c509d48e2f4..73b048748d2 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/random
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/random
@@ -1645,9 +1645,11 @@ class piecewise_linear_distribution
 #include <istream>
 #include <ostream>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_PUSH_MACROS
 #include <__undef_macros>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/ratio b/libcudacxx/include/cuda/std/detail/libcxx/include/ratio
index 9aba177c4be..672626f16fe 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/ratio
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/ratio
@@ -91,9 +91,11 @@ typedef ratio<1000000000000000000000000, 1> yotta;  // not supported
 #include <__pragma_push>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/regex b/libcudacxx/include/cuda/std/detail/libcxx/include/regex
index bff01cee3d3..46b9280b9d5 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/regex
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/regex
@@ -770,9 +770,11 @@ typedef regex_token_iterator<wstring::const_iterator> wsregex_token_iterator;
 #include <deque>
 #include <version>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_PUSH_MACROS
 #include <__undef_macros>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/scoped_allocator b/libcudacxx/include/cuda/std/detail/libcxx/include/scoped_allocator
index dc13e4ed98e..a249b9eeb6b 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/scoped_allocator
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/scoped_allocator
@@ -109,9 +109,11 @@ template <class OuterA1, class OuterA2, class... InnerAllocs>
 #include <memory>
 #include <version>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/semaphore b/libcudacxx/include/cuda/std/detail/libcxx/include/semaphore
index 5b6041a95ee..f6bfe4d41fa 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/semaphore
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/semaphore
@@ -58,9 +58,11 @@ using binary_semaphore = counting_semaphore<1>;
 #include <__pragma_push>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #ifdef _LIBCUDACXX_HAS_NO_THREADS
 # error <semaphore> is not supported on this single threaded system
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/set b/libcudacxx/include/cuda/std/detail/libcxx/include/set
index a9f458f1760..69508e9d090 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/set
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/set
@@ -429,9 +429,11 @@ template <class Key, class Compare, class Allocator, class Predicate>
 #include <functional>
 #include <version>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/setjmp.h b/libcudacxx/include/cuda/std/detail/libcxx/include/setjmp.h
index d609c166841..0c1e9ad54af 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/setjmp.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/setjmp.h
@@ -27,9 +27,11 @@ void longjmp(jmp_buf env, int val);
 
 #include <__config>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #include_next <setjmp.h>
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/shared_mutex b/libcudacxx/include/cuda/std/detail/libcxx/include/shared_mutex
index 31d0776312a..69a0f455e67 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/shared_mutex
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/shared_mutex
@@ -133,9 +133,11 @@ _LIBCUDACXX_PUSH_MACROS
 
 #include <__mutex_base>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #ifdef _LIBCUDACXX_HAS_NO_THREADS
 #error <shared_mutex> is not supported on this single threaded system
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/span b/libcudacxx/include/cuda/std/detail/libcxx/include/span
index 999d05b4633..42d880e632f 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/span
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/span
@@ -157,9 +157,11 @@ template<class R>
 #include "cstddef"      // for ptrdiff_t
 #include "version"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/sstream b/libcudacxx/include/cuda/std/detail/libcxx/include/sstream
index 6a2a01e42e7..ad25c7e7e1b 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/sstream
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/sstream
@@ -174,9 +174,11 @@ typedef basic_stringstream<wchar_t> wstringstream;
 #include <istream>
 #include <string>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_PUSH_MACROS
 #include <__undef_macros>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/stack b/libcudacxx/include/cuda/std/detail/libcxx/include/stack
index 26fdf711fec..3d0ca3223fe 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/stack
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/stack
@@ -90,9 +90,11 @@ template <class T, class Container>
 #include <__config>
 #include <deque>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/stdbool.h b/libcudacxx/include/cuda/std/detail/libcxx/include/stdbool.h
index 274609ad0d9..dae4aca58b9 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/stdbool.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/stdbool.h
@@ -21,9 +21,11 @@
 
 #include <__config>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #include_next <stdbool.h>
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/stddef.h b/libcudacxx/include/cuda/std/detail/libcxx/include/stddef.h
index fe0a0e9d8a1..1052ec45afc 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/stddef.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/stddef.h
@@ -10,9 +10,11 @@
 #if defined(__need_ptrdiff_t) || defined(__need_size_t) || \
     defined(__need_wchar_t) || defined(__need_NULL) || defined(__need_wint_t)
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #include_next <stddef.h>
 
@@ -40,9 +42,11 @@
 
 #include <__pragma_push>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #include_next <stddef.h>
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/stdexcept b/libcudacxx/include/cuda/std/detail/libcxx/include/stdexcept
index 51b7ef92cfe..02d678e510b 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/stdexcept
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/stdexcept
@@ -57,9 +57,11 @@ public:
 #include <__pragma_push>
 #endif //__cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/stdint.h b/libcudacxx/include/cuda/std/detail/libcxx/include/stdint.h
index 6a95b3362a5..189c9c087bc 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/stdint.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/stdint.h
@@ -107,9 +107,11 @@
 
 #include <__pragma_push>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 /* C99 stdlib (e.g. glibc < 2.18) does not provide macros needed
    for C++11 unless __STDC_LIMIT_MACROS and __STDC_CONSTANT_MACROS
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/stdio.h b/libcudacxx/include/cuda/std/detail/libcxx/include/stdio.h
index e793e93de9d..145319f36f7 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/stdio.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/stdio.h
@@ -9,9 +9,11 @@
 
 #if defined(__need_FILE) || defined(__need___FILE)
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #include_next <stdio.h>
 
@@ -100,9 +102,11 @@ void perror(const char* s);
 
 #include <__config>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #include_next <stdio.h>
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/stdlib.h b/libcudacxx/include/cuda/std/detail/libcxx/include/stdlib.h
index a980e7642c5..abc54deae08 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/stdlib.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/stdlib.h
@@ -9,9 +9,11 @@
 
 #if defined(__need_malloc_and_calloc) || defined(_LIBCUDACXX_STDLIB_INCLUDE_NEXT)
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #if defined(_LIBCUDACXX_STDLIB_INCLUDE_NEXT)
 #undef _LIBCUDACXX_STDLIB_INCLUDE_NEXT
@@ -90,9 +92,11 @@ void *aligned_alloc(size_t alignment, size_t size);                       // C11
 
 #include <__config>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #include_next <stdlib.h>
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/streambuf b/libcudacxx/include/cuda/std/detail/libcxx/include/streambuf
index 8a305dc2ad5..b1ddd4417d5 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/streambuf
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/streambuf
@@ -111,9 +111,11 @@ protected:
 #include <iosfwd>
 #include <ios>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_PUSH_MACROS
 #include <__undef_macros>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/string b/libcudacxx/include/cuda/std/detail/libcxx/include/string
index b338306a2e6..ee4ec891ca4 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/string
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/string
@@ -523,9 +523,11 @@ basic_string<char32_t> operator "" s( const char32_t *str, size_t len ); // C++1
 
 #include "__fwd/string.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_PUSH_MACROS
 #include <__undef_macros>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/string.h b/libcudacxx/include/cuda/std/detail/libcxx/include/string.h
index f0591447854..9f0f0a0f151 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/string.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/string.h
@@ -53,9 +53,11 @@ size_t strlen(const char* s);
 
 #include <__config>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #include_next <string.h>
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/string_view b/libcudacxx/include/cuda/std/detail/libcxx/include/string_view
index 16989dba1b3..167805405ef 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/string_view
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/string_view
@@ -183,9 +183,11 @@ namespace std {
 
 #include "__fwd/string_view.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_PUSH_MACROS
 #include <__undef_macros>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/strstream b/libcudacxx/include/cuda/std/detail/libcxx/include/strstream
index 6604c82aef2..aaeb4e063d6 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/strstream
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/strstream
@@ -130,9 +130,11 @@ private:
 #include <ostream>
 #include <istream>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/system_error b/libcudacxx/include/cuda/std/detail/libcxx/include/system_error
index 73130a53c5c..c7283f4f5b1 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/system_error
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/system_error
@@ -148,9 +148,11 @@ template <> struct hash<std::error_condition>;
 #include <__functional_base>
 #include <string>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/tgmath.h b/libcudacxx/include/cuda/std/detail/libcxx/include/tgmath.h
index a2f03bc906e..3e804cdbc28 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/tgmath.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/tgmath.h
@@ -19,9 +19,11 @@
 
 #include <__config>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #ifdef __cplusplus
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/thread b/libcudacxx/include/cuda/std/detail/libcxx/include/thread
index 8adcfd49a21..4b05aead9a4 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/thread
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/thread
@@ -96,9 +96,11 @@ void sleep_for(const chrono::duration<Rep, Period>& rel_time);
 #include <__threading_support>
 #include <__debug>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_PUSH_MACROS
 #include <__undef_macros>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/tuple b/libcudacxx/include/cuda/std/detail/libcxx/include/tuple
index 1e507edf140..bb2c8d0e241 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/tuple
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/tuple
@@ -187,9 +187,11 @@ template <class... Types>
 #include <__pragma_push>
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/type_traits b/libcudacxx/include/cuda/std/detail/libcxx/include/type_traits
index a4814ac2066..062c344a82e 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/type_traits
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/type_traits
@@ -561,9 +561,11 @@ namespace std
 #include "cstddef"
 #include "version"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/typeindex b/libcudacxx/include/cuda/std/detail/libcxx/include/typeindex
index 37f3c47ad1e..c28ce638ade 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/typeindex
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/typeindex
@@ -48,9 +48,11 @@ struct hash<type_index>
 #include <typeinfo>
 #include <__functional_base>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/typeinfo b/libcudacxx/include/cuda/std/detail/libcxx/include/typeinfo
index 81855efa921..83ac67ff077 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/typeinfo
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/typeinfo
@@ -64,9 +64,11 @@ public:
 #include <cstdlib>
 #endif
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #if defined(_LIBCUDACXX_ABI_VCRUNTIME)
 #include <vcruntime_typeinfo.h>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/unordered_map b/libcudacxx/include/cuda/std/detail/libcxx/include/unordered_map
index 06c6222525c..fa3f41a9f6b 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/unordered_map
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/unordered_map
@@ -415,9 +415,11 @@ template <class Key, class T, class Hash, class Pred, class Alloc>
 
 #include <__debug>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/unordered_set b/libcudacxx/include/cuda/std/detail/libcxx/include/unordered_set
index ed8a2a6044b..996aedbec2e 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/unordered_set
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/unordered_set
@@ -368,9 +368,11 @@ template <class Value, class Hash, class Pred, class Alloc>
 
 #include <__debug>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_BEGIN_NAMESPACE_STD
 
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/utility b/libcudacxx/include/cuda/std/detail/libcxx/include/utility
index fc5d846e897..fe81a993cc1 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/utility
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/utility
@@ -292,8 +292,10 @@ template <class T>
 #include "__tuple_dir/tuple_element.h"
 #include "__tuple_dir/tuple_size.h"
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #endif  // _LIBCUDACXX_UTILITY
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/valarray b/libcudacxx/include/cuda/std/detail/libcxx/include/valarray
index 82a34d70437..728f4a6134d 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/valarray
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/valarray
@@ -361,9 +361,11 @@ template <class T> unspecified2 end(const valarray<T>& v);
 // [valarray.syn]
 #include <initializer_list>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_PUSH_MACROS
 #include <__undef_macros>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/variant b/libcudacxx/include/cuda/std/detail/libcxx/include/variant
index a799e753b54..403d7122eff 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/variant
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/variant
@@ -208,9 +208,11 @@ namespace std {
 #include <limits>
 #include <version>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_PUSH_MACROS
 #include <__undef_macros>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/vector b/libcudacxx/include/cuda/std/detail/libcxx/include/vector
index 3b975c5dc8f..f18ce0d8602 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/vector
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/vector
@@ -286,9 +286,11 @@ template <class T, class Allocator, class Predicate>
 
 #include <__debug>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 _LIBCUDACXX_PUSH_MACROS
 #include <__undef_macros>
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/version b/libcudacxx/include/cuda/std/detail/libcxx/include/version
index d6b4b45fdba..a2c91e81cd2 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/version
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/version
@@ -215,9 +215,11 @@ __cpp_lib_void_t                                        201411L <type_traits>
 #endif
 #endif // __cuda_std__
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 // We define certain feature test macros depending on availability. When
 // __CUDA_MINIMUM_ARCH__ is not available, we define the following features
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/wchar.h b/libcudacxx/include/cuda/std/detail/libcxx/include/wchar.h
index 999837c3893..4981c861720 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/wchar.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/wchar.h
@@ -9,9 +9,11 @@
 
 #if defined(__need_wint_t) || defined(__need_mbstate_t)
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #include_next <wchar.h>
 
@@ -107,9 +109,11 @@ size_t wcsrtombs(char* restrict dst, const wchar_t** restrict src, size_t len,
 
 #include <__config>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #ifdef __cplusplus
 #define __CORRECT_ISO_CPP_WCHAR_H_PROTO
diff --git a/libcudacxx/include/cuda/std/detail/libcxx/include/wctype.h b/libcudacxx/include/cuda/std/detail/libcxx/include/wctype.h
index ad4a88a38ae..dd5c13cc85d 100644
--- a/libcudacxx/include/cuda/std/detail/libcxx/include/wctype.h
+++ b/libcudacxx/include/cuda/std/detail/libcxx/include/wctype.h
@@ -46,9 +46,11 @@ wctrans_t wctrans(const char* property);
 
 #include <__config>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #include_next <wctype.h>
 
diff --git a/libcudacxx/libcxx/utils/generate_feature_test_macro_components.py b/libcudacxx/libcxx/utils/generate_feature_test_macro_components.py
index b34d05501b7..2e1b6d8a42b 100755
--- a/libcudacxx/libcxx/utils/generate_feature_test_macro_components.py
+++ b/libcudacxx/libcxx/utils/generate_feature_test_macro_components.py
@@ -716,9 +716,11 @@ def produce_version_header():
 
 #include <__config>
 
-#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
 #pragma GCC system_header
-#endif
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #if _LIBCUDACXX_STD_VER > 11
 {cxx14_macros}
diff --git a/thrust/cmake/ThrustBuildCompilerTargets.cmake b/thrust/cmake/ThrustBuildCompilerTargets.cmake
index f65cb1bd95b..0473b3a5057 100644
--- a/thrust/cmake/ThrustBuildCompilerTargets.cmake
+++ b/thrust/cmake/ThrustBuildCompilerTargets.cmake
@@ -26,6 +26,9 @@ function(thrust_build_compiler_targets)
 
   thrust_update_system_found_flags()
 
+  # Ensure that we build our tests without treating ourself as system header
+  list(APPEND cxx_compile_definitions "_CCCL_NO_SYSTEM_HEADER")
+
   if ("MSVC" STREQUAL "${CMAKE_CXX_COMPILER_ID}")
     append_option_if_available("/W4" cxx_compile_options)
 
diff --git a/thrust/thrust/addressof.h b/thrust/thrust/addressof.h
index d21df0c76a5..ac1af8a3aec 100644
--- a/thrust/thrust/addressof.h
+++ b/thrust/thrust/addressof.h
@@ -7,6 +7,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_CPP_DIALECT >= 2011
 #  include <thrust/detail/memory_wrapper.h>
 #endif
@@ -19,7 +25,7 @@ THRUST_NAMESPACE_BEGIN
  */
 template <typename T>
 __host__ __device__
-T* addressof(T& arg) 
+T* addressof(T& arg)
 {
   return reinterpret_cast<T*>(
     &const_cast<char&>(reinterpret_cast<const volatile char&>(arg))
diff --git a/thrust/thrust/adjacent_difference.h b/thrust/thrust/adjacent_difference.h
index e8385c240dd..6aa67465554 100644
--- a/thrust/thrust/adjacent_difference.h
+++ b/thrust/thrust/adjacent_difference.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
@@ -81,7 +87,7 @@ THRUST_NAMESPACE_BEGIN
 template<typename DerivedPolicy, typename InputIterator, typename OutputIterator>
 __host__ __device__
 OutputIterator adjacent_difference(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
-                                   InputIterator first, InputIterator last, 
+                                   InputIterator first, InputIterator last,
                                    OutputIterator result);
 
 /*! \p adjacent_difference calculates the differences of adjacent elements in the
@@ -89,7 +95,7 @@ OutputIterator adjacent_difference(const thrust::detail::execution_policy_base<D
  *  <tt>\*result</tt>, and, for each iterator \p i in the range
  *  <tt>[first + 1, last)</tt>, <tt>binary_op(\*i, \*(i - 1))</tt> is assigned to
  *  <tt>\*(result + (i - first))</tt>.
- *  
+ *
  *  This version of \p adjacent_difference uses the binary function \p binary_op to
  *  calculate differences.
  *
@@ -183,7 +189,7 @@ OutputIterator adjacent_difference(const thrust::detail::execution_policy_base<D
  *  \see inclusive_scan
  */
 template <typename InputIterator, typename OutputIterator>
-OutputIterator adjacent_difference(InputIterator first, InputIterator last, 
+OutputIterator adjacent_difference(InputIterator first, InputIterator last,
                                    OutputIterator result);
 
 /*! \p adjacent_difference calculates the differences of adjacent elements in the
@@ -191,7 +197,7 @@ OutputIterator adjacent_difference(InputIterator first, InputIterator last,
  *  <tt>\*result</tt>, and, for each iterator \p i in the range
  *  <tt>[first + 1, last)</tt>, <tt>binary_op(\*i, \*(i - 1))</tt> is assigned to
  *  <tt>\*(result + (i - first))</tt>.
- *  
+ *
  *  This version of \p adjacent_difference uses the binary function \p binary_op to
  *  calculate differences.
  *
diff --git a/thrust/thrust/advance.h b/thrust/thrust/advance.h
index a5162e20300..e7b36236522 100644
--- a/thrust/thrust/advance.h
+++ b/thrust/thrust/advance.h
@@ -23,6 +23,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 THRUST_NAMESPACE_BEGIN
 
 /*! \addtogroup iterators
diff --git a/thrust/thrust/allocate_unique.h b/thrust/thrust/allocate_unique.h
index ff10cb51cd1..bd3ca33ded7 100644
--- a/thrust/thrust/allocate_unique.h
+++ b/thrust/thrust/allocate_unique.h
@@ -6,6 +6,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp11_required.h>
 
 #if THRUST_CPP_DIALECT >= 2011
@@ -245,7 +251,7 @@ struct array_allocator_delete final
   allocator_type alloc_;
   std::size_t    count_;
 };
-  
+
 template <typename T, typename Allocator>
 using uninitialized_array_allocator_delete
   = array_allocator_delete<T, Allocator, true>;
diff --git a/thrust/thrust/async/copy.h b/thrust/thrust/async/copy.h
index a8edc7411e5..5fc18424c74 100644
--- a/thrust/thrust/async/copy.h
+++ b/thrust/thrust/async/copy.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp14_required.h>
 
 #if THRUST_CPP_DIALECT >= 2014
diff --git a/thrust/thrust/async/for_each.h b/thrust/thrust/async/for_each.h
index 0d3b3a18905..8dd198c15b3 100644
--- a/thrust/thrust/async/for_each.h
+++ b/thrust/thrust/async/for_each.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp14_required.h>
 
 #if THRUST_CPP_DIALECT >= 2014
diff --git a/thrust/thrust/async/reduce.h b/thrust/thrust/async/reduce.h
index 8f4fe3133e6..893c3c5b132 100644
--- a/thrust/thrust/async/reduce.h
+++ b/thrust/thrust/async/reduce.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp14_required.h>
 
 #if THRUST_CPP_DIALECT >= 2014
diff --git a/thrust/thrust/async/scan.h b/thrust/thrust/async/scan.h
index 1bcf8125780..56ef43f317a 100644
--- a/thrust/thrust/async/scan.h
+++ b/thrust/thrust/async/scan.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp14_required.h>
 
 #if THRUST_CPP_DIALECT >= 2014
diff --git a/thrust/thrust/async/sort.h b/thrust/thrust/async/sort.h
index 8881793970b..551f599035e 100644
--- a/thrust/thrust/async/sort.h
+++ b/thrust/thrust/async/sort.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp14_required.h>
 
 #if THRUST_CPP_DIALECT >= 2014
diff --git a/thrust/thrust/async/transform.h b/thrust/thrust/async/transform.h
index de72549bf01..3f4d2af526c 100644
--- a/thrust/thrust/async/transform.h
+++ b/thrust/thrust/async/transform.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp14_required.h>
 
 #if THRUST_CPP_DIALECT >= 2014
diff --git a/thrust/thrust/binary_search.h b/thrust/thrust/binary_search.h
index 7a4746e0b16..c6633ce18fd 100644
--- a/thrust/thrust/binary_search.h
+++ b/thrust/thrust/binary_search.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 #include <thrust/pair.h>
 
@@ -43,18 +49,18 @@ THRUST_NAMESPACE_BEGIN
  */
 
 
-//////////////////////   
+//////////////////////
 // Scalar Functions //
 //////////////////////
 
 
 /*! \p lower_bound is a version of binary search: it attempts to find
- * the element value in an ordered range <tt>[first, last)</tt>. 
+ * the element value in an ordered range <tt>[first, last)</tt>.
  * Specifically, it returns the first position where value could be
- * inserted without violating the ordering. This version of 
+ * inserted without violating the ordering. This version of
  * \p lower_bound uses <tt>operator<</tt> for comparison and returns
  * the furthermost iterator \c i in <tt>[first, last)</tt> such that,
- * for every iterator \c j in <tt>[first, i)</tt>, <tt>*j < value</tt>. 
+ * for every iterator \c j in <tt>[first, i)</tt>, <tt>*j < value</tt>.
  *
  * The algorithm's execution is parallelized as determined by \p exec.
  *
@@ -63,10 +69,10 @@ THRUST_NAMESPACE_BEGIN
  *  \param last The end of the ordered sequence.
  *  \param value The value to be searched.
  *  \return The furthermost iterator \c i, such that <tt>*i < value</tt>.
- * 
+ *
  *  \tparam DerivedPolicy The name of the derived execution policy.
  *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
- *  \tparam LessThanComparable is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>. 
+ *  \tparam LessThanComparable is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>.
  *
  *  The following code snippet demonstrates how to use \p lower_bound
  *  to search for values in a ordered range using the \p thrust::device execution policy for parallelization:
@@ -106,20 +112,20 @@ ForwardIterator lower_bound(const thrust::detail::execution_policy_base<DerivedP
 
 
 /*! \p lower_bound is a version of binary search: it attempts to find
- * the element value in an ordered range <tt>[first, last)</tt>. 
+ * the element value in an ordered range <tt>[first, last)</tt>.
  * Specifically, it returns the first position where value could be
- * inserted without violating the ordering. This version of 
+ * inserted without violating the ordering. This version of
  * \p lower_bound uses <tt>operator<</tt> for comparison and returns
  * the furthermost iterator \c i in <tt>[first, last)</tt> such that,
- * for every iterator \c j in <tt>[first, i)</tt>, <tt>*j < value</tt>. 
+ * for every iterator \c j in <tt>[first, i)</tt>, <tt>*j < value</tt>.
  *
  *  \param first The beginning of the ordered sequence.
  *  \param last The end of the ordered sequence.
  *  \param value The value to be searched.
  *  \return The furthermost iterator \c i, such that <tt>*i < value</tt>.
- * 
+ *
  *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
- *  \tparam LessThanComparable is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>. 
+ *  \tparam LessThanComparable is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>.
  *
  *  The following code snippet demonstrates how to use \p lower_bound
  *  to search for values in a ordered range.
@@ -150,19 +156,19 @@ ForwardIterator lower_bound(const thrust::detail::execution_policy_base<DerivedP
  *  \see \p binary_search
  */
 template <class ForwardIterator, class LessThanComparable>
-ForwardIterator lower_bound(ForwardIterator first, 
+ForwardIterator lower_bound(ForwardIterator first,
                             ForwardIterator last,
                             const LessThanComparable& value);
 
 
 /*! \p lower_bound is a version of binary search: it attempts to find
- * the element value in an ordered range <tt>[first, last)</tt>. 
+ * the element value in an ordered range <tt>[first, last)</tt>.
  * Specifically, it returns the first position where value could be
- * inserted without violating the ordering. This version of 
- * \p lower_bound uses function object \c comp for comparison 
+ * inserted without violating the ordering. This version of
+ * \p lower_bound uses function object \c comp for comparison
  * and returns the furthermost iterator \c i in <tt>[first, last)</tt>
- * such that, for every iterator \c j in <tt>[first, i)</tt>, 
- * <tt>comp(*j, value)</tt> is \c true. 
+ * such that, for every iterator \c j in <tt>[first, i)</tt>,
+ * <tt>comp(*j, value)</tt> is \c true.
  *
  * The algorithm's execution is parallelized as determined by \p exec.
  *
@@ -172,7 +178,7 @@ ForwardIterator lower_bound(ForwardIterator first,
  *  \param value The value to be searched.
  *  \param comp The comparison operator.
  *  \return The furthermost iterator \c i, such that <tt>comp(*i, value)</tt> is \c true.
- * 
+ *
  *  \tparam DerivedPolicy The name of the derived execution policy.
  *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
  *  \tparam T is comparable to \p ForwardIterator's \c value_type.
@@ -218,20 +224,20 @@ ForwardIterator lower_bound(const thrust::detail::execution_policy_base<DerivedP
 
 
 /*! \p lower_bound is a version of binary search: it attempts to find
- * the element value in an ordered range <tt>[first, last)</tt>. 
+ * the element value in an ordered range <tt>[first, last)</tt>.
  * Specifically, it returns the first position where value could be
- * inserted without violating the ordering. This version of 
- * \p lower_bound uses function object \c comp for comparison 
+ * inserted without violating the ordering. This version of
+ * \p lower_bound uses function object \c comp for comparison
  * and returns the furthermost iterator \c i in <tt>[first, last)</tt>
- * such that, for every iterator \c j in <tt>[first, i)</tt>, 
- * <tt>comp(*j, value)</tt> is \c true. 
+ * such that, for every iterator \c j in <tt>[first, i)</tt>,
+ * <tt>comp(*j, value)</tt> is \c true.
  *
  *  \param first The beginning of the ordered sequence.
  *  \param last The end of the ordered sequence.
  *  \param value The value to be searched.
  *  \param comp The comparison operator.
  *  \return The furthermost iterator \c i, such that <tt>comp(*i, value)</tt> is \c true.
- * 
+ *
  *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
  *  \tparam T is comparable to \p ForwardIterator's \c value_type.
  *  \tparam StrictWeakOrdering is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
@@ -268,14 +274,14 @@ ForwardIterator lower_bound(const thrust::detail::execution_policy_base<DerivedP
 template <class ForwardIterator, class T, class StrictWeakOrdering>
 ForwardIterator lower_bound(ForwardIterator first,
                             ForwardIterator last,
-                            const T& value, 
+                            const T& value,
                             StrictWeakOrdering comp);
 
 
 /*! \p upper_bound is a version of binary search: it attempts to find
- * the element value in an ordered range <tt>[first, last)</tt>. 
+ * the element value in an ordered range <tt>[first, last)</tt>.
  * Specifically, it returns the last position where value could be
- * inserted without violating the ordering. This version of 
+ * inserted without violating the ordering. This version of
  * \p upper_bound uses <tt>operator<</tt> for comparison and returns
  * the furthermost iterator \c i in <tt>[first, last)</tt> such that,
  * for every iterator \c j in <tt>[first, i)</tt>, <tt>value < *j</tt>
@@ -288,10 +294,10 @@ ForwardIterator lower_bound(ForwardIterator first,
  *  \param last The end of the ordered sequence.
  *  \param value The value to be searched.
  *  \return The furthermost iterator \c i, such that <tt>value < *i</tt> is \c false.
- * 
+ *
  *  \tparam DerivedPolicy The name of the derived execution policy.
  *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
- *  \tparam LessThanComparable is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>. 
+ *  \tparam LessThanComparable is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>.
  *
  *  The following code snippet demonstrates how to use \p upper_bound
  *  to search for values in a ordered range using the \p thrust::device execution policy for parallelism:
@@ -331,9 +337,9 @@ ForwardIterator upper_bound(const thrust::detail::execution_policy_base<DerivedP
 
 
 /*! \p upper_bound is a version of binary search: it attempts to find
- * the element value in an ordered range <tt>[first, last)</tt>. 
+ * the element value in an ordered range <tt>[first, last)</tt>.
  * Specifically, it returns the last position where value could be
- * inserted without violating the ordering. This version of 
+ * inserted without violating the ordering. This version of
  * \p upper_bound uses <tt>operator<</tt> for comparison and returns
  * the furthermost iterator \c i in <tt>[first, last)</tt> such that,
  * for every iterator \c j in <tt>[first, i)</tt>, <tt>value < *j</tt>
@@ -343,9 +349,9 @@ ForwardIterator upper_bound(const thrust::detail::execution_policy_base<DerivedP
  *  \param last The end of the ordered sequence.
  *  \param value The value to be searched.
  *  \return The furthermost iterator \c i, such that <tt>value < *i</tt> is \c false.
- * 
+ *
  *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
- *  \tparam LessThanComparable is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>. 
+ *  \tparam LessThanComparable is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>.
  *
  *  The following code snippet demonstrates how to use \p upper_bound
  *  to search for values in a ordered range.
@@ -376,15 +382,15 @@ ForwardIterator upper_bound(const thrust::detail::execution_policy_base<DerivedP
  *  \see \p binary_search
  */
 template <class ForwardIterator, class LessThanComparable>
-ForwardIterator upper_bound(ForwardIterator first, 
+ForwardIterator upper_bound(ForwardIterator first,
                             ForwardIterator last,
                             const LessThanComparable& value);
 
 
 /*! \p upper_bound is a version of binary search: it attempts to find
- * the element value in an ordered range <tt>[first, last)</tt>. 
+ * the element value in an ordered range <tt>[first, last)</tt>.
  * Specifically, it returns the last position where value could be
- * inserted without violating the ordering. This version of 
+ * inserted without violating the ordering. This version of
  * \p upper_bound uses function object \c comp for comparison and returns
  * the furthermost iterator \c i in <tt>[first, last)</tt> such that,
  * for every iterator \c j in <tt>[first, i)</tt>, <tt>comp(value, *j)</tt>
@@ -398,7 +404,7 @@ ForwardIterator upper_bound(ForwardIterator first,
  *  \param value The value to be searched.
  *  \param comp The comparison operator.
  *  \return The furthermost iterator \c i, such that <tt>comp(value, *i)</tt> is \c false.
- * 
+ *
  *  \tparam DerivedPolicy The name of the derived execution policy.
  *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
  *  \tparam T is comparable to \p ForwardIterator's \c value_type.
@@ -443,9 +449,9 @@ ForwardIterator upper_bound(const thrust::detail::execution_policy_base<DerivedP
                             StrictWeakOrdering comp);
 
 /*! \p upper_bound is a version of binary search: it attempts to find
- * the element value in an ordered range <tt>[first, last)</tt>. 
+ * the element value in an ordered range <tt>[first, last)</tt>.
  * Specifically, it returns the last position where value could be
- * inserted without violating the ordering. This version of 
+ * inserted without violating the ordering. This version of
  * \p upper_bound uses function object \c comp for comparison and returns
  * the furthermost iterator \c i in <tt>[first, last)</tt> such that,
  * for every iterator \c j in <tt>[first, i)</tt>, <tt>comp(value, *j)</tt>
@@ -456,7 +462,7 @@ ForwardIterator upper_bound(const thrust::detail::execution_policy_base<DerivedP
  *  \param value The value to be searched.
  *  \param comp The comparison operator.
  *  \return The furthermost iterator \c i, such that <tt>comp(value, *i)</tt> is \c false.
- * 
+ *
  *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
  *  \tparam T is comparable to \p ForwardIterator's \c value_type.
  *  \tparam StrictWeakOrdering is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
@@ -493,16 +499,16 @@ ForwardIterator upper_bound(const thrust::detail::execution_policy_base<DerivedP
 template <class ForwardIterator, class T, class StrictWeakOrdering>
 ForwardIterator upper_bound(ForwardIterator first,
                             ForwardIterator last,
-                            const T& value, 
+                            const T& value,
                             StrictWeakOrdering comp);
 
 
 /*! \p binary_search is a version of binary search: it attempts to find
- * the element value in an ordered range <tt>[first, last)</tt>. 
- * It returns \c true if an element that is equivalent to \c value 
+ * the element value in an ordered range <tt>[first, last)</tt>.
+ * It returns \c true if an element that is equivalent to \c value
  * is present in <tt>[first, last)</tt> and \c false if no such element
- * exists.  Specifically, this version returns \c true if and only if 
- * there exists an iterator \c i in <tt>[first, last)</tt> such that 
+ * exists.  Specifically, this version returns \c true if and only if
+ * there exists an iterator \c i in <tt>[first, last)</tt> such that
  * <tt>*i < value</tt> and <tt>value < *i</tt> are both \c false.
  *
  * The algorithm's execution is parallelized as determined by \p exec.
@@ -512,10 +518,10 @@ ForwardIterator upper_bound(ForwardIterator first,
  *  \param last The end of the ordered sequence.
  *  \param value The value to be searched.
  *  \return \c true if an equivalent element exists in <tt>[first, last)</tt>, otherwise \c false.
- * 
+ *
  *  \tparam DerivedPolicy The name of the derived execution policy.
  *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
- *  \tparam LessThanComparable is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>. 
+ *  \tparam LessThanComparable is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>.
  *
  *  The following code snippet demonstrates how to use \p binary_search
  *  to search for values in a ordered range using the \p thrust::device execution policy for parallelization:
@@ -549,26 +555,26 @@ ForwardIterator upper_bound(ForwardIterator first,
 template <typename DerivedPolicy, typename ForwardIterator, typename LessThanComparable>
 __host__ __device__
 bool binary_search(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
-                   ForwardIterator first, 
+                   ForwardIterator first,
                    ForwardIterator last,
                    const LessThanComparable& value);
 
 
 /*! \p binary_search is a version of binary search: it attempts to find
- * the element value in an ordered range <tt>[first, last)</tt>. 
- * It returns \c true if an element that is equivalent to \c value 
+ * the element value in an ordered range <tt>[first, last)</tt>.
+ * It returns \c true if an element that is equivalent to \c value
  * is present in <tt>[first, last)</tt> and \c false if no such element
- * exists.  Specifically, this version returns \c true if and only if 
- * there exists an iterator \c i in <tt>[first, last)</tt> such that 
+ * exists.  Specifically, this version returns \c true if and only if
+ * there exists an iterator \c i in <tt>[first, last)</tt> such that
  * <tt>*i < value</tt> and <tt>value < *i</tt> are both \c false.
  *
  *  \param first The beginning of the ordered sequence.
  *  \param last The end of the ordered sequence.
  *  \param value The value to be searched.
  *  \return \c true if an equivalent element exists in <tt>[first, last)</tt>, otherwise \c false.
- * 
+ *
  *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
- *  \tparam LessThanComparable is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>. 
+ *  \tparam LessThanComparable is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>.
  *
  *  The following code snippet demonstrates how to use \p binary_search
  *  to search for values in a ordered range.
@@ -599,17 +605,17 @@ bool binary_search(const thrust::detail::execution_policy_base<DerivedPolicy> &e
  *  \see \p equal_range
  */
 template <class ForwardIterator, class LessThanComparable>
-bool binary_search(ForwardIterator first, 
+bool binary_search(ForwardIterator first,
                    ForwardIterator last,
                    const LessThanComparable& value);
 
 
 /*! \p binary_search is a version of binary search: it attempts to find
- * the element value in an ordered range <tt>[first, last)</tt>. 
- * It returns \c true if an element that is equivalent to \c value 
+ * the element value in an ordered range <tt>[first, last)</tt>.
+ * It returns \c true if an element that is equivalent to \c value
  * is present in <tt>[first, last)</tt> and \c false if no such element
- * exists.  Specifically, this version returns \c true if and only if 
- * there exists an iterator \c i in <tt>[first, last)</tt> such that 
+ * exists.  Specifically, this version returns \c true if and only if
+ * there exists an iterator \c i in <tt>[first, last)</tt> such that
  * <tt>comp(*i, value)</tt> and <tt>comp(value, *i)</tt> are both \c false.
  *
  * The algorithm's execution is parallelized as determined by \p exec.
@@ -620,7 +626,7 @@ bool binary_search(ForwardIterator first,
  *  \param value The value to be searched.
  *  \param comp The comparison operator.
  *  \return \c true if an equivalent element exists in <tt>[first, last)</tt>, otherwise \c false.
- * 
+ *
  *  \tparam DerivedPolicy The name of the derived execution policy.
  *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
  *  \tparam T is comparable to \p ForwardIterator's \c value_type.
@@ -661,16 +667,16 @@ __host__ __device__
 bool binary_search(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
                    ForwardIterator first,
                    ForwardIterator last,
-                   const T& value, 
+                   const T& value,
                    StrictWeakOrdering comp);
 
 
 /*! \p binary_search is a version of binary search: it attempts to find
- * the element value in an ordered range <tt>[first, last)</tt>. 
- * It returns \c true if an element that is equivalent to \c value 
+ * the element value in an ordered range <tt>[first, last)</tt>.
+ * It returns \c true if an element that is equivalent to \c value
  * is present in <tt>[first, last)</tt> and \c false if no such element
- * exists.  Specifically, this version returns \c true if and only if 
- * there exists an iterator \c i in <tt>[first, last)</tt> such that 
+ * exists.  Specifically, this version returns \c true if and only if
+ * there exists an iterator \c i in <tt>[first, last)</tt> such that
  * <tt>comp(*i, value)</tt> and <tt>comp(value, *i)</tt> are both \c false.
  *
  *  \param first The beginning of the ordered sequence.
@@ -678,7 +684,7 @@ bool binary_search(const thrust::detail::execution_policy_base<DerivedPolicy> &e
  *  \param value The value to be searched.
  *  \param comp The comparison operator.
  *  \return \c true if an equivalent element exists in <tt>[first, last)</tt>, otherwise \c false.
- * 
+ *
  *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
  *  \tparam T is comparable to \p ForwardIterator's \c value_type.
  *  \tparam StrictWeakOrdering is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
@@ -715,29 +721,29 @@ bool binary_search(const thrust::detail::execution_policy_base<DerivedPolicy> &e
 template <class ForwardIterator, class T, class StrictWeakOrdering>
 bool binary_search(ForwardIterator first,
                    ForwardIterator last,
-                   const T& value, 
+                   const T& value,
                    StrictWeakOrdering comp);
 
 
 /*! \p equal_range is a version of binary search: it attempts to find
- * the element value in an ordered range <tt>[first, last)</tt>. The 
+ * the element value in an ordered range <tt>[first, last)</tt>. The
  * value returned by \p equal_range is essentially a combination of
  * the values returned by \p lower_bound and \p upper_bound: it returns
  * a \p pair of iterators \c i and \c j such that \c i is the first
- * position where value could be inserted without violating the 
+ * position where value could be inserted without violating the
  * ordering and \c j is the last position where value could be inserted
  * without violating the ordering. It follows that every element in the
- * range <tt>[i, j)</tt> is equivalent to value, and that 
+ * range <tt>[i, j)</tt> is equivalent to value, and that
  * <tt>[i, j)</tt> is the largest subrange of <tt>[first, last)</tt> that
- * has this property. 
+ * has this property.
  *
- * This version of \p equal_range returns a \p pair of iterators 
- * <tt>[i, j)</tt>, where \c i is the furthermost iterator in 
- * <tt>[first, last)</tt> such that, for every iterator \c k in 
+ * This version of \p equal_range returns a \p pair of iterators
+ * <tt>[i, j)</tt>, where \c i is the furthermost iterator in
+ * <tt>[first, last)</tt> such that, for every iterator \c k in
  * <tt>[first, i)</tt>, <tt>*k < value</tt>.  \c j is the furthermost
- * iterator in <tt>[first, last)</tt> such that, for every iterator 
- * \c k in <tt>[first, j)</tt>, <tt>value < *k</tt> is \c false. 
- * For every iterator \c k in <tt>[i, j)</tt>, neither 
+ * iterator in <tt>[first, last)</tt> such that, for every iterator
+ * \c k in <tt>[first, j)</tt>, <tt>value < *k</tt> is \c false.
+ * For every iterator \c k in <tt>[i, j)</tt>, neither
  * <tt>value < *k</tt> nor <tt>*k < value</tt> is \c true.
  *
  * The algorithm's execution is parallelized as determined by \p exec.
@@ -747,10 +753,10 @@ bool binary_search(ForwardIterator first,
  *  \param last The end of the ordered sequence.
  *  \param value The value to be searched.
  *  \return A \p pair of iterators <tt>[i, j)</tt> that define the range of equivalent elements.
- * 
+ *
  *  \tparam DerivedPolicy The name of the derived execution policy.
  *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
- *  \tparam LessThanComparable is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>. 
+ *  \tparam LessThanComparable is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>.
  *
  *  The following code snippet demonstrates how to use \p equal_range
  *  to search for values in a ordered range using the \p thrust::device execution policy for parallelization:
@@ -791,33 +797,33 @@ equal_range(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
 
 
 /*! \p equal_range is a version of binary search: it attempts to find
- * the element value in an ordered range <tt>[first, last)</tt>. The 
+ * the element value in an ordered range <tt>[first, last)</tt>. The
  * value returned by \p equal_range is essentially a combination of
  * the values returned by \p lower_bound and \p upper_bound: it returns
  * a \p pair of iterators \c i and \c j such that \c i is the first
- * position where value could be inserted without violating the 
+ * position where value could be inserted without violating the
  * ordering and \c j is the last position where value could be inserted
  * without violating the ordering. It follows that every element in the
- * range <tt>[i, j)</tt> is equivalent to value, and that 
+ * range <tt>[i, j)</tt> is equivalent to value, and that
  * <tt>[i, j)</tt> is the largest subrange of <tt>[first, last)</tt> that
- * has this property. 
+ * has this property.
  *
- * This version of \p equal_range returns a \p pair of iterators 
- * <tt>[i, j)</tt>, where \c i is the furthermost iterator in 
- * <tt>[first, last)</tt> such that, for every iterator \c k in 
+ * This version of \p equal_range returns a \p pair of iterators
+ * <tt>[i, j)</tt>, where \c i is the furthermost iterator in
+ * <tt>[first, last)</tt> such that, for every iterator \c k in
  * <tt>[first, i)</tt>, <tt>*k < value</tt>.  \c j is the furthermost
- * iterator in <tt>[first, last)</tt> such that, for every iterator 
- * \c k in <tt>[first, j)</tt>, <tt>value < *k</tt> is \c false. 
- * For every iterator \c k in <tt>[i, j)</tt>, neither 
+ * iterator in <tt>[first, last)</tt> such that, for every iterator
+ * \c k in <tt>[first, j)</tt>, <tt>value < *k</tt> is \c false.
+ * For every iterator \c k in <tt>[i, j)</tt>, neither
  * <tt>value < *k</tt> nor <tt>*k < value</tt> is \c true.
  *
  *  \param first The beginning of the ordered sequence.
  *  \param last The end of the ordered sequence.
  *  \param value The value to be searched.
  *  \return A \p pair of iterators <tt>[i, j)</tt> that define the range of equivalent elements.
- * 
+ *
  *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
- *  \tparam LessThanComparable is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>. 
+ *  \tparam LessThanComparable is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>.
  *
  *  The following code snippet demonstrates how to use \p equal_range
  *  to search for values in a ordered range.
@@ -855,25 +861,25 @@ equal_range(ForwardIterator first,
 
 
 /*! \p equal_range is a version of binary search: it attempts to find
- * the element value in an ordered range <tt>[first, last)</tt>. The 
+ * the element value in an ordered range <tt>[first, last)</tt>. The
  * value returned by \p equal_range is essentially a combination of
  * the values returned by \p lower_bound and \p upper_bound: it returns
  * a \p pair of iterators \c i and \c j such that \c i is the first
- * position where value could be inserted without violating the 
+ * position where value could be inserted without violating the
  * ordering and \c j is the last position where value could be inserted
  * without violating the ordering. It follows that every element in the
- * range <tt>[i, j)</tt> is equivalent to value, and that 
+ * range <tt>[i, j)</tt> is equivalent to value, and that
  * <tt>[i, j)</tt> is the largest subrange of <tt>[first, last)</tt> that
- * has this property. 
+ * has this property.
  *
- * This version of \p equal_range returns a \p pair of iterators 
- * <tt>[i, j)</tt>. \c i is the furthermost iterator in 
- * <tt>[first, last)</tt> such that, for every iterator \c k in 
+ * This version of \p equal_range returns a \p pair of iterators
+ * <tt>[i, j)</tt>. \c i is the furthermost iterator in
+ * <tt>[first, last)</tt> such that, for every iterator \c k in
  * <tt>[first, i)</tt>, <tt>comp(*k, value)</tt> is \c true.
  * \c j is the furthermost iterator in <tt>[first, last)</tt> such
- * that, for every iterator \c k in <tt>[first, last)</tt>, 
- * <tt>comp(value, *k)</tt> is \c false. For every iterator \c k 
- * in <tt>[i, j)</tt>, neither <tt>comp(value, *k)</tt> nor 
+ * that, for every iterator \c k in <tt>[first, last)</tt>,
+ * <tt>comp(value, *k)</tt> is \c false. For every iterator \c k
+ * in <tt>[i, j)</tt>, neither <tt>comp(value, *k)</tt> nor
  * <tt>comp(*k, value)</tt> is \c true.
  *
  * The algorithm's execution is parallelized as determined by \p exec.
@@ -884,7 +890,7 @@ equal_range(ForwardIterator first,
  *  \param value The value to be searched.
  *  \param comp The comparison operator.
  *  \return A \p pair of iterators <tt>[i, j)</tt> that define the range of equivalent elements.
- * 
+ *
  *  \tparam DerivedPolicy The name of the derived execution policy.
  *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
  *  \tparam T is comparable to \p ForwardIterator's \c value_type.
@@ -931,25 +937,25 @@ equal_range(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
 
 
 /*! \p equal_range is a version of binary search: it attempts to find
- * the element value in an ordered range <tt>[first, last)</tt>. The 
+ * the element value in an ordered range <tt>[first, last)</tt>. The
  * value returned by \p equal_range is essentially a combination of
  * the values returned by \p lower_bound and \p upper_bound: it returns
  * a \p pair of iterators \c i and \c j such that \c i is the first
- * position where value could be inserted without violating the 
+ * position where value could be inserted without violating the
  * ordering and \c j is the last position where value could be inserted
  * without violating the ordering. It follows that every element in the
- * range <tt>[i, j)</tt> is equivalent to value, and that 
+ * range <tt>[i, j)</tt> is equivalent to value, and that
  * <tt>[i, j)</tt> is the largest subrange of <tt>[first, last)</tt> that
- * has this property. 
+ * has this property.
  *
- * This version of \p equal_range returns a \p pair of iterators 
- * <tt>[i, j)</tt>. \c i is the furthermost iterator in 
- * <tt>[first, last)</tt> such that, for every iterator \c k in 
+ * This version of \p equal_range returns a \p pair of iterators
+ * <tt>[i, j)</tt>. \c i is the furthermost iterator in
+ * <tt>[first, last)</tt> such that, for every iterator \c k in
  * <tt>[first, i)</tt>, <tt>comp(*k, value)</tt> is \c true.
  * \c j is the furthermost iterator in <tt>[first, last)</tt> such
- * that, for every iterator \c k in <tt>[first, last)</tt>, 
- * <tt>comp(value, *k)</tt> is \c false. For every iterator \c k 
- * in <tt>[i, j)</tt>, neither <tt>comp(value, *k)</tt> nor 
+ * that, for every iterator \c k in <tt>[first, last)</tt>,
+ * <tt>comp(value, *k)</tt> is \c false. For every iterator \c k
+ * in <tt>[i, j)</tt>, neither <tt>comp(value, *k)</tt> nor
  * <tt>comp(*k, value)</tt> is \c true.
  *
  *  \param first The beginning of the ordered sequence.
@@ -957,7 +963,7 @@ equal_range(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
  *  \param value The value to be searched.
  *  \param comp The comparison operator.
  *  \return A \p pair of iterators <tt>[i, j)</tt> that define the range of equivalent elements.
- * 
+ *
  *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
  *  \tparam T is comparable to \p ForwardIterator's \c value_type.
  *  \tparam StrictWeakOrdering is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
@@ -1010,7 +1016,7 @@ equal_range(ForwardIterator first,
 //////////////////////
 
 
-/*! \p lower_bound is a vectorized version of binary search: for each 
+/*! \p lower_bound is a vectorized version of binary search: for each
  * iterator \c v in <tt>[values_first, values_last)</tt> it attempts to
  * find the value <tt>*v</tt> in an ordered range <tt>[first, last)</tt>.
  * Specifically, it returns the index of first position where value could
@@ -1024,7 +1030,7 @@ equal_range(ForwardIterator first,
  *  \param values_first The beginning of the search values sequence.
  *  \param values_last The end of the search values sequence.
  *  \param result The beginning of the output sequence.
- * 
+ *
  *  \tparam DerivedPolicy The name of the derived execution policy.
  *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
  *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
@@ -1052,7 +1058,7 @@ equal_range(ForwardIterator first,
  *  input[4] = 8;
  *
  *  thrust::device_vector<int> values(6);
- *  values[0] = 0; 
+ *  values[0] = 0;
  *  values[1] = 1;
  *  values[2] = 2;
  *  values[3] = 3;
@@ -1077,14 +1083,14 @@ equal_range(ForwardIterator first,
 template <typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator>
 __host__ __device__
 OutputIterator lower_bound(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
-                           ForwardIterator first, 
+                           ForwardIterator first,
                            ForwardIterator last,
-                           InputIterator values_first, 
+                           InputIterator values_first,
                            InputIterator values_last,
                            OutputIterator result);
 
 
-/*! \p lower_bound is a vectorized version of binary search: for each 
+/*! \p lower_bound is a vectorized version of binary search: for each
  * iterator \c v in <tt>[values_first, values_last)</tt> it attempts to
  * find the value <tt>*v</tt> in an ordered range <tt>[first, last)</tt>.
  * Specifically, it returns the index of first position where value could
@@ -1095,7 +1101,7 @@ OutputIterator lower_bound(const thrust::detail::execution_policy_base<DerivedPo
  *  \param values_first The beginning of the search values sequence.
  *  \param values_last The end of the search values sequence.
  *  \param result The beginning of the output sequence.
- * 
+ *
  *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
  *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
  *                        and \c InputIterator's \c value_type is <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>.
@@ -1120,7 +1126,7 @@ OutputIterator lower_bound(const thrust::detail::execution_policy_base<DerivedPo
  *  input[4] = 8;
  *
  *  thrust::device_vector<int> values(6);
- *  values[0] = 0; 
+ *  values[0] = 0;
  *  values[1] = 1;
  *  values[2] = 2;
  *  values[3] = 3;
@@ -1142,18 +1148,18 @@ OutputIterator lower_bound(const thrust::detail::execution_policy_base<DerivedPo
  *  \see \p binary_search
  */
 template <class ForwardIterator, class InputIterator, class OutputIterator>
-OutputIterator lower_bound(ForwardIterator first, 
+OutputIterator lower_bound(ForwardIterator first,
                            ForwardIterator last,
-                           InputIterator values_first, 
+                           InputIterator values_first,
                            InputIterator values_last,
                            OutputIterator result);
 
 
-/*! \p lower_bound is a vectorized version of binary search: for each 
+/*! \p lower_bound is a vectorized version of binary search: for each
  * iterator \c v in <tt>[values_first, values_last)</tt> it attempts to
  * find the value <tt>*v</tt> in an ordered range <tt>[first, last)</tt>.
  * Specifically, it returns the index of first position where value could
- * be inserted without violating the ordering.  This version of 
+ * be inserted without violating the ordering.  This version of
  * \p lower_bound uses function object \c comp for comparison.
  *
  * The algorithm's execution is parallelized as determined by \p exec.
@@ -1165,7 +1171,7 @@ OutputIterator lower_bound(ForwardIterator first,
  *  \param values_last The end of the search values sequence.
  *  \param result The beginning of the output sequence.
  *  \param comp The comparison operator.
- * 
+ *
  *  \tparam DerivedPolicy The name of the derived execution policy.
  *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
  *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
@@ -1194,7 +1200,7 @@ OutputIterator lower_bound(ForwardIterator first,
  *  input[4] = 8;
  *
  *  thrust::device_vector<int> values(6);
- *  values[0] = 0; 
+ *  values[0] = 0;
  *  values[1] = 1;
  *  values[2] = 2;
  *  values[3] = 3;
@@ -1204,7 +1210,7 @@ OutputIterator lower_bound(ForwardIterator first,
  *  thrust::device_vector<unsigned int> output(6);
  *
  *  thrust::lower_bound(input.begin(), input.end(),
- *                      values.begin(), values.end(), 
+ *                      values.begin(), values.end(),
  *                      output.begin(),
  *                      thrust::less<int>());
  *
@@ -1219,19 +1225,19 @@ OutputIterator lower_bound(ForwardIterator first,
 template <typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator, typename StrictWeakOrdering>
 __host__ __device__
 OutputIterator lower_bound(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
-                           ForwardIterator first, 
+                           ForwardIterator first,
                            ForwardIterator last,
-                           InputIterator values_first, 
+                           InputIterator values_first,
                            InputIterator values_last,
                            OutputIterator result,
                            StrictWeakOrdering comp);
 
 
-/*! \p lower_bound is a vectorized version of binary search: for each 
+/*! \p lower_bound is a vectorized version of binary search: for each
  * iterator \c v in <tt>[values_first, values_last)</tt> it attempts to
  * find the value <tt>*v</tt> in an ordered range <tt>[first, last)</tt>.
  * Specifically, it returns the index of first position where value could
- * be inserted without violating the ordering.  This version of 
+ * be inserted without violating the ordering.  This version of
  * \p lower_bound uses function object \c comp for comparison.
  *
  *  \param first The beginning of the ordered sequence.
@@ -1240,7 +1246,7 @@ OutputIterator lower_bound(const thrust::detail::execution_policy_base<DerivedPo
  *  \param values_last The end of the search values sequence.
  *  \param result The beginning of the output sequence.
  *  \param comp The comparison operator.
- * 
+ *
  *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
  *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
  *                        and \c InputIterator's \c value_type is comparable to \p ForwardIterator's \c value_type.
@@ -1267,7 +1273,7 @@ OutputIterator lower_bound(const thrust::detail::execution_policy_base<DerivedPo
  *  input[4] = 8;
  *
  *  thrust::device_vector<int> values(6);
- *  values[0] = 0; 
+ *  values[0] = 0;
  *  values[1] = 1;
  *  values[2] = 2;
  *  values[3] = 3;
@@ -1277,7 +1283,7 @@ OutputIterator lower_bound(const thrust::detail::execution_policy_base<DerivedPo
  *  thrust::device_vector<unsigned int> output(6);
  *
  *  thrust::lower_bound(input.begin(), input.end(),
- *                      values.begin(), values.end(), 
+ *                      values.begin(), values.end(),
  *                      output.begin(),
  *                      thrust::less<int>());
  *
@@ -1290,15 +1296,15 @@ OutputIterator lower_bound(const thrust::detail::execution_policy_base<DerivedPo
  *  \see \p binary_search
  */
 template <class ForwardIterator, class InputIterator, class OutputIterator, class StrictWeakOrdering>
-OutputIterator lower_bound(ForwardIterator first, 
+OutputIterator lower_bound(ForwardIterator first,
                            ForwardIterator last,
-                           InputIterator values_first, 
+                           InputIterator values_first,
                            InputIterator values_last,
                            OutputIterator result,
                            StrictWeakOrdering comp);
 
 
-/*! \p upper_bound is a vectorized version of binary search: for each 
+/*! \p upper_bound is a vectorized version of binary search: for each
  * iterator \c v in <tt>[values_first, values_last)</tt> it attempts to
  * find the value <tt>*v</tt> in an ordered range <tt>[first, last)</tt>.
  * Specifically, it returns the index of last position where value could
@@ -1312,7 +1318,7 @@ OutputIterator lower_bound(ForwardIterator first,
  *  \param values_first The beginning of the search values sequence.
  *  \param values_last The end of the search values sequence.
  *  \param result The beginning of the output sequence.
- * 
+ *
  *  \tparam DerivedPolicy The name of the derived execution policy.
  *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
  *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
@@ -1340,7 +1346,7 @@ OutputIterator lower_bound(ForwardIterator first,
  *  input[4] = 8;
  *
  *  thrust::device_vector<int> values(6);
- *  values[0] = 0; 
+ *  values[0] = 0;
  *  values[1] = 1;
  *  values[2] = 2;
  *  values[3] = 3;
@@ -1365,14 +1371,14 @@ OutputIterator lower_bound(ForwardIterator first,
 template <typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator>
 __host__ __device__
 OutputIterator upper_bound(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
-                           ForwardIterator first, 
+                           ForwardIterator first,
                            ForwardIterator last,
-                           InputIterator values_first, 
+                           InputIterator values_first,
                            InputIterator values_last,
                            OutputIterator result);
 
 
-/*! \p upper_bound is a vectorized version of binary search: for each 
+/*! \p upper_bound is a vectorized version of binary search: for each
  * iterator \c v in <tt>[values_first, values_last)</tt> it attempts to
  * find the value <tt>*v</tt> in an ordered range <tt>[first, last)</tt>.
  * Specifically, it returns the index of last position where value could
@@ -1383,7 +1389,7 @@ OutputIterator upper_bound(const thrust::detail::execution_policy_base<DerivedPo
  *  \param values_first The beginning of the search values sequence.
  *  \param values_last The end of the search values sequence.
  *  \param result The beginning of the output sequence.
- * 
+ *
  *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
  *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
  *                        and \c InputIterator's \c value_type is <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>.
@@ -1408,7 +1414,7 @@ OutputIterator upper_bound(const thrust::detail::execution_policy_base<DerivedPo
  *  input[4] = 8;
  *
  *  thrust::device_vector<int> values(6);
- *  values[0] = 0; 
+ *  values[0] = 0;
  *  values[1] = 1;
  *  values[2] = 2;
  *  values[3] = 3;
@@ -1430,18 +1436,18 @@ OutputIterator upper_bound(const thrust::detail::execution_policy_base<DerivedPo
  *  \see \p binary_search
  */
 template <class ForwardIterator, class InputIterator, class OutputIterator>
-OutputIterator upper_bound(ForwardIterator first, 
+OutputIterator upper_bound(ForwardIterator first,
                            ForwardIterator last,
-                           InputIterator values_first, 
+                           InputIterator values_first,
                            InputIterator values_last,
                            OutputIterator result);
 
 
-/*! \p upper_bound is a vectorized version of binary search: for each 
+/*! \p upper_bound is a vectorized version of binary search: for each
  * iterator \c v in <tt>[values_first, values_last)</tt> it attempts to
  * find the value <tt>*v</tt> in an ordered range <tt>[first, last)</tt>.
  * Specifically, it returns the index of first position where value could
- * be inserted without violating the ordering.  This version of 
+ * be inserted without violating the ordering.  This version of
  * \p upper_bound uses function object \c comp for comparison.
  *
  * The algorithm's execution is parallelized as determined by \p exec.
@@ -1453,7 +1459,7 @@ OutputIterator upper_bound(ForwardIterator first,
  *  \param values_last The end of the search values sequence.
  *  \param result The beginning of the output sequence.
  *  \param comp The comparison operator.
- * 
+ *
  *  \tparam DerivedPolicy The name of the derived execution policy.
  *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
  *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
@@ -1483,7 +1489,7 @@ OutputIterator upper_bound(ForwardIterator first,
  *  input[4] = 8;
  *
  *  thrust::device_vector<int> values(6);
- *  values[0] = 0; 
+ *  values[0] = 0;
  *  values[1] = 1;
  *  values[2] = 2;
  *  values[3] = 3;
@@ -1494,7 +1500,7 @@ OutputIterator upper_bound(ForwardIterator first,
  *
  *  thrust::upper_bound(thrust::device,
  *                      input.begin(), input.end(),
- *                      values.begin(), values.end(), 
+ *                      values.begin(), values.end(),
  *                      output.begin(),
  *                      thrust::less<int>());
  *
@@ -1509,19 +1515,19 @@ OutputIterator upper_bound(ForwardIterator first,
 template <typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator, typename StrictWeakOrdering>
 __host__ __device__
 OutputIterator upper_bound(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
-                           ForwardIterator first, 
+                           ForwardIterator first,
                            ForwardIterator last,
-                           InputIterator values_first, 
+                           InputIterator values_first,
                            InputIterator values_last,
                            OutputIterator result,
                            StrictWeakOrdering comp);
 
 
-/*! \p upper_bound is a vectorized version of binary search: for each 
+/*! \p upper_bound is a vectorized version of binary search: for each
  * iterator \c v in <tt>[values_first, values_last)</tt> it attempts to
  * find the value <tt>*v</tt> in an ordered range <tt>[first, last)</tt>.
  * Specifically, it returns the index of first position where value could
- * be inserted without violating the ordering.  This version of 
+ * be inserted without violating the ordering.  This version of
  * \p upper_bound uses function object \c comp for comparison.
  *
  *  \param first The beginning of the ordered sequence.
@@ -1530,7 +1536,7 @@ OutputIterator upper_bound(const thrust::detail::execution_policy_base<DerivedPo
  *  \param values_last The end of the search values sequence.
  *  \param result The beginning of the output sequence.
  *  \param comp The comparison operator.
- * 
+ *
  *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
  *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
  *                        and \c InputIterator's \c value_type is comparable to \p ForwardIterator's \c value_type.
@@ -1557,7 +1563,7 @@ OutputIterator upper_bound(const thrust::detail::execution_policy_base<DerivedPo
  *  input[4] = 8;
  *
  *  thrust::device_vector<int> values(6);
- *  values[0] = 0; 
+ *  values[0] = 0;
  *  values[1] = 1;
  *  values[2] = 2;
  *  values[3] = 3;
@@ -1567,7 +1573,7 @@ OutputIterator upper_bound(const thrust::detail::execution_policy_base<DerivedPo
  *  thrust::device_vector<unsigned int> output(6);
  *
  *  thrust::upper_bound(input.begin(), input.end(),
- *                      values.begin(), values.end(), 
+ *                      values.begin(), values.end(),
  *                      output.begin(),
  *                      thrust::less<int>());
  *
@@ -1580,18 +1586,18 @@ OutputIterator upper_bound(const thrust::detail::execution_policy_base<DerivedPo
  *  \see \p binary_search
  */
 template <class ForwardIterator, class InputIterator, class OutputIterator, class StrictWeakOrdering>
-OutputIterator upper_bound(ForwardIterator first, 
+OutputIterator upper_bound(ForwardIterator first,
                            ForwardIterator last,
-                           InputIterator values_first, 
+                           InputIterator values_first,
                            InputIterator values_last,
                            OutputIterator result,
                            StrictWeakOrdering comp);
 
 
-/*! \p binary_search is a vectorized version of binary search: for each 
+/*! \p binary_search is a vectorized version of binary search: for each
  * iterator \c v in <tt>[values_first, values_last)</tt> it attempts to
  * find the value <tt>*v</tt> in an ordered range <tt>[first, last)</tt>.
- * It returns \c true if an element that is equivalent to \c value 
+ * It returns \c true if an element that is equivalent to \c value
  * is present in <tt>[first, last)</tt> and \c false if no such element
  * exists.
  *
@@ -1603,7 +1609,7 @@ OutputIterator upper_bound(ForwardIterator first,
  *  \param values_first The beginning of the search values sequence.
  *  \param values_last The end of the search values sequence.
  *  \param result The beginning of the output sequence.
- * 
+ *
  *  \tparam DerivedPolicy The name of the derived execution policy.
  *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
  *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
@@ -1631,7 +1637,7 @@ OutputIterator upper_bound(ForwardIterator first,
  *  input[4] = 8;
  *
  *  thrust::device_vector<int> values(6);
- *  values[0] = 0; 
+ *  values[0] = 0;
  *  values[1] = 1;
  *  values[2] = 2;
  *  values[3] = 3;
@@ -1656,17 +1662,17 @@ OutputIterator upper_bound(ForwardIterator first,
 template <typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator>
 __host__ __device__
 OutputIterator binary_search(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
-                             ForwardIterator first, 
+                             ForwardIterator first,
                              ForwardIterator last,
-                             InputIterator values_first, 
+                             InputIterator values_first,
                              InputIterator values_last,
                              OutputIterator result);
 
 
-/*! \p binary_search is a vectorized version of binary search: for each 
+/*! \p binary_search is a vectorized version of binary search: for each
  * iterator \c v in <tt>[values_first, values_last)</tt> it attempts to
  * find the value <tt>*v</tt> in an ordered range <tt>[first, last)</tt>.
- * It returns \c true if an element that is equivalent to \c value 
+ * It returns \c true if an element that is equivalent to \c value
  * is present in <tt>[first, last)</tt> and \c false if no such element
  * exists.
  *
@@ -1675,7 +1681,7 @@ OutputIterator binary_search(const thrust::detail::execution_policy_base<Derived
  *  \param values_first The beginning of the search values sequence.
  *  \param values_last The end of the search values sequence.
  *  \param result The beginning of the output sequence.
- * 
+ *
  *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
  *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
  *                        and \c InputIterator's \c value_type is <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>.
@@ -1700,7 +1706,7 @@ OutputIterator binary_search(const thrust::detail::execution_policy_base<Derived
  *  input[4] = 8;
  *
  *  thrust::device_vector<int> values(6);
- *  values[0] = 0; 
+ *  values[0] = 0;
  *  values[1] = 1;
  *  values[2] = 2;
  *  values[3] = 3;
@@ -1722,19 +1728,19 @@ OutputIterator binary_search(const thrust::detail::execution_policy_base<Derived
  *  \see \p equal_range
  */
 template <class ForwardIterator, class InputIterator, class OutputIterator>
-OutputIterator binary_search(ForwardIterator first, 
+OutputIterator binary_search(ForwardIterator first,
                              ForwardIterator last,
-                             InputIterator values_first, 
+                             InputIterator values_first,
                              InputIterator values_last,
                              OutputIterator result);
 
 
-/*! \p binary_search is a vectorized version of binary search: for each 
+/*! \p binary_search is a vectorized version of binary search: for each
  * iterator \c v in <tt>[values_first, values_last)</tt> it attempts to
  * find the value <tt>*v</tt> in an ordered range <tt>[first, last)</tt>.
- * It returns \c true if an element that is equivalent to \c value 
+ * It returns \c true if an element that is equivalent to \c value
  * is present in <tt>[first, last)</tt> and \c false if no such element
- * exists.  This version of \p binary_search uses function object 
+ * exists.  This version of \p binary_search uses function object
  * \c comp for comparison.
  *
  * The algorithm's execution is parallelized as determined by \p exec.
@@ -1746,7 +1752,7 @@ OutputIterator binary_search(ForwardIterator first,
  *  \param values_last The end of the search values sequence.
  *  \param result The beginning of the output sequence.
  *  \param comp The comparison operator.
- * 
+ *
  *  \tparam DerivedPolicy The name of the derived execution policy.
  *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
  *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
@@ -1776,7 +1782,7 @@ OutputIterator binary_search(ForwardIterator first,
  *  input[4] = 8;
  *
  *  thrust::device_vector<int> values(6);
- *  values[0] = 0; 
+ *  values[0] = 0;
  *  values[1] = 1;
  *  values[2] = 2;
  *  values[3] = 3;
@@ -1802,20 +1808,20 @@ OutputIterator binary_search(ForwardIterator first,
 template <typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator, typename StrictWeakOrdering>
 __host__ __device__
 OutputIterator binary_search(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
-                             ForwardIterator first, 
+                             ForwardIterator first,
                              ForwardIterator last,
-                             InputIterator values_first, 
+                             InputIterator values_first,
                              InputIterator values_last,
                              OutputIterator result,
                              StrictWeakOrdering comp);
 
 
-/*! \p binary_search is a vectorized version of binary search: for each 
+/*! \p binary_search is a vectorized version of binary search: for each
  * iterator \c v in <tt>[values_first, values_last)</tt> it attempts to
  * find the value <tt>*v</tt> in an ordered range <tt>[first, last)</tt>.
- * It returns \c true if an element that is equivalent to \c value 
+ * It returns \c true if an element that is equivalent to \c value
  * is present in <tt>[first, last)</tt> and \c false if no such element
- * exists.  This version of \p binary_search uses function object 
+ * exists.  This version of \p binary_search uses function object
  * \c comp for comparison.
  *
  *  \param first The beginning of the ordered sequence.
@@ -1824,7 +1830,7 @@ OutputIterator binary_search(const thrust::detail::execution_policy_base<Derived
  *  \param values_last The end of the search values sequence.
  *  \param result The beginning of the output sequence.
  *  \param comp The comparison operator.
- * 
+ *
  *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
  *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
  *                        and \c InputIterator's \c value_type is <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>.
@@ -1851,7 +1857,7 @@ OutputIterator binary_search(const thrust::detail::execution_policy_base<Derived
  *  input[4] = 8;
  *
  *  thrust::device_vector<int> values(6);
- *  values[0] = 0; 
+ *  values[0] = 0;
  *  values[1] = 1;
  *  values[2] = 2;
  *  values[3] = 3;
@@ -1874,9 +1880,9 @@ OutputIterator binary_search(const thrust::detail::execution_policy_base<Derived
  *  \see \p equal_range
  */
 template <class ForwardIterator, class InputIterator, class OutputIterator, class StrictWeakOrdering>
-OutputIterator binary_search(ForwardIterator first, 
+OutputIterator binary_search(ForwardIterator first,
                              ForwardIterator last,
-                             InputIterator values_first, 
+                             InputIterator values_first,
                              InputIterator values_last,
                              OutputIterator result,
                              StrictWeakOrdering comp);
diff --git a/thrust/thrust/complex.h b/thrust/thrust/complex.h
index 7f1c601c8f4..a400a6dbd7d 100644
--- a/thrust/thrust/complex.h
+++ b/thrust/thrust/complex.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/type_traits.h>
 #include <thrust/type_traits/is_trivially_relocatable.h>
 
diff --git a/thrust/thrust/copy.h b/thrust/thrust/copy.h
index 99d488174dd..24f1ac78cfc 100644
--- a/thrust/thrust/copy.h
+++ b/thrust/thrust/copy.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
@@ -140,7 +146,7 @@ __host__ __device__
                         OutputIterator result);
 
 
-	
+
 /*! \p copy copies elements from the range [\p first, \p last) to the range
  *  [\p result, \p result + (\p last - \p first)). That is, it performs
  *  the assignments *\p result = *\p first, *(\p result + \c 1) = *(\p first + \c 1),
diff --git a/thrust/thrust/count.h b/thrust/thrust/count.h
index abf8b2d6c30..0512933cbbe 100644
--- a/thrust/thrust/count.h
+++ b/thrust/thrust/count.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 #include <thrust/iterator/iterator_traits.h>
 
@@ -57,7 +63,7 @@ THRUST_NAMESPACE_BEGIN
  *  \tparam InputIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator's \c value_type must be a model of must be a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a>.
  *  \tparam EqualityComparable must be a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a> and can be compared for equality with \c InputIterator's \c value_type
  *
- *  The following code snippet demonstrates how to use \p count to 
+ *  The following code snippet demonstrates how to use \p count to
  *  count the number of instances in a range of a value of interest using the \p thrust::device execution policy:
  *
  *  \code
@@ -70,7 +76,7 @@ THRUST_NAMESPACE_BEGIN
  *  vec[1] = 1;
  *  vec[3] = 1;
  *  vec[4] = 1;
- *  
+ *
  *  // count the 1s
  *  int result = thrust::count(thrust::device, vec.begin(), vec.end(), 1);
  *  // result == 3
@@ -97,7 +103,7 @@ __host__ __device__
  *  \tparam InputIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator's \c value_type must be a model of must be a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a>.
  *  \tparam EqualityComparable must be a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a> and can be compared for equality with \c InputIterator's \c value_type
  *
- *  The following code snippet demonstrates how to use \p count to 
+ *  The following code snippet demonstrates how to use \p count to
  *  count the number of instances in a range of a value of interest.
  *  \code
  *  #include <thrust/count.h>
@@ -108,7 +114,7 @@ __host__ __device__
  *  vec[1] = 1;
  *  vec[3] = 1;
  *  vec[4] = 1;
- *  
+ *
  *  // count the 1s
  *  int result = thrust::count(vec.begin(), vec.end(), 1);
  *  // result == 3
@@ -121,7 +127,7 @@ template <typename InputIterator, typename EqualityComparable>
     count(InputIterator first, InputIterator last, const EqualityComparable& value);
 
 
-/*! \p count_if finds the number of elements in <tt>[first,last)</tt> for which 
+/*! \p count_if finds the number of elements in <tt>[first,last)</tt> for which
  *  a predicate is \c true. More precisely, \p count_if returns the number of iterators
  *  \c i in <tt>[first, last)</tt> such that <tt>pred(*i) == true</tt>.
  *
@@ -175,7 +181,7 @@ __host__ __device__
     count_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, InputIterator first, InputIterator last, Predicate pred);
 
 
-/*! \p count_if finds the number of elements in <tt>[first,last)</tt> for which 
+/*! \p count_if finds the number of elements in <tt>[first,last)</tt> for which
  *  a predicate is \c true. More precisely, \p count_if returns the number of iterators
  *  \c i in <tt>[first, last)</tt> such that <tt>pred(*i) == true</tt>.
  *
diff --git a/thrust/thrust/detail/adjacent_difference.inl b/thrust/thrust/detail/adjacent_difference.inl
index 844687cff4c..1f604c07118 100644
--- a/thrust/thrust/detail/adjacent_difference.inl
+++ b/thrust/thrust/detail/adjacent_difference.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/select_system.h>
 #include <thrust/system/detail/generic/adjacent_difference.h>
 #include <thrust/system/detail/adl/adjacent_difference.h>
diff --git a/thrust/thrust/detail/advance.inl b/thrust/thrust/detail/advance.inl
index 7b5f261bdce..e73e9cd8035 100644
--- a/thrust/thrust/detail/advance.inl
+++ b/thrust/thrust/detail/advance.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/advance.h>
 #include <thrust/system/detail/generic/advance.h>
 #include <thrust/iterator/iterator_traits.h>
diff --git a/thrust/thrust/detail/algorithm_wrapper.h b/thrust/thrust/detail/algorithm_wrapper.h
index c09b9a0a0b4..ad5efce8b06 100644
--- a/thrust/thrust/detail/algorithm_wrapper.h
+++ b/thrust/thrust/detail/algorithm_wrapper.h
@@ -16,6 +16,14 @@
 
 #pragma once
 
+#include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // When a compiler uses Thrust as part of its implementation of Standard C++
 // algorithms, a cycle of included files may result when Thrust code tries to
 // use a standard algorithm.  Having a macro that is defined only when Thrust
diff --git a/thrust/thrust/detail/alignment.h b/thrust/thrust/detail/alignment.h
index 96fb8436a75..6b8f11b333f 100644
--- a/thrust/thrust/detail/alignment.h
+++ b/thrust/thrust/detail/alignment.h
@@ -21,12 +21,18 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/type_traits.h> // For `integral_constant`.
 
 #include <cstddef> // For `std::size_t` and `std::max_align_t`.
 
 #if THRUST_CPP_DIALECT >= 2011
-    #include <type_traits> // For `std::alignment_of`. 
+    #include <type_traits> // For `std::alignment_of`.
 #endif
 
 THRUST_NAMESPACE_BEGIN
@@ -36,7 +42,7 @@ namespace detail
 
 /// \p THRUST_ALIGNOF is a macro that takes a single type-id as a parameter,
 /// and returns the alignment requirement of the type in bytes.
-/// 
+///
 /// It is an approximation of C++11's `alignof` operator.
 ///
 /// Note: MSVC does not allow the builtin used to implement this to be placed
@@ -44,7 +50,7 @@ namespace detail
 /// assign the result of \p THRUST_ALIGNOF to a variable and pass the variable
 /// as the argument to `__declspec(align(#))`.
 #if THRUST_CPP_DIALECT >= 2011
-    #define THRUST_ALIGNOF(x) alignof(x) 
+    #define THRUST_ALIGNOF(x) alignof(x)
 #else
     #define THRUST_ALIGNOF(x) __alignof(x)
 #endif
@@ -52,7 +58,7 @@ namespace detail
 /// \p alignment_of provides the member constant `value` which is equal to the
 /// alignment requirement of the type `T`, as if obtained by a C++11 `alignof`
 /// expression.
-/// 
+///
 /// It is an implementation of C++11's \p std::alignment_of.
 #if THRUST_CPP_DIALECT >= 2011
     template <typename T>
@@ -110,9 +116,9 @@ struct aligned_type;
     || (   (THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_GCC)                 \
         && (THRUST_GCC_VERSION < 40600))
     // C++03 implementation for MSVC and GCC <= 4.5.
-    // 
+    //
     // We have to implement `aligned_type` with specializations for MSVC
-    // and GCC 4.2.x and older because they require literals as arguments to 
+    // and GCC 4.2.x and older because they require literals as arguments to
     // their alignment attribute.
 
     #if (THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_MSVC)
@@ -134,7 +140,7 @@ struct aligned_type;
             };                                                                \
             /**/
     #endif
-    
+
     THRUST_DEFINE_ALIGNED_TYPE_SPECIALIZATION(1);
     THRUST_DEFINE_ALIGNED_TYPE_SPECIALIZATION(2);
     THRUST_DEFINE_ALIGNED_TYPE_SPECIALIZATION(4);
diff --git a/thrust/thrust/detail/allocator/allocator_traits.h b/thrust/thrust/detail/allocator/allocator_traits.h
index 3a5af3661f3..d59f38591c2 100644
--- a/thrust/thrust/detail/allocator/allocator_traits.h
+++ b/thrust/thrust/detail/allocator/allocator_traits.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/type_traits/pointer_traits.h>
 #include <thrust/detail/type_traits/has_nested_type.h>
 #include <thrust/detail/type_traits/has_member_function.h>
@@ -384,7 +390,7 @@ template<typename Alloc>
 
   template<typename T>
   inline __host__ __device__ static void construct(allocator_type &a, T *p);
-  
+
   template<typename T, typename Arg1>
   inline __host__ __device__ static void construct(allocator_type &a, T *p, const Arg1 &arg1);
 
diff --git a/thrust/thrust/detail/allocator/allocator_traits.inl b/thrust/thrust/detail/allocator/allocator_traits.inl
index 27533009494..1548373a6eb 100644
--- a/thrust/thrust/detail/allocator/allocator_traits.inl
+++ b/thrust/thrust/detail/allocator/allocator_traits.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/allocator/allocator_traits.h>
 #include <thrust/detail/type_traits/is_call_possible.h>
 #include <thrust/detail/integer_traits.h>
diff --git a/thrust/thrust/detail/allocator/copy_construct_range.h b/thrust/thrust/detail/allocator/copy_construct_range.h
index b3c2de32462..25139925efe 100644
--- a/thrust/thrust/detail/allocator/copy_construct_range.h
+++ b/thrust/thrust/detail/allocator/copy_construct_range.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/detail/allocator/copy_construct_range.inl b/thrust/thrust/detail/allocator/copy_construct_range.inl
index a71cca1f720..54ae413a100 100644
--- a/thrust/thrust/detail/allocator/copy_construct_range.inl
+++ b/thrust/thrust/detail/allocator/copy_construct_range.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/allocator/allocator_traits.h>
 #include <thrust/detail/type_traits/pointer_traits.h>
 #include <thrust/iterator/iterator_traits.h>
diff --git a/thrust/thrust/detail/allocator/default_construct_range.h b/thrust/thrust/detail/allocator/default_construct_range.h
index 8b5026c05b3..b53534936cc 100644
--- a/thrust/thrust/detail/allocator/default_construct_range.h
+++ b/thrust/thrust/detail/allocator/default_construct_range.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 THRUST_NAMESPACE_BEGIN
 namespace detail
 {
diff --git a/thrust/thrust/detail/allocator/default_construct_range.inl b/thrust/thrust/detail/allocator/default_construct_range.inl
index 6d26578fa23..6edc4f7a2cb 100644
--- a/thrust/thrust/detail/allocator/default_construct_range.inl
+++ b/thrust/thrust/detail/allocator/default_construct_range.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/allocator/allocator_traits.h>
 #include <thrust/detail/type_traits.h>
 #include <thrust/detail/type_traits/pointer_traits.h>
diff --git a/thrust/thrust/detail/allocator/destroy_range.h b/thrust/thrust/detail/allocator/destroy_range.h
index cfc7e3f6ef5..cfedf0c4f32 100644
--- a/thrust/thrust/detail/allocator/destroy_range.h
+++ b/thrust/thrust/detail/allocator/destroy_range.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 THRUST_NAMESPACE_BEGIN
 namespace detail
 {
diff --git a/thrust/thrust/detail/allocator/destroy_range.inl b/thrust/thrust/detail/allocator/destroy_range.inl
index 662177f3a72..a288d3d663b 100644
--- a/thrust/thrust/detail/allocator/destroy_range.inl
+++ b/thrust/thrust/detail/allocator/destroy_range.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/allocator/destroy_range.h>
 #include <thrust/detail/allocator/allocator_traits.h>
 #include <thrust/detail/type_traits/pointer_traits.h>
diff --git a/thrust/thrust/detail/allocator/fill_construct_range.h b/thrust/thrust/detail/allocator/fill_construct_range.h
index a7572cb2d2b..5aa6f84bee8 100644
--- a/thrust/thrust/detail/allocator/fill_construct_range.h
+++ b/thrust/thrust/detail/allocator/fill_construct_range.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 THRUST_NAMESPACE_BEGIN
 namespace detail
 {
diff --git a/thrust/thrust/detail/allocator/fill_construct_range.inl b/thrust/thrust/detail/allocator/fill_construct_range.inl
index 876b5ddd264..4019d9d9b46 100644
--- a/thrust/thrust/detail/allocator/fill_construct_range.inl
+++ b/thrust/thrust/detail/allocator/fill_construct_range.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/type_traits.h>
 #include <thrust/detail/allocator/allocator_traits.h>
 #include <thrust/detail/type_traits/pointer_traits.h>
diff --git a/thrust/thrust/detail/allocator/malloc_allocator.h b/thrust/thrust/detail/allocator/malloc_allocator.h
index af3d0fccbd7..be230a986f5 100644
--- a/thrust/thrust/detail/allocator/malloc_allocator.h
+++ b/thrust/thrust/detail/allocator/malloc_allocator.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/type_traits/pointer_traits.h>
 #include <thrust/detail/allocator/tagged_allocator.h>
 
diff --git a/thrust/thrust/detail/allocator/malloc_allocator.inl b/thrust/thrust/detail/allocator/malloc_allocator.inl
index d03d33305cb..0e64abf73fa 100644
--- a/thrust/thrust/detail/allocator/malloc_allocator.inl
+++ b/thrust/thrust/detail/allocator/malloc_allocator.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/allocator/malloc_allocator.h>
 #include <thrust/system/detail/generic/select_system.h>
 #include <thrust/system/detail/bad_alloc.h>
diff --git a/thrust/thrust/detail/allocator/no_throw_allocator.h b/thrust/thrust/detail/allocator/no_throw_allocator.h
index a6c16985b11..7a35b3d1160 100644
--- a/thrust/thrust/detail/allocator/no_throw_allocator.h
+++ b/thrust/thrust/detail/allocator/no_throw_allocator.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <nv/target>
 
 THRUST_NAMESPACE_BEGIN
@@ -29,7 +35,7 @@ template<typename BaseAllocator>
 {
   private:
     typedef BaseAllocator super_t;
-  
+
   public:
     inline __host__ __device__
     no_throw_allocator(const BaseAllocator &other = BaseAllocator())
diff --git a/thrust/thrust/detail/allocator/tagged_allocator.h b/thrust/thrust/detail/allocator/tagged_allocator.h
index 804c4e42e51..fa7a716f119 100644
--- a/thrust/thrust/detail/allocator/tagged_allocator.h
+++ b/thrust/thrust/detail/allocator/tagged_allocator.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/type_traits/pointer_traits.h>
 #include <thrust/iterator/iterator_traits.h>
 
diff --git a/thrust/thrust/detail/allocator/tagged_allocator.inl b/thrust/thrust/detail/allocator/tagged_allocator.inl
index bcd534cbcee..4f02e59c840 100644
--- a/thrust/thrust/detail/allocator/tagged_allocator.inl
+++ b/thrust/thrust/detail/allocator/tagged_allocator.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/allocator/tagged_allocator.h>
 #include <limits>
 
diff --git a/thrust/thrust/detail/allocator/temporary_allocator.h b/thrust/thrust/detail/allocator/temporary_allocator.h
index c8ef606258a..81110e3f2e7 100644
--- a/thrust/thrust/detail/allocator/temporary_allocator.h
+++ b/thrust/thrust/detail/allocator/temporary_allocator.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/allocator/tagged_allocator.h>
 #include <thrust/detail/allocator/allocator_traits.h>
 #include <thrust/pair.h>
diff --git a/thrust/thrust/detail/allocator/temporary_allocator.inl b/thrust/thrust/detail/allocator/temporary_allocator.inl
index ef5d1afa503..23f96cfb996 100644
--- a/thrust/thrust/detail/allocator/temporary_allocator.inl
+++ b/thrust/thrust/detail/allocator/temporary_allocator.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/allocator/temporary_allocator.h>
 #include <thrust/detail/temporary_buffer.h>
 #include <thrust/system/detail/bad_alloc.h>
diff --git a/thrust/thrust/detail/allocator_aware_execution_policy.h b/thrust/thrust/detail/allocator_aware_execution_policy.h
index eea93c035ea..bfc29736bfd 100644
--- a/thrust/thrust/detail/allocator_aware_execution_policy.h
+++ b/thrust/thrust/detail/allocator_aware_execution_policy.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execute_with_allocator_fwd.h>
 #include <thrust/detail/alignment.h>
 
diff --git a/thrust/thrust/detail/binary_search.inl b/thrust/thrust/detail/binary_search.inl
index 90350ced4bb..d2a42fdf6af 100644
--- a/thrust/thrust/detail/binary_search.inl
+++ b/thrust/thrust/detail/binary_search.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/binary_search.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
diff --git a/thrust/thrust/detail/caching_allocator.h b/thrust/thrust/detail/caching_allocator.h
index 941f52755ca..f306f0f1220 100644
--- a/thrust/thrust/detail/caching_allocator.h
+++ b/thrust/thrust/detail/caching_allocator.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/mr/allocator.h>
 #include <thrust/mr/disjoint_tls_pool.h>
 #include <thrust/mr/new.h>
diff --git a/thrust/thrust/detail/config.h b/thrust/thrust/detail/config.h
index 5a5573a410e..12df82f6944 100644
--- a/thrust/thrust/detail/config.h
+++ b/thrust/thrust/detail/config.h
@@ -22,3 +22,8 @@
 #include <thrust/version.h>
 #include <thrust/detail/config/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
diff --git a/thrust/thrust/detail/config/compiler.h b/thrust/thrust/detail/config/compiler.h
index e35652f6a41..3359808f6c2 100644
--- a/thrust/thrust/detail/config/compiler.h
+++ b/thrust/thrust/detail/config/compiler.h
@@ -20,6 +20,14 @@
 
 #pragma once
 
+// Internal config header that is only included through thrust/detail/config/config.h
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // enumerate host compilers we know about
 #define THRUST_HOST_COMPILER_UNKNOWN 0
 #define THRUST_HOST_COMPILER_MSVC    1
diff --git a/thrust/thrust/detail/config/compiler_fence.h b/thrust/thrust/detail/config/compiler_fence.h
index c379abaf364..629e0a549f4 100644
--- a/thrust/thrust/detail/config/compiler_fence.h
+++ b/thrust/thrust/detail/config/compiler_fence.h
@@ -17,6 +17,13 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/preprocessor.h>
 
 // TODO: Enable this or remove this file once nvGRAPH/CUSP migrates off of it.
diff --git a/thrust/thrust/detail/config/config.h b/thrust/thrust/detail/config/config.h
index 797f6605b08..692d356c5f1 100644
--- a/thrust/thrust/detail/config/config.h
+++ b/thrust/thrust/detail/config/config.h
@@ -20,6 +20,15 @@
 
 #pragma once
 
+// For _CCCL_IMPLICIT_SYSTEM_HEADER
+#include <cuda/__cccl_config>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // NOTE: The order of these #includes matters.
 
 #include <thrust/detail/config/simple_defines.h>
diff --git a/thrust/thrust/detail/config/cpp_compatibility.h b/thrust/thrust/detail/config/cpp_compatibility.h
index 18b9cbdcf59..ba6c9b65b6d 100644
--- a/thrust/thrust/detail/config/cpp_compatibility.h
+++ b/thrust/thrust/detail/config/cpp_compatibility.h
@@ -16,6 +16,14 @@
 
 #pragma once
 
+// Internal config header that is only included through thrust/detail/config/config.h
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/config/cpp_dialect.h>
 
 #include <cstddef>
diff --git a/thrust/thrust/detail/config/cpp_dialect.h b/thrust/thrust/detail/config/cpp_dialect.h
index 97c39bea574..87905893fc9 100644
--- a/thrust/thrust/detail/config/cpp_dialect.h
+++ b/thrust/thrust/detail/config/cpp_dialect.h
@@ -20,6 +20,14 @@
 
 #pragma once
 
+#include <thrust/detail/config/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/config/compiler.h>
 
 // Deprecation warnings may be silenced by defining the following macros. These
diff --git a/thrust/thrust/detail/config/debug.h b/thrust/thrust/detail/config/debug.h
index 16f65d67c90..be903b1f36b 100644
--- a/thrust/thrust/detail/config/debug.h
+++ b/thrust/thrust/detail/config/debug.h
@@ -16,6 +16,14 @@
 
 #pragma once
 
+// Internal config header that is only included through thrust/detail/config/config.h
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #ifndef THRUST_DEBUG
 #  ifndef NDEBUG
 #    if defined(DEBUG) || defined(_DEBUG)
diff --git a/thrust/thrust/detail/config/deprecated.h b/thrust/thrust/detail/config/deprecated.h
index 05851c6768f..0ec83a15dc7 100644
--- a/thrust/thrust/detail/config/deprecated.h
+++ b/thrust/thrust/detail/config/deprecated.h
@@ -20,6 +20,14 @@
 
 #pragma once
 
+// Internal config header that is only included through thrust/detail/config/config.h
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/config/compiler.h>
 #include <thrust/detail/config/cpp_dialect.h>
 
diff --git a/thrust/thrust/detail/config/device_system.h b/thrust/thrust/detail/config/device_system.h
index 29418c90308..188202276ae 100644
--- a/thrust/thrust/detail/config/device_system.h
+++ b/thrust/thrust/detail/config/device_system.h
@@ -16,6 +16,14 @@
 
 #pragma once
 
+// Internal config header that is only included through thrust/detail/config/config.h
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // reserve 0 for undefined
 #define THRUST_DEVICE_SYSTEM_CUDA    1
 #define THRUST_DEVICE_SYSTEM_OMP     2
diff --git a/thrust/thrust/detail/config/exec_check_disable.h b/thrust/thrust/detail/config/exec_check_disable.h
index 9b25b375dfb..8a9811f86d3 100644
--- a/thrust/thrust/detail/config/exec_check_disable.h
+++ b/thrust/thrust/detail/config/exec_check_disable.h
@@ -20,7 +20,13 @@
 
 #pragma once
 
-#include <thrust/detail/config.h>
+#include <thrust/detail/config/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 // #pragma nv_exec_check_disable is only recognized by NVCC.  Having a macro
 // expand to a #pragma (rather than _Pragma) only works with NVCC's compilation
diff --git a/thrust/thrust/detail/config/forceinline.h b/thrust/thrust/detail/config/forceinline.h
index b001fd4b114..c90124cb8e7 100644
--- a/thrust/thrust/detail/config/forceinline.h
+++ b/thrust/thrust/detail/config/forceinline.h
@@ -20,7 +20,13 @@
 
 #pragma once
 
-#include <thrust/detail/config.h>
+// Internal config header that is only included through thrust/detail/config/config.h
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #if defined(__CUDACC__) || defined(_NVHPC_CUDA)
 
@@ -28,7 +34,7 @@
 
 #else
 
-// TODO add 
+// TODO add
 
 #define __thrust_forceinline__
 
diff --git a/thrust/thrust/detail/config/global_workarounds.h b/thrust/thrust/detail/config/global_workarounds.h
index 9800f03593a..e16de143b55 100644
--- a/thrust/thrust/detail/config/global_workarounds.h
+++ b/thrust/thrust/detail/config/global_workarounds.h
@@ -16,6 +16,14 @@
 
 #pragma once
 
+// Internal config header that is only included through thrust/detail/config/config.h
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/config/compiler.h>
 
 // XXX workaround gcc 4.8+'s complaints about unused local typedefs by silencing them globally
diff --git a/thrust/thrust/detail/config/host_device.h b/thrust/thrust/detail/config/host_device.h
index 5540f91260d..b4b4a26be9f 100644
--- a/thrust/thrust/detail/config/host_device.h
+++ b/thrust/thrust/detail/config/host_device.h
@@ -20,7 +20,13 @@
 
 #pragma once
 
-#include <thrust/detail/config.h>
+// Internal config header that is only included through thrust/detail/config/config.h
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 // since nvcc defines __host__ and __device__ for us,
 // and only nvcc knows what to do with __host__ and __device__,
diff --git a/thrust/thrust/detail/config/host_system.h b/thrust/thrust/detail/config/host_system.h
index f216f649237..aab787a9ffd 100644
--- a/thrust/thrust/detail/config/host_system.h
+++ b/thrust/thrust/detail/config/host_system.h
@@ -16,6 +16,14 @@
 
 #pragma once
 
+// Internal config header that is only included through thrust/detail/config/config.h
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // reserve 0 for undefined
 #define THRUST_HOST_SYSTEM_CPP    1
 #define THRUST_HOST_SYSTEM_OMP    2
diff --git a/thrust/thrust/detail/config/memory_resource.h b/thrust/thrust/detail/config/memory_resource.h
index ab719c9bde9..a8a4f18971e 100644
--- a/thrust/thrust/detail/config/memory_resource.h
+++ b/thrust/thrust/detail/config/memory_resource.h
@@ -16,9 +16,16 @@
 
 #pragma once
 
+#include <thrust/detail/config/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cstddef>
 
-#include <thrust/detail/config.h>
 #include <thrust/detail/alignment.h>
 #include <thrust/detail/config/cpp_compatibility.h>
 
diff --git a/thrust/thrust/detail/config/namespace.h b/thrust/thrust/detail/config/namespace.h
index 91b9f879cdd..94cbbbbe6eb 100644
--- a/thrust/thrust/detail/config/namespace.h
+++ b/thrust/thrust/detail/config/namespace.h
@@ -16,6 +16,14 @@
 
 #pragma once
 
+// Internal config header that is only included through thrust/detail/config/config.h
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/config/device_system.h>
 #include <thrust/version.h>
 
diff --git a/thrust/thrust/detail/config/simple_defines.h b/thrust/thrust/detail/config/simple_defines.h
index e3ea2eb64e7..c99d7ab2a40 100644
--- a/thrust/thrust/detail/config/simple_defines.h
+++ b/thrust/thrust/detail/config/simple_defines.h
@@ -20,6 +20,14 @@
 
 #pragma once
 
+// Internal config header that is only included through thrust/detail/config/config.h
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #define THRUST_UNKNOWN 0
 #define THRUST_FALSE   0
 #define THRUST_TRUE    1
diff --git a/thrust/thrust/detail/contiguous_storage.h b/thrust/thrust/detail/contiguous_storage.h
index 536c1c27c2e..3ed000a26b6 100644
--- a/thrust/thrust/detail/contiguous_storage.h
+++ b/thrust/thrust/detail/contiguous_storage.h
@@ -16,10 +16,17 @@
 
 #pragma once
 
+#include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/iterator/detail/normal_iterator.h>
 #include <thrust/detail/execution_policy.h>
 #include <thrust/detail/allocator/allocator_traits.h>
-#include <thrust/detail/config.h>
 
 THRUST_NAMESPACE_BEGIN
 
diff --git a/thrust/thrust/detail/contiguous_storage.inl b/thrust/thrust/detail/contiguous_storage.inl
index 7ae8657f081..ded03e01710 100644
--- a/thrust/thrust/detail/contiguous_storage.inl
+++ b/thrust/thrust/detail/contiguous_storage.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/contiguous_storage.h>
 #include <thrust/detail/swap.h>
 #include <thrust/detail/allocator/allocator_traits.h>
diff --git a/thrust/thrust/detail/copy.h b/thrust/thrust/detail/copy.h
index d6c5bc805c4..59d0f264b0b 100644
--- a/thrust/thrust/detail/copy.h
+++ b/thrust/thrust/detail/copy.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/detail/copy.inl b/thrust/thrust/detail/copy.inl
index 4d62798c705..ada0aa5d2bd 100644
--- a/thrust/thrust/detail/copy.inl
+++ b/thrust/thrust/detail/copy.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/copy.h>
 #include <thrust/system/detail/generic/select_system.h>
 #include <thrust/system/detail/generic/copy.h>
diff --git a/thrust/thrust/detail/copy_if.h b/thrust/thrust/detail/copy_if.h
index 32eb5e08323..a7a72734665 100644
--- a/thrust/thrust/detail/copy_if.h
+++ b/thrust/thrust/detail/copy_if.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/detail/copy_if.inl b/thrust/thrust/detail/copy_if.inl
index 952541c5180..a013328bf85 100644
--- a/thrust/thrust/detail/copy_if.inl
+++ b/thrust/thrust/detail/copy_if.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/copy_if.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/copy_if.h>
diff --git a/thrust/thrust/detail/count.h b/thrust/thrust/detail/count.h
index 7c48bc54606..cc513f9bb75 100644
--- a/thrust/thrust/detail/count.h
+++ b/thrust/thrust/detail/count.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/detail/count.inl b/thrust/thrust/detail/count.inl
index 5d1f628a93c..c896e740859 100644
--- a/thrust/thrust/detail/count.inl
+++ b/thrust/thrust/detail/count.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/count.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
diff --git a/thrust/thrust/detail/cpp11_required.h b/thrust/thrust/detail/cpp11_required.h
index a7fb4fb12cd..6447d0eba7d 100644
--- a/thrust/thrust/detail/cpp11_required.h
+++ b/thrust/thrust/detail/cpp11_required.h
@@ -18,8 +18,14 @@
 
 #include <thrust/detail/config/cpp_dialect.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #ifndef THRUST_CPP11_REQUIRED_NO_ERROR
-#  if THRUST_CPP_DIALECT < 2011 
+#  if THRUST_CPP_DIALECT < 2011
 #    error C++11 is required for this Thrust feature; please upgrade your compiler or pass the appropriate -std=c++XX flag to it.
 #  endif
 #endif
diff --git a/thrust/thrust/detail/cpp14_required.h b/thrust/thrust/detail/cpp14_required.h
index 083c8a1ad47..27ed1bd7d96 100644
--- a/thrust/thrust/detail/cpp14_required.h
+++ b/thrust/thrust/detail/cpp14_required.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config/cpp_dialect.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #ifndef THRUST_CPP14_REQUIRED_NO_ERROR
 #  if THRUST_CPP_DIALECT < 2014
 #    error C++14 is required for this Thrust feature; please upgrade your compiler or pass the appropriate -std=c++14 flag to it.
diff --git a/thrust/thrust/detail/cstdint.h b/thrust/thrust/detail/cstdint.h
index f41e1147576..d9f5fd595db 100644
--- a/thrust/thrust/detail/cstdint.h
+++ b/thrust/thrust/detail/cstdint.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if (THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_GCC) || \
     (THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_CLANG) || \
     (THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_INTEL)
diff --git a/thrust/thrust/detail/dependencies_aware_execution_policy.h b/thrust/thrust/detail/dependencies_aware_execution_policy.h
index a7567a3fab3..d781daf2428 100644
--- a/thrust/thrust/detail/dependencies_aware_execution_policy.h
+++ b/thrust/thrust/detail/dependencies_aware_execution_policy.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp11_required.h>
 
 #if THRUST_CPP_DIALECT >= 2011
diff --git a/thrust/thrust/detail/device_delete.inl b/thrust/thrust/detail/device_delete.inl
index 87f73aad902..6b86808633d 100644
--- a/thrust/thrust/detail/device_delete.inl
+++ b/thrust/thrust/detail/device_delete.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/device_delete.h>
 #include <thrust/device_free.h>
 #include <thrust/detail/allocator/destroy_range.h>
diff --git a/thrust/thrust/detail/device_free.inl b/thrust/thrust/detail/device_free.inl
index 806802e1651..bc3ccbb30dd 100644
--- a/thrust/thrust/detail/device_free.inl
+++ b/thrust/thrust/detail/device_free.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/device_free.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
diff --git a/thrust/thrust/detail/device_malloc.inl b/thrust/thrust/detail/device_malloc.inl
index f4222f51da3..aaf5887b02f 100644
--- a/thrust/thrust/detail/device_malloc.inl
+++ b/thrust/thrust/detail/device_malloc.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/device_malloc.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
diff --git a/thrust/thrust/detail/device_new.inl b/thrust/thrust/detail/device_new.inl
index c66e2cbff6c..b6578a662ca 100644
--- a/thrust/thrust/detail/device_new.inl
+++ b/thrust/thrust/detail/device_new.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/device_new.h>
 #include <thrust/device_malloc.h>
 #include <thrust/uninitialized_fill.h>
diff --git a/thrust/thrust/detail/device_ptr.inl b/thrust/thrust/detail/device_ptr.inl
index 361c61f3314..d980f45dda1 100644
--- a/thrust/thrust/detail/device_ptr.inl
+++ b/thrust/thrust/detail/device_ptr.inl
@@ -16,9 +16,16 @@
 
 #pragma once
 
+#include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/device_ptr.h>
 #include <thrust/device_reference.h>
-#include <thrust/detail/config.h>
 #include <thrust/detail/type_traits.h>
 #include <thrust/iterator/iterator_traits.h>
 
diff --git a/thrust/thrust/detail/distance.inl b/thrust/thrust/detail/distance.inl
index 6702c2b6f73..452469cb4e1 100644
--- a/thrust/thrust/detail/distance.inl
+++ b/thrust/thrust/detail/distance.inl
@@ -16,8 +16,15 @@
 
 #pragma once
 
-#include <thrust/advance.h>
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
+#include <thrust/advance.h>
 #include <thrust/system/detail/generic/distance.h>
 #include <thrust/iterator/iterator_traits.h>
 
diff --git a/thrust/thrust/detail/equal.inl b/thrust/thrust/detail/equal.inl
index e21ddfa5abb..02d9686691c 100644
--- a/thrust/thrust/detail/equal.inl
+++ b/thrust/thrust/detail/equal.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/equal.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
diff --git a/thrust/thrust/detail/event_error.h b/thrust/thrust/detail/event_error.h
index b928e0650c2..90bfafcd887 100644
--- a/thrust/thrust/detail/event_error.h
+++ b/thrust/thrust/detail/event_error.h
@@ -20,6 +20,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp14_required.h>
 
 #if THRUST_CPP_DIALECT >= 2014
diff --git a/thrust/thrust/detail/execute_with_allocator.h b/thrust/thrust/detail/execute_with_allocator.h
index 430fe739c3f..a41a2c205fc 100644
--- a/thrust/thrust/detail/execute_with_allocator.h
+++ b/thrust/thrust/detail/execute_with_allocator.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/execute_with_allocator_fwd.h>
 #include <thrust/pair.h>
 #include <thrust/detail/raw_pointer_cast.h>
diff --git a/thrust/thrust/detail/execute_with_allocator_fwd.h b/thrust/thrust/detail/execute_with_allocator_fwd.h
index 1d5899a7db7..8c5cec162c9 100644
--- a/thrust/thrust/detail/execute_with_allocator_fwd.h
+++ b/thrust/thrust/detail/execute_with_allocator_fwd.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/type_traits.h>
 
 #if THRUST_CPP_DIALECT >= 2011
diff --git a/thrust/thrust/detail/execute_with_dependencies.h b/thrust/thrust/detail/execute_with_dependencies.h
index 9f82f28badf..834f8868ce6 100644
--- a/thrust/thrust/detail/execute_with_dependencies.h
+++ b/thrust/thrust/detail/execute_with_dependencies.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp11_required.h>
 
 #if THRUST_CPP_DIALECT >= 2011
@@ -96,7 +102,7 @@ struct execute_with_dependencies
 
     std::tuple<remove_cvref_t<Dependencies>...>
     __host__
-    extract_dependencies() 
+    extract_dependencies()
     {
         return std::move(dependencies);
     }
@@ -184,7 +190,7 @@ struct execute_with_allocator_and_dependencies
 
     std::tuple<remove_cvref_t<Dependencies>...>
     __host__
-    extract_dependencies() 
+    extract_dependencies()
     {
         return std::move(dependencies);
     }
diff --git a/thrust/thrust/detail/execution_policy.h b/thrust/thrust/detail/execution_policy.h
index dcc11a77033..87857a51d0a 100644
--- a/thrust/thrust/detail/execution_policy.h
+++ b/thrust/thrust/detail/execution_policy.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 THRUST_NAMESPACE_BEGIN
 
 namespace detail
diff --git a/thrust/thrust/detail/extrema.inl b/thrust/thrust/detail/extrema.inl
index 2c1750e7d48..cfa21741109 100644
--- a/thrust/thrust/detail/extrema.inl
+++ b/thrust/thrust/detail/extrema.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/extrema.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
diff --git a/thrust/thrust/detail/fill.inl b/thrust/thrust/detail/fill.inl
index e68672bbebe..0ba85a6f1e0 100644
--- a/thrust/thrust/detail/fill.inl
+++ b/thrust/thrust/detail/fill.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/fill.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
diff --git a/thrust/thrust/detail/find.inl b/thrust/thrust/detail/find.inl
index 5b494f61a60..621a2cd8e4b 100644
--- a/thrust/thrust/detail/find.inl
+++ b/thrust/thrust/detail/find.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
 #include <thrust/system/detail/generic/find.h>
diff --git a/thrust/thrust/detail/for_each.inl b/thrust/thrust/detail/for_each.inl
index 4ba39c71a40..1f5111ade7c 100644
--- a/thrust/thrust/detail/for_each.inl
+++ b/thrust/thrust/detail/for_each.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/for_each.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
diff --git a/thrust/thrust/detail/function.h b/thrust/thrust/detail/function.h
index ba20507a547..14b288bf94e 100644
--- a/thrust/thrust/detail/function.h
+++ b/thrust/thrust/detail/function.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/raw_reference_cast.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/detail/functional.inl b/thrust/thrust/detail/functional.inl
index bdf8e041529..bf56fce23a2 100644
--- a/thrust/thrust/detail/functional.inl
+++ b/thrust/thrust/detail/functional.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/functional.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/detail/functional/actor.h b/thrust/thrust/detail/functional/actor.h
index 208a8706180..52d0c6660ae 100644
--- a/thrust/thrust/detail/functional/actor.h
+++ b/thrust/thrust/detail/functional/actor.h
@@ -26,6 +26,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/tuple.h>
 #include <thrust/detail/functional/value.h>
 #include <thrust/detail/functional/composite.h>
diff --git a/thrust/thrust/detail/functional/actor.inl b/thrust/thrust/detail/functional/actor.inl
index 5272101b7f4..9043062436f 100644
--- a/thrust/thrust/detail/functional/actor.inl
+++ b/thrust/thrust/detail/functional/actor.inl
@@ -26,6 +26,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/functional/composite.h>
 #include <thrust/detail/functional/operators/assignment_operator.h>
 #include <thrust/functional.h>
diff --git a/thrust/thrust/detail/functional/argument.h b/thrust/thrust/detail/functional/argument.h
index b1ceb220555..be3824e3671 100644
--- a/thrust/thrust/detail/functional/argument.h
+++ b/thrust/thrust/detail/functional/argument.h
@@ -26,6 +26,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/tuple.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/detail/functional/composite.h b/thrust/thrust/detail/functional/composite.h
index 40466b3b8ea..129d8320831 100644
--- a/thrust/thrust/detail/functional/composite.h
+++ b/thrust/thrust/detail/functional/composite.h
@@ -27,6 +27,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/functional/actor.h>
 #include <thrust/tuple.h>
 
diff --git a/thrust/thrust/detail/functional/operators.h b/thrust/thrust/detail/functional/operators.h
index f86ea205218..74804745f82 100644
--- a/thrust/thrust/detail/functional/operators.h
+++ b/thrust/thrust/detail/functional/operators.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/functional/operators/arithmetic_operators.h>
 #include <thrust/detail/functional/operators/relational_operators.h>
 #include <thrust/detail/functional/operators/logical_operators.h>
diff --git a/thrust/thrust/detail/functional/operators/arithmetic_operators.h b/thrust/thrust/detail/functional/operators/arithmetic_operators.h
index 443d307cbd6..238f2e5698c 100644
--- a/thrust/thrust/detail/functional/operators/arithmetic_operators.h
+++ b/thrust/thrust/detail/functional/operators/arithmetic_operators.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/functional/actor.h>
 #include <thrust/detail/functional/composite.h>
 #include <thrust/detail/functional/operators/operator_adaptors.h>
diff --git a/thrust/thrust/detail/functional/operators/assignment_operator.h b/thrust/thrust/detail/functional/operators/assignment_operator.h
index 870354b6f0b..62f93ce710b 100644
--- a/thrust/thrust/detail/functional/operators/assignment_operator.h
+++ b/thrust/thrust/detail/functional/operators/assignment_operator.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/functional/actor.h>
 #include <thrust/detail/functional/composite.h>
 #include <thrust/detail/functional/operators/operator_adaptors.h>
diff --git a/thrust/thrust/detail/functional/operators/bitwise_operators.h b/thrust/thrust/detail/functional/operators/bitwise_operators.h
index 065cd15403c..b49a464aba6 100644
--- a/thrust/thrust/detail/functional/operators/bitwise_operators.h
+++ b/thrust/thrust/detail/functional/operators/bitwise_operators.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/functional/actor.h>
 #include <thrust/detail/functional/composite.h>
 #include <thrust/detail/functional/operators/operator_adaptors.h>
diff --git a/thrust/thrust/detail/functional/operators/compound_assignment_operators.h b/thrust/thrust/detail/functional/operators/compound_assignment_operators.h
index b5ba77fb48d..61fc6f5971e 100644
--- a/thrust/thrust/detail/functional/operators/compound_assignment_operators.h
+++ b/thrust/thrust/detail/functional/operators/compound_assignment_operators.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/functional/actor.h>
 #include <thrust/detail/functional/composite.h>
 #include <thrust/detail/functional/operators/operator_adaptors.h>
diff --git a/thrust/thrust/detail/functional/operators/logical_operators.h b/thrust/thrust/detail/functional/operators/logical_operators.h
index e1e4ff71994..a30850b5886 100644
--- a/thrust/thrust/detail/functional/operators/logical_operators.h
+++ b/thrust/thrust/detail/functional/operators/logical_operators.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/functional/actor.h>
 #include <thrust/detail/functional/composite.h>
 #include <thrust/detail/functional/operators/operator_adaptors.h>
diff --git a/thrust/thrust/detail/functional/operators/operator_adaptors.h b/thrust/thrust/detail/functional/operators/operator_adaptors.h
index addfbd3dce2..6f69e6f91ec 100644
--- a/thrust/thrust/detail/functional/operators/operator_adaptors.h
+++ b/thrust/thrust/detail/functional/operators/operator_adaptors.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/functional/argument.h>
 #include <thrust/detail/type_deduction.h>
 #include <thrust/tuple.h>
diff --git a/thrust/thrust/detail/functional/operators/relational_operators.h b/thrust/thrust/detail/functional/operators/relational_operators.h
index 6c58325e225..fb55c34627d 100644
--- a/thrust/thrust/detail/functional/operators/relational_operators.h
+++ b/thrust/thrust/detail/functional/operators/relational_operators.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/functional/actor.h>
 #include <thrust/detail/functional/composite.h>
 #include <thrust/detail/functional/operators/operator_adaptors.h>
diff --git a/thrust/thrust/detail/functional/placeholder.h b/thrust/thrust/detail/functional/placeholder.h
index e3c0835531f..19b40de8722 100644
--- a/thrust/thrust/detail/functional/placeholder.h
+++ b/thrust/thrust/detail/functional/placeholder.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/functional/actor.h>
 #include <thrust/detail/functional/argument.h>
 
diff --git a/thrust/thrust/detail/functional/value.h b/thrust/thrust/detail/functional/value.h
index d6b1563b15c..0b30a0cf39b 100644
--- a/thrust/thrust/detail/functional/value.h
+++ b/thrust/thrust/detail/functional/value.h
@@ -26,6 +26,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/functional/actor.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/detail/gather.inl b/thrust/thrust/detail/gather.inl
index 3812702f61c..9488c4caf15 100644
--- a/thrust/thrust/detail/gather.inl
+++ b/thrust/thrust/detail/gather.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/gather.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
diff --git a/thrust/thrust/detail/generate.inl b/thrust/thrust/detail/generate.inl
index 2ecb65d5843..8e3924c2810 100644
--- a/thrust/thrust/detail/generate.inl
+++ b/thrust/thrust/detail/generate.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/generate.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
diff --git a/thrust/thrust/detail/get_iterator_value.h b/thrust/thrust/detail/get_iterator_value.h
index 27e0a4e47a0..c5828ff2a6b 100644
--- a/thrust/thrust/detail/get_iterator_value.h
+++ b/thrust/thrust/detail/get_iterator_value.h
@@ -1,4 +1,3 @@
-#pragma once
 /*
  *  Copyright 2008-2016 NVIDIA Corporation
  *
@@ -14,8 +13,16 @@
  *  See the License for the specific language governing permissions and
  *  limitations under the License.
  */
+#pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/execution_policy.h>
 #include <thrust/detail/type_traits/pointer_traits.h>
@@ -39,12 +46,12 @@ get_iterator_value(thrust::execution_policy<DerivedPolicy> &, Iterator it)
 // get_iterator_value specialization on pointer
 // ----------------------------------------------
 // we can't just dereference a pointer in the usual way, because
-// it may point to a location in the device memory. 
+// it may point to a location in the device memory.
 // we use get_value(exec,pointer*) function
 // to perform a dereferencing consistent with the execution policy
 template<typename DerivedPolicy, typename Pointer>
 __host__ __device__
-typename thrust::detail::pointer_traits<Pointer*>::element_type 
+typename thrust::detail::pointer_traits<Pointer*>::element_type
 get_iterator_value(thrust::execution_policy<DerivedPolicy> &exec, Pointer* ptr)
 {
   return get_value(derived_cast(exec),ptr);
diff --git a/thrust/thrust/detail/inner_product.inl b/thrust/thrust/detail/inner_product.inl
index 97cd2b0b52e..373d15af0be 100644
--- a/thrust/thrust/detail/inner_product.inl
+++ b/thrust/thrust/detail/inner_product.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/inner_product.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
diff --git a/thrust/thrust/detail/integer_math.h b/thrust/thrust/detail/integer_math.h
index 0f8c8aac168..c5640493f95 100644
--- a/thrust/thrust/detail/integer_math.h
+++ b/thrust/thrust/detail/integer_math.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/type_deduction.h>
 
 #include <nv/target>
diff --git a/thrust/thrust/detail/integer_traits.h b/thrust/thrust/detail/integer_traits.h
index 853af20b860..cafea573f23 100644
--- a/thrust/thrust/detail/integer_traits.h
+++ b/thrust/thrust/detail/integer_traits.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <limits>
 #include <limits.h>
 
diff --git a/thrust/thrust/detail/internal_functional.h b/thrust/thrust/detail/internal_functional.h
index a0c4056feba..585592ebc14 100644
--- a/thrust/thrust/detail/internal_functional.h
+++ b/thrust/thrust/detail/internal_functional.h
@@ -21,9 +21,16 @@
 
 #pragma once
 
+#include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/tuple.h>
 #include <thrust/iterator/iterator_traits.h>
-#include <thrust/detail/config.h>
 #include <thrust/detail/static_assert.h>
 #include <thrust/detail/type_traits.h>
 #include <thrust/iterator/detail/tuple_of_iterator_references.h>
diff --git a/thrust/thrust/detail/logical.inl b/thrust/thrust/detail/logical.inl
index 3d39cac9268..30992bbaa46 100644
--- a/thrust/thrust/detail/logical.inl
+++ b/thrust/thrust/detail/logical.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
 #include <thrust/system/detail/generic/logical.h>
diff --git a/thrust/thrust/detail/malloc_and_free.h b/thrust/thrust/detail/malloc_and_free.h
index 14351889319..a03df639d44 100644
--- a/thrust/thrust/detail/malloc_and_free.h
+++ b/thrust/thrust/detail/malloc_and_free.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 #include <thrust/detail/pointer.h>
 #include <thrust/detail/raw_pointer_cast.h>
diff --git a/thrust/thrust/detail/memory_algorithms.h b/thrust/thrust/detail/memory_algorithms.h
index 2f6b3a81d71..05d68f043fa 100644
--- a/thrust/thrust/detail/memory_algorithms.h
+++ b/thrust/thrust/detail/memory_algorithms.h
@@ -9,6 +9,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/type_traits.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/detail/allocator/allocator_traits.h>
diff --git a/thrust/thrust/detail/memory_wrapper.h b/thrust/thrust/detail/memory_wrapper.h
index bfc9056fa15..1df1dfcb813 100644
--- a/thrust/thrust/detail/memory_wrapper.h
+++ b/thrust/thrust/detail/memory_wrapper.h
@@ -16,6 +16,14 @@
 
 #pragma once
 
+#include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // When a compiler uses Thrust as part of its implementation of Standard C++
 // algorithms, a cycle of included files may result when Thrust code tries to
 // use a standard algorithm.  Having a macro that is defined only when Thrust
diff --git a/thrust/thrust/detail/merge.inl b/thrust/thrust/detail/merge.inl
index 1595cc1a145..64b1f5f2ee1 100644
--- a/thrust/thrust/detail/merge.inl
+++ b/thrust/thrust/detail/merge.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/merge.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
diff --git a/thrust/thrust/detail/minmax.h b/thrust/thrust/detail/minmax.h
index c565a74bdff..bbc54d47537 100644
--- a/thrust/thrust/detail/minmax.h
+++ b/thrust/thrust/detail/minmax.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 THRUST_NAMESPACE_BEGIN
 
 template<typename T, typename BinaryPredicate>
diff --git a/thrust/thrust/detail/mismatch.inl b/thrust/thrust/detail/mismatch.inl
index 16c579d8069..e0dee618395 100644
--- a/thrust/thrust/detail/mismatch.inl
+++ b/thrust/thrust/detail/mismatch.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/mismatch.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
diff --git a/thrust/thrust/detail/modern_gcc_required.h b/thrust/thrust/detail/modern_gcc_required.h
index a8c3d98ba99..517d285a2c8 100644
--- a/thrust/thrust/detail/modern_gcc_required.h
+++ b/thrust/thrust/detail/modern_gcc_required.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config/cpp_dialect.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #ifndef THRUST_MODERN_GCC_REQUIRED_NO_ERROR
 #  if defined(THRUST_GCC_VERSION) && !defined(THRUST_MODERN_GCC)
 #    error GCC 5 or later is required for this Thrust feature; please upgrade your compiler.
diff --git a/thrust/thrust/detail/mpl/math.h b/thrust/thrust/detail/mpl/math.h
index bda98003cd5..ccdba5ce3b2 100644
--- a/thrust/thrust/detail/mpl/math.h
+++ b/thrust/thrust/detail/mpl/math.h
@@ -24,6 +24,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 THRUST_NAMESPACE_BEGIN
 
 namespace detail
diff --git a/thrust/thrust/detail/numeric_traits.h b/thrust/thrust/detail/numeric_traits.h
index e728adcaf3e..8e1fa8e5936 100644
--- a/thrust/thrust/detail/numeric_traits.h
+++ b/thrust/thrust/detail/numeric_traits.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/type_traits.h>
 #include <limits>
 
@@ -46,7 +52,7 @@ template<typename T>
         >,
         integral_constant<
           int,
-          sizeof(T) * std::numeric_limits<unsigned char>::digits - (is_signed<T>::value ? 1 : 0)  
+          sizeof(T) * std::numeric_limits<unsigned char>::digits - (is_signed<T>::value ? 1 : 0)
         >
       >::type
 {}; // end num_digits
diff --git a/thrust/thrust/detail/numeric_wrapper.h b/thrust/thrust/detail/numeric_wrapper.h
index 9ebc6e23b71..0460fba98d6 100644
--- a/thrust/thrust/detail/numeric_wrapper.h
+++ b/thrust/thrust/detail/numeric_wrapper.h
@@ -16,6 +16,14 @@
 
 #pragma once
 
+#include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // When a compiler uses Thrust as part of its implementation of Standard C++
 // algorithms, a cycle of included files may result when Thrust code tries to
 // use a standard algorithm.  Having a macro that is defined only when Thrust
diff --git a/thrust/thrust/detail/overlapped_copy.h b/thrust/thrust/detail/overlapped_copy.h
index 418497de8db..7d4f5dc2ada 100644
--- a/thrust/thrust/detail/overlapped_copy.h
+++ b/thrust/thrust/detail/overlapped_copy.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/iterator/detail/minimum_system.h>
 #include <thrust/detail/copy.h>
diff --git a/thrust/thrust/detail/partition.inl b/thrust/thrust/detail/partition.inl
index 5c51bca8095..5a72aad321e 100644
--- a/thrust/thrust/detail/partition.inl
+++ b/thrust/thrust/detail/partition.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/partition.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
diff --git a/thrust/thrust/detail/pointer.h b/thrust/thrust/detail/pointer.h
index aed1fcc2444..3dab11a58b1 100644
--- a/thrust/thrust/detail/pointer.h
+++ b/thrust/thrust/detail/pointer.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/iterator/iterator_adaptor.h>
 #include <thrust/iterator/detail/iterator_traversal_tags.h>
 #include <thrust/type_traits/remove_cvref.h>
diff --git a/thrust/thrust/detail/pointer.inl b/thrust/thrust/detail/pointer.inl
index de05ff20f0c..f8d9f20e8f0 100644
--- a/thrust/thrust/detail/pointer.inl
+++ b/thrust/thrust/detail/pointer.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/pointer.h>
 #include <thrust/detail/type_traits.h>
 
diff --git a/thrust/thrust/detail/preprocessor.h b/thrust/thrust/detail/preprocessor.h
index 2e850c764dd..a077405d84d 100644
--- a/thrust/thrust/detail/preprocessor.h
+++ b/thrust/thrust/detail/preprocessor.h
@@ -9,6 +9,14 @@
 
 #pragma once
 
+#include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 ///////////////////////////////////////////////////////////////////////////////
 
 /// \def THRUST_PP_STRINGIZE(expr)
diff --git a/thrust/thrust/detail/range/head_flags.h b/thrust/thrust/detail/range/head_flags.h
index b755840c991..188c0fc67f8 100644
--- a/thrust/thrust/detail/range/head_flags.h
+++ b/thrust/thrust/detail/range/head_flags.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/iterator/transform_iterator.h>
 #include <thrust/iterator/zip_iterator.h>
 #include <thrust/iterator/counting_iterator.h>
diff --git a/thrust/thrust/detail/range/tail_flags.h b/thrust/thrust/detail/range/tail_flags.h
index 41ee5dd29cc..10126e73f76 100644
--- a/thrust/thrust/detail/range/tail_flags.h
+++ b/thrust/thrust/detail/range/tail_flags.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/iterator/transform_iterator.h>
 #include <thrust/iterator/zip_iterator.h>
 #include <thrust/iterator/counting_iterator.h>
diff --git a/thrust/thrust/detail/raw_pointer_cast.h b/thrust/thrust/detail/raw_pointer_cast.h
index 53a77861ef0..7f72b30c293 100644
--- a/thrust/thrust/detail/raw_pointer_cast.h
+++ b/thrust/thrust/detail/raw_pointer_cast.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/type_traits/pointer_traits.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/detail/raw_reference_cast.h b/thrust/thrust/detail/raw_reference_cast.h
index 99af0da183d..c01fe9c04d9 100644
--- a/thrust/thrust/detail/raw_reference_cast.h
+++ b/thrust/thrust/detail/raw_reference_cast.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/raw_pointer_cast.h>
 #include <thrust/detail/type_traits/has_nested_type.h>
 #include <thrust/detail/type_traits.h>
diff --git a/thrust/thrust/detail/reduce.inl b/thrust/thrust/detail/reduce.inl
index 448a4b38c4a..86a82ab30a1 100644
--- a/thrust/thrust/detail/reduce.inl
+++ b/thrust/thrust/detail/reduce.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/reduce.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
diff --git a/thrust/thrust/detail/reference.h b/thrust/thrust/detail/reference.h
index 5cc13625daf..f1be715a9f5 100644
--- a/thrust/thrust/detail/reference.h
+++ b/thrust/thrust/detail/reference.h
@@ -14,7 +14,7 @@
  *  limitations under the License.
  */
 
-/*! \file 
+/*! \file
  *  \brief A pointer to a variable which resides in memory associated with a
  *  system.
  */
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/reference_forward_declaration.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
diff --git a/thrust/thrust/detail/reference_forward_declaration.h b/thrust/thrust/detail/reference_forward_declaration.h
index 6f2b999497a..a412dc5bc4f 100644
--- a/thrust/thrust/detail/reference_forward_declaration.h
+++ b/thrust/thrust/detail/reference_forward_declaration.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/use_default.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/detail/remove.inl b/thrust/thrust/detail/remove.inl
index 7ccc0cc4656..efd99651d1d 100644
--- a/thrust/thrust/detail/remove.inl
+++ b/thrust/thrust/detail/remove.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/remove.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
diff --git a/thrust/thrust/detail/replace.inl b/thrust/thrust/detail/replace.inl
index 629287bee08..7dd797b12ae 100644
--- a/thrust/thrust/detail/replace.inl
+++ b/thrust/thrust/detail/replace.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/replace.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
diff --git a/thrust/thrust/detail/reverse.inl b/thrust/thrust/detail/reverse.inl
index dc316d18f52..128af196874 100644
--- a/thrust/thrust/detail/reverse.inl
+++ b/thrust/thrust/detail/reverse.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/reverse.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
diff --git a/thrust/thrust/detail/scan.inl b/thrust/thrust/detail/scan.inl
index b781b0e280c..7300ff4c3a4 100644
--- a/thrust/thrust/detail/scan.inl
+++ b/thrust/thrust/detail/scan.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/scan.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
diff --git a/thrust/thrust/detail/scatter.inl b/thrust/thrust/detail/scatter.inl
index 30dd611d11f..115020afb47 100644
--- a/thrust/thrust/detail/scatter.inl
+++ b/thrust/thrust/detail/scatter.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/scatter.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
diff --git a/thrust/thrust/detail/select_system.h b/thrust/thrust/detail/select_system.h
index 968446162dd..45afb93e33d 100644
--- a/thrust/thrust/detail/select_system.h
+++ b/thrust/thrust/detail/select_system.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp11_required.h>
 
 #if THRUST_CPP_DIALECT >= 2011
diff --git a/thrust/thrust/detail/seq.h b/thrust/thrust/detail/seq.h
index ba18c2dbf51..13df612cb20 100644
--- a/thrust/thrust/detail/seq.h
+++ b/thrust/thrust/detail/seq.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/allocator_aware_execution_policy.h>
 #include <thrust/system/detail/sequential/execution_policy.h>
 
diff --git a/thrust/thrust/detail/sequence.inl b/thrust/thrust/detail/sequence.inl
index ffc9b968b97..6ae4180284b 100644
--- a/thrust/thrust/detail/sequence.inl
+++ b/thrust/thrust/detail/sequence.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/sequence.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
diff --git a/thrust/thrust/detail/set_operations.inl b/thrust/thrust/detail/set_operations.inl
index 7915f7b3eab..e993b583188 100644
--- a/thrust/thrust/detail/set_operations.inl
+++ b/thrust/thrust/detail/set_operations.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
 #include <thrust/system/detail/generic/set_operations.h>
diff --git a/thrust/thrust/detail/shuffle.inl b/thrust/thrust/detail/shuffle.inl
index 48f5ba63929..d34454cb405 100644
--- a/thrust/thrust/detail/shuffle.inl
+++ b/thrust/thrust/detail/shuffle.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp11_required.h>
 
 #if THRUST_CPP_DIALECT >= 2011
diff --git a/thrust/thrust/detail/sort.inl b/thrust/thrust/detail/sort.inl
index 53f8bad93b1..6b6f5ba639e 100644
--- a/thrust/thrust/detail/sort.inl
+++ b/thrust/thrust/detail/sort.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/sort.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
diff --git a/thrust/thrust/detail/static_assert.h b/thrust/thrust/detail/static_assert.h
index 0e613279055..dad671caf2b 100644
--- a/thrust/thrust/detail/static_assert.h
+++ b/thrust/thrust/detail/static_assert.h
@@ -16,7 +16,7 @@
 
 /*
  * (C) Copyright John Maddock 2000.
- * 
+ *
  * Distributed under the Boost Software License, Version 1.0.
  * (See accompanying NOTICE file for the complete license)
  *
@@ -26,6 +26,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/type_traits.h>
 #include <thrust/detail/preprocessor.h>
 
@@ -70,14 +76,14 @@ template <int x> struct static_assert_test {};
     >                                                                         \
       THRUST_PP_CAT2(thrust_static_assert_typedef_, __LINE__)                 \
       __attribute__((unused))                                                 \
-    /**/      
+    /**/
 #else
 #  define THRUST_STATIC_ASSERT(B)                                             \
     typedef THRUST_NS_QUALIFIER::detail::static_assert_test<                  \
       sizeof(THRUST_NS_QUALIFIER::detail::STATIC_ASSERTION_FAILURE<(bool)(B)>)\
     >                                                                         \
       THRUST_PP_CAT2(thrust_static_assert_typedef_, __LINE__)                 \
-    /**/      
+    /**/
 #endif
 
 #define THRUST_STATIC_ASSERT_MSG(B, msg) THRUST_STATIC_ASSERT(B)
diff --git a/thrust/thrust/detail/static_map.h b/thrust/thrust/detail/static_map.h
index 9f0d79e83e2..c6a69981714 100644
--- a/thrust/thrust/detail/static_map.h
+++ b/thrust/thrust/detail/static_map.h
@@ -16,9 +16,13 @@
 
 #pragma once
 
-
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 THRUST_NAMESPACE_BEGIN
 namespace detail
diff --git a/thrust/thrust/detail/swap.h b/thrust/thrust/detail/swap.h
index 305750f8ae7..41782b12ef2 100644
--- a/thrust/thrust/detail/swap.h
+++ b/thrust/thrust/detail/swap.h
@@ -19,6 +19,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 THRUST_NAMESPACE_BEGIN
 
 __thrust_exec_check_disable__
diff --git a/thrust/thrust/detail/swap.inl b/thrust/thrust/detail/swap.inl
index 196c34f41ee..3d11608821e 100644
--- a/thrust/thrust/detail/swap.inl
+++ b/thrust/thrust/detail/swap.inl
@@ -16,6 +16,14 @@
 
 #pragma once
 
+#include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/swap.h>
 #include <thrust/detail/swap.h>
 #include <thrust/detail/swap_ranges.inl>
diff --git a/thrust/thrust/detail/swap_ranges.inl b/thrust/thrust/detail/swap_ranges.inl
index 1f35c1ff318..0e834ac36ad 100644
--- a/thrust/thrust/detail/swap_ranges.inl
+++ b/thrust/thrust/detail/swap_ranges.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/swap.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
diff --git a/thrust/thrust/detail/tabulate.inl b/thrust/thrust/detail/tabulate.inl
index 308be061faf..7f56d95f1aa 100644
--- a/thrust/thrust/detail/tabulate.inl
+++ b/thrust/thrust/detail/tabulate.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/tabulate.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
diff --git a/thrust/thrust/detail/temporary_array.h b/thrust/thrust/detail/temporary_array.h
index cf4bc7d2d2e..5e938a96dc0 100644
--- a/thrust/thrust/detail/temporary_array.h
+++ b/thrust/thrust/detail/temporary_array.h
@@ -22,6 +22,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 THRUST_NAMESPACE_BEGIN
 namespace detail
 {
@@ -34,7 +40,6 @@ template<typename T, typename System>
 } // end detail
 THRUST_NAMESPACE_END
 
-#include <thrust/detail/config.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/iterator/detail/tagged_iterator.h>
 #include <thrust/detail/contiguous_storage.h>
diff --git a/thrust/thrust/detail/temporary_array.inl b/thrust/thrust/detail/temporary_array.inl
index 90b7279ac6d..6edde810008 100644
--- a/thrust/thrust/detail/temporary_array.inl
+++ b/thrust/thrust/detail/temporary_array.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/temporary_array.h>
 #include <thrust/distance.h>
 #include <thrust/system/detail/generic/select_system.h>
diff --git a/thrust/thrust/detail/temporary_buffer.h b/thrust/thrust/detail/temporary_buffer.h
index be95e7180f5..a32c014635d 100644
--- a/thrust/thrust/detail/temporary_buffer.h
+++ b/thrust/thrust/detail/temporary_buffer.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 #include <thrust/pair.h>
 #include <thrust/detail/pointer.h>
diff --git a/thrust/thrust/detail/transform.inl b/thrust/thrust/detail/transform.inl
index 62bafd35ef0..0999bca04ac 100644
--- a/thrust/thrust/detail/transform.inl
+++ b/thrust/thrust/detail/transform.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/transform.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
diff --git a/thrust/thrust/detail/transform_reduce.inl b/thrust/thrust/detail/transform_reduce.inl
index 702dd9f73b5..16b41010e61 100644
--- a/thrust/thrust/detail/transform_reduce.inl
+++ b/thrust/thrust/detail/transform_reduce.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
 #include <thrust/system/detail/generic/transform_reduce.h>
diff --git a/thrust/thrust/detail/transform_scan.inl b/thrust/thrust/detail/transform_scan.inl
index 957001cefd6..f78fc055689 100644
--- a/thrust/thrust/detail/transform_scan.inl
+++ b/thrust/thrust/detail/transform_scan.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/scan.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
diff --git a/thrust/thrust/detail/trivial_sequence.h b/thrust/thrust/detail/trivial_sequence.h
index 2cf98e7872a..df70462c265 100644
--- a/thrust/thrust/detail/trivial_sequence.h
+++ b/thrust/thrust/detail/trivial_sequence.h
@@ -25,6 +25,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/detail/type_traits.h>
 #include <thrust/detail/execution_policy.h>
@@ -64,7 +70,7 @@ struct _trivial_sequence<Iterator, DerivedPolicy, thrust::detail::false_type>
 {
     typedef typename thrust::iterator_value<Iterator>::type iterator_value;
     typedef typename thrust::detail::temporary_array<iterator_value, DerivedPolicy>::iterator iterator_type;
-    
+
     thrust::detail::temporary_array<iterator_value, DerivedPolicy> buffer;
 
     __host__ __device__
diff --git a/thrust/thrust/detail/tuple_algorithms.h b/thrust/thrust/detail/tuple_algorithms.h
index 2e49f4281f3..f75f3f63ac8 100644
--- a/thrust/thrust/detail/tuple_algorithms.h
+++ b/thrust/thrust/detail/tuple_algorithms.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp11_required.h>
 
 #if THRUST_CPP_DIALECT >= 2011
diff --git a/thrust/thrust/detail/tuple_meta_transform.h b/thrust/thrust/detail/tuple_meta_transform.h
index 285cae8b4c7..9ecfe113ab8 100644
--- a/thrust/thrust/detail/tuple_meta_transform.h
+++ b/thrust/thrust/detail/tuple_meta_transform.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/tuple.h>
 #include <thrust/type_traits/integer_sequence.h>
 
diff --git a/thrust/thrust/detail/tuple_transform.h b/thrust/thrust/detail/tuple_transform.h
index 1011d5179e3..0be5814cd00 100644
--- a/thrust/thrust/detail/tuple_transform.h
+++ b/thrust/thrust/detail/tuple_transform.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/tuple.h>
 #include <thrust/detail/tuple_meta_transform.h>
 
diff --git a/thrust/thrust/detail/type_deduction.h b/thrust/thrust/detail/type_deduction.h
index 6f240711d0a..d861400b6ec 100644
--- a/thrust/thrust/detail/type_deduction.h
+++ b/thrust/thrust/detail/type_deduction.h
@@ -8,6 +8,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp11_required.h>
 
 #if THRUST_CPP_DIALECT >= 2011
diff --git a/thrust/thrust/detail/type_traits.h b/thrust/thrust/detail/type_traits.h
index 605db3f5d40..1d19fc7b838 100644
--- a/thrust/thrust/detail/type_traits.h
+++ b/thrust/thrust/detail/type_traits.h
@@ -24,6 +24,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cuda/std/type_traits>
 
 #include <type_traits>
diff --git a/thrust/thrust/detail/type_traits/function_traits.h b/thrust/thrust/detail/type_traits/function_traits.h
index 1098201366e..7565b5a5b47 100644
--- a/thrust/thrust/detail/type_traits/function_traits.h
+++ b/thrust/thrust/detail/type_traits/function_traits.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/type_traits.h>
 #include <thrust/detail/type_traits/has_nested_type.h>
 
diff --git a/thrust/thrust/detail/type_traits/has_member_function.h b/thrust/thrust/detail/type_traits/has_member_function.h
index c33fe28f654..f12802d9d8a 100644
--- a/thrust/thrust/detail/type_traits/has_member_function.h
+++ b/thrust/thrust/detail/type_traits/has_member_function.h
@@ -16,6 +16,14 @@
 
 #pragma once
 
+#include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/type_traits.h>
 
 #include <utility> // for std::declval
diff --git a/thrust/thrust/detail/type_traits/has_nested_type.h b/thrust/thrust/detail/type_traits/has_nested_type.h
index 78bb4b7f57a..860514f35d2 100644
--- a/thrust/thrust/detail/type_traits/has_nested_type.h
+++ b/thrust/thrust/detail/type_traits/has_nested_type.h
@@ -16,6 +16,14 @@
 
 #pragma once
 
+#include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/type_traits.h>
 
 #define __THRUST_DEFINE_HAS_NESTED_TYPE(trait_name, nested_type_name) \
diff --git a/thrust/thrust/detail/type_traits/has_trivial_assign.h b/thrust/thrust/detail/type_traits/has_trivial_assign.h
index 7222ce593f5..8d057b72033 100644
--- a/thrust/thrust/detail/type_traits/has_trivial_assign.h
+++ b/thrust/thrust/detail/type_traits/has_trivial_assign.h
@@ -23,6 +23,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/type_traits.h>
 
 #include <cuda/std/type_traits>
@@ -32,11 +38,11 @@ THRUST_NAMESPACE_BEGIN
 namespace detail
 {
 
-template<typename T> 
+template<typename T>
 struct has_trivial_assign
   : public integral_constant<
-      bool, 
-      (is_pod<T>::value && !is_const<T>::value) 
+      bool,
+      (is_pod<T>::value && !is_const<T>::value)
       || ::cuda::std::is_trivially_copy_assignable<T>::value
     >
 {};
diff --git a/thrust/thrust/detail/type_traits/is_call_possible.h b/thrust/thrust/detail/type_traits/is_call_possible.h
index 58c1aca4df0..cbc231e8760 100644
--- a/thrust/thrust/detail/type_traits/is_call_possible.h
+++ b/thrust/thrust/detail/type_traits/is_call_possible.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/type_traits.h>
 #include <thrust/detail/type_traits/has_member_function.h>
 
@@ -30,24 +36,24 @@ namespace detail
 namespace is_call_possible_detail
 {
 
-template<typename T> class void_exp_result {}; 
+template<typename T> class void_exp_result {};
 
-template<typename T, typename U> 
-U const& operator,(U const&, void_exp_result<T>); 
+template<typename T, typename U>
+U const& operator,(U const&, void_exp_result<T>);
 
-template<typename T, typename U> 
-U& operator,(U&, void_exp_result<T>); 
+template<typename T, typename U>
+U& operator,(U&, void_exp_result<T>);
 
-template<typename src_type, typename dest_type> 
-struct clone_constness 
+template<typename src_type, typename dest_type>
+struct clone_constness
 {
-  typedef dest_type type; 
-}; 
+  typedef dest_type type;
+};
 
-template<typename src_type, typename dest_type> 
-struct clone_constness<const src_type, dest_type> 
-{ 
-  typedef const dest_type type; 
+template<typename src_type, typename dest_type>
+struct clone_constness<const src_type, dest_type>
+{
+  typedef const dest_type type;
 };
 
 } // end is_call_possible_detail
@@ -158,5 +164,5 @@ struct trait_name
   public:                                                                                                                                 \
     static const bool value = impl<trait_name##_has_member<T,Signature>::value, Signature>::value;                                        \
     typedef thrust::detail::integral_constant<bool,value> type;                                                                           \
-}; 
+};
 
diff --git a/thrust/thrust/detail/type_traits/is_metafunction_defined.h b/thrust/thrust/detail/type_traits/is_metafunction_defined.h
index 2c7a4be52df..5e2c217322e 100644
--- a/thrust/thrust/detail/type_traits/is_metafunction_defined.h
+++ b/thrust/thrust/detail/type_traits/is_metafunction_defined.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/type_traits/has_nested_type.h>
 #include <thrust/detail/type_traits.h>
 
diff --git a/thrust/thrust/detail/type_traits/iterator/is_discard_iterator.h b/thrust/thrust/detail/type_traits/iterator/is_discard_iterator.h
index 210409d6269..218273ff438 100644
--- a/thrust/thrust/detail/type_traits/iterator/is_discard_iterator.h
+++ b/thrust/thrust/detail/type_traits/iterator/is_discard_iterator.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/type_traits.h>
 #include <thrust/iterator/discard_iterator.h>
 
diff --git a/thrust/thrust/detail/type_traits/iterator/is_output_iterator.h b/thrust/thrust/detail/type_traits/iterator/is_output_iterator.h
index 555b67400cd..fdf95e0fae7 100644
--- a/thrust/thrust/detail/type_traits/iterator/is_output_iterator.h
+++ b/thrust/thrust/detail/type_traits/iterator/is_output_iterator.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/type_traits.h>
 #include <thrust/detail/type_traits/is_metafunction_defined.h>
 #include <thrust/iterator/iterator_traits.h>
diff --git a/thrust/thrust/detail/type_traits/minimum_type.h b/thrust/thrust/detail/type_traits/minimum_type.h
index 2417e327df8..d520fd5f827 100644
--- a/thrust/thrust/detail/type_traits/minimum_type.h
+++ b/thrust/thrust/detail/type_traits/minimum_type.h
@@ -18,12 +18,18 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/type_traits.h>
 
 THRUST_NAMESPACE_BEGIN
 
 namespace detail
-{ 
+{
 
 namespace minimum_type_detail
 {
@@ -33,7 +39,7 @@ namespace minimum_type_detail
 // if T1 and T2 are unrelated.
 //
 template <typename T1, typename T2, bool GreaterEqual, bool LessEqual> struct minimum_type_impl {};
-  
+
 template <typename T1, typename T2>
 struct minimum_type_impl<T1,T2,true,false>
 {
diff --git a/thrust/thrust/detail/type_traits/pointer_traits.h b/thrust/thrust/detail/type_traits/pointer_traits.h
index 90a8bc29dec..e35047ba2b0 100644
--- a/thrust/thrust/detail/type_traits/pointer_traits.h
+++ b/thrust/thrust/detail/type_traits/pointer_traits.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/type_traits.h>
 #include <thrust/detail/type_traits/is_metafunction_defined.h>
 #include <thrust/detail/type_traits/has_nested_type.h>
diff --git a/thrust/thrust/detail/type_traits/result_of_adaptable_function.h b/thrust/thrust/detail/type_traits/result_of_adaptable_function.h
index edf797f141b..4baf724035e 100644
--- a/thrust/thrust/detail/type_traits/result_of_adaptable_function.h
+++ b/thrust/thrust/detail/type_traits/result_of_adaptable_function.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/type_traits.h>
 #include <thrust/detail/type_traits/function_traits.h>
 
diff --git a/thrust/thrust/detail/uninitialized_copy.inl b/thrust/thrust/detail/uninitialized_copy.inl
index 2778693ad3c..bf8f6c5cfb0 100644
--- a/thrust/thrust/detail/uninitialized_copy.inl
+++ b/thrust/thrust/detail/uninitialized_copy.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/uninitialized_copy.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
diff --git a/thrust/thrust/detail/uninitialized_fill.inl b/thrust/thrust/detail/uninitialized_fill.inl
index e013dac7b07..1dc65edb916 100644
--- a/thrust/thrust/detail/uninitialized_fill.inl
+++ b/thrust/thrust/detail/uninitialized_fill.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/uninitialized_fill.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
diff --git a/thrust/thrust/detail/unique.inl b/thrust/thrust/detail/unique.inl
index ac5475f0238..1a5b827d16c 100644
--- a/thrust/thrust/detail/unique.inl
+++ b/thrust/thrust/detail/unique.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/unique.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/generic/select_system.h>
diff --git a/thrust/thrust/detail/use_default.h b/thrust/thrust/detail/use_default.h
index f25b6274c44..745547ca5f7 100644
--- a/thrust/thrust/detail/use_default.h
+++ b/thrust/thrust/detail/use_default.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 THRUST_NAMESPACE_BEGIN
 
 struct use_default {};
diff --git a/thrust/thrust/detail/util/align.h b/thrust/thrust/detail/util/align.h
index a3aa75bfe41..c00a000f598 100644
--- a/thrust/thrust/detail/util/align.h
+++ b/thrust/thrust/detail/util/align.h
@@ -19,6 +19,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/cstdint.h>
 
 // functions to handle memory alignment
diff --git a/thrust/thrust/detail/vector_base.h b/thrust/thrust/detail/vector_base.h
index 0c4da449eab..0657cbe25b2 100644
--- a/thrust/thrust/detail/vector_base.h
+++ b/thrust/thrust/detail/vector_base.h
@@ -22,11 +22,18 @@
 
 #pragma once
 
+#include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/iterator/detail/normal_iterator.h>
 #include <thrust/iterator/reverse_iterator.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/detail/type_traits.h>
-#include <thrust/detail/config.h>
 #include <thrust/detail/contiguous_storage.h>
 
 #include <initializer_list>
@@ -134,13 +141,13 @@ template<typename T, typename Alloc>
      *  \param il The intializer_list.
      */
     vector_base(std::initializer_list<T> il);
-      
+
     /*! This constructor builds a \p vector_base from an intializer_list.
      *  \param il The intializer_list.
      *  \param alloc The allocator to use by this device_vector.
      */
     vector_base(std::initializer_list<T> il, const Alloc &alloc);
-    
+
     /*! Assign operator copies from an initializer_list
      *  \param il The initializer_list.
      */
@@ -445,7 +452,7 @@ template<typename T, typename Alloc>
      *  \param x The exemplar element to copy & insert.
      *  \return An iterator pointing to the newly inserted element.
      */
-    iterator insert(iterator position, const T &x); 
+    iterator insert(iterator position, const T &x);
 
     /*! This method inserts a copy of an exemplar value to a range at the
      *  specified position in this vector_base.
@@ -498,10 +505,10 @@ template<typename T, typename Alloc>
   private:
     // these methods resolve the ambiguity of the constructor template of form (Iterator, Iterator)
     template<typename IteratorOrIntegralType>
-      void init_dispatch(IteratorOrIntegralType begin, IteratorOrIntegralType end, false_type); 
+      void init_dispatch(IteratorOrIntegralType begin, IteratorOrIntegralType end, false_type);
 
     template<typename IteratorOrIntegralType>
-      void init_dispatch(IteratorOrIntegralType n, IteratorOrIntegralType value, true_type); 
+      void init_dispatch(IteratorOrIntegralType n, IteratorOrIntegralType value, true_type);
 
     template<typename InputIterator>
       void range_init(InputIterator first, InputIterator last);
@@ -589,7 +596,7 @@ template<typename T1, typename Alloc1,
          typename T2, typename Alloc2>
 bool operator==(const detail::vector_base<T1,Alloc1>& lhs,
                 const detail::vector_base<T2,Alloc2>& rhs);
-    
+
 template<typename T1, typename Alloc1,
          typename T2, typename Alloc2>
 bool operator==(const detail::vector_base<T1,Alloc1>& lhs,
@@ -610,7 +617,7 @@ template<typename T1, typename Alloc1,
          typename T2, typename Alloc2>
 bool operator!=(const detail::vector_base<T1,Alloc1>& lhs,
                 const detail::vector_base<T2,Alloc2>& rhs);
-    
+
 template<typename T1, typename Alloc1,
          typename T2, typename Alloc2>
 bool operator!=(const detail::vector_base<T1,Alloc1>& lhs,
diff --git a/thrust/thrust/detail/vector_base.inl b/thrust/thrust/detail/vector_base.inl
index bdd6c1c7ac1..7d6bf5dd2a9 100644
--- a/thrust/thrust/detail/vector_base.inl
+++ b/thrust/thrust/detail/vector_base.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/vector_base.h>
 #include <thrust/detail/copy.h>
 #include <thrust/detail/overlapped_copy.h>
diff --git a/thrust/thrust/device_allocator.h b/thrust/thrust/device_allocator.h
index f64c3854f0e..a0b240a66cc 100644
--- a/thrust/thrust/device_allocator.h
+++ b/thrust/thrust/device_allocator.h
@@ -23,6 +23,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/device_ptr.h>
 #include <thrust/mr/allocator.h>
 #include <thrust/mr/device_memory_resource.h>
diff --git a/thrust/thrust/device_delete.h b/thrust/thrust/device_delete.h
index 0811936fbbe..fb3af0316fe 100644
--- a/thrust/thrust/device_delete.h
+++ b/thrust/thrust/device_delete.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/device_ptr.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/device_free.h b/thrust/thrust/device_free.h
index 1cd305045e4..410201cbbdd 100644
--- a/thrust/thrust/device_free.h
+++ b/thrust/thrust/device_free.h
@@ -14,13 +14,19 @@
  *  limitations under the License.
  */
 
-/*! \file 
+/*! \file
  *  \brief Deallocates storage allocated by \p device_malloc.
  */
 
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/device_ptr.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/device_make_unique.h b/thrust/thrust/device_make_unique.h
index cdb8c31d82f..6f63f4cfb70 100644
--- a/thrust/thrust/device_make_unique.h
+++ b/thrust/thrust/device_make_unique.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp11_required.h>
 
 #if THRUST_CPP_DIALECT >= 2011
diff --git a/thrust/thrust/device_malloc.h b/thrust/thrust/device_malloc.h
index 790ddbac3a7..ed8276dee6f 100644
--- a/thrust/thrust/device_malloc.h
+++ b/thrust/thrust/device_malloc.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/device_ptr.h>
 #include <cstddef> // for std::size_t
 
diff --git a/thrust/thrust/device_malloc_allocator.h b/thrust/thrust/device_malloc_allocator.h
index 1b15045f2f8..e34f3be3b14 100644
--- a/thrust/thrust/device_malloc_allocator.h
+++ b/thrust/thrust/device_malloc_allocator.h
@@ -14,13 +14,19 @@
  *  limitations under the License.
  */
 
-/*! \file 
+/*! \file
  *  \brief An allocator which allocates storage with \p device_malloc.
  */
 
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/device_ptr.h>
 #include <thrust/device_reference.h>
 #include <thrust/device_malloc.h>
@@ -34,7 +40,7 @@ THRUST_NAMESPACE_BEGIN
 template<typename> class device_ptr;
 template<typename T> device_ptr<T> device_malloc(const std::size_t n);
 
-/*! \addtogroup allocators Allocators 
+/*! \addtogroup allocators Allocators
  *  \ingroup memory_management
  *  \{
  */
diff --git a/thrust/thrust/device_new.h b/thrust/thrust/device_new.h
index c615e58f2d6..bfd252f0b3c 100644
--- a/thrust/thrust/device_new.h
+++ b/thrust/thrust/device_new.h
@@ -23,6 +23,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // #include this for size_t
 #include <cstddef>
 #include <thrust/device_ptr.h>
diff --git a/thrust/thrust/device_new_allocator.h b/thrust/thrust/device_new_allocator.h
index c9c6b0e9507..a34888140ae 100644
--- a/thrust/thrust/device_new_allocator.h
+++ b/thrust/thrust/device_new_allocator.h
@@ -14,13 +14,19 @@
  *  limitations under the License.
  */
 
-/*! \file 
+/*! \file
  *  \brief An allocator which allocates storage with \p device_new.
  */
 
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/device_ptr.h>
 #include <thrust/device_reference.h>
 #include <thrust/device_new.h>
@@ -105,7 +111,7 @@ template<typename T>
      */
     __host__ __device__
     inline pointer address(reference r) { return &r; }
-    
+
     /*! Returns the address an allocated object.
      *  \return <tt>&r</tt>.
      */
diff --git a/thrust/thrust/device_ptr.h b/thrust/thrust/device_ptr.h
index 5ef4aa464fc..8d3890c0cc8 100644
--- a/thrust/thrust/device_ptr.h
+++ b/thrust/thrust/device_ptr.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/memory.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/device_reference.h b/thrust/thrust/device_reference.h
index 512ab4c60c3..50601ec5179 100644
--- a/thrust/thrust/device_reference.h
+++ b/thrust/thrust/device_reference.h
@@ -14,7 +14,7 @@
  *  limitations under the License.
  */
 
-/*! \file 
+/*! \file
  *  \brief A reference to an object which resides in memory associated with the
  *  device system.
  */
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/device_ptr.h>
 #include <thrust/detail/type_traits.h>
 #include <thrust/detail/reference.h>
diff --git a/thrust/thrust/device_vector.h b/thrust/thrust/device_vector.h
index 9b97e8d7033..47d67b412d7 100644
--- a/thrust/thrust/device_vector.h
+++ b/thrust/thrust/device_vector.h
@@ -23,6 +23,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/vector_base.h>
 #include <thrust/device_allocator.h>
 
@@ -203,14 +209,14 @@ template<typename T, typename Alloc = thrust::device_allocator<T> >
      */
     device_vector(std::initializer_list<T> il)
       :Parent(il) {}
-      
+
     /*! This constructor builds a \p device_vector from an intializer_list.
      *  \param il The intializer_list.
      *  \param alloc The allocator to use by this device_vector.
      */
     device_vector(std::initializer_list<T> il, const Alloc &alloc)
       :Parent(il, alloc) {}
-      
+
     /*! Assign an \p intializer_list with a matching element type
      *  \param il The intializer_list.
      */
diff --git a/thrust/thrust/distance.h b/thrust/thrust/distance.h
index 89087911512..82f045a06bb 100644
--- a/thrust/thrust/distance.h
+++ b/thrust/thrust/distance.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/iterator/iterator_traits.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/equal.h b/thrust/thrust/equal.h
index 2f351890789..d85a36cc6b1 100644
--- a/thrust/thrust/equal.h
+++ b/thrust/thrust/equal.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
@@ -150,7 +156,7 @@ bool equal(InputIterator1 first1, InputIterator1 last1,
  *  #include <thrust/equal.h>
  *  #include <thrust/execution_policy.h>
  *  ...
- *  
+ *
  *  struct compare_modulo_two
  *  {
  *    __host__ __device__
@@ -200,7 +206,7 @@ bool equal(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, Inp
  *
  *  \code
  *  #include <thrust/equal.h>
- *  
+ *
  *  struct compare_modulo_two
  *  {
  *    __host__ __device__
@@ -220,7 +226,7 @@ bool equal(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, Inp
  *
  *  \see https://en.cppreference.com/w/cpp/algorithm/equal
  */
-template <typename InputIterator1, typename InputIterator2, 
+template <typename InputIterator1, typename InputIterator2,
           typename BinaryPredicate>
 bool equal(InputIterator1 first1, InputIterator1 last1,
            InputIterator2 first2, BinaryPredicate binary_pred);
diff --git a/thrust/thrust/event.h b/thrust/thrust/event.h
index 75578d96443..d7b6a5d393b 100644
--- a/thrust/thrust/event.h
+++ b/thrust/thrust/event.h
@@ -20,6 +20,14 @@
 
 #pragma once
 
+#include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/future.h>
 
 // TODO: Actually separate `<thrust/future.h>` into two headers.
diff --git a/thrust/thrust/execution_policy.h b/thrust/thrust/execution_policy.h
index ecf14413f60..fe0ca84a728 100644
--- a/thrust/thrust/execution_policy.h
+++ b/thrust/thrust/execution_policy.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 #include <thrust/detail/execute_with_allocator.h>
 #include <thrust/detail/seq.h>
diff --git a/thrust/thrust/extrema.h b/thrust/thrust/extrema.h
index ca419a0aa5b..f680eb35ebf 100644
--- a/thrust/thrust/extrema.h
+++ b/thrust/thrust/extrema.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 #include <thrust/pair.h>
 
@@ -220,7 +226,7 @@ __host__ __device__
  *  // *result is 0
  *  \endcode
  *
- *  \see https://en.cppreference.com/w/cpp/algorithm/min_element 
+ *  \see https://en.cppreference.com/w/cpp/algorithm/min_element
  */
 template<typename DerivedPolicy, typename ForwardIterator>
 __host__ __device__
@@ -258,7 +264,7 @@ ForwardIterator min_element(const thrust::detail::execution_policy_base<DerivedP
  *  // *result is 0
  *  \endcode
  *
- *  \see https://en.cppreference.com/w/cpp/algorithm/min_element 
+ *  \see https://en.cppreference.com/w/cpp/algorithm/min_element
  */
 template <typename ForwardIterator>
 ForwardIterator min_element(ForwardIterator first, ForwardIterator last);
@@ -323,7 +329,7 @@ ForwardIterator min_element(ForwardIterator first, ForwardIterator last);
  *  // *smallest == {0,7}
  *  \endcode
  *
- *  \see https://en.cppreference.com/w/cpp/algorithm/min_element 
+ *  \see https://en.cppreference.com/w/cpp/algorithm/min_element
  */
 template<typename DerivedPolicy, typename ForwardIterator, typename BinaryPredicate>
 __host__ __device__
@@ -383,7 +389,7 @@ ForwardIterator min_element(const thrust::detail::execution_policy_base<DerivedP
  *  // *smallest == {0,7}
  *  \endcode
  *
- *  \see https://en.cppreference.com/w/cpp/algorithm/min_element 
+ *  \see https://en.cppreference.com/w/cpp/algorithm/min_element
  */
 template <typename ForwardIterator, typename BinaryPredicate>
 ForwardIterator min_element(ForwardIterator first, ForwardIterator last,
@@ -425,7 +431,7 @@ ForwardIterator min_element(ForwardIterator first, ForwardIterator last,
  *  // *result == 3
  *  \endcode
  *
- *  \see https://en.cppreference.com/w/cpp/algorithm/max_element 
+ *  \see https://en.cppreference.com/w/cpp/algorithm/max_element
  */
 template<typename DerivedPolicy, typename ForwardIterator>
 __host__ __device__
@@ -462,7 +468,7 @@ ForwardIterator max_element(const thrust::detail::execution_policy_base<DerivedP
  *  // *result == 3
  *  \endcode
  *
- *  \see https://en.cppreference.com/w/cpp/algorithm/max_element 
+ *  \see https://en.cppreference.com/w/cpp/algorithm/max_element
  */
 template <typename ForwardIterator>
 ForwardIterator max_element(ForwardIterator first, ForwardIterator last);
@@ -527,7 +533,7 @@ ForwardIterator max_element(ForwardIterator first, ForwardIterator last);
  *  // *largest == {6,1}
  *  \endcode
  *
- *  \see https://en.cppreference.com/w/cpp/algorithm/max_element 
+ *  \see https://en.cppreference.com/w/cpp/algorithm/max_element
  */
 template<typename DerivedPolicy, typename ForwardIterator, typename BinaryPredicate>
 __host__ __device__
@@ -587,7 +593,7 @@ ForwardIterator max_element(const thrust::detail::execution_policy_base<DerivedP
  *  // *largest == {6,1}
  *  \endcode
  *
- *  \see https://en.cppreference.com/w/cpp/algorithm/max_element 
+ *  \see https://en.cppreference.com/w/cpp/algorithm/max_element
  */
 template <typename ForwardIterator, typename BinaryPredicate>
 ForwardIterator max_element(ForwardIterator first, ForwardIterator last,
@@ -665,7 +671,7 @@ thrust::pair<ForwardIterator,ForwardIterator> minmax_element(const thrust::detai
  *  \see http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2005/n1840.pdf
  */
 template <typename ForwardIterator>
-thrust::pair<ForwardIterator,ForwardIterator> minmax_element(ForwardIterator first, 
+thrust::pair<ForwardIterator,ForwardIterator> minmax_element(ForwardIterator first,
                                                              ForwardIterator last);
 
 
@@ -787,7 +793,7 @@ thrust::pair<ForwardIterator,ForwardIterator> minmax_element(const thrust::detai
  *  \see http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2005/n1840.pdf
  */
 template <typename ForwardIterator, typename BinaryPredicate>
-thrust::pair<ForwardIterator,ForwardIterator> minmax_element(ForwardIterator first, 
+thrust::pair<ForwardIterator,ForwardIterator> minmax_element(ForwardIterator first,
                                                              ForwardIterator last,
                                                              BinaryPredicate comp);
 
diff --git a/thrust/thrust/fill.h b/thrust/thrust/fill.h
index bd9e4026897..33ff6d614c3 100644
--- a/thrust/thrust/fill.h
+++ b/thrust/thrust/fill.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/find.h b/thrust/thrust/find.h
index 5ab9b0a2df8..4bfa9952203 100644
--- a/thrust/thrust/find.h
+++ b/thrust/thrust/find.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
@@ -35,7 +41,7 @@ THRUST_NAMESPACE_BEGIN
  */
 
 
-/*! \p find returns the first iterator \c i in the range 
+/*! \p find returns the first iterator \c i in the range
  *  <tt>[first, last)</tt> such that <tt>*i == value</tt>
  *  or \c last if no such iterator exists.
  *
@@ -50,7 +56,7 @@ THRUST_NAMESPACE_BEGIN
  *  \tparam DerivedPolicy The name of the derived execution policy.
  *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
  *          and \p InputIterator's \c value_type is equality comparable to type \c T.
- *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">EqualityComparable</a>. 
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">EqualityComparable</a>.
  *
  *  \code
  *  #include <thrust/find.h>
@@ -82,7 +88,7 @@ InputIterator find(const thrust::detail::execution_policy_base<DerivedPolicy> &e
                    const T& value);
 
 
-/*! \p find returns the first iterator \c i in the range 
+/*! \p find returns the first iterator \c i in the range
  *  <tt>[first, last)</tt> such that <tt>*i == value</tt>
  *  or \c last if no such iterator exists.
  *
@@ -93,7 +99,7 @@ InputIterator find(const thrust::detail::execution_policy_base<DerivedPolicy> &e
  *
  *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
  *          and \p InputIterator's \c value_type is equality comparable to type \c T.
- *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">EqualityComparable</a>. 
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">EqualityComparable</a>.
  *
  *  \code
  *  #include <thrust/find.h>
@@ -122,7 +128,7 @@ InputIterator find(InputIterator first,
                    const T& value);
 
 
-/*! \p find_if returns the first iterator \c i in the range 
+/*! \p find_if returns the first iterator \c i in the range
  *  <tt>[first, last)</tt> such that <tt>pred(*i)</tt> is \c true
  *  or \c last if no such iterator exists.
  *
@@ -189,7 +195,7 @@ InputIterator find_if(const thrust::detail::execution_policy_base<DerivedPolicy>
                       Predicate pred);
 
 
-/*! \p find_if returns the first iterator \c i in the range 
+/*! \p find_if returns the first iterator \c i in the range
  *  <tt>[first, last)</tt> such that <tt>pred(*i)</tt> is \c true
  *  or \c last if no such iterator exists.
  *
@@ -248,7 +254,7 @@ InputIterator find_if(InputIterator first,
                       Predicate pred);
 
 
-/*! \p find_if_not returns the first iterator \c i in the range 
+/*! \p find_if_not returns the first iterator \c i in the range
  *  <tt>[first, last)</tt> such that <tt>pred(*i)</tt> is \c false
  *  or \c last if no such iterator exists.
  *
@@ -315,7 +321,7 @@ InputIterator find_if_not(const thrust::detail::execution_policy_base<DerivedPol
                           Predicate pred);
 
 
-/*! \p find_if_not returns the first iterator \c i in the range 
+/*! \p find_if_not returns the first iterator \c i in the range
  *  <tt>[first, last)</tt> such that <tt>pred(*i)</tt> is \c false
  *  or \c last if no such iterator exists.
  *
diff --git a/thrust/thrust/for_each.h b/thrust/thrust/for_each.h
index 7d05e3ea16b..1f11ea1a696 100644
--- a/thrust/thrust/for_each.h
+++ b/thrust/thrust/for_each.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/type_traits.h>
 #include <thrust/detail/execution_policy.h>
 
diff --git a/thrust/thrust/functional.h b/thrust/thrust/functional.h
index 0608f4b3d45..009b5b96f44 100644
--- a/thrust/thrust/functional.h
+++ b/thrust/thrust/functional.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <functional>
 #include <thrust/detail/functional/placeholder.h>
 
diff --git a/thrust/thrust/future.h b/thrust/thrust/future.h
index d8fb7544b17..5a896c23705 100644
--- a/thrust/thrust/future.h
+++ b/thrust/thrust/future.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp14_required.h>
 
 #if THRUST_CPP_DIALECT >= 2014
@@ -70,14 +76,14 @@ namespace unimplemented
 
 struct no_unique_eager_event_type_found {};
 
-inline __host__ 
+inline __host__
 no_unique_eager_event_type_found
 unique_eager_event_type(...) noexcept;
 
 struct no_unique_eager_future_type_found {};
 
 template <typename T>
-__host__ 
+__host__
 no_unique_eager_future_type_found
 unique_eager_future_type(...) noexcept;
 
diff --git a/thrust/thrust/gather.h b/thrust/thrust/gather.h
index 41acc22a34e..b0d4553f812 100644
--- a/thrust/thrust/gather.h
+++ b/thrust/thrust/gather.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
@@ -32,7 +38,7 @@ THRUST_NAMESPACE_BEGIN
  */
 
 
-/*! \p gather copies elements from a source array into a destination range according 
+/*! \p gather copies elements from a source array into a destination range according
  *  to a map. For each input iterator \c i in the range <tt>[map_first, map_last)</tt>, the
  *  value <tt>input_first[\*i]</tt> is assigned to <tt>*(result + (i - map_first))</tt>.
  *  \p RandomAccessIterator must permit random access.
@@ -92,7 +98,7 @@ __host__ __device__
                         OutputIterator                                              result);
 
 
-/*! \p gather copies elements from a source array into a destination range according 
+/*! \p gather copies elements from a source array into a destination range according
  *  to a map. For each input iterator \c i in the range <tt>[map_first, map_last)</tt>, the
  *  value <tt>input_first[\*i]</tt> is assigned to <tt>*(result + (i - map_first))</tt>.
  *  \p RandomAccessIterator must permit random access.
@@ -143,7 +149,7 @@ template<typename InputIterator,
                         OutputIterator       result);
 
 
-/*! \p gather_if conditionally copies elements from a source array into a destination 
+/*! \p gather_if conditionally copies elements from a source array into a destination
  *  range according to a map. For each input iterator \c i in the range <tt>[map_first, map_last)</tt>,
  *  such that the value of <tt>\*(stencil + (i - map_first))</tt> is \c true, the value
  *  <tt>input_first[\*i]</tt> is assigned to <tt>*(result + (i - map_first))</tt>.
@@ -214,7 +220,7 @@ __host__ __device__
                            OutputIterator                                              result);
 
 
-/*! \p gather_if conditionally copies elements from a source array into a destination 
+/*! \p gather_if conditionally copies elements from a source array into a destination
  *  range according to a map. For each input iterator \c i in the range <tt>[map_first, map_last)</tt>,
  *  such that the value of <tt>\*(stencil + (i - map_first))</tt> is \c true, the value
  *  <tt>input_first[\*i]</tt> is assigned to <tt>*(result + (i - map_first))</tt>.
@@ -276,7 +282,7 @@ template<typename InputIterator1,
                            OutputIterator       result);
 
 
-/*! \p gather_if conditionally copies elements from a source array into a destination 
+/*! \p gather_if conditionally copies elements from a source array into a destination
  *  range according to a map. For each input iterator \c i in the range <tt>[map_first, map_last)</tt>
  *  such that the value of <tt>pred(\*(stencil + (i - map_first)))</tt> is \c true,
  *  the value <tt>input_first[\*i]</tt> is assigned to <tt>*(result + (i - map_first))</tt>.
@@ -312,7 +318,7 @@ template<typename InputIterator1,
  *  #include <thrust/gather.h>
  *  #include <thrust/device_vector.h>
  *  #include <thrust/execution_policy.h>
- *  
+ *
  *  struct is_even
  *  {
  *    __host__ __device__
@@ -361,7 +367,7 @@ __host__ __device__
                            Predicate                                                   pred);
 
 
-/*! \p gather_if conditionally copies elements from a source array into a destination 
+/*! \p gather_if conditionally copies elements from a source array into a destination
  *  range according to a map. For each input iterator \c i in the range <tt>[map_first, map_last)</tt>
  *  such that the value of <tt>pred(\*(stencil + (i - map_first)))</tt> is \c true,
  *  the value <tt>input_first[\*i]</tt> is assigned to <tt>*(result + (i - map_first))</tt>.
@@ -392,7 +398,7 @@ __host__ __device__
  *  \code
  *  #include <thrust/gather.h>
  *  #include <thrust/device_vector.h>
- *  
+ *
  *  struct is_even
  *  {
  *    __host__ __device__
diff --git a/thrust/thrust/generate.h b/thrust/thrust/generate.h
index d4729534460..671383018b8 100644
--- a/thrust/thrust/generate.h
+++ b/thrust/thrust/generate.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/host_vector.h b/thrust/thrust/host_vector.h
index bb925ea9ce8..2dd00cd59c7 100644
--- a/thrust/thrust/host_vector.h
+++ b/thrust/thrust/host_vector.h
@@ -23,6 +23,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/memory_wrapper.h>
 #include <thrust/detail/vector_base.h>
 
@@ -218,20 +224,20 @@ template<typename T, typename Alloc = std::allocator<T> >
     __host__
     host_vector &operator=(const detail::vector_base<OtherT,OtherAlloc> &v)
     { Parent::operator=(v); return *this; }
-    
+
     /*! This constructor builds a \p host_vector from an intializer_list.
      *  \param il The intializer_list.
      */
     host_vector(std::initializer_list<T> il)
       :Parent(il) {}
-      
+
     /*! This constructor builds a \p host_vector from an intializer_list.
      *  \param il The intializer_list.
      *  \param alloc The allocator to use by this host_vector.
      */
     host_vector(std::initializer_list<T> il, const Alloc &alloc)
       :Parent(il, alloc) {}
-      
+
     /*! Assign an \p intializer_list with a matching element type
      *  \param il The intializer_list.
      */
diff --git a/thrust/thrust/inner_product.h b/thrust/thrust/inner_product.h
index 80068cf0c83..a4cda16172d 100644
--- a/thrust/thrust/inner_product.h
+++ b/thrust/thrust/inner_product.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
@@ -162,7 +168,7 @@ OutputType inner_product(InputIterator1 first1, InputIterator1 last1,
  *          and \p BinaryFunction1's \c return_type is convertible to \p OutputType.
  *  \tparam BinaryFunction2 is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/binary_function">Binary Function</a>,
  *          and \p BinaryFunction2's \c return_type is convertible to \p BinaryFunction1's \c second_argument_type.
- * 
+ *
  *  \code
  *  #include <thrust/inner_product.h>
  *  #include <thrust/execution_policy.h>
@@ -192,7 +198,7 @@ OutputType inner_product(const thrust::detail::execution_policy_base<DerivedPoli
                          InputIterator1 first1,
                          InputIterator1 last1,
                          InputIterator2 first2,
-                         OutputType init, 
+                         OutputType init,
                          BinaryFunction1 binary_op1,
                          BinaryFunction2 binary_op2);
 
@@ -227,7 +233,7 @@ OutputType inner_product(const thrust::detail::execution_policy_base<DerivedPoli
  *          and \p BinaryFunction1's \c return_type is convertible to \p OutputType.
  *  \tparam BinaryFunction2 is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/binary_function">Binary Function</a>,
  *          and \p BinaryFunction2's \c return_type is convertible to \p BinaryFunction1's \c second_argument_type.
- * 
+ *
  *  \code
  *  #include <thrust/inner_product.h>
  *  ...
@@ -248,7 +254,7 @@ OutputType inner_product(const thrust::detail::execution_policy_base<DerivedPoli
 template<typename InputIterator1, typename InputIterator2, typename OutputType,
          typename BinaryFunction1, typename BinaryFunction2>
 OutputType inner_product(InputIterator1 first1, InputIterator1 last1,
-                         InputIterator2 first2, OutputType init, 
+                         InputIterator2 first2, OutputType init,
                          BinaryFunction1 binary_op1, BinaryFunction2 binary_op2);
 
 
diff --git a/thrust/thrust/iterator/constant_iterator.h b/thrust/thrust/iterator/constant_iterator.h
index c6eec28e70e..dc773489299 100644
--- a/thrust/thrust/iterator/constant_iterator.h
+++ b/thrust/thrust/iterator/constant_iterator.h
@@ -23,6 +23,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/iterator/detail/constant_iterator_base.h>
 #include <thrust/iterator/iterator_facade.h>
 
@@ -77,15 +83,15 @@ THRUST_NAMESPACE_BEGIN
  *    data[1] = 7;
  *    data[2] = 2;
  *    data[3] = 5;
- *    
+ *
  *    // add 10 to all values in data
  *    thrust::transform(data.begin(), data.end(),
  *                      thrust::make_constant_iterator(10),
  *                      data.begin(),
  *                      thrust::plus<int>());
- *    
+ *
  *    // data is now [13, 17, 12, 15]
- *    
+ *
  *    return 0;
  *  }
  *  \endcode
@@ -145,7 +151,7 @@ template<typename Value,
     /*! This constructor receives a value to use as the constant value of this
      *  \p constant_iterator and an index specifying the location of this
      *  \p constant_iterator in a sequence.
-     *  
+     *
      *  \p v The value of this \p constant_iterator's constant value.
      *  \p i The index of this \p constant_iterator in a sequence. Defaults to the
      *       value returned by \c Incrementable's null constructor. For example,
@@ -186,7 +192,7 @@ template<typename Value,
     __host__ __device__
     Value & value_reference()
     { return m_value; }
-  
+
   private: // Core iterator interface
     __host__ __device__
     reference dereference() const
diff --git a/thrust/thrust/iterator/counting_iterator.h b/thrust/thrust/iterator/counting_iterator.h
index f66cb97ef82..70cf6d12732 100644
--- a/thrust/thrust/iterator/counting_iterator.h
+++ b/thrust/thrust/iterator/counting_iterator.h
@@ -32,6 +32,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/iterator/iterator_adaptor.h>
 #include <thrust/iterator/iterator_facade.h>
 #include <thrust/iterator/iterator_categories.h>
diff --git a/thrust/thrust/iterator/detail/any_assign.h b/thrust/thrust/iterator/detail/any_assign.h
index 87192215c3a..f07ec00d86c 100644
--- a/thrust/thrust/iterator/detail/any_assign.h
+++ b/thrust/thrust/iterator/detail/any_assign.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 THRUST_NAMESPACE_BEGIN
 namespace detail
 {
diff --git a/thrust/thrust/iterator/detail/any_system_tag.h b/thrust/thrust/iterator/detail/any_system_tag.h
index 2c5ce6448be..e53323f807f 100644
--- a/thrust/thrust/iterator/detail/any_system_tag.h
+++ b/thrust/thrust/iterator/detail/any_system_tag.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/iterator/detail/constant_iterator_base.h b/thrust/thrust/iterator/detail/constant_iterator_base.h
index 56bb7a5d01f..22065dc1489 100644
--- a/thrust/thrust/iterator/detail/constant_iterator_base.h
+++ b/thrust/thrust/iterator/detail/constant_iterator_base.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/iterator/counting_iterator.h>
 #include <thrust/iterator/iterator_adaptor.h>
 
@@ -66,6 +72,6 @@ template<typename Value,
 }; // end constant_iterator_base
 
 } // end detail
-  
+
 THRUST_NAMESPACE_END
 
diff --git a/thrust/thrust/iterator/detail/counting_iterator.inl b/thrust/thrust/iterator/detail/counting_iterator.inl
index ee4a9df1556..0924ae8a9dc 100644
--- a/thrust/thrust/iterator/detail/counting_iterator.inl
+++ b/thrust/thrust/iterator/detail/counting_iterator.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/iterator/counting_iterator.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/detail/numeric_traits.h>
diff --git a/thrust/thrust/iterator/detail/device_system_tag.h b/thrust/thrust/iterator/detail/device_system_tag.h
index b86109d216b..f9518536927 100644
--- a/thrust/thrust/iterator/detail/device_system_tag.h
+++ b/thrust/thrust/iterator/detail/device_system_tag.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // #include the device system's execution_policy header
 #define __THRUST_DEVICE_SYSTEM_TAG_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/execution_policy.h>
 #include __THRUST_DEVICE_SYSTEM_TAG_HEADER
diff --git a/thrust/thrust/iterator/detail/discard_iterator_base.h b/thrust/thrust/iterator/detail/discard_iterator_base.h
index 38f77b37812..aa213cd5169 100644
--- a/thrust/thrust/iterator/detail/discard_iterator_base.h
+++ b/thrust/thrust/iterator/detail/discard_iterator_base.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/iterator/counting_iterator.h>
 #include <thrust/iterator/iterator_adaptor.h>
 #include <thrust/iterator/detail/any_assign.h>
@@ -58,7 +64,7 @@ template<typename System>
 
 
 } // end detail
-  
+
 THRUST_NAMESPACE_END
 
 
diff --git a/thrust/thrust/iterator/detail/distance_from_result.h b/thrust/thrust/iterator/detail/distance_from_result.h
index fe140344d49..478cdf00b98 100644
--- a/thrust/thrust/iterator/detail/distance_from_result.h
+++ b/thrust/thrust/iterator/detail/distance_from_result.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/type_traits.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/iterator/detail/host_system_tag.h b/thrust/thrust/iterator/detail/host_system_tag.h
index 58478f8d926..28523ff4ef6 100644
--- a/thrust/thrust/iterator/detail/host_system_tag.h
+++ b/thrust/thrust/iterator/detail/host_system_tag.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // #include the host system's execution_policy header
 #define __THRUST_HOST_SYSTEM_TAG_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/execution_policy.h>
 #include __THRUST_HOST_SYSTEM_TAG_HEADER
diff --git a/thrust/thrust/iterator/detail/is_iterator_category.h b/thrust/thrust/iterator/detail/is_iterator_category.h
index e520452a3ca..d6ccad460e9 100644
--- a/thrust/thrust/iterator/detail/is_iterator_category.h
+++ b/thrust/thrust/iterator/detail/is_iterator_category.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/iterator/iterator_categories.h>
 #include <thrust/detail/type_traits.h>
 
diff --git a/thrust/thrust/iterator/detail/iterator_adaptor_base.h b/thrust/thrust/iterator/detail/iterator_adaptor_base.h
index 1173e414ca5..b7abacf35b7 100644
--- a/thrust/thrust/iterator/detail/iterator_adaptor_base.h
+++ b/thrust/thrust/iterator/detail/iterator_adaptor_base.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/type_traits.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/detail/use_default.h>
diff --git a/thrust/thrust/iterator/detail/iterator_category_to_system.h b/thrust/thrust/iterator/detail/iterator_category_to_system.h
index e6103b539ae..68bfcfb6fae 100644
--- a/thrust/thrust/iterator/detail/iterator_category_to_system.h
+++ b/thrust/thrust/iterator/detail/iterator_category_to_system.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/iterator/iterator_categories.h>
 #include <thrust/iterator/detail/iterator_traversal_tags.h>
 #include <thrust/iterator/detail/host_system_tag.h>
@@ -46,7 +52,7 @@ template<typename Category>
         >::value,
 
         detail::identity_<thrust::host_system_tag>,
-        
+
         // convertible to device iterator?
         eval_if<
           or_<
diff --git a/thrust/thrust/iterator/detail/iterator_category_to_traversal.h b/thrust/thrust/iterator/detail/iterator_category_to_traversal.h
index d8c736c50a4..39bd5d325ce 100644
--- a/thrust/thrust/iterator/detail/iterator_category_to_traversal.h
+++ b/thrust/thrust/iterator/detail/iterator_category_to_traversal.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/iterator/iterator_categories.h>
 #include <thrust/iterator/detail/iterator_traversal_tags.h>
 #include <thrust/iterator/detail/iterator_category_to_system.h>
diff --git a/thrust/thrust/iterator/detail/iterator_category_with_system_and_traversal.h b/thrust/thrust/iterator/detail/iterator_category_with_system_and_traversal.h
index cdd8a6d366a..a06f303f377 100644
--- a/thrust/thrust/iterator/detail/iterator_category_with_system_and_traversal.h
+++ b/thrust/thrust/iterator/detail/iterator_category_with_system_and_traversal.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 THRUST_NAMESPACE_BEGIN
 namespace detail
 {
diff --git a/thrust/thrust/iterator/detail/iterator_facade_category.h b/thrust/thrust/iterator/detail/iterator_facade_category.h
index 81b518002ba..5b3af7ecbb0 100644
--- a/thrust/thrust/iterator/detail/iterator_facade_category.h
+++ b/thrust/thrust/iterator/detail/iterator_facade_category.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/type_traits.h>
 #include <thrust/iterator/detail/host_system_tag.h>
 #include <thrust/iterator/detail/device_system_tag.h>
diff --git a/thrust/thrust/iterator/detail/iterator_traits.inl b/thrust/thrust/iterator/detail/iterator_traits.inl
index 544c24f0b42..cdc99473776 100644
--- a/thrust/thrust/iterator/detail/iterator_traits.inl
+++ b/thrust/thrust/iterator/detail/iterator_traits.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/iterator/iterator_categories.h>
 #include <thrust/iterator/detail/iterator_category_to_traversal.h>
 #include <thrust/detail/type_traits.h>
diff --git a/thrust/thrust/iterator/detail/iterator_traversal_tags.h b/thrust/thrust/iterator/detail/iterator_traversal_tags.h
index 1fbc8a1e49c..88efd921649 100644
--- a/thrust/thrust/iterator/detail/iterator_traversal_tags.h
+++ b/thrust/thrust/iterator/detail/iterator_traversal_tags.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 THRUST_NAMESPACE_BEGIN
 
 // define Boost's traversal tags
diff --git a/thrust/thrust/iterator/detail/join_iterator.h b/thrust/thrust/iterator/detail/join_iterator.h
index 83f143dc00a..d62099abb6b 100644
--- a/thrust/thrust/iterator/detail/join_iterator.h
+++ b/thrust/thrust/iterator/detail/join_iterator.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/iterator/iterator_facade.h>
 #include <thrust/iterator/detail/minimum_system.h>
 #include <thrust/iterator/counting_iterator.h>
diff --git a/thrust/thrust/iterator/detail/minimum_category.h b/thrust/thrust/iterator/detail/minimum_category.h
index 01e7e82c535..bbbcbe08454 100644
--- a/thrust/thrust/iterator/detail/minimum_category.h
+++ b/thrust/thrust/iterator/detail/minimum_category.h
@@ -18,12 +18,18 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/type_traits/minimum_type.h>
 
 THRUST_NAMESPACE_BEGIN
 
 namespace detail
-{ 
+{
 
 template<typename T1,
          typename T2  = minimum_type_detail::any_conversion,
diff --git a/thrust/thrust/iterator/detail/minimum_system.h b/thrust/thrust/iterator/detail/minimum_system.h
index dcb29ccd254..e94fbc5e3d5 100644
--- a/thrust/thrust/iterator/detail/minimum_system.h
+++ b/thrust/thrust/iterator/detail/minimum_system.h
@@ -17,13 +17,19 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/type_traits.h>
 #include <thrust/detail/type_traits/is_metafunction_defined.h>
 #include <thrust/detail/type_traits/minimum_type.h>
 
 THRUST_NAMESPACE_BEGIN
 namespace detail
-{ 
+{
 
 
 template<typename T1,
diff --git a/thrust/thrust/iterator/detail/normal_iterator.h b/thrust/thrust/iterator/detail/normal_iterator.h
index eb5d336044f..1d0511eb2bf 100644
--- a/thrust/thrust/iterator/detail/normal_iterator.h
+++ b/thrust/thrust/iterator/detail/normal_iterator.h
@@ -24,6 +24,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/iterator/iterator_adaptor.h>
 #include <thrust/detail/type_traits.h>
 #include <thrust/type_traits/is_contiguous_iterator.h>
@@ -49,7 +55,7 @@ template<typename Pointer>
     __host__ __device__
     normal_iterator(Pointer p)
       : super_t(p) {}
-    
+
     template<typename OtherPointer>
     __host__ __device__
     normal_iterator(const normal_iterator<OtherPointer> &other,
diff --git a/thrust/thrust/iterator/detail/permutation_iterator_base.h b/thrust/thrust/iterator/detail/permutation_iterator_base.h
index d586cabb780..33d271588e2 100644
--- a/thrust/thrust/iterator/detail/permutation_iterator_base.h
+++ b/thrust/thrust/iterator/detail/permutation_iterator_base.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/iterator/iterator_adaptor.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/detail/type_traits.h>
diff --git a/thrust/thrust/iterator/detail/retag.h b/thrust/thrust/iterator/detail/retag.h
index d277d8b6f2b..fc3e96b1785 100644
--- a/thrust/thrust/iterator/detail/retag.h
+++ b/thrust/thrust/iterator/detail/retag.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/type_traits.h>
 #include <thrust/iterator/detail/tagged_iterator.h>
 #include <thrust/detail/pointer.h>
diff --git a/thrust/thrust/iterator/detail/reverse_iterator.inl b/thrust/thrust/iterator/detail/reverse_iterator.inl
index 9182ac3e896..36b8705abba 100644
--- a/thrust/thrust/iterator/detail/reverse_iterator.inl
+++ b/thrust/thrust/iterator/detail/reverse_iterator.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/iterator/reverse_iterator.h>
 #include <thrust/iterator/iterator_traits.h>
 
diff --git a/thrust/thrust/iterator/detail/reverse_iterator_base.h b/thrust/thrust/iterator/detail/reverse_iterator_base.h
index de3bafde979..88fd0e4e0ba 100644
--- a/thrust/thrust/iterator/detail/reverse_iterator_base.h
+++ b/thrust/thrust/iterator/detail/reverse_iterator_base.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/iterator/iterator_adaptor.h>
 #include <thrust/iterator/iterator_traits.h>
 
diff --git a/thrust/thrust/iterator/detail/tagged_iterator.h b/thrust/thrust/iterator/detail/tagged_iterator.h
index 24cbbb7361d..f461ada8a80 100644
--- a/thrust/thrust/iterator/detail/tagged_iterator.h
+++ b/thrust/thrust/iterator/detail/tagged_iterator.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/iterator/iterator_adaptor.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/detail/use_default.h>
diff --git a/thrust/thrust/iterator/detail/transform_input_output_iterator.inl b/thrust/thrust/iterator/detail/transform_input_output_iterator.inl
index b4792f724df..1020c3b8086 100644
--- a/thrust/thrust/iterator/detail/transform_input_output_iterator.inl
+++ b/thrust/thrust/iterator/detail/transform_input_output_iterator.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/iterator/iterator_adaptor.h>
 #include <thrust/detail/type_traits.h>
 
diff --git a/thrust/thrust/iterator/detail/transform_iterator.inl b/thrust/thrust/iterator/detail/transform_iterator.inl
index 0dc6f9854de..4616e021e06 100644
--- a/thrust/thrust/iterator/detail/transform_iterator.inl
+++ b/thrust/thrust/iterator/detail/transform_iterator.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/type_traits.h>
 #include <thrust/detail/type_traits/result_of_adaptable_function.h>
 #include <thrust/iterator/iterator_adaptor.h>
diff --git a/thrust/thrust/iterator/detail/transform_output_iterator.inl b/thrust/thrust/iterator/detail/transform_output_iterator.inl
index d5033f10589..94aa6288f3a 100644
--- a/thrust/thrust/iterator/detail/transform_output_iterator.inl
+++ b/thrust/thrust/iterator/detail/transform_output_iterator.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/iterator/transform_output_iterator.h>
 #include <thrust/iterator/iterator_adaptor.h>
 
diff --git a/thrust/thrust/iterator/detail/tuple_of_iterator_references.h b/thrust/thrust/iterator/detail/tuple_of_iterator_references.h
index eb171338d27..c04c9705a5f 100644
--- a/thrust/thrust/iterator/detail/tuple_of_iterator_references.h
+++ b/thrust/thrust/iterator/detail/tuple_of_iterator_references.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cuda/std/type_traits>
 #include <cuda/std/tuple>
 
diff --git a/thrust/thrust/iterator/detail/universal_categories.h b/thrust/thrust/iterator/detail/universal_categories.h
index d2abd7f55bd..b5410b2c442 100644
--- a/thrust/thrust/iterator/detail/universal_categories.h
+++ b/thrust/thrust/iterator/detail/universal_categories.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/iterator/iterator_categories.h>
 
 // XXX eliminate this file
diff --git a/thrust/thrust/iterator/detail/zip_iterator.inl b/thrust/thrust/iterator/detail/zip_iterator.inl
index a2bc98afec2..9419737d4fd 100644
--- a/thrust/thrust/iterator/detail/zip_iterator.inl
+++ b/thrust/thrust/iterator/detail/zip_iterator.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/iterator/zip_iterator.h>
 #include <thrust/detail/tuple_transform.h>
 
diff --git a/thrust/thrust/iterator/detail/zip_iterator_base.h b/thrust/thrust/iterator/detail/zip_iterator_base.h
index 030153b65ea..cfecd742ac0 100644
--- a/thrust/thrust/iterator/detail/zip_iterator_base.h
+++ b/thrust/thrust/iterator/detail/zip_iterator_base.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/advance.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/iterator/iterator_facade.h>
@@ -84,7 +90,7 @@ struct dereference_iterator
 {
   template<typename Iterator>
   struct apply
-  { 
+  {
     typedef typename
       iterator_traits<Iterator>::reference
     type;
@@ -122,8 +128,8 @@ template<typename UnaryMetaFunctionClass, class Arg1, class Arg2>
 }; // end apply2
 
 
-// Meta-accumulate algorithm for tuples. Note: The template 
-// parameter StartType corresponds to the initial value in 
+// Meta-accumulate algorithm for tuples. Note: The template
+// parameter StartType corresponds to the initial value in
 // ordinary accumulation.
 //
 template<class Tuple, class BinaryMetaFun, class StartType>
@@ -153,7 +159,7 @@ template<
      , typename tuple_meta_accumulate<
            thrust::tuple<Ts...>
          , BinaryMetaFun
-         , StartType 
+         , StartType
        >::type
    >::type type;
 };
@@ -187,7 +193,7 @@ Fun tuple_for_each(thrust::tuple<Ts...>& t, Fun f, thrust::index_sequence<Is...>
 template<typename Fun, typename... Ts>
 inline __host__ __device__
 Fun tuple_for_each(thrust::tuple<Ts...>& t, Fun f)
-{ 
+{
   return tuple_for_each(t, f, thrust::make_index_sequence<thrust::tuple_size<thrust::tuple<Ts...>>::value>{});
 }
 
@@ -227,7 +233,7 @@ struct minimum_traversal_category_in_iterator_tuple
       IteratorTuple
     , thrust::iterator_traversal
   >::type tuple_of_traversal_tags;
-      
+
   typedef typename tuple_impl_specific::tuple_meta_accumulate<
       tuple_of_traversal_tags
     , minimum_category_lambda
@@ -302,7 +308,7 @@ template<typename IteratorTuple>
 //
 // Class zip_iterator_base
 //
-// Builds and exposes the iterator facade type from which the zip 
+// Builds and exposes the iterator facade type from which the zip
 // iterator will be derived.
 //
 template<typename IteratorTuple>
@@ -331,14 +337,14 @@ template<typename IteratorTuple>
     // iterator tuple
     typedef typename
     minimum_traversal_category_in_iterator_tuple<IteratorTuple>::type traversal_category;
-  
+
  public:
-  
+
     // The iterator facade type from which the zip iterator will
     // be derived.
     typedef thrust::iterator_facade<
         zip_iterator<IteratorTuple>,
-        value_type,  
+        value_type,
         system,
         traversal_category,
         reference,
diff --git a/thrust/thrust/iterator/discard_iterator.h b/thrust/thrust/iterator/discard_iterator.h
index eb5156edac4..7d3b5467e62 100644
--- a/thrust/thrust/iterator/discard_iterator.h
+++ b/thrust/thrust/iterator/discard_iterator.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/iterator/detail/discard_iterator_base.h>
 #include <thrust/iterator/iterator_facade.h>
 
diff --git a/thrust/thrust/iterator/iterator_adaptor.h b/thrust/thrust/iterator/iterator_adaptor.h
index 36ba0f63f7b..ff5af899984 100644
--- a/thrust/thrust/iterator/iterator_adaptor.h
+++ b/thrust/thrust/iterator/iterator_adaptor.h
@@ -33,6 +33,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/iterator/iterator_facade.h>
 #include <thrust/detail/use_default.h>
 #include <thrust/iterator/detail/iterator_adaptor_base.h>
diff --git a/thrust/thrust/iterator/iterator_categories.h b/thrust/thrust/iterator/iterator_categories.h
index 9a6f3f4ae52..4b02ec6be25 100644
--- a/thrust/thrust/iterator/iterator_categories.h
+++ b/thrust/thrust/iterator/iterator_categories.h
@@ -21,7 +21,7 @@
 
 /*
  * (C) Copyright Jeremy Siek 2002.
- * 
+ *
  * Distributed under the Boost Software License, Version 1.0.
  * (See accompanying NOTICE file for the complete license)
  *
@@ -32,6 +32,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/iterator/detail/iterator_category_with_system_and_traversal.h>
 #include <thrust/iterator/detail/iterator_traversal_tags.h>
 #include <thrust/iterator/detail/device_system_tag.h>
@@ -111,7 +117,7 @@ struct forward_device_iterator_tag
  *  representation of the Bidirectional Device Iterator concept within the C++
  *  type system.
  *
- *  \see https://en.cppreference.com/w/cpp/iterator/iterator_tags 
+ *  \see https://en.cppreference.com/w/cpp/iterator/iterator_tags
  *  iterator_traits, input_device_iterator_tag, output_device_iterator_tag,
  *  forward_device_iterator_tag, random_access_device_iterator_tag,
  *  input_host_iterator_tag, output_host_iterator_tag, forward_host_iterator_tag,
@@ -130,7 +136,7 @@ struct bidirectional_device_iterator_tag
  *  representation of the Random Access Device Iterator concept within the C++
  *  type system.
  *
- *  \see https://en.cppreference.com/w/cpp/iterator/iterator_tags 
+ *  \see https://en.cppreference.com/w/cpp/iterator/iterator_tags
  *  iterator_traits, input_device_iterator_tag, output_device_iterator_tag,
  *  forward_device_iterator_tag, bidirectional_device_iterator_tag,
  *  input_host_iterator_tag, output_host_iterator_tag, forward_host_iterator_tag,
@@ -149,7 +155,7 @@ struct random_access_device_iterator_tag
  *  representation of the Input Host Iterator concept within the C++
  *  type system.
  *
- *  \see https://en.cppreference.com/w/cpp/iterator/iterator_tags 
+ *  \see https://en.cppreference.com/w/cpp/iterator/iterator_tags
  *  iterator_traits, input_device_iterator_tag, output_device_iterator_tag,
  *  forward_device_iterator_tag, bidirectional_device_iterator_tag,
  *  random_access_device_iterator_tag,
@@ -163,7 +169,7 @@ typedef std::input_iterator_tag input_host_iterator_tag;
  *  representation of the Output Host Iterator concept within the C++
  *  type system.
  *
- *  \see https://en.cppreference.com/w/cpp/iterator/iterator_tags 
+ *  \see https://en.cppreference.com/w/cpp/iterator/iterator_tags
  *  iterator_traits, input_device_iterator_tag, output_device_iterator_tag,
  *  forward_device_iterator_tag, bidirectional_device_iterator_tag,
  *  random_access_device_iterator_tag,
@@ -177,7 +183,7 @@ typedef std::output_iterator_tag output_host_iterator_tag;
  *  representation of the Forward Host Iterator concept within the C++
  *  type system.
  *
- *  \see https://en.cppreference.com/w/cpp/iterator/iterator_tags 
+ *  \see https://en.cppreference.com/w/cpp/iterator/iterator_tags
  *  iterator_traits, input_device_iterator_tag, output_device_iterator_tag,
  *  forward_device_iterator_tag, bidirectional_device_iterator_tag,
  *  random_access_device_iterator_tag,
@@ -191,7 +197,7 @@ typedef std::forward_iterator_tag forward_host_iterator_tag;
  *  representation of the Forward Host Iterator concept within the C++
  *  type system.
  *
- *  \see https://en.cppreference.com/w/cpp/iterator/iterator_tags 
+ *  \see https://en.cppreference.com/w/cpp/iterator/iterator_tags
  *  iterator_traits, input_device_iterator_tag, output_device_iterator_tag,
  *  forward_device_iterator_tag, bidirectional_device_iterator_tag,
  *  random_access_device_iterator_tag,
@@ -205,7 +211,7 @@ typedef std::bidirectional_iterator_tag bidirectional_host_iterator_tag;
  *  representation of the Forward Host Iterator concept within the C++
  *  type system.
  *
- *  \see https://en.cppreference.com/w/cpp/iterator/iterator_tags 
+ *  \see https://en.cppreference.com/w/cpp/iterator/iterator_tags
  *  iterator_traits, input_device_iterator_tag, output_device_iterator_tag,
  *  forward_device_iterator_tag, bidirectional_device_iterator_tag,
  *  random_access_device_iterator_tag,
diff --git a/thrust/thrust/iterator/iterator_facade.h b/thrust/thrust/iterator/iterator_facade.h
index f6920c5c85c..e1d4e0889d8 100644
--- a/thrust/thrust/iterator/iterator_facade.h
+++ b/thrust/thrust/iterator/iterator_facade.h
@@ -22,7 +22,7 @@
  * (C) Copyright David Abrahams 2002.
  * (C) Copyright Jeremy Siek    2002.
  * (C) Copyright Thomas Witt    2002.
- * 
+ *
  * Distributed under the Boost Software License, Version 1.0.
  * (See accompanying NOTICE file for the complete license)
  *
@@ -33,6 +33,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/type_traits.h>
 #include <thrust/iterator/detail/iterator_facade_category.h>
 #include <thrust/iterator/detail/distance_from_result.h>
@@ -189,7 +195,7 @@ class iterator_core_access
     {
       return f2.distance_to(f1);
     }
-    
+
     template <class Facade1, class Facade2>
     __host__ __device__
     static typename thrust::detail::distance_from_result<Facade1,Facade2>::type
diff --git a/thrust/thrust/iterator/iterator_traits.h b/thrust/thrust/iterator/iterator_traits.h
index b2f4b175a3f..7d9cf29c942 100644
--- a/thrust/thrust/iterator/iterator_traits.h
+++ b/thrust/thrust/iterator/iterator_traits.h
@@ -21,7 +21,7 @@
 
 /*
  * (C) Copyright David Abrahams 2003.
- * 
+ *
  * Distributed under the Boost Software License, Version 1.0.
  * (See accompanying NOTICE file for the complete license)
  *
@@ -31,6 +31,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/type_traits/void_t.h>
 
 #include <iterator>
diff --git a/thrust/thrust/iterator/permutation_iterator.h b/thrust/thrust/iterator/permutation_iterator.h
index 60e9c8e8ada..6a615eaff2a 100644
--- a/thrust/thrust/iterator/permutation_iterator.h
+++ b/thrust/thrust/iterator/permutation_iterator.h
@@ -32,6 +32,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/type_traits.h>
 #include <thrust/iterator/detail/permutation_iterator_base.h>
 #include <thrust/iterator/iterator_facade.h>
diff --git a/thrust/thrust/iterator/retag.h b/thrust/thrust/iterator/retag.h
index 1eb770ae321..17740d3f1bd 100644
--- a/thrust/thrust/iterator/retag.h
+++ b/thrust/thrust/iterator/retag.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/iterator/detail/retag.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/iterator/reverse_iterator.h b/thrust/thrust/iterator/reverse_iterator.h
index fe8bbe0cf83..0f903ca4886 100644
--- a/thrust/thrust/iterator/reverse_iterator.h
+++ b/thrust/thrust/iterator/reverse_iterator.h
@@ -23,7 +23,7 @@
  * (C) Copyright David Abrahams 2002.
  * (C) Copyright Jeremy Siek    2002.
  * (C) Copyright Thomas Witt    2002.
- * 
+ *
  * Distributed under the Boost Software License, Version 1.0.
  * (See accompanying NOTICE file for the complete license)
  *
@@ -33,6 +33,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/type_traits.h>
 #include <thrust/iterator/detail/reverse_iterator_base.h>
 #include <thrust/iterator/iterator_facade.h>
@@ -216,7 +222,7 @@ template<typename BidirectionalIterator>
 
 /*! \p make_reverse_iterator creates a \p reverse_iterator
  *  from a \c BidirectionalIterator pointing to a range of elements to reverse.
- *  
+ *
  *  \param x A \c BidirectionalIterator pointing to a range to reverse.
  *  \return A new \p reverse_iterator which reverses the range \p x.
  */
diff --git a/thrust/thrust/iterator/transform_input_output_iterator.h b/thrust/thrust/iterator/transform_input_output_iterator.h
index a5f725dc537..f050f509bdf 100644
--- a/thrust/thrust/iterator/transform_input_output_iterator.h
+++ b/thrust/thrust/iterator/transform_input_output_iterator.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/iterator/detail/transform_input_output_iterator.inl>
 
 THRUST_NAMESPACE_BEGIN
@@ -62,7 +68,7 @@ THRUST_NAMESPACE_BEGIN
  *    // Iterator that returns negated values and writes squared values
  *    auto iter = thrust::make_transform_input_output_iterator(v.begin(),
  *        thrust::negate<float>{}, thrust::square<float>{});
- * 
+ *
  *    // Iterator negates values when reading
  *    std::cout << iter[0] << " ";  // -1.0f;
  *    std::cout << iter[1] << " ";  // -2.0f;
diff --git a/thrust/thrust/iterator/transform_iterator.h b/thrust/thrust/iterator/transform_iterator.h
index f777a9baf27..128ddf3f73d 100644
--- a/thrust/thrust/iterator/transform_iterator.h
+++ b/thrust/thrust/iterator/transform_iterator.h
@@ -34,6 +34,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // #include the details first
 #include <thrust/iterator/detail/transform_iterator.inl>
 #include <thrust/iterator/iterator_facade.h>
diff --git a/thrust/thrust/iterator/transform_output_iterator.h b/thrust/thrust/iterator/transform_output_iterator.h
index 3ac4b85729e..d5b1f272f74 100644
--- a/thrust/thrust/iterator/transform_output_iterator.h
+++ b/thrust/thrust/iterator/transform_output_iterator.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/iterator/detail/transform_output_iterator.inl>
 
 THRUST_NAMESPACE_BEGIN
@@ -38,7 +44,7 @@ THRUST_NAMESPACE_BEGIN
 /*! \p transform_output_iterator is a special kind of output iterator which
  * transforms a value written upon dereference. This iterator is useful
  * for transforming an output from algorithms without explicitly storing the
- * intermediate result in the memory and applying subsequent transformation, 
+ * intermediate result in the memory and applying subsequent transformation,
  * thereby avoiding wasting memory capacity and bandwidth.
  * Using \p transform_iterator facilitates kernel fusion by deferring execution
  * of transformation until the value is written while saving both memory
@@ -61,7 +67,7 @@ THRUST_NAMESPACE_BEGIN
  *      return sqrtf(x);
  *    }
  *  };
- *  
+ *
  *  int main()
  *  {
  *    thrust::device_vector<float> v(4);
@@ -69,17 +75,17 @@ THRUST_NAMESPACE_BEGIN
  *    typedef thrust::device_vector<float>::iterator FloatIterator;
  *    thrust::transform_output_iterator<square_root, FloatIterator> iter(v.begin(), square_root());
  *
- *    iter[0] =  1.0f;    // stores sqrtf( 1.0f) 
+ *    iter[0] =  1.0f;    // stores sqrtf( 1.0f)
  *    iter[1] =  4.0f;    // stores sqrtf( 4.0f)
  *    iter[2] =  9.0f;    // stores sqrtf( 9.0f)
  *    iter[3] = 16.0f;    // stores sqrtf(16.0f)
  *    // iter[4] is an out-of-bounds error
- *                                                                                           
+ *
  *    v[0]; // returns 1.0f;
  *    v[1]; // returns 2.0f;
  *    v[2]; // returns 3.0f;
  *    v[3]; // returns 4.0f;
- *                                                                                           
+ *
  *  }
  *  \endcode
  *
@@ -109,7 +115,7 @@ template <typename UnaryFunction, typename OutputIterator>
   /*! This constructor takes as argument an \c OutputIterator and an \c
    * UnaryFunction and copies them to a new \p transform_output_iterator
    *
-   * \param out An \c OutputIterator pointing to the output range whereto the result of 
+   * \param out An \c OutputIterator pointing to the output range whereto the result of
    *            \p transform_output_iterator's \c UnaryFunction will be written.
    * \param fun An \c UnaryFunction used to transform the objects assigned to
    *            this \p transform_output_iterator.
diff --git a/thrust/thrust/iterator/zip_iterator.h b/thrust/thrust/iterator/zip_iterator.h
index c2dd5ddc456..fb55295aaa9 100644
--- a/thrust/thrust/iterator/zip_iterator.h
+++ b/thrust/thrust/iterator/zip_iterator.h
@@ -22,7 +22,7 @@
 
 /*
  * Copyright David Abrahams and Thomas Becker 2000-2006.
- * 
+ *
  * Distributed under the Boost Software License, Version 1.0.
  * (See accompanying NOTICE file for the complete license)
  *
@@ -32,6 +32,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/iterator/detail/zip_iterator_base.h>
 #include <thrust/iterator/iterator_facade.h>
 #include <thrust/detail/type_traits.h>
@@ -56,7 +62,7 @@ THRUST_NAMESPACE_BEGIN
  *
  *  The following code snippet demonstrates how to create a \p zip_iterator
  *  which represents the result of "zipping" multiple ranges together.
- *  
+ *
  *  \code
  *  #include <thrust/iterator/zip_iterator.h>
  *  #include <thrust/tuple.h>
@@ -147,7 +153,7 @@ template <typename IteratorTuple>
 
     /*! This constructor creates a new \p zip_iterator from a
      *  \p tuple of iterators.
-     *  
+     *
      *  \param iterator_tuple The \p tuple of iterators to copy from.
      */
     inline __host__ __device__
diff --git a/thrust/thrust/limits.h b/thrust/thrust/limits.h
index 52f38b1fc9e..92281ea0a59 100644
--- a/thrust/thrust/limits.h
+++ b/thrust/thrust/limits.h
@@ -5,6 +5,14 @@
 
 #pragma once
 
+#include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <limits>
 
 #include <thrust/detail/config.h>
diff --git a/thrust/thrust/logical.h b/thrust/thrust/logical.h
index 5a8dbbecf6e..6e4e7d4e45a 100644
--- a/thrust/thrust/logical.h
+++ b/thrust/thrust/logical.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
@@ -36,7 +42,7 @@ THRUST_NAMESPACE_BEGIN
 
 /*! \p all_of determines whether all elements in a range satify a predicate.
  *  Specifically, \p all_of returns \c true if <tt>pred(*i)</tt> is \c true
- *  for every iterator \c i in the range <tt>[first, last)</tt> and 
+ *  for every iterator \c i in the range <tt>[first, last)</tt> and
  *  \c false otherwise.
  *
  *  The algorithm's execution is parallelized as determined by \p exec.
@@ -63,7 +69,7 @@ THRUST_NAMESPACE_BEGIN
  *
  *  // empty range
  *  thrust::all_of(thrust::host, A, A, thrust::identity<bool>()); // returns false
- *  
+ *
  *  \endcode
  *
  *  \see any_of
@@ -77,7 +83,7 @@ bool all_of(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, In
 
 /*! \p all_of determines whether all elements in a range satify a predicate.
  * Specifically, \p all_of returns \c true if <tt>pred(*i)</tt> is \c true
- * for every iterator \c i in the range <tt>[first, last)</tt> and 
+ * for every iterator \c i in the range <tt>[first, last)</tt> and
  * \c false otherwise.
  *
  *  \param first The beginning of the sequence.
@@ -99,7 +105,7 @@ bool all_of(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, In
  *
  *  // empty range
  *  thrust::all_of(A, A, thrust::identity<bool>()); // returns false
- *  
+ *
  *  \endcode
  *
  *  \see any_of
@@ -112,7 +118,7 @@ bool all_of(InputIterator first, InputIterator last, Predicate pred);
 
 /*! \p any_of determines whether any element in a range satifies a predicate.
  *  Specifically, \p any_of returns \c true if <tt>pred(*i)</tt> is \c true
- *  for any iterator \c i in the range <tt>[first, last)</tt> and 
+ *  for any iterator \c i in the range <tt>[first, last)</tt> and
  *  \c false otherwise.
  *
  *  The algorithm's execution is parallelized as determined by \p exec.
@@ -150,11 +156,11 @@ bool all_of(InputIterator first, InputIterator last, Predicate pred);
 template<typename DerivedPolicy, typename InputIterator, typename Predicate>
 __host__ __device__
 bool any_of(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, InputIterator first, InputIterator last, Predicate pred);
-   
+
 
 /*! \p any_of determines whether any element in a range satifies a predicate.
  * Specifically, \p any_of returns \c true if <tt>pred(*i)</tt> is \c true
- * for any iterator \c i in the range <tt>[first, last)</tt> and 
+ * for any iterator \c i in the range <tt>[first, last)</tt> and
  * \c false otherwise.
  *
  *  \param first The beginning of the sequence.
@@ -189,7 +195,7 @@ bool any_of(InputIterator first, InputIterator last, Predicate pred);
 
 
 /*! \p none_of determines whether no element in a range satifies a predicate.
- *  Specifically, \p none_of returns \c true if there is no iterator \c i in 
+ *  Specifically, \p none_of returns \c true if there is no iterator \c i in
  *  the range <tt>[first, last)</tt> such that <tt>pred(*i)</tt> is \c true,
  *  and \c false otherwise.
  *
@@ -231,7 +237,7 @@ bool none_of(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, I
 
 
 /*! \p none_of determines whether no element in a range satifies a predicate.
- *  Specifically, \p none_of returns \c true if there is no iterator \c i in 
+ *  Specifically, \p none_of returns \c true if there is no iterator \c i in
  *  the range <tt>[first, last)</tt> such that <tt>pred(*i)</tt> is \c true,
  *  and \c false otherwise.
  *
diff --git a/thrust/thrust/memory.h b/thrust/thrust/memory.h
index 819ac251301..39557fe9f48 100644
--- a/thrust/thrust/memory.h
+++ b/thrust/thrust/memory.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/type_traits/pointer_traits.h>
 #include <thrust/detail/pointer.h>
 #include <thrust/detail/reference.h>
diff --git a/thrust/thrust/merge.h b/thrust/thrust/merge.h
index 724f4c167f1..85139edc52d 100644
--- a/thrust/thrust/merge.h
+++ b/thrust/thrust/merge.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 #include <thrust/pair.h>
 
diff --git a/thrust/thrust/mismatch.h b/thrust/thrust/mismatch.h
index bbdf2923a09..ed614df340a 100644
--- a/thrust/thrust/mismatch.h
+++ b/thrust/thrust/mismatch.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 #include <thrust/pair.h>
 
@@ -37,7 +43,7 @@ THRUST_NAMESPACE_BEGIN
 
 
 /*! \p mismatch finds the first position where the two ranges <tt>[first1, last1)</tt>
- *  and <tt>[first2, first2 + (last1 - first1))</tt> differ. The two versions of 
+ *  and <tt>[first2, first2 + (last1 - first1))</tt> differ. The two versions of
  *  \p mismatch use different tests for whether elements differ.
  *
  *  This version of \p mismatch finds the first iterator \c i in <tt>[first1, last1)</tt>
@@ -67,7 +73,7 @@ THRUST_NAMESPACE_BEGIN
  *  thrust::device_vector<int> vec1(4);
  *  thrust::device_vector<int> vec2(4);
  *
- *  vec1[0] = 0;  vec2[0] = 0; 
+ *  vec1[0] = 0;  vec2[0] = 0;
  *  vec1[1] = 5;  vec2[1] = 5;
  *  vec1[2] = 3;  vec2[2] = 8;
  *  vec1[3] = 7;  vec2[3] = 7;
@@ -93,7 +99,7 @@ thrust::pair<InputIterator1, InputIterator2> mismatch(const thrust::detail::exec
 
 
 /*! \p mismatch finds the first position where the two ranges <tt>[first1, last1)</tt>
- * and <tt>[first2, first2 + (last1 - first1))</tt> differ. The two versions of 
+ * and <tt>[first2, first2 + (last1 - first1))</tt> differ. The two versions of
  * \p mismatch use different tests for whether elements differ.
  *
  * This version of \p mismatch finds the first iterator \c i in <tt>[first1, last1)</tt>
@@ -118,7 +124,7 @@ thrust::pair<InputIterator1, InputIterator2> mismatch(const thrust::detail::exec
  *  thrust::device_vector<int> vec1(4);
  *  thrust::device_vector<int> vec2(4);
  *
- *  vec1[0] = 0;  vec2[0] = 0; 
+ *  vec1[0] = 0;  vec2[0] = 0;
  *  vec1[1] = 5;  vec2[1] = 5;
  *  vec1[2] = 3;  vec2[2] = 8;
  *  vec1[3] = 7;  vec2[3] = 7;
@@ -142,7 +148,7 @@ thrust::pair<InputIterator1, InputIterator2> mismatch(InputIterator1 first1,
 
 
 /*! \p mismatch finds the first position where the two ranges <tt>[first1, last1)</tt>
- *  and <tt>[first2, first2 + (last1 - first1))</tt> differ. The two versions of 
+ *  and <tt>[first2, first2 + (last1 - first1))</tt> differ. The two versions of
  *  \p mismatch use different tests for whether elements differ.
  *
  *  This version of \p mismatch finds the first iterator \c i in <tt>[first1, last1)</tt>
@@ -173,7 +179,7 @@ thrust::pair<InputIterator1, InputIterator2> mismatch(InputIterator1 first1,
  *  thrust::device_vector<int> vec1(4);
  *  thrust::device_vector<int> vec2(4);
  *
- *  vec1[0] = 0;  vec2[0] = 0; 
+ *  vec1[0] = 0;  vec2[0] = 0;
  *  vec1[1] = 5;  vec2[1] = 5;
  *  vec1[2] = 3;  vec2[2] = 8;
  *  vec1[3] = 7;  vec2[3] = 7;
@@ -200,7 +206,7 @@ thrust::pair<InputIterator1, InputIterator2> mismatch(const thrust::detail::exec
 
 
 /*! \p mismatch finds the first position where the two ranges <tt>[first1, last1)</tt>
- * and <tt>[first2, first2 + (last1 - first1))</tt> differ. The two versions of 
+ * and <tt>[first2, first2 + (last1 - first1))</tt> differ. The two versions of
  * \p mismatch use different tests for whether elements differ.
  *
  * This version of \p mismatch finds the first iterator \c i in <tt>[first1, last1)</tt>
@@ -226,7 +232,7 @@ thrust::pair<InputIterator1, InputIterator2> mismatch(const thrust::detail::exec
  *  thrust::device_vector<int> vec1(4);
  *  thrust::device_vector<int> vec2(4);
  *
- *  vec1[0] = 0;  vec2[0] = 0; 
+ *  vec1[0] = 0;  vec2[0] = 0;
  *  vec1[1] = 5;  vec2[1] = 5;
  *  vec1[2] = 3;  vec2[2] = 8;
  *  vec1[3] = 7;  vec2[3] = 7;
diff --git a/thrust/thrust/mr/allocator.h b/thrust/thrust/mr/allocator.h
index 67adbe87c34..e6dd57b763b 100644
--- a/thrust/thrust/mr/allocator.h
+++ b/thrust/thrust/mr/allocator.h
@@ -14,16 +14,23 @@
  *  limitations under the License.
  */
 
-/*! \file 
+/*! \file
  *  \brief Allocator types usable with \ref Memory Resources.
  */
 
 #pragma once
 
+#include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <limits>
 
 #include <thrust/detail/config.h>
-#include <thrust/detail/config/exec_check_disable.h>
 #include <thrust/detail/config/memory_resource.h>
 #include <thrust/detail/type_traits/pointer_traits.h>
 
diff --git a/thrust/thrust/mr/device_memory_resource.h b/thrust/thrust/mr/device_memory_resource.h
index 3a671142aaf..6dce1a6e09b 100644
--- a/thrust/thrust/mr/device_memory_resource.h
+++ b/thrust/thrust/mr/device_memory_resource.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // #include the device system's memory_resource header
 #define __THRUST_DEVICE_SYSTEM_MEMORY_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/memory_resource.h>
 #include __THRUST_DEVICE_SYSTEM_MEMORY_HEADER
diff --git a/thrust/thrust/mr/disjoint_pool.h b/thrust/thrust/mr/disjoint_pool.h
index b00a8644cca..245cc91a8a0 100644
--- a/thrust/thrust/mr/disjoint_pool.h
+++ b/thrust/thrust/mr/disjoint_pool.h
@@ -14,13 +14,21 @@
  *  limitations under the License.
  */
 
-/*! \file 
+/*! \file
  *  \brief A caching and pooling memory resource adaptor which uses separate upstream resources for memory allocation
  *      and bookkeeping.
  */
 
 #pragma once
 
+#include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/algorithm_wrapper.h>
 #include <thrust/detail/config.h>
 
diff --git a/thrust/thrust/mr/disjoint_sync_pool.h b/thrust/thrust/mr/disjoint_sync_pool.h
index ed81ae4cb84..aea7fc2a109 100644
--- a/thrust/thrust/mr/disjoint_sync_pool.h
+++ b/thrust/thrust/mr/disjoint_sync_pool.h
@@ -14,13 +14,19 @@
  *  limitations under the License.
  */
 
-/*! \file 
+/*! \file
  *  \brief A mutex-synchronized version of \p disjoint_unsynchronized_pool_resource.
  */
 
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp11_required.h>
 
 #if THRUST_CPP_DIALECT >= 2011
diff --git a/thrust/thrust/mr/disjoint_tls_pool.h b/thrust/thrust/mr/disjoint_tls_pool.h
index 9fc7917cae1..730f1c2a116 100644
--- a/thrust/thrust/mr/disjoint_tls_pool.h
+++ b/thrust/thrust/mr/disjoint_tls_pool.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp11_required.h>
 
 #if THRUST_CPP_DIALECT >= 2011
diff --git a/thrust/thrust/mr/fancy_pointer_resource.h b/thrust/thrust/mr/fancy_pointer_resource.h
index b8810756409..3282e81ac98 100644
--- a/thrust/thrust/mr/fancy_pointer_resource.h
+++ b/thrust/thrust/mr/fancy_pointer_resource.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/type_traits/pointer_traits.h>
 
 #include <thrust/mr/memory_resource.h>
diff --git a/thrust/thrust/mr/host_memory_resource.h b/thrust/thrust/mr/host_memory_resource.h
index 9359a97a7de..6460148d4fb 100644
--- a/thrust/thrust/mr/host_memory_resource.h
+++ b/thrust/thrust/mr/host_memory_resource.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // #include the host system's memory_resource header
 #define __THRUST_HOST_SYSTEM_MEMORY_HEADER <__THRUST_HOST_SYSTEM_ROOT/memory_resource.h>
 #include __THRUST_HOST_SYSTEM_MEMORY_HEADER
diff --git a/thrust/thrust/mr/memory_resource.h b/thrust/thrust/mr/memory_resource.h
index 6af2f167c64..bda6d54ce61 100644
--- a/thrust/thrust/mr/memory_resource.h
+++ b/thrust/thrust/mr/memory_resource.h
@@ -22,6 +22,13 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/config/memory_resource.h>
 #ifdef THRUST_MR_STD_MR_HEADER
 #  include THRUST_MR_STD_MR_HEADER
diff --git a/thrust/thrust/mr/new.h b/thrust/thrust/mr/new.h
index 644e25169ed..ccc788e9aca 100644
--- a/thrust/thrust/mr/new.h
+++ b/thrust/thrust/mr/new.h
@@ -22,6 +22,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/mr/memory_resource.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/mr/polymorphic_adaptor.h b/thrust/thrust/mr/polymorphic_adaptor.h
index 0562a8f8269..9114be766d9 100644
--- a/thrust/thrust/mr/polymorphic_adaptor.h
+++ b/thrust/thrust/mr/polymorphic_adaptor.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/mr/memory_resource.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/mr/pool.h b/thrust/thrust/mr/pool.h
index 6259a23f172..2eee32ced63 100644
--- a/thrust/thrust/mr/pool.h
+++ b/thrust/thrust/mr/pool.h
@@ -14,7 +14,7 @@
  *  limitations under the License.
  */
 
-/*! \file 
+/*! \file
  *  \brief A caching and pooling memory resource adaptor which uses a single
  *  upstream resource for memory allocation, and embeds bookkeeping information
  *  in allocated blocks.
@@ -24,6 +24,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/algorithm_wrapper.h>
 
 #include <thrust/host_vector.h>
diff --git a/thrust/thrust/mr/pool_options.h b/thrust/thrust/mr/pool_options.h
index 13a8fe67400..c0346b13c95 100644
--- a/thrust/thrust/mr/pool_options.h
+++ b/thrust/thrust/mr/pool_options.h
@@ -14,18 +14,24 @@
  *  limitations under the License.
  */
 
-/*! \file 
+/*! \file
  *  \brief A type used by the pooling resource adaptors to fine-tune their
  *  behavior.
  */
 
 #pragma once
 
+#include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cstddef>
 
 #include <thrust/detail/integer_math.h>
-
-#include <thrust/detail/config.h>
 #include <thrust/detail/config/memory_resource.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/mr/sync_pool.h b/thrust/thrust/mr/sync_pool.h
index 46c0e844137..6e1a338226e 100644
--- a/thrust/thrust/mr/sync_pool.h
+++ b/thrust/thrust/mr/sync_pool.h
@@ -14,13 +14,19 @@
  *  limitations under the License.
  */
 
-/*! \file 
+/*! \file
  *  \brief A mutex-synchronized version of \p unsynchronized_pool_resource.
  */
 
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp11_required.h>
 
 #if THRUST_CPP_DIALECT >= 2011
diff --git a/thrust/thrust/mr/tls_pool.h b/thrust/thrust/mr/tls_pool.h
index 8ee8127a3d7..f8984d97be6 100644
--- a/thrust/thrust/mr/tls_pool.h
+++ b/thrust/thrust/mr/tls_pool.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp11_required.h>
 
 #if THRUST_CPP_DIALECT >= 2011
diff --git a/thrust/thrust/mr/universal_memory_resource.h b/thrust/thrust/mr/universal_memory_resource.h
index b7f1ebd6f46..f056fda5efc 100644
--- a/thrust/thrust/mr/universal_memory_resource.h
+++ b/thrust/thrust/mr/universal_memory_resource.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/mr/device_memory_resource.h>
 
diff --git a/thrust/thrust/mr/validator.h b/thrust/thrust/mr/validator.h
index 10e964821a5..3a9c9c3da99 100644
--- a/thrust/thrust/mr/validator.h
+++ b/thrust/thrust/mr/validator.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/config/memory_resource.h>
 #include <thrust/mr/memory_resource.h>
 
diff --git a/thrust/thrust/optional.h b/thrust/thrust/optional.h
index a1ca4f46554..746f202db24 100644
--- a/thrust/thrust/optional.h
+++ b/thrust/thrust/optional.h
@@ -14,6 +14,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp11_required.h>
 #include <thrust/detail/type_traits.h>
 
diff --git a/thrust/thrust/pair.h b/thrust/thrust/pair.h
index 8b1b6b91f87..15ebc1cc5ee 100644
--- a/thrust/thrust/pair.h
+++ b/thrust/thrust/pair.h
@@ -22,6 +22,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cuda/std/utility>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/partition.h b/thrust/thrust/partition.h
index 90768f24676..c98116c7fc3 100644
--- a/thrust/thrust/partition.h
+++ b/thrust/thrust/partition.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 #include <thrust/pair.h>
 
@@ -1246,7 +1252,7 @@ template<typename InputIterator1,
  *      return (x % 2) == 0;
  *    }
  *  };
- *  
+ *
  *  ...
  *
  *  int A[] = {2, 4, 6, 8, 10, 1, 3, 5, 7, 9};
@@ -1298,7 +1304,7 @@ __host__ __device__
  *      return (x % 2) == 0;
  *    }
  *  };
- *  
+ *
  *  ...
  *
  *  int A[] = {2, 4, 6, 8, 10, 1, 3, 5, 7, 9};
@@ -1325,7 +1331,7 @@ template<typename ForwardIterator, typename Predicate>
  */
 
 
-/*! \p is_partitioned returns \c true if the given range 
+/*! \p is_partitioned returns \c true if the given range
  *  is partitioned with respect to a predicate, and \c false otherwise.
  *
  *  Specifically, \p is_partitioned returns \c true if <tt>[first, last)</tt>
@@ -1346,7 +1352,7 @@ template<typename ForwardIterator, typename Predicate>
  *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Input Iterator</a>,
  *          and \p InputIterator's \c value_type is convertible to \p Predicate's \c argument_type.
  *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
- *  
+ *
  *  \code
  *  #include <thrust/partition.h>
  *  #include <thrust/execution_policy.h>
@@ -1359,7 +1365,7 @@ template<typename ForwardIterator, typename Predicate>
  *      return (x % 2) == 0;
  *    }
  *  };
- *  
+ *
  *  ...
  *
  *  int A[] = {2, 4, 6, 8, 10, 1, 3, 5, 7, 9};
@@ -1379,7 +1385,7 @@ __host__ __device__
                       Predicate pred);
 
 
-/*! \p is_partitioned returns \c true if the given range 
+/*! \p is_partitioned returns \c true if the given range
  *  is partitioned with respect to a predicate, and \c false otherwise.
  *
  *  Specifically, \p is_partitioned returns \c true if <tt>[first, last)</tt>
@@ -1396,7 +1402,7 @@ __host__ __device__
  *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Input Iterator</a>,
  *          and \p InputIterator's \c value_type is convertible to \p Predicate's \c argument_type.
  *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
- *  
+ *
  *  \code
  *  #include <thrust/partition.h>
  *
@@ -1408,7 +1414,7 @@ __host__ __device__
  *      return (x % 2) == 0;
  *    }
  *  };
- *  
+ *
  *  ...
  *
  *  int A[] = {2, 4, 6, 8, 10, 1, 3, 5, 7, 9};
diff --git a/thrust/thrust/per_device_resource.h b/thrust/thrust/per_device_resource.h
index a6d620f85ec..8ee9b5d5b19 100644
--- a/thrust/thrust/per_device_resource.h
+++ b/thrust/thrust/per_device_resource.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp11_required.h>
 
 #if THRUST_CPP_DIALECT >= 2011
diff --git a/thrust/thrust/random.h b/thrust/thrust/random.h
index 7463620b7bc..223566dae01 100644
--- a/thrust/thrust/random.h
+++ b/thrust/thrust/random.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cstdint.h>
 
 // RNGs
diff --git a/thrust/thrust/random/detail/discard_block_engine.inl b/thrust/thrust/random/detail/discard_block_engine.inl
index 31128e25061..fc95d7e29ea 100644
--- a/thrust/thrust/random/detail/discard_block_engine.inl
+++ b/thrust/thrust/random/detail/discard_block_engine.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/random/discard_block_engine.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/random/detail/linear_congruential_engine.inl b/thrust/thrust/random/detail/linear_congruential_engine.inl
index fa9fd7d0d66..5058d1a912d 100644
--- a/thrust/thrust/random/detail/linear_congruential_engine.inl
+++ b/thrust/thrust/random/detail/linear_congruential_engine.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/random/linear_congruential_engine.h>
 #include <thrust/random/detail/mod.h>
 #include <thrust/random/detail/random_core_access.h>
diff --git a/thrust/thrust/random/detail/linear_congruential_engine_discard.h b/thrust/thrust/random/detail/linear_congruential_engine_discard.h
index c8103d9dc94..c6c6fd54411 100644
--- a/thrust/thrust/random/detail/linear_congruential_engine_discard.h
+++ b/thrust/thrust/random/detail/linear_congruential_engine_discard.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/cstdint.h>
 #include <thrust/random/detail/mod.h>
 
@@ -59,7 +65,7 @@ template<thrust::detail::uint32_t a, thrust::detail::uint32_t m>
     //     figure out a robust implementation of this later
     unsigned long long multiplier = a;
     unsigned long long multiplier_to_z = 1;
-    
+
     // see http://en.wikipedia.org/wiki/Modular_exponentiation
     while(z > 0)
     {
@@ -89,7 +95,7 @@ struct linear_congruential_engine_discard
     const result_type c = LinearCongruentialEngine::increment;
     const result_type a = LinearCongruentialEngine::multiplier;
     const result_type m = LinearCongruentialEngine::modulus;
-    
+
     // XXX WAR unused variable warnings
     (void) c;
     (void) a;
diff --git a/thrust/thrust/random/detail/linear_feedback_shift_engine.inl b/thrust/thrust/random/detail/linear_feedback_shift_engine.inl
index ac3ca86736b..90be369f33e 100644
--- a/thrust/thrust/random/detail/linear_feedback_shift_engine.inl
+++ b/thrust/thrust/random/detail/linear_feedback_shift_engine.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/random/linear_feedback_shift_engine.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/random/detail/linear_feedback_shift_engine_wordmask.h b/thrust/thrust/random/detail/linear_feedback_shift_engine_wordmask.h
index 73c8ae83efb..627ce8afd42 100644
--- a/thrust/thrust/random/detail/linear_feedback_shift_engine_wordmask.h
+++ b/thrust/thrust/random/detail/linear_feedback_shift_engine_wordmask.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 THRUST_NAMESPACE_BEGIN
 
 namespace random
diff --git a/thrust/thrust/random/detail/mod.h b/thrust/thrust/random/detail/mod.h
index f0637582de5..b3f0de9445e 100644
--- a/thrust/thrust/random/detail/mod.h
+++ b/thrust/thrust/random/detail/mod.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 THRUST_NAMESPACE_BEGIN
 
 namespace random
diff --git a/thrust/thrust/random/detail/normal_distribution.inl b/thrust/thrust/random/detail/normal_distribution.inl
index 4b69bab2139..46c3667f95c 100644
--- a/thrust/thrust/random/detail/normal_distribution.inl
+++ b/thrust/thrust/random/detail/normal_distribution.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/random/normal_distribution.h>
 #include <thrust/random/uniform_real_distribution.h>
 #include <thrust/detail/cstdint.h>
diff --git a/thrust/thrust/random/detail/normal_distribution_base.h b/thrust/thrust/random/detail/normal_distribution_base.h
index a42e8001403..fed4f718ec7 100644
--- a/thrust/thrust/random/detail/normal_distribution_base.h
+++ b/thrust/thrust/random/detail/normal_distribution_base.h
@@ -24,6 +24,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/pair.h>
 #include <thrust/random/uniform_real_distribution.h>
 #include <limits>
diff --git a/thrust/thrust/random/detail/random_core_access.h b/thrust/thrust/random/detail/random_core_access.h
index a3e34e02b93..549e268e42f 100644
--- a/thrust/thrust/random/detail/random_core_access.h
+++ b/thrust/thrust/random/detail/random_core_access.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 THRUST_NAMESPACE_BEGIN
 
 namespace random
diff --git a/thrust/thrust/random/detail/subtract_with_carry_engine.inl b/thrust/thrust/random/detail/subtract_with_carry_engine.inl
index 21c22fe775c..e9518fdf0de 100644
--- a/thrust/thrust/random/detail/subtract_with_carry_engine.inl
+++ b/thrust/thrust/random/detail/subtract_with_carry_engine.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/random/linear_congruential_engine.h>
 #include <thrust/random/subtract_with_carry_engine.h>
 #include <thrust/random/detail/mod.h>
diff --git a/thrust/thrust/random/detail/uniform_int_distribution.inl b/thrust/thrust/random/detail/uniform_int_distribution.inl
index 064bfcc733b..00e9cfe1b64 100644
--- a/thrust/thrust/random/detail/uniform_int_distribution.inl
+++ b/thrust/thrust/random/detail/uniform_int_distribution.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/random/uniform_int_distribution.h>
 #include <thrust/random/uniform_real_distribution.h>
 #include <thrust/detail/type_traits.h>
diff --git a/thrust/thrust/random/detail/uniform_real_distribution.inl b/thrust/thrust/random/detail/uniform_real_distribution.inl
index 119f82c1eba..f7a7ec4074c 100644
--- a/thrust/thrust/random/detail/uniform_real_distribution.inl
+++ b/thrust/thrust/random/detail/uniform_real_distribution.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/random/uniform_real_distribution.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/random/detail/xor_combine_engine.inl b/thrust/thrust/random/detail/xor_combine_engine.inl
index c948214438f..3fbce644ac7 100644
--- a/thrust/thrust/random/detail/xor_combine_engine.inl
+++ b/thrust/thrust/random/detail/xor_combine_engine.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/random/xor_combine_engine.h>
 #include <thrust/random/detail/random_core_access.h>
 
diff --git a/thrust/thrust/random/detail/xor_combine_engine_max.h b/thrust/thrust/random/detail/xor_combine_engine_max.h
index 0756ff9e04c..f450aaeebc3 100644
--- a/thrust/thrust/random/detail/xor_combine_engine_max.h
+++ b/thrust/thrust/random/detail/xor_combine_engine_max.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/type_traits.h>
 #include <thrust/detail/mpl/math.h>
 #include <limits>
@@ -219,7 +225,7 @@ template<typename result_type, result_type a, result_type b, int d>
 {
   typedef xor_combine_engine_max_aux_constants<result_type,a,b,d> constants;
 
-  static const result_type value = 
+  static const result_type value =
     thrust::detail::eval_if<
       // if k is odd...
       math::is_odd<result_type, constants::k>::value,
@@ -287,7 +293,7 @@ template<typename Engine1, size_t s1, typename Engine2, size_t s2, typename resu
     math::min<
       result_type,
       result_type(Engine1::max - Engine1::min),
-      two_to_the_power<result_type, w-s1>::value - 1 
+      two_to_the_power<result_type, w-s1>::value - 1
     >::value;
 
   static const result_type m2 =
diff --git a/thrust/thrust/random/discard_block_engine.h b/thrust/thrust/random/discard_block_engine.h
index 88e1155861e..96c50abbd49 100644
--- a/thrust/thrust/random/discard_block_engine.h
+++ b/thrust/thrust/random/discard_block_engine.h
@@ -24,7 +24,12 @@
 
 #include <thrust/detail/config.h>
 
-#include <thrust/detail/config.h>
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <iostream>
 #include <thrust/detail/cstdint.h>
 #include <thrust/random/detail/random_core_access.h>
@@ -71,7 +76,7 @@ namespace random
  *    return 0;
  *  }
  *  \endcode
- */         
+ */
 template<typename Engine, size_t p, size_t r>
   class discard_block_engine
 {
@@ -124,7 +129,7 @@ template<typename Engine, size_t p, size_t r>
     explicit discard_block_engine(const base_type &urng);
 
     /*! This constructor initializes a new \p discard_block_engine with a given seed.
-     *  
+     *
      *  \param s The seed used to intialize this \p discard_block_engine's adapted base engine.
      */
     __host__ __device__
@@ -145,7 +150,7 @@ template<typename Engine, size_t p, size_t r>
     void seed(result_type s);
 
     // generating functions
-    
+
     /*! This member function produces a new random value and updates this \p discard_block_engine's state.
      *  \return A new random number.
      */
diff --git a/thrust/thrust/random/linear_congruential_engine.h b/thrust/thrust/random/linear_congruential_engine.h
index dac03d90eb2..63e31a2dcc3 100644
--- a/thrust/thrust/random/linear_congruential_engine.h
+++ b/thrust/thrust/random/linear_congruential_engine.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <iostream>
 #include <thrust/detail/cstdint.h>
 #include <thrust/random/detail/random_core_access.h>
@@ -113,7 +119,7 @@ template<typename UIntType, UIntType a, UIntType c, UIntType m>
 {
   public:
     // types
-    
+
     /*! \typedef result_type
      *  \brief The type of the unsigned integer produced by this \p linear_congruential_engine.
      */
@@ -149,7 +155,7 @@ template<typename UIntType, UIntType a, UIntType c, UIntType m>
 
     /*! This constructor, which optionally accepts a seed, initializes a new
      *  \p linear_congruential_engine.
-     *  
+     *
      *  \param s The seed used to intialize this \p linear_congruential_engine's state.
      */
     __host__ __device__
@@ -280,7 +286,7 @@ typedef linear_congruential_engine<thrust::detail::uint32_t, 48271, 0, 214748364
 
 /*! \} // predefined_random
  */
-  
+
 } // end random
 
 // import names into thrust::
diff --git a/thrust/thrust/random/linear_feedback_shift_engine.h b/thrust/thrust/random/linear_feedback_shift_engine.h
index a46c6d8ab8a..ea25a883266 100644
--- a/thrust/thrust/random/linear_feedback_shift_engine.h
+++ b/thrust/thrust/random/linear_feedback_shift_engine.h
@@ -30,6 +30,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/random/detail/linear_feedback_shift_engine_wordmask.h>
 #include <iostream>
 #include <cstddef> // for size_t
@@ -116,7 +122,7 @@ template<typename UIntType, size_t w, size_t k, size_t q, size_t s>
 
     /*! This constructor, which optionally accepts a seed, initializes a new
      *  \p linear_feedback_shift_engine.
-     *  
+     *
      *  \param value The seed used to intialize this \p linear_feedback_shift_engine's state.
      */
     __host__ __device__
@@ -131,7 +137,7 @@ template<typename UIntType, size_t w, size_t k, size_t q, size_t s>
     void seed(result_type value = default_seed);
 
     // generating functions
-    
+
     /*! This member function produces a new random value and updates this \p linear_feedback_shift_engine's state.
      *  \return A new random number.
      */
diff --git a/thrust/thrust/random/normal_distribution.h b/thrust/thrust/random/normal_distribution.h
index 36b985cb685..e781e8a0272 100644
--- a/thrust/thrust/random/normal_distribution.h
+++ b/thrust/thrust/random/normal_distribution.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/pair.h>
 #include <thrust/random/detail/random_core_access.h>
 #include <thrust/random/detail/normal_distribution_base.h>
@@ -43,7 +49,7 @@ namespace random
  *
  *  \tparam RealType The type of floating point number to produce.
  *
- *  The following code snippet demonstrates examples of using a \p normal_distribution with a 
+ *  The following code snippet demonstrates examples of using a \p normal_distribution with a
  *  random number engine to produce random values drawn from the Normal distribution with a given
  *  mean and variance:
  *
@@ -86,7 +92,7 @@ template<typename RealType = double>
 
   public:
     // types
-    
+
     /*! \typedef result_type
      *  \brief The type of the floating point number produced by this \p normal_distribution.
      */
@@ -98,10 +104,10 @@ template<typename RealType = double>
     typedef thrust::pair<RealType,RealType> param_type;
 
     // constructors and reset functions
-    
+
     /*! This constructor creates a new \p normal_distribution from two values defining the
      *  half-open interval of the distribution.
-     *  
+     *
      *  \param mean The mean (expected value) of the distribution. Defaults to \c 0.0.
      *  \param stddev The standard deviation of the distribution. Defaults to \c 1.0.
      */
@@ -110,7 +116,7 @@ template<typename RealType = double>
 
     /*! This constructor creates a new \p normal_distribution from a \p param_type object
      *  encapsulating the range of the distribution.
-     *  
+     *
      *  \param parm A \p param_type object encapsulating the parameters (i.e., the mean and standard deviation) of the distribution.
      */
     __host__ __device__
@@ -134,7 +140,7 @@ template<typename RealType = double>
     __host__ __device__
     result_type operator()(UniformRandomNumberGenerator &urng);
 
-    /*! This method produces a new Normal random integer as if by creating a new \p normal_distribution 
+    /*! This method produces a new Normal random integer as if by creating a new \p normal_distribution
      *  from the given \p param_type object, and calling its <tt>operator()</tt> method with the given
      *  \p UniformRandomNumberGenerator as a source of randomness.
      *
diff --git a/thrust/thrust/random/subtract_with_carry_engine.h b/thrust/thrust/random/subtract_with_carry_engine.h
index 69ee841fda3..7f62615fc29 100644
--- a/thrust/thrust/random/subtract_with_carry_engine.h
+++ b/thrust/thrust/random/subtract_with_carry_engine.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/random/detail/random_core_access.h>
 
 #include <thrust/detail/cstdint.h>
@@ -74,7 +80,7 @@ template<typename UIntType, size_t w, size_t s, size_t r>
 
   public:
     // types
-    
+
     /*! \typedef result_type
      *  \brief The type of the unsigned integer produced by this \p subtract_with_carry_engine.
      */
@@ -110,7 +116,7 @@ template<typename UIntType, size_t w, size_t s, size_t r>
 
     /*! This constructor, which optionally accepts a seed, initializes a new
      *  \p subtract_with_carry_engine.
-     *  
+     *
      *  \param value The seed used to intialize this \p subtract_with_carry_engine's state.
      */
     __host__ __device__
@@ -125,7 +131,7 @@ template<typename UIntType, size_t w, size_t s, size_t r>
     void seed(result_type value = default_seed);
 
     // generating functions
-    
+
     /*! This member function produces a new random value and updates this \p subtract_with_carry_engine's state.
      *  \return A new random number.
      */
diff --git a/thrust/thrust/random/uniform_int_distribution.h b/thrust/thrust/random/uniform_int_distribution.h
index 18f369fc2a4..7d2db7e4b07 100644
--- a/thrust/thrust/random/uniform_int_distribution.h
+++ b/thrust/thrust/random/uniform_int_distribution.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/pair.h>
 #include <thrust/detail/integer_traits.h>
 #include <thrust/random/detail/random_core_access.h>
@@ -43,7 +49,7 @@ namespace random
  *
  *  \tparam IntType The type of integer to produce.
  *
- *  The following code snippet demonstrates examples of using a \p uniform_int_distribution with a 
+ *  The following code snippet demonstrates examples of using a \p uniform_int_distribution with a
  *  random number engine to produce random integers drawn from a given range:
  *
  *  \code
@@ -103,7 +109,7 @@ template<typename IntType = int>
 
     /*! This constructor creates a new \p uniform_int_distribution from two values defining the
      *  range of the distribution.
-     *  
+     *
      *  \param a The smallest integer to potentially produce. Defaults to \c 0.
      *  \param b The largest integer to potentially produce. Defaults to the largest representable integer in
      *           the platform.
@@ -114,7 +120,7 @@ template<typename IntType = int>
 
     /*! This constructor creates a new \p uniform_int_distribution from a \p param_type object
      *  encapsulating the range of the distribution.
-     *  
+     *
      *  \param parm A \p param_type object encapsulating the parameters (i.e., the range) of the distribution.
      */
     __host__ __device__
@@ -136,7 +142,7 @@ template<typename IntType = int>
     __host__ __device__
     result_type operator()(UniformRandomNumberGenerator &urng);
 
-    /*! This method produces a new uniform random integer as if by creating a new \p uniform_int_distribution 
+    /*! This method produces a new uniform random integer as if by creating a new \p uniform_int_distribution
      *  from the given \p param_type object, and calling its <tt>operator()</tt> method with the given
      *  \p UniformRandomNumberGenerator as a source of randomness.
      *
@@ -149,7 +155,7 @@ template<typename IntType = int>
     result_type operator()(UniformRandomNumberGenerator &urng, const param_type &parm);
 
     // property functions
-    
+
     /*! This method returns the value of the parameter with which this \p uniform_int_distribution
      *  was constructed.
      *
diff --git a/thrust/thrust/random/uniform_real_distribution.h b/thrust/thrust/random/uniform_real_distribution.h
index e6c5a7d88a5..bfc703a5e83 100644
--- a/thrust/thrust/random/uniform_real_distribution.h
+++ b/thrust/thrust/random/uniform_real_distribution.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/pair.h>
 #include <thrust/random/detail/random_core_access.h>
 #include <iostream>
@@ -42,7 +48,7 @@ namespace random
  *
  *  \tparam RealType The type of floating point number to produce.
  *
- *  The following code snippet demonstrates examples of using a \p uniform_real_distribution with a 
+ *  The following code snippet demonstrates examples of using a \p uniform_real_distribution with a
  *  random number engine to produce random integers drawn from a given range:
  *
  *  \code
@@ -87,7 +93,7 @@ template<typename RealType = double>
 {
   public:
     // types
-    
+
     /*! \typedef result_type
      *  \brief The type of the floating point number produced by this \p uniform_real_distribution.
      */
@@ -99,10 +105,10 @@ template<typename RealType = double>
     typedef thrust::pair<RealType,RealType> param_type;
 
     // constructors and reset functions
-    
+
     /*! This constructor creates a new \p uniform_real_distribution from two values defining the
      *  half-open interval of the distribution.
-     *  
+     *
      *  \param a The smallest floating point number to potentially produce. Defaults to \c 0.0.
      *  \param b The smallest number larger than the largest floating point number to potentially produce. Defaults to \c 1.0.
      */
@@ -111,7 +117,7 @@ template<typename RealType = double>
 
     /*! This constructor creates a new \p uniform_real_distribution from a \p param_type object
      *  encapsulating the range of the distribution.
-     *  
+     *
      *  \param parm A \p param_type object encapsulating the parameters (i.e., the range) of the distribution.
      */
     __host__ __device__
@@ -133,7 +139,7 @@ template<typename RealType = double>
     __host__ __device__
     result_type operator()(UniformRandomNumberGenerator &urng);
 
-    /*! This method produces a new uniform random integer as if by creating a new \p uniform_real_distribution 
+    /*! This method produces a new uniform random integer as if by creating a new \p uniform_real_distribution
      *  from the given \p param_type object, and calling its <tt>operator()</tt> method with the given
      *  \p UniformRandomNumberGenerator as a source of randomness.
      *
diff --git a/thrust/thrust/random/xor_combine_engine.h b/thrust/thrust/random/xor_combine_engine.h
index 321f040333d..b560b85fcf4 100644
--- a/thrust/thrust/random/xor_combine_engine.h
+++ b/thrust/thrust/random/xor_combine_engine.h
@@ -23,6 +23,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/type_traits.h>
 #include <thrust/random/detail/xor_combine_engine_max.h>
 #include <thrust/random/detail/random_core_access.h>
@@ -92,7 +98,7 @@ template<typename Engine1, size_t s1,
       thrust::detail::identity_<typename base2_type::result_type>,
       thrust::detail::identity_<typename base1_type::result_type>
     >::type result_type;
-    
+
     /*! The size of the first shift used in the generation algorithm.
      */
     static const size_t shift1 = s1;
@@ -132,7 +138,7 @@ template<typename Engine1, size_t s1,
     xor_combine_engine(const base1_type &urng1, const base2_type &urng2);
 
     /*! This constructor initializes a new \p xor_combine_engine with a given seed.
-     *  
+     *
      *  \param s The seed used to intialize this \p xor_combine_engine's adapted base engines.
      */
     __host__ __device__
diff --git a/thrust/thrust/reduce.h b/thrust/thrust/reduce.h
index c7b378f7297..8c9cc5e5172 100644
--- a/thrust/thrust/reduce.h
+++ b/thrust/thrust/reduce.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/pair.h>
@@ -44,7 +50,7 @@ THRUST_NAMESPACE_BEGIN
  *
  *  Note that \p reduce also assumes that the binary reduction operator (in this
  *  case operator+) is commutative.  If the reduction operator is not commutative
- *  then \p thrust::reduce should not be used.  Instead, one could use 
+ *  then \p thrust::reduce should not be used.  Instead, one could use
  *  \p inclusive_scan (which does not require commutativity) and select the
  *  last element of the output array.
  *
@@ -94,7 +100,7 @@ __host__ __device__
  *
  *  Note that \p reduce also assumes that the binary reduction operator (in this
  *  case operator+) is commutative.  If the reduction operator is not commutative
- *  then \p thrust::reduce should not be used.  Instead, one could use 
+ *  then \p thrust::reduce should not be used.  Instead, one could use
  *  \p inclusive_scan (which does not require commutativity) and select the
  *  last element of the output array.
  *
@@ -137,7 +143,7 @@ template<typename InputIterator> typename
  *
  *  Note that \p reduce also assumes that the binary reduction operator (in this
  *  case operator+) is commutative.  If the reduction operator is not commutative
- *  then \p thrust::reduce should not be used.  Instead, one could use 
+ *  then \p thrust::reduce should not be used.  Instead, one could use
  *  \p inclusive_scan (which does not require commutativity) and select the
  *  last element of the output array.
  *
@@ -190,7 +196,7 @@ __host__ __device__
  *
  *  Note that \p reduce also assumes that the binary reduction operator (in this
  *  case operator+) is commutative.  If the reduction operator is not commutative
- *  then \p thrust::reduce should not be used.  Instead, one could use 
+ *  then \p thrust::reduce should not be used.  Instead, one could use
  *  \p inclusive_scan (which does not require commutativity) and select the
  *  last element of the output array.
  *
@@ -235,7 +241,7 @@ template<typename InputIterator, typename T>
  *
  *  Note that \p reduce also assumes that the binary reduction operator (in this
  *  case \p binary_op) is commutative.  If the reduction operator is not commutative
- *  then \p thrust::reduce should not be used.  Instead, one could use 
+ *  then \p thrust::reduce should not be used.  Instead, one could use
  *  \p inclusive_scan (which does not require commutativity) and select the
  *  last element of the output array.
  *
@@ -299,7 +305,7 @@ __host__ __device__
  *
  *  Note that \p reduce also assumes that the binary reduction operator (in this
  *  case \p binary_op) is commutative.  If the reduction operator is not commutative
- *  then \p thrust::reduce should not be used.  Instead, one could use 
+ *  then \p thrust::reduce should not be used.  Instead, one could use
  *  \p inclusive_scan (which does not require commutativity) and select the
  *  last element of the output array.
  *
@@ -346,7 +352,7 @@ template<typename InputIterator,
  *  For each group of consecutive keys in the range <tt>[keys_first, keys_last)</tt>
  *  that are equal, \p reduce_by_key copies the first element of the group to the
  *  \c keys_output. The corresponding values in the range are reduced using the
- *  \c plus and the result copied to \c values_output. 
+ *  \c plus and the result copied to \c values_output.
  *
  *  This version of \p reduce_by_key uses the function object \c equal_to
  *  to test for equality and \c plus to reduce values with equal keys.
@@ -391,7 +397,7 @@ template<typename InputIterator,
  *  // The first four keys in C are now {1, 3, 2, 1} and new_end.first - C is 4.
  *  // The first four values in D are now {9, 21, 9, 3} and new_end.second - D is 4.
  *  \endcode
- *  
+ *
  *  \see reduce
  *  \see unique_copy
  *  \see unique_by_key
@@ -405,7 +411,7 @@ template<typename DerivedPolicy,
 __host__ __device__
   thrust::pair<OutputIterator1,OutputIterator2>
   reduce_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
-                InputIterator1 keys_first, 
+                InputIterator1 keys_first,
                 InputIterator1 keys_last,
                 InputIterator2 values_first,
                 OutputIterator1 keys_output,
@@ -416,7 +422,7 @@ __host__ __device__
  *  For each group of consecutive keys in the range <tt>[keys_first, keys_last)</tt>
  *  that are equal, \p reduce_by_key copies the first element of the group to the
  *  \c keys_output. The corresponding values in the range are reduced using the
- *  \c plus and the result copied to \c values_output. 
+ *  \c plus and the result copied to \c values_output.
  *
  *  This version of \p reduce_by_key uses the function object \c equal_to
  *  to test for equality and \c plus to reduce values with equal keys.
@@ -455,7 +461,7 @@ __host__ __device__
  *  // The first four keys in C are now {1, 3, 2, 1} and new_end.first - C is 4.
  *  // The first four values in D are now {9, 21, 9, 3} and new_end.second - D is 4.
  *  \endcode
- *  
+ *
  *  \see reduce
  *  \see unique_copy
  *  \see unique_by_key
@@ -466,7 +472,7 @@ template<typename InputIterator1,
          typename OutputIterator1,
          typename OutputIterator2>
   thrust::pair<OutputIterator1,OutputIterator2>
-  reduce_by_key(InputIterator1 keys_first, 
+  reduce_by_key(InputIterator1 keys_first,
                 InputIterator1 keys_last,
                 InputIterator2 values_first,
                 OutputIterator1 keys_output,
@@ -477,7 +483,7 @@ template<typename InputIterator1,
  *  For each group of consecutive keys in the range <tt>[keys_first, keys_last)</tt>
  *  that are equal, \p reduce_by_key copies the first element of the group to the
  *  \c keys_output. The corresponding values in the range are reduced using the
- *  \c plus and the result copied to \c values_output. 
+ *  \c plus and the result copied to \c values_output.
  *
  *  This version of \p reduce_by_key uses the function object \c binary_pred
  *  to test for equality and \c plus to reduce values with equal keys.
@@ -525,7 +531,7 @@ template<typename InputIterator1,
  *  // The first four keys in C are now {1, 3, 2, 1} and new_end.first - C is 4.
  *  // The first four values in D are now {9, 21, 9, 3} and new_end.second - D is 4.
  *  \endcode
- *  
+ *
  *  \see reduce
  *  \see unique_copy
  *  \see unique_by_key
@@ -540,7 +546,7 @@ template<typename DerivedPolicy,
 __host__ __device__
   thrust::pair<OutputIterator1,OutputIterator2>
   reduce_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
-                InputIterator1 keys_first, 
+                InputIterator1 keys_first,
                 InputIterator1 keys_last,
                 InputIterator2 values_first,
                 OutputIterator1 keys_output,
@@ -552,7 +558,7 @@ __host__ __device__
  *  For each group of consecutive keys in the range <tt>[keys_first, keys_last)</tt>
  *  that are equal, \p reduce_by_key copies the first element of the group to the
  *  \c keys_output. The corresponding values in the range are reduced using the
- *  \c plus and the result copied to \c values_output. 
+ *  \c plus and the result copied to \c values_output.
  *
  *  This version of \p reduce_by_key uses the function object \c binary_pred
  *  to test for equality and \c plus to reduce values with equal keys.
@@ -594,7 +600,7 @@ __host__ __device__
  *  // The first four keys in C are now {1, 3, 2, 1} and new_end.first - C is 4.
  *  // The first four values in D are now {9, 21, 9, 3} and new_end.second - D is 4.
  *  \endcode
- *  
+ *
  *  \see reduce
  *  \see unique_copy
  *  \see unique_by_key
@@ -606,7 +612,7 @@ template<typename InputIterator1,
          typename OutputIterator2,
          typename BinaryPredicate>
   thrust::pair<OutputIterator1,OutputIterator2>
-  reduce_by_key(InputIterator1 keys_first, 
+  reduce_by_key(InputIterator1 keys_first,
                 InputIterator1 keys_last,
                 InputIterator2 values_first,
                 OutputIterator1 keys_output,
@@ -618,8 +624,8 @@ template<typename InputIterator1,
  *  For each group of consecutive keys in the range <tt>[keys_first, keys_last)</tt>
  *  that are equal, \p reduce_by_key copies the first element of the group to the
  *  \c keys_output. The corresponding values in the range are reduced using the
- *  \c BinaryFunction \c binary_op and the result copied to \c values_output. 
- *  Specifically, if consecutive key iterators \c i and \c (i + 1) are 
+ *  \c BinaryFunction \c binary_op and the result copied to \c values_output.
+ *  Specifically, if consecutive key iterators \c i and \c (i + 1) are
  *  such that <tt>binary_pred(*i, *(i+1))</tt> is \c true, then the corresponding
  *  values are reduced to a single value with \c binary_op.
  *
@@ -673,7 +679,7 @@ template<typename InputIterator1,
  *  // The first four keys in C are now {1, 3, 2, 1} and new_end.first - C is 4.
  *  // The first four values in D are now {9, 21, 9, 3} and new_end.second - D is 4.
  *  \endcode
- *  
+ *
  *  \see reduce
  *  \see unique_copy
  *  \see unique_by_key
@@ -689,7 +695,7 @@ template<typename DerivedPolicy,
 __host__ __device__
   thrust::pair<OutputIterator1,OutputIterator2>
   reduce_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
-                InputIterator1 keys_first, 
+                InputIterator1 keys_first,
                 InputIterator1 keys_last,
                 InputIterator2 values_first,
                 OutputIterator1 keys_output,
@@ -702,8 +708,8 @@ __host__ __device__
  *  For each group of consecutive keys in the range <tt>[keys_first, keys_last)</tt>
  *  that are equal, \p reduce_by_key copies the first element of the group to the
  *  \c keys_output. The corresponding values in the range are reduced using the
- *  \c BinaryFunction \c binary_op and the result copied to \c values_output. 
- *  Specifically, if consecutive key iterators \c i and \c (i + 1) are 
+ *  \c BinaryFunction \c binary_op and the result copied to \c values_output.
+ *  Specifically, if consecutive key iterators \c i and \c (i + 1) are
  *  such that <tt>binary_pred(*i, *(i+1))</tt> is \c true, then the corresponding
  *  values are reduced to a single value with \c binary_op.
  *
@@ -751,7 +757,7 @@ __host__ __device__
  *  // The first four keys in C are now {1, 3, 2, 1} and new_end.first - C is 4.
  *  // The first four values in D are now {9, 21, 9, 3} and new_end.second - D is 4.
  *  \endcode
- *  
+ *
  *  \see reduce
  *  \see unique_copy
  *  \see unique_by_key
@@ -764,7 +770,7 @@ template<typename InputIterator1,
          typename BinaryPredicate,
          typename BinaryFunction>
   thrust::pair<OutputIterator1,OutputIterator2>
-  reduce_by_key(InputIterator1 keys_first, 
+  reduce_by_key(InputIterator1 keys_first,
                 InputIterator1 keys_last,
                 InputIterator2 values_first,
                 OutputIterator1 keys_output,
diff --git a/thrust/thrust/remove.h b/thrust/thrust/remove.h
index a57fcf211cf..c571d16e1ff 100644
--- a/thrust/thrust/remove.h
+++ b/thrust/thrust/remove.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
@@ -549,7 +555,7 @@ template<typename InputIterator,
 /*! \p remove_if removes from the range <tt>[first, last)</tt> every element \p x
  *  such that <tt>pred(x)</tt> is \c true. That is, \p remove_if returns an
  *  iterator \c new_last such that the range <tt>[first, new_last)</tt> contains
- *  no elements for which \p pred of the corresponding stencil value is \c true. 
+ *  no elements for which \p pred of the corresponding stencil value is \c true.
  *  The iterators in the range <tt>[new_last,last)</tt> are all still dereferenceable,
  *  but the elements that they point to are unspecified. \p remove_if is stable,
  *  meaning that the relative order of elements that are not removed is unchanged.
@@ -615,7 +621,7 @@ __host__ __device__
 /*! \p remove_if removes from the range <tt>[first, last)</tt> every element \p x
  *  such that <tt>pred(x)</tt> is \c true. That is, \p remove_if returns an
  *  iterator \c new_last such that the range <tt>[first, new_last)</tt> contains
- *  no elements for which \p pred of the corresponding stencil value is \c true. 
+ *  no elements for which \p pred of the corresponding stencil value is \c true.
  *  The iterators in the range <tt>[new_last,last)</tt> are all still dereferenceable,
  *  but the elements that they point to are unspecified. \p remove_if is stable,
  *  meaning that the relative order of elements that are not removed is unchanged.
@@ -670,10 +676,10 @@ template<typename ForwardIterator,
 
 
 /*! \p remove_copy_if copies elements from the range <tt>[first,last)</tt> to a
- *  range beginning at \p result, except that elements for which \p pred of the 
- *  corresponding stencil value is \c true are not copied. The return value is 
+ *  range beginning at \p result, except that elements for which \p pred of the
+ *  corresponding stencil value is \c true are not copied. The return value is
  *  the end of the resulting range.  This operation is stable, meaning that the
- *  relative order of the elements that are copied is the same as the 
+ *  relative order of the elements that are copied is the same as the
  *  range <tt>[first,last)</tt>.
  *
  *  The algorithm's execution policy is parallelized as determined by \p exec.
@@ -737,10 +743,10 @@ __host__ __device__
 
 
 /*! \p remove_copy_if copies elements from the range <tt>[first,last)</tt> to a
- *  range beginning at \p result, except that elements for which \p pred of the 
- *  corresponding stencil value is \c true are not copied. The return value is 
+ *  range beginning at \p result, except that elements for which \p pred of the
+ *  corresponding stencil value is \c true are not copied. The return value is
  *  the end of the resulting range.  This operation is stable, meaning that the
- *  relative order of the elements that are copied is the same as the 
+ *  relative order of the elements that are copied is the same as the
  *  range <tt>[first,last)</tt>.
  *
  *  \param first The beginning of the range of interest.
diff --git a/thrust/thrust/replace.h b/thrust/thrust/replace.h
index a5c0320c423..4488f780691 100644
--- a/thrust/thrust/replace.h
+++ b/thrust/thrust/replace.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
@@ -64,7 +70,7 @@ THRUST_NAMESPACE_BEGIN
  *  #include <thrust/execution_policy.h>
  *
  *  ...
- *  
+ *
  *  thrust::device_vector<int> A(4);
  *  A[0] = 1;
  *  A[1] = 2;
@@ -114,7 +120,7 @@ __host__ __device__
  *  #include <thrust/device_vector.h>
  *
  *  ...
- *  
+ *
  *  thrust::device_vector<int> A(4);
  *  A[0] = 1;
  *  A[1] = 2;
@@ -176,7 +182,7 @@ template<typename ForwardIterator, typename T>
  *  };
  *
  *  ...
- *  
+ *
  *  thrust::device_vector<int> A(4);
  *  A[0] =  1;
  *  A[1] = -3;
@@ -237,7 +243,7 @@ __host__ __device__
  *  };
  *
  *  ...
- *  
+ *
  *  thrust::device_vector<int> A(4);
  *  A[0] =  1;
  *  A[1] = -3;
@@ -303,9 +309,9 @@ template<typename ForwardIterator, typename Predicate, typename T>
  *      return x < 0;
  *    }
  *  };
- *  
+ *
  *  ...
- *  
+ *
  *  thrust::device_vector<int> A(4);
  *  A[0] =  10;
  *  A[1] =  20;
@@ -373,9 +379,9 @@ __host__ __device__
  *      return x < 0;
  *    }
  *  };
- *  
+ *
  *  ...
- *  
+ *
  *  thrust::device_vector<int> A(4);
  *  A[0] =  10;
  *  A[1] =  20;
@@ -563,13 +569,13 @@ template<typename InputIterator, typename OutputIterator, typename T>
  *  };
  *
  *  ...
- *  
+ *
  *  thrust::device_vector<int> A(4);
  *  A[0] =  1;
  *  A[1] = -3;
  *  A[2] =  2;
  *  A[3] = -1;
- 
+
  *  thrust::device_vector<int> B(4);
  *  is_less_than_zero pred;
  *
@@ -630,13 +636,13 @@ __host__ __device__
  *  };
  *
  *  ...
- *  
+ *
  *  thrust::device_vector<int> A(4);
  *  A[0] =  1;
  *  A[1] = -3;
  *  A[2] =  2;
  *  A[3] = -1;
- 
+
  *  thrust::device_vector<int> B(4);
  *  is_less_than_zero pred;
  *
@@ -673,7 +679,7 @@ template<typename InputIterator, typename OutputIterator, typename Predicate, ty
  *  \param stencil The beginning of the stencil sequence.
  *  \param result The beginning of the sequence to copy to.
  *  \param pred The predicate to test on every value of the range <tt>[stencil, stencil + (last - first))</tt>.
- *  \param new_value The replacement value to assign when <tt>pred(*s)</tt> evaluates to \c true. 
+ *  \param new_value The replacement value to assign when <tt>pred(*s)</tt> evaluates to \c true.
  *  \return <tt>result + (last-first)</tt>
  *
  *  \tparam DerivedPolicy The name of the derived execution policy.
@@ -701,9 +707,9 @@ template<typename InputIterator, typename OutputIterator, typename Predicate, ty
  *      return x < 0;
  *    }
  *  };
- *  
+ *
  *  ...
- *  
+ *
  *  thrust::device_vector<int> A(4);
  *  A[0] =  10;
  *  A[1] =  20;
@@ -750,7 +756,7 @@ __host__ __device__
  *  \param stencil The beginning of the stencil sequence.
  *  \param result The beginning of the sequence to copy to.
  *  \param pred The predicate to test on every value of the range <tt>[stencil, stencil + (last - first))</tt>.
- *  \param new_value The replacement value to assign when <tt>pred(*s)</tt> evaluates to \c true. 
+ *  \param new_value The replacement value to assign when <tt>pred(*s)</tt> evaluates to \c true.
  *  \return <tt>result + (last-first)</tt>
  *
  *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
@@ -776,9 +782,9 @@ __host__ __device__
  *      return x < 0;
  *    }
  *  };
- *  
+ *
  *  ...
- *  
+ *
  *  thrust::device_vector<int> A(4);
  *  A[0] =  10;
  *  A[1] =  20;
diff --git a/thrust/thrust/reverse.h b/thrust/thrust/reverse.h
index 056be200adc..ac40b2dfa49 100644
--- a/thrust/thrust/reverse.h
+++ b/thrust/thrust/reverse.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
@@ -59,7 +65,7 @@ THRUST_NAMESPACE_BEGIN
  *  thrust::reverse(thrust::device, v.begin(), v.end());
  *  // v is now {5, 4, 3, 2, 1, 0}
  *  \endcode
- *  
+ *
  *  \see https://en.cppreference.com/w/cpp/algorithm/reverse
  *  \see \p reverse_copy
  *  \see \p reverse_iterator
@@ -93,7 +99,7 @@ __host__ __device__
  *  thrust::reverse(v.begin(), v.end());
  *  // v is now {5, 4, 3, 2, 1, 0}
  *  \endcode
- *  
+ *
  *  \see https://en.cppreference.com/w/cpp/algorithm/reverse
  *  \see \p reverse_copy
  *  \see \p reverse_iterator
@@ -107,7 +113,7 @@ template<typename BidirectionalIterator>
  *  is written to a different output range, rather than inplace.
  *
  *  \p reverse_copy copies elements from the range <tt>[first, last)</tt> to the
- *  range <tt>[result, result + (last - first))</tt> such that the copy is a 
+ *  range <tt>[result, result + (last - first))</tt> such that the copy is a
  *  reverse of the original range. Specifically: for every <tt>i</tt> such that
  *  <tt>0 <= i < (last - first)</tt>, \p reverse_copy performs the assignment
  *  <tt>*(result + (last - first) - i) = *(first + i)</tt>.
@@ -144,7 +150,7 @@ template<typename BidirectionalIterator>
  *  // input is still {0, 1, 2, 3, 4, 5}
  *  // output is now  {5, 4, 3, 2, 1, 0}
  *  \endcode
- *  
+ *
  *  \see https://en.cppreference.com/w/cpp/algorithm/reverse_copy
  *  \see \p reverse
  *  \see \p reverse_iterator
@@ -161,7 +167,7 @@ __host__ __device__
  *  is written to a different output range, rather than inplace.
  *
  *  \p reverse_copy copies elements from the range <tt>[first, last)</tt> to the
- *  range <tt>[result, result + (last - first))</tt> such that the copy is a 
+ *  range <tt>[result, result + (last - first))</tt> such that the copy is a
  *  reverse of the original range. Specifically: for every <tt>i</tt> such that
  *  <tt>0 <= i < (last - first)</tt>, \p reverse_copy performs the assignment
  *  <tt>*(result + (last - first) - i) = *(first + i)</tt>.
@@ -192,7 +198,7 @@ __host__ __device__
  *  // input is still {0, 1, 2, 3, 4, 5}
  *  // output is now  {5, 4, 3, 2, 1, 0}
  *  \endcode
- *  
+ *
  *  \see https://en.cppreference.com/w/cpp/algorithm/reverse_copy
  *  \see \p reverse
  *  \see \p reverse_iterator
diff --git a/thrust/thrust/scan.h b/thrust/thrust/scan.h
index 9b381422323..79d2a85621a 100644
--- a/thrust/thrust/scan.h
+++ b/thrust/thrust/scan.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
@@ -38,16 +44,16 @@ THRUST_NAMESPACE_BEGIN
 
 /*! \p inclusive_scan computes an inclusive prefix sum operation. The
  *  term 'inclusive' means that each result includes the corresponding
- *  input operand in the partial sum. More precisely, <tt>*first</tt> is 
- *  assigned to <tt>*result</tt> and the sum of <tt>*first</tt> and 
- *  <tt>*(first + 1)</tt> is assigned to <tt>*(result + 1)</tt>, and so on. 
- *  This version of \p inclusive_scan assumes plus as the associative operator.  
- *  When the input and output sequences are the same, the scan is performed 
+ *  input operand in the partial sum. More precisely, <tt>*first</tt> is
+ *  assigned to <tt>*result</tt> and the sum of <tt>*first</tt> and
+ *  <tt>*(first + 1)</tt> is assigned to <tt>*(result + 1)</tt>, and so on.
+ *  This version of \p inclusive_scan assumes plus as the associative operator.
+ *  When the input and output sequences are the same, the scan is performed
  *  in-place.
  *
  *  \p inclusive_scan is similar to \c std::partial_sum in the STL.  The primary
  *  difference between the two functions is that \c std::partial_sum guarantees
- *  a serial summation order, while \p inclusive_scan requires associativity of 
+ *  a serial summation order, while \p inclusive_scan requires associativity of
  *  the binary operation to parallelize the prefix sum.
  *
  *  Results are not deterministic for pseudo-associative operators (e.g.,
@@ -55,7 +61,7 @@ THRUST_NAMESPACE_BEGIN
  *  operators may vary from run to run.
  *
  *  The algorithm's execution is parallelized as determined by \p exec.
- *    
+ *
  *  \param exec The execution policy to use for parallelization.
  *  \param first The beginning of the input sequence.
  *  \param last The end of the input sequence.
@@ -81,7 +87,7 @@ THRUST_NAMESPACE_BEGIN
  *  #include <thrust/scan.h>
  *  #include <thrust/execution_policy.h>
  *  ...
- *  
+ *
  *  int data[6] = {1, 0, 2, 2, 1, 3};
  *
  *  thrust::inclusive_scan(thrust::host, data, data + 6, data); // in-place scan
@@ -104,16 +110,16 @@ __host__ __device__
 
 /*! \p inclusive_scan computes an inclusive prefix sum operation. The
  *  term 'inclusive' means that each result includes the corresponding
- *  input operand in the partial sum. More precisely, <tt>*first</tt> is 
- *  assigned to <tt>*result</tt> and the sum of <tt>*first</tt> and 
- *  <tt>*(first + 1)</tt> is assigned to <tt>*(result + 1)</tt>, and so on. 
- *  This version of \p inclusive_scan assumes plus as the associative operator.  
- *  When the input and output sequences are the same, the scan is performed 
+ *  input operand in the partial sum. More precisely, <tt>*first</tt> is
+ *  assigned to <tt>*result</tt> and the sum of <tt>*first</tt> and
+ *  <tt>*(first + 1)</tt> is assigned to <tt>*(result + 1)</tt>, and so on.
+ *  This version of \p inclusive_scan assumes plus as the associative operator.
+ *  When the input and output sequences are the same, the scan is performed
  *  in-place.
  *
  *  \p inclusive_scan is similar to \c std::partial_sum in the STL.  The primary
  *  difference between the two functions is that \c std::partial_sum guarantees
- *  a serial summation order, while \p inclusive_scan requires associativity of 
+ *  a serial summation order, while \p inclusive_scan requires associativity of
  *  the binary operation to parallelize the prefix sum.
  *
  *  Results are not deterministic for pseudo-associative operators (e.g.,
@@ -140,7 +146,7 @@ __host__ __device__
  *
  *  \code
  *  #include <thrust/scan.h>
- *  
+ *
  *  int data[6] = {1, 0, 2, 2, 1, 3};
  *
  *  thrust::inclusive_scan(data, data + 6, data); // in-place scan
@@ -160,12 +166,12 @@ template<typename InputIterator,
 
 /*! \p inclusive_scan computes an inclusive prefix sum operation. The
  *  term 'inclusive' means that each result includes the corresponding
- *  input operand in the partial sum.  When the input and output sequences 
+ *  input operand in the partial sum.  When the input and output sequences
  *  are the same, the scan is performed in-place.
  *
  *  \p inclusive_scan is similar to \c std::partial_sum in the STL.  The primary
  *  difference between the two functions is that \c std::partial_sum guarantees
- *  a serial summation order, while \p inclusive_scan requires associativity of 
+ *  a serial summation order, while \p inclusive_scan requires associativity of
  *  the binary operation to parallelize the prefix sum.
  *
  *  Results are not deterministic for pseudo-associative operators (e.g.,
@@ -200,7 +206,7 @@ template<typename InputIterator,
  *
  *  \code
  *  int data[10] = {-5, 0, 2, -3, 2, 4, 0, -1, 2, 8};
- * 
+ *
  *  thrust::maximum<int> binary_op;
  *
  *  thrust::inclusive_scan(thrust::host, data, data + 10, data, binary_op); // in-place scan
@@ -224,12 +230,12 @@ __host__ __device__
 
 /*! \p inclusive_scan computes an inclusive prefix sum operation. The
  *  term 'inclusive' means that each result includes the corresponding
- *  input operand in the partial sum.  When the input and output sequences 
+ *  input operand in the partial sum.  When the input and output sequences
  *  are the same, the scan is performed in-place.
- *    
+ *
  *  \p inclusive_scan is similar to \c std::partial_sum in the STL.  The primary
  *  difference between the two functions is that \c std::partial_sum guarantees
- *  a serial summation order, while \p inclusive_scan requires associativity of 
+ *  a serial summation order, while \p inclusive_scan requires associativity of
  *  the binary operation to parallelize the prefix sum.
  *
  *  Results are not deterministic for pseudo-associative operators (e.g.,
@@ -259,7 +265,7 @@ __host__ __device__
  *
  *  \code
  *  int data[10] = {-5, 0, 2, -3, 2, 4, 0, -1, 2, 8};
- * 
+ *
  *  thrust::maximum<int> binary_op;
  *
  *  thrust::inclusive_scan(data, data + 10, data, binary_op); // in-place scan
@@ -279,12 +285,12 @@ template<typename InputIterator,
 
 
 /*! \p exclusive_scan computes an exclusive prefix sum operation. The
- *  term 'exclusive' means that each result does not include the 
+ *  term 'exclusive' means that each result does not include the
  *  corresponding input operand in the partial sum.  More precisely,
- *  <tt>0</tt> is assigned to <tt>*result</tt> and the sum of 
+ *  <tt>0</tt> is assigned to <tt>*result</tt> and the sum of
  *  <tt>0</tt> and <tt>*first</tt> is assigned to <tt>*(result + 1)</tt>,
- *  and so on. This version of \p exclusive_scan assumes plus as the 
- *  associative operator and \c 0 as the initial value.  When the input and 
+ *  and so on. This version of \p exclusive_scan assumes plus as the
+ *  associative operator and \c 0 as the initial value.  When the input and
  *  output sequences are the same, the scan is performed in-place.
  *
  *  Results are not deterministic for pseudo-associative operators (e.g.,
@@ -292,7 +298,7 @@ template<typename InputIterator,
  *  operators may vary from run to run.
  *
  *  The algorithm's execution is parallelized as determined by \p exec.
- *    
+ *
  *  \param exec The execution policy to use for parallelization.
  *  \param first The beginning of the input sequence.
  *  \param last The end of the input sequence.
@@ -318,7 +324,7 @@ template<typename InputIterator,
  *  #include <thrust/scan.h>
  *  #include <thrust/execution_policy.h>
  *  ...
- *  
+ *
  *  int data[6] = {1, 0, 2, 2, 1, 3};
  *
  *  thrust::exclusive_scan(thrust::host, data, data + 6, data); // in-place scan
@@ -339,12 +345,12 @@ __host__ __device__
 
 
 /*! \p exclusive_scan computes an exclusive prefix sum operation. The
- *  term 'exclusive' means that each result does not include the 
+ *  term 'exclusive' means that each result does not include the
  *  corresponding input operand in the partial sum.  More precisely,
- *  <tt>0</tt> is assigned to <tt>*result</tt> and the sum of 
+ *  <tt>0</tt> is assigned to <tt>*result</tt> and the sum of
  *  <tt>0</tt> and <tt>*first</tt> is assigned to <tt>*(result + 1)</tt>,
- *  and so on. This version of \p exclusive_scan assumes plus as the 
- *  associative operator and \c 0 as the initial value.  When the input and 
+ *  and so on. This version of \p exclusive_scan assumes plus as the
+ *  associative operator and \c 0 as the initial value.  When the input and
  *  output sequences are the same, the scan is performed in-place.
  *
  *  Results are not deterministic for pseudo-associative operators (e.g.,
@@ -371,7 +377,7 @@ __host__ __device__
  *
  *  \code
  *  #include <thrust/scan.h>
- *  
+ *
  *  int data[6] = {1, 0, 2, 2, 1, 3};
  *
  *  thrust::exclusive_scan(data, data + 6, data); // in-place scan
@@ -389,12 +395,12 @@ template<typename InputIterator,
 
 
 /*! \p exclusive_scan computes an exclusive prefix sum operation. The
- *  term 'exclusive' means that each result does not include the 
+ *  term 'exclusive' means that each result does not include the
  *  corresponding input operand in the partial sum.  More precisely,
- *  \p init is assigned to <tt>*result</tt> and the sum of \p init and 
- *  <tt>*first</tt> is assigned to <tt>*(result + 1)</tt>, and so on. 
- *  This version of \p exclusive_scan assumes plus as the associative 
- *  operator but requires an initial value \p init.  When the input and 
+ *  \p init is assigned to <tt>*result</tt> and the sum of \p init and
+ *  <tt>*first</tt> is assigned to <tt>*(result + 1)</tt>, and so on.
+ *  This version of \p exclusive_scan assumes plus as the associative
+ *  operator but requires an initial value \p init.  When the input and
  *  output sequences are the same, the scan is performed in-place.
  *
  *  Results are not deterministic for pseudo-associative operators (e.g.,
@@ -427,7 +433,7 @@ template<typename InputIterator,
  *  \code
  *  #include <thrust/scan.h>
  *  #include <thrust/execution_policy.h>
- *  
+ *
  *  int data[6] = {1, 0, 2, 2, 1, 3};
  *
  *  thrust::exclusive_scan(thrust::host, data, data + 6, data, 4); // in-place scan
@@ -450,12 +456,12 @@ __host__ __device__
 
 
 /*! \p exclusive_scan computes an exclusive prefix sum operation. The
- *  term 'exclusive' means that each result does not include the 
+ *  term 'exclusive' means that each result does not include the
  *  corresponding input operand in the partial sum.  More precisely,
- *  \p init is assigned to <tt>*result</tt> and the sum of \p init and 
- *  <tt>*first</tt> is assigned to <tt>*(result + 1)</tt>, and so on. 
- *  This version of \p exclusive_scan assumes plus as the associative 
- *  operator but requires an initial value \p init.  When the input and 
+ *  \p init is assigned to <tt>*result</tt> and the sum of \p init and
+ *  <tt>*first</tt> is assigned to <tt>*(result + 1)</tt>, and so on.
+ *  This version of \p exclusive_scan assumes plus as the associative
+ *  operator but requires an initial value \p init.  When the input and
  *  output sequences are the same, the scan is performed in-place.
  *
  *  Results are not deterministic for pseudo-associative operators (e.g.,
@@ -482,7 +488,7 @@ __host__ __device__
  *
  *  \code
  *  #include <thrust/scan.h>
- *  
+ *
  *  int data[6] = {1, 0, 2, 2, 1, 3};
  *
  *  thrust::exclusive_scan(data, data + 6, data, 4); // in-place scan
@@ -502,11 +508,11 @@ template<typename InputIterator,
 
 
 /*! \p exclusive_scan computes an exclusive prefix sum operation. The
- *  term 'exclusive' means that each result does not include the 
+ *  term 'exclusive' means that each result does not include the
  *  corresponding input operand in the partial sum.  More precisely,
  *  \p init is assigned to <tt>\*result</tt> and the value
  *  <tt>binary_op(init, \*first)</tt> is assigned to <tt>\*(result + 1)</tt>,
- *  and so on. This version of the function requires both an associative 
+ *  and so on. This version of the function requires both an associative
  *  operator and an initial value \p init.  When the input and output
  *  sequences are the same, the scan is performed in-place.
  *
@@ -515,7 +521,7 @@ template<typename InputIterator,
  *  operators may vary from run to run.
  *
  *  The algorithm's execution is parallelized as determined by \p exec.
- *    
+ *
  *  \param exec The execution policy to use for parallelization.
  *  \param first The beginning of the input sequence.
  *  \param last The end of the input sequence.
@@ -547,16 +553,16 @@ template<typename InputIterator,
  *  #include <thrust/functional.h>
  *  #include <thrust/execution_policy.h>
  *  ...
- *  
+ *
  *  int data[10] = {-5, 0, 2, -3, 2, 4, 0, -1, 2, 8};
- * 
+ *
  *  thrust::maximum<int> binary_op;
  *
  *  thrust::exclusive_scan(thrust::host, data, data + 10, data, 1, binary_op); // in-place scan
  *
  *  // data is now {1, 1, 1, 2, 2, 2, 4, 4, 4, 4 }
  *  \endcode
- *  
+ *
  *  \see https://en.cppreference.com/w/cpp/algorithm/partial_sum
  */
 template<typename DerivedPolicy,
@@ -574,11 +580,11 @@ __host__ __device__
 
 
 /*! \p exclusive_scan computes an exclusive prefix sum operation. The
- *  term 'exclusive' means that each result does not include the 
+ *  term 'exclusive' means that each result does not include the
  *  corresponding input operand in the partial sum.  More precisely,
  *  \p init is assigned to <tt>\*result</tt> and the value
  *  <tt>binary_op(init, \*first)</tt> is assigned to <tt>\*(result + 1)</tt>,
- *  and so on. This version of the function requires both an associative 
+ *  and so on. This version of the function requires both an associative
  *  operator and an initial value \p init.  When the input and output
  *  sequences are the same, the scan is performed in-place.
  *
@@ -612,16 +618,16 @@ __host__ __device__
  *  \code
  *  #include <thrust/scan.h>
  *  #include <thrust/functional.h>
- *  
+ *
  *  int data[10] = {-5, 0, 2, -3, 2, 4, 0, -1, 2, 8};
- * 
+ *
  *  thrust::maximum<int> binary_op;
  *
  *  thrust::exclusive_scan(data, data + 10, data, 1, binary_op); // in-place scan
  *
  *  // data is now {1, 1, 1, 2, 2, 2, 4, 4, 4, 4 }
  *  \endcode
- *  
+ *
  *  \see https://en.cppreference.com/w/cpp/algorithm/partial_sum
  */
 template<typename InputIterator,
@@ -641,8 +647,8 @@ template<typename InputIterator,
  */
 
 
-/*! \p inclusive_scan_by_key computes an inclusive key-value or 'segmented' prefix 
- *  sum operation. The term 'inclusive' means that each result includes 
+/*! \p inclusive_scan_by_key computes an inclusive key-value or 'segmented' prefix
+ *  sum operation. The term 'inclusive' means that each result includes
  *  the corresponding input operand in the partial sum. The term 'segmented'
  *  means that the partial sums are broken into distinct segments.  In other
  *  words, within each segment a separate inclusive scan operation is computed.
@@ -676,7 +682,7 @@ template<typename InputIterator,
  *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
  *                         and \c InputIterator2's \c value_type is convertible to \c OutputIterator's \c value_type.
  *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>,
- *                         and if \c x and \c y are objects of \c OutputIterator's \c value_type, then 
+ *                         and if \c x and \c y are objects of \c OutputIterator's \c value_type, then
  *                         <tt>binary_op(x,y)</tt> is defined.
  *
  *  \pre \p first1 may equal \p result but the range <tt>[first1, last1)</tt> and the range <tt>[result, result + (last1 - first1))</tt> shall not overlap otherwise.
@@ -689,7 +695,7 @@ template<typename InputIterator,
  *  #include <thrust/scan.h>
  *  #include <thrust/execution_policy.h>
  *  ...
- *  
+ *
  *  int data[10] = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1};
  *  int keys[10] = {0, 0, 0, 1, 1, 2, 3, 3, 3, 3};
  *
@@ -712,10 +718,10 @@ __host__ __device__
                                        InputIterator1 last1,
                                        InputIterator2 first2,
                                        OutputIterator result);
- 
 
-/*! \p inclusive_scan_by_key computes an inclusive key-value or 'segmented' prefix 
- *  sum operation. The term 'inclusive' means that each result includes 
+
+/*! \p inclusive_scan_by_key computes an inclusive key-value or 'segmented' prefix
+ *  sum operation. The term 'inclusive' means that each result includes
  *  the corresponding input operand in the partial sum. The term 'segmented'
  *  means that the partial sums are broken into distinct segments.  In other
  *  words, within each segment a separate inclusive scan operation is computed.
@@ -745,7 +751,7 @@ __host__ __device__
  *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
  *                         and \c InputIterator2's \c value_type is convertible to \c OutputIterator's \c value_type.
  *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>,
- *                         and if \c x and \c y are objects of \c OutputIterator's \c value_type, then 
+ *                         and if \c x and \c y are objects of \c OutputIterator's \c value_type, then
  *                         <tt>binary_op(x,y)</tt> is defined.
  *
  *  \pre \p first1 may equal \p result but the range <tt>[first1, last1)</tt> and the range <tt>[result, result + (last1 - first1))</tt> shall not overlap otherwise.
@@ -755,7 +761,7 @@ __host__ __device__
  *
  *  \code
  *  #include <thrust/scan.h>
- *  
+ *
  *  int data[10] = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1};
  *  int keys[10] = {0, 0, 0, 1, 1, 2, 3, 3, 3, 3};
  *
@@ -777,17 +783,17 @@ template<typename InputIterator1,
                                        OutputIterator result);
 
 
-/*! \p inclusive_scan_by_key computes an inclusive key-value or 'segmented' prefix 
- *  sum operation. The term 'inclusive' means that each result includes 
+/*! \p inclusive_scan_by_key computes an inclusive key-value or 'segmented' prefix
+ *  sum operation. The term 'inclusive' means that each result includes
  *  the corresponding input operand in the partial sum. The term 'segmented'
  *  means that the partial sums are broken into distinct segments.  In other
  *  words, within each segment a separate inclusive scan operation is computed.
  *  Refer to the code sample below for example usage.
  *
- *  This version of \p inclusive_scan_by_key uses the binary predicate 
+ *  This version of \p inclusive_scan_by_key uses the binary predicate
  *  \c pred to compare adjacent keys.  Specifically, consecutive iterators
  *  <tt>i</tt> and <tt>i+1</tt> in the range <tt>[first1, last1)</tt>
- *  belong to the same segment if <tt>binary_pred(*i, *(i+1))</tt> is true, and belong to 
+ *  belong to the same segment if <tt>binary_pred(*i, *(i+1))</tt> is true, and belong to
  *  different segments otherwise.
  *
  *  This version of \p inclusive_scan_by_key assumes \c plus as the associative
@@ -798,7 +804,7 @@ template<typename InputIterator1,
  *  addition of floating-point types). Results for pseudo-associative
  *  operators may vary from run to run.
  *
- *  The algorithm's execution is parallelized as determined by \p exec. 
+ *  The algorithm's execution is parallelized as determined by \p exec.
  *
  *  \param exec The execution policy to use for parallelization.
  *  \param first1 The beginning of the key sequence.
@@ -813,7 +819,7 @@ template<typename InputIterator1,
  *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
  *                         and \c InputIterator2's \c value_type is convertible to \c OutputIterator's \c value_type.
  *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>,
- *                         and if \c x and \c y are objects of \c OutputIterator's \c value_type, then 
+ *                         and if \c x and \c y are objects of \c OutputIterator's \c value_type, then
  *                         <tt>binary_op(x,y)</tt> is defined.
  *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
  *
@@ -828,7 +834,7 @@ template<typename InputIterator1,
  *  #include <thrust/functional.h>
  *  #include <thrust/execution_policy.h>
  *  ...
- *  
+ *
  *  int data[10] = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1};
  *  int keys[10] = {0, 0, 0, 1, 1, 2, 3, 3, 3, 3};
  *
@@ -857,17 +863,17 @@ __host__ __device__
                                        BinaryPredicate binary_pred);
 
 
-/*! \p inclusive_scan_by_key computes an inclusive key-value or 'segmented' prefix 
- *  sum operation. The term 'inclusive' means that each result includes 
+/*! \p inclusive_scan_by_key computes an inclusive key-value or 'segmented' prefix
+ *  sum operation. The term 'inclusive' means that each result includes
  *  the corresponding input operand in the partial sum. The term 'segmented'
  *  means that the partial sums are broken into distinct segments.  In other
  *  words, within each segment a separate inclusive scan operation is computed.
  *  Refer to the code sample below for example usage.
  *
- *  This version of \p inclusive_scan_by_key uses the binary predicate 
+ *  This version of \p inclusive_scan_by_key uses the binary predicate
  *  \c pred to compare adjacent keys.  Specifically, consecutive iterators
  *  <tt>i</tt> and <tt>i+1</tt> in the range <tt>[first1, last1)</tt>
- *  belong to the same segment if <tt>binary_pred(*i, *(i+1))</tt> is true, and belong to 
+ *  belong to the same segment if <tt>binary_pred(*i, *(i+1))</tt> is true, and belong to
  *  different segments otherwise.
  *
  *  This version of \p inclusive_scan_by_key assumes \c plus as the associative
@@ -889,7 +895,7 @@ __host__ __device__
  *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
  *                         and \c InputIterator2's \c value_type is convertible to \c OutputIterator's \c value_type.
  *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>,
- *                         and if \c x and \c y are objects of \c OutputIterator's \c value_type, then 
+ *                         and if \c x and \c y are objects of \c OutputIterator's \c value_type, then
  *                         <tt>binary_op(x,y)</tt> is defined.
  *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
  *
@@ -901,7 +907,7 @@ __host__ __device__
  *  \code
  *  #include <thrust/scan.h>
  *  #include <thrust/functional.h>
- *  
+ *
  *  int data[10] = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1};
  *  int keys[10] = {0, 0, 0, 1, 1, 2, 3, 3, 3, 3};
  *
@@ -927,20 +933,20 @@ template<typename InputIterator1,
                                        BinaryPredicate binary_pred);
 
 
-/*! \p inclusive_scan_by_key computes an inclusive key-value or 'segmented' prefix 
- *  sum operation. The term 'inclusive' means that each result includes 
+/*! \p inclusive_scan_by_key computes an inclusive key-value or 'segmented' prefix
+ *  sum operation. The term 'inclusive' means that each result includes
  *  the corresponding input operand in the partial sum. The term 'segmented'
  *  means that the partial sums are broken into distinct segments.  In other
  *  words, within each segment a separate inclusive scan operation is computed.
  *  Refer to the code sample below for example usage.
  *
- *  This version of \p inclusive_scan_by_key uses the binary predicate 
+ *  This version of \p inclusive_scan_by_key uses the binary predicate
  *  \c pred to compare adjacent keys.  Specifically, consecutive iterators
  *  <tt>i</tt> and <tt>i+1</tt> in the range <tt>[first1, last1)</tt>
- *  belong to the same segment if <tt>binary_pred(*i, *(i+1))</tt> is true, and belong to 
+ *  belong to the same segment if <tt>binary_pred(*i, *(i+1))</tt> is true, and belong to
  *  different segments otherwise.
  *
- *  This version of \p inclusive_scan_by_key uses the associative operator 
+ *  This version of \p inclusive_scan_by_key uses the associative operator
  *  \c binary_op to perform the prefix sum. When the input and output sequences
  *  are the same, the scan is performed in-place.
  *
@@ -964,7 +970,7 @@ template<typename InputIterator1,
  *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
  *                         and \c InputIterator2's \c value_type is convertible to \c OutputIterator's \c value_type.
  *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>,
- *                         and if \c x and \c y are objects of \c OutputIterator's \c value_type, then 
+ *                         and if \c x and \c y are objects of \c OutputIterator's \c value_type, then
  *                         <tt>binary_op(x,y)</tt> is defined.
  *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
  *  \tparam AssociativeOperator is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/binary_function">Binary Function</a>
@@ -982,7 +988,7 @@ template<typename InputIterator1,
  *  #include <thrust/functional.h>
  *  #include <thrust/execution_policy.h>
  *  ...
- *  
+ *
  *  int data[10] = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1};
  *  int keys[10] = {0, 0, 0, 1, 1, 2, 3, 3, 3, 3};
  *
@@ -1014,24 +1020,24 @@ __host__ __device__
                                        AssociativeOperator binary_op);
 
 
-/*! \p inclusive_scan_by_key computes an inclusive key-value or 'segmented' prefix 
- *  sum operation. The term 'inclusive' means that each result includes 
+/*! \p inclusive_scan_by_key computes an inclusive key-value or 'segmented' prefix
+ *  sum operation. The term 'inclusive' means that each result includes
  *  the corresponding input operand in the partial sum. The term 'segmented'
  *  means that the partial sums are broken into distinct segments.  In other
  *  words, within each segment a separate inclusive scan operation is computed.
  *  Refer to the code sample below for example usage.
  *
- *  This version of \p inclusive_scan_by_key uses the binary predicate 
+ *  This version of \p inclusive_scan_by_key uses the binary predicate
  *  \c pred to compare adjacent keys.  Specifically, consecutive iterators
  *  <tt>i</tt> and <tt>i+1</tt> in the range <tt>[first1, last1)</tt>
- *  belong to the same segment if <tt>binary_pred(*i, *(i+1))</tt> is true, and belong to 
+ *  belong to the same segment if <tt>binary_pred(*i, *(i+1))</tt> is true, and belong to
  *  different segments otherwise.
  *
  *  Results are not deterministic for pseudo-associative operators (e.g.,
  *  addition of floating-point types). Results for pseudo-associative
  *  operators may vary from run to run.
  *
- *  This version of \p inclusive_scan_by_key uses the associative operator 
+ *  This version of \p inclusive_scan_by_key uses the associative operator
  *  \c binary_op to perform the prefix sum. When the input and output sequences
  *  are the same, the scan is performed in-place.
  *
@@ -1047,7 +1053,7 @@ __host__ __device__
  *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
  *                         and \c InputIterator2's \c value_type is convertible to \c OutputIterator's \c value_type.
  *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>,
- *                         and if \c x and \c y are objects of \c OutputIterator's \c value_type, then 
+ *                         and if \c x and \c y are objects of \c OutputIterator's \c value_type, then
  *                         <tt>binary_op(x,y)</tt> is defined.
  *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
  *  \tparam AssociativeOperator is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/binary_function">Binary Function</a>
@@ -1062,7 +1068,7 @@ __host__ __device__
  *  \code
  *  #include <thrust/scan.h>
  *  #include <thrust/functional.h>
- *  
+ *
  *  int data[10] = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1};
  *  int keys[10] = {0, 0, 0, 1, 1, 2, 3, 3, 3, 3};
  *
@@ -1091,7 +1097,7 @@ template<typename InputIterator1,
                                        AssociativeOperator binary_op);
 
 
-/*! \p exclusive_scan_by_key computes an exclusive segmented prefix 
+/*! \p exclusive_scan_by_key computes an exclusive segmented prefix
  *
  *  This version of \p exclusive_scan_by_key uses the value \c 0 to
  *  initialize the exclusive scan operation.
@@ -1099,11 +1105,11 @@ template<typename InputIterator1,
  *  This version of \p exclusive_scan_by_key assumes \c plus as the associative
  *  operator used to perform the prefix sum. When the input and output sequences
  *  are the same, the scan is performed in-place.
- * 
+ *
  *  This version of \p exclusive_scan_by_key assumes \c equal_to as the binary
  *  predicate used to compare adjacent keys.  Specifically, consecutive iterators
  *  <tt>i</tt> and <tt>i+1</tt> in the range <tt>[first1, last1</tt>
- *  belong to the same segment if <tt>*i == *(i+1)</tt>, and belong to 
+ *  belong to the same segment if <tt>*i == *(i+1)</tt>, and belong to
  *  different segments otherwise.
  *
  *  Results are not deterministic for pseudo-associative operators (e.g.,
@@ -1130,7 +1136,7 @@ template<typename InputIterator1,
  *  #include <thrust/scan.h>
  *  #include <thrust/execution_policy.h>
  *  ...
- *  
+ *
  *  int keys[10] = {0, 0, 0, 1, 1, 2, 3, 3, 3, 3};
  *  int vals[10] = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1};
  *
@@ -1154,7 +1160,7 @@ __host__ __device__
                                        OutputIterator result);
 
 
-/*! \p exclusive_scan_by_key computes an exclusive segmented prefix 
+/*! \p exclusive_scan_by_key computes an exclusive segmented prefix
  *
  *  This version of \p exclusive_scan_by_key uses the value \c 0 to
  *  initialize the exclusive scan operation.
@@ -1162,11 +1168,11 @@ __host__ __device__
  *  This version of \p exclusive_scan_by_key assumes \c plus as the associative
  *  operator used to perform the prefix sum. When the input and output sequences
  *  are the same, the scan is performed in-place.
- * 
+ *
  *  This version of \p exclusive_scan_by_key assumes \c equal_to as the binary
  *  predicate used to compare adjacent keys.  Specifically, consecutive iterators
  *  <tt>i</tt> and <tt>i+1</tt> in the range <tt>[first1, last1</tt>
- *  belong to the same segment if <tt>*i == *(i+1)</tt>, and belong to 
+ *  belong to the same segment if <tt>*i == *(i+1)</tt>, and belong to
  *  different segments otherwise.
  *
  *  Results are not deterministic for pseudo-associative operators (e.g.,
@@ -1187,7 +1193,7 @@ __host__ __device__
  *
  *  \code
  *  #include <thrust/scan.h>
- *  
+ *
  *  int keys[10] = {0, 0, 0, 1, 1, 2, 3, 3, 3, 3};
  *  int vals[10] = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1};
  *
@@ -1208,7 +1214,7 @@ template<typename InputIterator1,
                                        OutputIterator result);
 
 
-/*! \p exclusive_scan_by_key computes an exclusive key-value or 'segmented' prefix 
+/*! \p exclusive_scan_by_key computes an exclusive key-value or 'segmented' prefix
  *  sum operation. The term 'exclusive' means that each result does not include
  *  the corresponding input operand in the partial sum. The term 'segmented'
  *  means that the partial sums are broken into distinct segments.  In other
@@ -1243,7 +1249,7 @@ template<typename InputIterator1,
  *  #include <thrust/functional.h>
  *  #include <thrust/execution_policy.h>
  *  ...
- *  
+ *
  *  int keys[10] = {0, 0, 0, 1, 1, 2, 3, 3, 3, 3};
  *  int vals[10] = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1};
  *
@@ -1272,7 +1278,7 @@ __host__ __device__
                                        T init);
 
 
-/*! \p exclusive_scan_by_key computes an exclusive key-value or 'segmented' prefix 
+/*! \p exclusive_scan_by_key computes an exclusive key-value or 'segmented' prefix
  *  sum operation. The term 'exclusive' means that each result does not include
  *  the corresponding input operand in the partial sum. The term 'segmented'
  *  means that the partial sums are broken into distinct segments.  In other
@@ -1301,7 +1307,7 @@ __host__ __device__
  *  \code
  *  #include <thrust/scan.h>
  *  #include <thrust/functional.h>
- *  
+ *
  *  int keys[10] = {0, 0, 0, 1, 1, 2, 3, 3, 3, 3};
  *  int vals[10] = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1};
  *
@@ -1327,7 +1333,7 @@ template<typename InputIterator1,
                                        T init);
 
 
-/*! \p exclusive_scan_by_key computes an exclusive key-value or 'segmented' prefix 
+/*! \p exclusive_scan_by_key computes an exclusive key-value or 'segmented' prefix
  *  sum operation. The term 'exclusive' means that each result does not include
  *  the corresponding input operand in the partial sum. The term 'segmented'
  *  means that the partial sums are broken into distinct segments.  In other
@@ -1368,7 +1374,7 @@ template<typename InputIterator1,
  *  #include <thrust/functional.h>
  *  #include <thrust/execution_policy.h>
  *  ...
- *  
+ *
  *  int keys[10] = {0, 0, 0, 1, 1, 2, 3, 3, 3, 3};
  *  int vals[10] = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1};
  *
@@ -1401,7 +1407,7 @@ __host__ __device__
                                        BinaryPredicate binary_pred);
 
 
-/*! \p exclusive_scan_by_key computes an exclusive key-value or 'segmented' prefix 
+/*! \p exclusive_scan_by_key computes an exclusive key-value or 'segmented' prefix
  *  sum operation. The term 'exclusive' means that each result does not include
  *  the corresponding input operand in the partial sum. The term 'segmented'
  *  means that the partial sums are broken into distinct segments.  In other
@@ -1436,7 +1442,7 @@ __host__ __device__
  *  \code
  *  #include <thrust/scan.h>
  *  #include <thrust/functional.h>
- *  
+ *
  *  int keys[10] = {0, 0, 0, 1, 1, 2, 3, 3, 3, 3};
  *  int vals[10] = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1};
  *
@@ -1466,7 +1472,7 @@ template<typename InputIterator1,
                                        BinaryPredicate binary_pred);
 
 
-/*! \p exclusive_scan_by_key computes an exclusive key-value or 'segmented' prefix 
+/*! \p exclusive_scan_by_key computes an exclusive key-value or 'segmented' prefix
  *  sum operation. The term 'exclusive' means that each result does not include
  *  the corresponding input operand in the partial sum. The term 'segmented'
  *  means that the partial sums are broken into distinct segments.  In other
@@ -1478,10 +1484,10 @@ template<typename InputIterator1,
  *
  *  This version of \p exclusive_scan_by_key uses the binary predicate \c binary_pred
  *  to compare adjacent keys.  Specifically, consecutive iterators <tt>i</tt> and
- *  <tt>i+1</tt> in the range <tt>[first1, last1)</tt> belong to the same segment if 
+ *  <tt>i+1</tt> in the range <tt>[first1, last1)</tt> belong to the same segment if
  *  <tt>binary_pred(*i, *(i+1))</tt> is true, and belong to different segments otherwise.
  *
- *  This version of \p exclusive_scan_by_key uses the associative operator 
+ *  This version of \p exclusive_scan_by_key uses the associative operator
  *  \c binary_op to perform the prefix sum. When the input and output sequences
  *  are the same, the scan is performed in-place.
  *
@@ -1506,7 +1512,7 @@ template<typename InputIterator1,
  *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
  *                         and \c InputIterator2's \c value_type is convertible to \c OutputIterator's \c value_type.
  *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>,
- *                         and if \c x and \c y are objects of \c OutputIterator's \c value_type, then 
+ *                         and if \c x and \c y are objects of \c OutputIterator's \c value_type, then
  *                         <tt>binary_op(x,y)</tt> is defined.
  *  \tparam T is convertible to \c OutputIterator's \c value_type.
  *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
@@ -1524,7 +1530,7 @@ template<typename InputIterator1,
  *  #include <thrust/functional.h>
  *  #include <thrust/execution_policy.h>
  *  ...
- *  
+ *
  *  int keys[10] = {0, 0, 0, 1, 1, 2, 3, 3, 3, 3};
  *  int vals[10] = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1};
  *
@@ -1560,7 +1566,7 @@ __host__ __device__
                                        AssociativeOperator binary_op);
 
 
-/*! \p exclusive_scan_by_key computes an exclusive key-value or 'segmented' prefix 
+/*! \p exclusive_scan_by_key computes an exclusive key-value or 'segmented' prefix
  *  sum operation. The term 'exclusive' means that each result does not include
  *  the corresponding input operand in the partial sum. The term 'segmented'
  *  means that the partial sums are broken into distinct segments.  In other
@@ -1572,10 +1578,10 @@ __host__ __device__
  *
  *  This version of \p exclusive_scan_by_key uses the binary predicate \c binary_pred
  *  to compare adjacent keys.  Specifically, consecutive iterators <tt>i</tt> and
- *  <tt>i+1</tt> in the range <tt>[first1, last1)</tt> belong to the same segment if 
+ *  <tt>i+1</tt> in the range <tt>[first1, last1)</tt> belong to the same segment if
  *  <tt>binary_pred(*i, *(i+1))</tt> is true, and belong to different segments otherwise.
  *
- *  This version of \p exclusive_scan_by_key uses the associative operator 
+ *  This version of \p exclusive_scan_by_key uses the associative operator
  *  \c binary_op to perform the prefix sum. When the input and output sequences
  *  are the same, the scan is performed in-place.
  *
@@ -1596,7 +1602,7 @@ __host__ __device__
  *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
  *                         and \c InputIterator2's \c value_type is convertible to \c OutputIterator's \c value_type.
  *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>,
- *                         and if \c x and \c y are objects of \c OutputIterator's \c value_type, then 
+ *                         and if \c x and \c y are objects of \c OutputIterator's \c value_type, then
  *                         <tt>binary_op(x,y)</tt> is defined.
  *  \tparam T is convertible to \c OutputIterator's \c value_type.
  *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
@@ -1611,7 +1617,7 @@ __host__ __device__
  *  \code
  *  #include <thrust/scan.h>
  *  #include <thrust/functional.h>
- *  
+ *
  *  int keys[10] = {0, 0, 0, 1, 1, 2, 3, 3, 3, 3};
  *  int vals[10] = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1};
  *
diff --git a/thrust/thrust/scatter.h b/thrust/thrust/scatter.h
index b8b0bd84f52..40a5e45264c 100644
--- a/thrust/thrust/scatter.h
+++ b/thrust/thrust/scatter.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
@@ -34,8 +40,8 @@ THRUST_NAMESPACE_BEGIN
 
 /*! \p scatter copies elements from a source range into an output array
  *  according to a map. For each iterator \c i in the range [\p first, \p last),
- *  the value \c *i is assigned to <tt>output[*(map + (i - first))]</tt>. The 
- *  output iterator must permit random access. If the same index 
+ *  the value \c *i is assigned to <tt>output[*(map + (i - first))]</tt>. The
+ *  output iterator must permit random access. If the same index
  *  appears more than once in the range <tt>[map, map + (last - first))</tt>,
  *  the result is undefined.
  *
@@ -98,8 +104,8 @@ __host__ __device__
 
 /*! \p scatter copies elements from a source range into an output array
  *  according to a map. For each iterator \c i in the range [\p first, \p last),
- *  the value \c *i is assigned to <tt>output[*(map + (i - first))]</tt>. The 
- *  output iterator must permit random access. If the same index 
+ *  the value \c *i is assigned to <tt>output[*(map + (i - first))]</tt>. The
+ *  output iterator must permit random access. If the same index
  *  appears more than once in the range <tt>[map, map + (last - first))</tt>,
  *  the result is undefined.
  *
@@ -151,11 +157,11 @@ template<typename InputIterator1,
                RandomAccessIterator result);
 
 
-/*! \p scatter_if conditionally copies elements from a source range into an 
- *  output array according to a map. For each iterator \c i in the 
+/*! \p scatter_if conditionally copies elements from a source range into an
+ *  output array according to a map. For each iterator \c i in the
  *  range <tt>[first, last)</tt> such that <tt>*(stencil + (i - first))</tt> is
  *  true, the value \c *i is assigned to <tt>output[*(map + (i - first))]</tt>.
- *  The output iterator must permit random access. If the same index 
+ *  The output iterator must permit random access. If the same index
  *  appears more than once in the range <tt>[map, map + (last - first))</tt>
  *  the result is undefined.
  *
@@ -190,9 +196,9 @@ template<typename InputIterator1,
  *  int M[8] = {0, 5, 1, 6, 2, 7, 3, 4};
  *  int S[8] = {1, 0, 1, 0, 1, 0, 1, 0};
  *  int D[8] = {0, 0, 0, 0, 0, 0, 0, 0};
- * 
+ *
  *  thrust::scatter_if(thrust::host, V, V + 8, M, S, D);
- * 
+ *
  *  // D contains [10, 30, 50, 70, 0, 0, 0, 0];
  *  \endcode
  *
@@ -212,11 +218,11 @@ __host__ __device__
                   RandomAccessIterator output);
 
 
-/*! \p scatter_if conditionally copies elements from a source range into an 
- *  output array according to a map. For each iterator \c i in the 
+/*! \p scatter_if conditionally copies elements from a source range into an
+ *  output array according to a map. For each iterator \c i in the
  *  range <tt>[first, last)</tt> such that <tt>*(stencil + (i - first))</tt> is
  *  true, the value \c *i is assigned to <tt>output[*(map + (i - first))]</tt>.
- *  The output iterator must permit random access. If the same index 
+ *  The output iterator must permit random access. If the same index
  *  appears more than once in the range <tt>[map, map + (last - first))</tt>
  *  the result is undefined.
  *
@@ -246,9 +252,9 @@ __host__ __device__
  *  int M[8] = {0, 5, 1, 6, 2, 7, 3, 4};
  *  int S[8] = {1, 0, 1, 0, 1, 0, 1, 0};
  *  int D[8] = {0, 0, 0, 0, 0, 0, 0, 0};
- * 
+ *
  *  thrust::scatter_if(V, V + 8, M, S, D);
- * 
+ *
  *  // D contains [10, 30, 50, 70, 0, 0, 0, 0];
  *  \endcode
  *
@@ -265,11 +271,11 @@ template<typename InputIterator1,
                   RandomAccessIterator output);
 
 
-/*! \p scatter_if conditionally copies elements from a source range into an 
- *  output array according to a map. For each iterator \c i in the 
+/*! \p scatter_if conditionally copies elements from a source range into an
+ *  output array according to a map. For each iterator \c i in the
  *  range <tt>[first, last)</tt> such that <tt>pred(*(stencil + (i - first)))</tt> is
  *  \c true, the value \c *i is assigned to <tt>output[*(map + (i - first))]</tt>.
- *  The output iterator must permit random access. If the same index 
+ *  The output iterator must permit random access. If the same index
  *  appears more than once in the range <tt>[map, map + (last - first))</tt>
  *  the result is undefined.
  *
@@ -317,13 +323,13 @@ template<typename InputIterator1,
  *  int M[8] = {0, 5, 1, 6, 2, 7, 3, 4};
  *  int S[8] = {2, 1, 2, 1, 2, 1, 2, 1};
  *  int D[8] = {0, 0, 0, 0, 0, 0, 0, 0};
- * 
+ *
  *  is_even pred;
  *  thrust::scatter_if(thrust::host, V, V + 8, M, S, D, pred);
- * 
+ *
  *  // D contains [10, 30, 50, 70, 0, 0, 0, 0];
  *  \endcode
- *  
+ *
  *  \note \p scatter_if is the inverse of thrust::gather_if.
  */
 template<typename DerivedPolicy,
@@ -340,13 +346,13 @@ __host__ __device__
                   InputIterator3 stencil,
                   RandomAccessIterator output,
                   Predicate pred);
-                  
 
-/*! \p scatter_if conditionally copies elements from a source range into an 
- *  output array according to a map. For each iterator \c i in the 
+
+/*! \p scatter_if conditionally copies elements from a source range into an
+ *  output array according to a map. For each iterator \c i in the
  *  range <tt>[first, last)</tt> such that <tt>pred(*(stencil + (i - first)))</tt> is
  *  \c true, the value \c *i is assigned to <tt>output[*(map + (i - first))]</tt>.
- *  The output iterator must permit random access. If the same index 
+ *  The output iterator must permit random access. If the same index
  *  appears more than once in the range <tt>[map, map + (last - first))</tt>
  *  the result is undefined.
  *
@@ -389,13 +395,13 @@ __host__ __device__
  *  int M[8] = {0, 5, 1, 6, 2, 7, 3, 4};
  *  int S[8] = {2, 1, 2, 1, 2, 1, 2, 1};
  *  int D[8] = {0, 0, 0, 0, 0, 0, 0, 0};
- * 
+ *
  *  is_even pred;
  *  thrust::scatter_if(V, V + 8, M, S, D, pred);
- * 
+ *
  *  // D contains [10, 30, 50, 70, 0, 0, 0, 0];
  *  \endcode
- *  
+ *
  *  \note \p scatter_if is the inverse of thrust::gather_if.
  */
 template<typename InputIterator1,
diff --git a/thrust/thrust/sequence.h b/thrust/thrust/sequence.h
index fb3959e3cab..fe5f6f64396 100644
--- a/thrust/thrust/sequence.h
+++ b/thrust/thrust/sequence.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
@@ -33,7 +39,7 @@ THRUST_NAMESPACE_BEGIN
 
 /*! \p sequence fills the range <tt>[first, last)</tt> with a sequence of numbers.
  *
- *  For each iterator \c i in the range <tt>[first, last)</tt>, this version of 
+ *  For each iterator \c i in the range <tt>[first, last)</tt>, this version of
  *  \p sequence performs the assignment <tt>*i =  (i - first)</tt>.
  *
  *  The algorithm's execution is parallelized as determined by \p exec.
@@ -75,7 +81,7 @@ __host__ __device__
 
 /*! \p sequence fills the range <tt>[first, last)</tt> with a sequence of numbers.
  *
- *  For each iterator \c i in the range <tt>[first, last)</tt>, this version of 
+ *  For each iterator \c i in the range <tt>[first, last)</tt>, this version of
  *  \p sequence performs the assignment <tt>*i =  (i - first)</tt>.
  *
  *  \param first The beginning of the sequence.
@@ -110,7 +116,7 @@ template<typename ForwardIterator>
 
 /*! \p sequence fills the range <tt>[first, last)</tt> with a sequence of numbers.
  *
- *  For each iterator \c i in the range <tt>[first, last)</tt>, this version of 
+ *  For each iterator \c i in the range <tt>[first, last)</tt>, this version of
  *  \p sequence performs the assignment <tt>*i =  init + (i - first)</tt>.
  *
  *  The algorithm's execution is parallelized as determined by \p exec.
@@ -157,7 +163,7 @@ __host__ __device__
 
 /*! \p sequence fills the range <tt>[first, last)</tt> with a sequence of numbers.
  *
- *  For each iterator \c i in the range <tt>[first, last)</tt>, this version of 
+ *  For each iterator \c i in the range <tt>[first, last)</tt>, this version of
  *  \p sequence performs the assignment <tt>*i =  init + (i - first)</tt>.
  *
  *  \param first The beginning of the sequence.
@@ -196,7 +202,7 @@ template<typename ForwardIterator, typename T>
 
 /*! \p sequence fills the range <tt>[first, last)</tt> with a sequence of numbers.
  *
- *  For each iterator \c i in the range <tt>[first, last)</tt>, this version of 
+ *  For each iterator \c i in the range <tt>[first, last)</tt>, this version of
  *  \p sequence performs the assignment <tt>*i =  init + step * (i - first)</tt>.
  *
  *  The algorithm's execution is parallelized as determined by \p exec.
@@ -245,7 +251,7 @@ __host__ __device__
 
 /*! \p sequence fills the range <tt>[first, last)</tt> with a sequence of numbers.
  *
- *  For each iterator \c i in the range <tt>[first, last)</tt>, this version of 
+ *  For each iterator \c i in the range <tt>[first, last)</tt>, this version of
  *  \p sequence performs the assignment <tt>*i =  init + step * (i - first)</tt>.
  *
  *  \param first The beginning of the sequence.
diff --git a/thrust/thrust/set_operations.h b/thrust/thrust/set_operations.h
index 65a48d1b661..223907a4605 100644
--- a/thrust/thrust/set_operations.h
+++ b/thrust/thrust/set_operations.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 #include <thrust/pair.h>
 
@@ -1771,7 +1777,7 @@ template<typename InputIterator1,
  *  ...
  *  int A_keys[6] = {1, 3, 5, 7, 9, 11};
  *  int A_vals[6] = {0, 0, 0, 0, 0,  0};
- *  
+ *
  *  int B_keys[7] = {1, 1, 2, 3, 5,  8, 13};
  *
  *  int keys_result[7];
@@ -1869,7 +1875,7 @@ __host__ __device__
  *  ...
  *  int A_keys[6] = {1, 3, 5, 7, 9, 11};
  *  int A_vals[6] = {0, 0, 0, 0, 0,  0};
- *  
+ *
  *  int B_keys[7] = {1, 1, 2, 3, 5,  8, 13};
  *
  *  int keys_result[7];
@@ -1973,7 +1979,7 @@ template<typename InputIterator1,
  *  ...
  *  int A_keys[6] = {11, 9, 7, 5, 3, 1};
  *  int A_vals[6] = { 0, 0, 0, 0, 0, 0};
- *  
+ *
  *  int B_keys[7] = {13, 8, 5, 3, 2, 1, 1};
  *
  *  int keys_result[7];
@@ -2076,7 +2082,7 @@ __host__ __device__
  *  ...
  *  int A_keys[6] = {11, 9, 7, 5, 3, 1};
  *  int A_vals[6] = { 0, 0, 0, 0, 0, 0};
- *  
+ *
  *  int B_keys[7] = {13, 8, 5, 3, 2, 1, 1};
  *
  *  int keys_result[7];
diff --git a/thrust/thrust/shuffle.h b/thrust/thrust/shuffle.h
index d95327e29ce..6986d0fdfe8 100644
--- a/thrust/thrust/shuffle.h
+++ b/thrust/thrust/shuffle.h
@@ -21,11 +21,17 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/cpp11_required.h>
 
 #if THRUST_CPP_DIALECT >= 2011
 
-#include <thrust/detail/config.h>
 #include <thrust/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/sort.h b/thrust/thrust/sort.h
index 5cf9d6217ab..8c4f0d49b94 100644
--- a/thrust/thrust/sort.h
+++ b/thrust/thrust/sort.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
@@ -1055,7 +1061,7 @@ template<typename ForwardIterator>
                  ForwardIterator last);
 
 
-/*! \p is_sorted returns \c true if the range <tt>[first, last)</tt> is sorted in ascending 
+/*! \p is_sorted returns \c true if the range <tt>[first, last)</tt> is sorted in ascending
  *  order accoring to a user-defined comparison operation, and \c false otherwise.
  *
  *  Specifically, this version of \p is_sorted returns \c false if for some iterator \c i in
@@ -1117,7 +1123,7 @@ __host__ __device__
                  Compare comp);
 
 
-/*! \p is_sorted returns \c true if the range <tt>[first, last)</tt> is sorted in ascending 
+/*! \p is_sorted returns \c true if the range <tt>[first, last)</tt> is sorted in ascending
  *  order accoring to a user-defined comparison operation, and \c false otherwise.
  *
  *  Specifically, this version of \p is_sorted returns \c false if for some iterator \c i in
@@ -1189,17 +1195,17 @@ template<typename ForwardIterator, typename Compare>
  *  The following code snippet demonstrates how to use \p is_sorted_until to find the first position
  *  in an array where the data becomes unsorted using the \p thrust::host execution policy for
  *  parallelization:
- *  
+ *
  *  \code
  *  #include <thrust/sort.h>
  *  #include <thrust/execution_policy.h>
  *
  *  ...
- *   
+ *
  *  int A[8] = {0, 1, 2, 3, 0, 1, 2, 3};
- *  
+ *
  *  int * B = thrust::is_sorted_until(thrust::host, A, A + 8);
- *  
+ *
  *  // B - A is 4
  *  // [A, B) is sorted
  *  \endcode
@@ -1230,16 +1236,16 @@ __host__ __device__
  *
  *  The following code snippet demonstrates how to use \p is_sorted_until to find the first position
  *  in an array where the data becomes unsorted:
- *  
+ *
  *  \code
  *  #include <thrust/sort.h>
  *
  *  ...
- *   
+ *
  *  int A[8] = {0, 1, 2, 3, 0, 1, 2, 3};
- *  
+ *
  *  int * B = thrust::is_sorted_until(A, A + 8);
- *  
+ *
  *  // B - A is 4
  *  // [A, B) is sorted
  *  \endcode
@@ -1282,12 +1288,12 @@ template<typename ForwardIterator>
  *  #include <thrust/execution_policy.h>
  *
  *  ...
- *   
+ *
  *  int A[8] = {3, 2, 1, 0, 3, 2, 1, 0};
- *  
+ *
  *  thrust::greater<int> comp;
  *  int * B = thrust::is_sorted_until(thrust::host, A, A + 8, comp);
- *  
+ *
  *  // B - A is 4
  *  // [A, B) is sorted in descending order
  *  \endcode
@@ -1327,12 +1333,12 @@ __host__ __device__
  *  #include <thrust/functional.h>
  *
  *  ...
- *   
+ *
  *  int A[8] = {3, 2, 1, 0, 3, 2, 1, 0};
- *  
+ *
  *  thrust::greater<int> comp;
  *  int * B = thrust::is_sorted_until(A, A + 8, comp);
- *  
+ *
  *  // B - A is 4
  *  // [A, B) is sorted in descending order
  *  \endcode
diff --git a/thrust/thrust/swap.h b/thrust/thrust/swap.h
index d8a8be73c31..7716458b004 100644
--- a/thrust/thrust/swap.h
+++ b/thrust/thrust/swap.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
@@ -36,7 +42,7 @@ THRUST_NAMESPACE_BEGIN
 /*! \p swap assigns the contents of \c a to \c b and the
  *  contents of \c b to \c a. This is used as a primitive operation
  *  by many other algorithms.
- *  
+ *
  *  \param a The first value of interest. After completion,
  *           the value of b will be returned here.
  *  \param b The second value of interest. After completion,
@@ -58,7 +64,7 @@ THRUST_NAMESPACE_BEGIN
  *  \endcode
  */
 template<typename Assignable1, typename Assignable2>
-__host__ __device__ 
+__host__ __device__
 inline void swap(Assignable1 &a, Assignable2 &b);
 
 /*! \} // swap
diff --git a/thrust/thrust/system/cpp/detail/adjacent_difference.h b/thrust/thrust/system/cpp/detail/adjacent_difference.h
index b82242c7c07..dc050ffb801 100644
--- a/thrust/thrust/system/cpp/detail/adjacent_difference.h
+++ b/thrust/thrust/system/cpp/detail/adjacent_difference.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits adjacent_difference
 #include <thrust/system/detail/sequential/adjacent_difference.h>
 
diff --git a/thrust/thrust/system/cpp/detail/assign_value.h b/thrust/thrust/system/cpp/detail/assign_value.h
index d5f14bd1636..6596d8a9aae 100644
--- a/thrust/thrust/system/cpp/detail/assign_value.h
+++ b/thrust/thrust/system/cpp/detail/assign_value.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits assign_value
 #include <thrust/system/detail/sequential/assign_value.h>
 
diff --git a/thrust/thrust/system/cpp/detail/binary_search.h b/thrust/thrust/system/cpp/detail/binary_search.h
index a2c33f32aad..10129190061 100644
--- a/thrust/thrust/system/cpp/detail/binary_search.h
+++ b/thrust/thrust/system/cpp/detail/binary_search.h
@@ -16,6 +16,14 @@
 
 #pragma once
 
+#include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/system/cpp/detail/execution_policy.h>
 
 // this system inherits the binary search algorithms
diff --git a/thrust/thrust/system/cpp/detail/copy.h b/thrust/thrust/system/cpp/detail/copy.h
index d2f5b1bd45c..6edd9af7755 100644
--- a/thrust/thrust/system/cpp/detail/copy.h
+++ b/thrust/thrust/system/cpp/detail/copy.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits copy
 #include <thrust/system/detail/sequential/copy.h>
 
diff --git a/thrust/thrust/system/cpp/detail/copy_if.h b/thrust/thrust/system/cpp/detail/copy_if.h
index 10869e2a902..3da9e682187 100644
--- a/thrust/thrust/system/cpp/detail/copy_if.h
+++ b/thrust/thrust/system/cpp/detail/copy_if.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits copy_if
 #include <thrust/system/detail/sequential/copy_if.h>
 
diff --git a/thrust/thrust/system/cpp/detail/count.h b/thrust/thrust/system/cpp/detail/count.h
index c6ae90664ad..c3f89646b4e 100644
--- a/thrust/thrust/system/cpp/detail/count.h
+++ b/thrust/thrust/system/cpp/detail/count.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
-// this system has no special version of this algorithm 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
+// this system has no special version of this algorithm
 
diff --git a/thrust/thrust/system/cpp/detail/equal.h b/thrust/thrust/system/cpp/detail/equal.h
index c6ae90664ad..c3f89646b4e 100644
--- a/thrust/thrust/system/cpp/detail/equal.h
+++ b/thrust/thrust/system/cpp/detail/equal.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
-// this system has no special version of this algorithm 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
+// this system has no special version of this algorithm
 
diff --git a/thrust/thrust/system/cpp/detail/execution_policy.h b/thrust/thrust/system/cpp/detail/execution_policy.h
index 1a8193bf3ac..39d19c3b222 100644
--- a/thrust/thrust/system/cpp/detail/execution_policy.h
+++ b/thrust/thrust/system/cpp/detail/execution_policy.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/sequential/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
@@ -55,7 +61,7 @@ template<typename Derived>
   struct execution_policy
     : thrust::system::detail::sequential::execution_policy<Derived>
 {
-  typedef tag tag_type; 
+  typedef tag tag_type;
   operator tag() const { return tag(); }
 };
 
diff --git a/thrust/thrust/system/cpp/detail/extrema.h b/thrust/thrust/system/cpp/detail/extrema.h
index 5fbb8c55c28..a5aa41669f2 100644
--- a/thrust/thrust/system/cpp/detail/extrema.h
+++ b/thrust/thrust/system/cpp/detail/extrema.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits extrema algorithms
 #include <thrust/system/detail/sequential/extrema.h>
 
diff --git a/thrust/thrust/system/cpp/detail/fill.h b/thrust/thrust/system/cpp/detail/fill.h
index c6ae90664ad..c3f89646b4e 100644
--- a/thrust/thrust/system/cpp/detail/fill.h
+++ b/thrust/thrust/system/cpp/detail/fill.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
-// this system has no special version of this algorithm 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
+// this system has no special version of this algorithm
 
diff --git a/thrust/thrust/system/cpp/detail/find.h b/thrust/thrust/system/cpp/detail/find.h
index 29c0dafc8ce..f246e9be742 100644
--- a/thrust/thrust/system/cpp/detail/find.h
+++ b/thrust/thrust/system/cpp/detail/find.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits find
 #include <thrust/system/detail/sequential/find.h>
 
diff --git a/thrust/thrust/system/cpp/detail/for_each.h b/thrust/thrust/system/cpp/detail/for_each.h
index 9ef45dfe62e..505ffd3285f 100644
--- a/thrust/thrust/system/cpp/detail/for_each.h
+++ b/thrust/thrust/system/cpp/detail/for_each.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits for_each
 #include <thrust/system/detail/sequential/for_each.h>
 
diff --git a/thrust/thrust/system/cpp/detail/gather.h b/thrust/thrust/system/cpp/detail/gather.h
index c6ae90664ad..c3f89646b4e 100644
--- a/thrust/thrust/system/cpp/detail/gather.h
+++ b/thrust/thrust/system/cpp/detail/gather.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
-// this system has no special version of this algorithm 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
+// this system has no special version of this algorithm
 
diff --git a/thrust/thrust/system/cpp/detail/generate.h b/thrust/thrust/system/cpp/detail/generate.h
index c6ae90664ad..c3f89646b4e 100644
--- a/thrust/thrust/system/cpp/detail/generate.h
+++ b/thrust/thrust/system/cpp/detail/generate.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
-// this system has no special version of this algorithm 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
+// this system has no special version of this algorithm
 
diff --git a/thrust/thrust/system/cpp/detail/get_value.h b/thrust/thrust/system/cpp/detail/get_value.h
index 915001d37f4..a39301e3988 100644
--- a/thrust/thrust/system/cpp/detail/get_value.h
+++ b/thrust/thrust/system/cpp/detail/get_value.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits get_value
 #include <thrust/system/detail/sequential/get_value.h>
 
diff --git a/thrust/thrust/system/cpp/detail/inner_product.h b/thrust/thrust/system/cpp/detail/inner_product.h
index c6ae90664ad..c3f89646b4e 100644
--- a/thrust/thrust/system/cpp/detail/inner_product.h
+++ b/thrust/thrust/system/cpp/detail/inner_product.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
-// this system has no special version of this algorithm 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
+// this system has no special version of this algorithm
 
diff --git a/thrust/thrust/system/cpp/detail/iter_swap.h b/thrust/thrust/system/cpp/detail/iter_swap.h
index 7a982becb39..3e419612d4a 100644
--- a/thrust/thrust/system/cpp/detail/iter_swap.h
+++ b/thrust/thrust/system/cpp/detail/iter_swap.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits iter_swap
 #include <thrust/system/detail/sequential/iter_swap.h>
 
diff --git a/thrust/thrust/system/cpp/detail/logical.h b/thrust/thrust/system/cpp/detail/logical.h
index c6ae90664ad..c3f89646b4e 100644
--- a/thrust/thrust/system/cpp/detail/logical.h
+++ b/thrust/thrust/system/cpp/detail/logical.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
-// this system has no special version of this algorithm 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
+// this system has no special version of this algorithm
 
diff --git a/thrust/thrust/system/cpp/detail/malloc_and_free.h b/thrust/thrust/system/cpp/detail/malloc_and_free.h
index b1ad7a7341b..2d3fb189ade 100644
--- a/thrust/thrust/system/cpp/detail/malloc_and_free.h
+++ b/thrust/thrust/system/cpp/detail/malloc_and_free.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits malloc & free
 #include <thrust/system/detail/sequential/malloc_and_free.h>
 
diff --git a/thrust/thrust/system/cpp/detail/memory.inl b/thrust/thrust/system/cpp/detail/memory.inl
index 650aa1cb501..f8ed48b32c8 100644
--- a/thrust/thrust/system/cpp/detail/memory.inl
+++ b/thrust/thrust/system/cpp/detail/memory.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/cpp/memory.h>
 #include <thrust/system/cpp/detail/malloc_and_free.h>
 #include <limits>
diff --git a/thrust/thrust/system/cpp/detail/merge.h b/thrust/thrust/system/cpp/detail/merge.h
index 20e17f2d40d..f7c60de2eea 100644
--- a/thrust/thrust/system/cpp/detail/merge.h
+++ b/thrust/thrust/system/cpp/detail/merge.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits merge
 #include <thrust/system/detail/sequential/merge.h>
 
diff --git a/thrust/thrust/system/cpp/detail/mismatch.h b/thrust/thrust/system/cpp/detail/mismatch.h
index c6ae90664ad..c3f89646b4e 100644
--- a/thrust/thrust/system/cpp/detail/mismatch.h
+++ b/thrust/thrust/system/cpp/detail/mismatch.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
-// this system has no special version of this algorithm 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
+// this system has no special version of this algorithm
 
diff --git a/thrust/thrust/system/cpp/detail/par.h b/thrust/thrust/system/cpp/detail/par.h
index c56921327f5..03a1c4d2b9a 100644
--- a/thrust/thrust/system/cpp/detail/par.h
+++ b/thrust/thrust/system/cpp/detail/par.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/allocator_aware_execution_policy.h>
 #include <thrust/system/cpp/detail/execution_policy.h>
 
diff --git a/thrust/thrust/system/cpp/detail/partition.h b/thrust/thrust/system/cpp/detail/partition.h
index 50d48e222aa..742b7c806f7 100644
--- a/thrust/thrust/system/cpp/detail/partition.h
+++ b/thrust/thrust/system/cpp/detail/partition.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits partition
 #include <thrust/system/detail/sequential/partition.h>
 
diff --git a/thrust/thrust/system/cpp/detail/per_device_resource.h b/thrust/thrust/system/cpp/detail/per_device_resource.h
index 1b8d61f9216..64c721210a7 100644
--- a/thrust/thrust/system/cpp/detail/per_device_resource.h
+++ b/thrust/thrust/system/cpp/detail/per_device_resource.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system has no special per device resource functions
 
diff --git a/thrust/thrust/system/cpp/detail/reduce.h b/thrust/thrust/system/cpp/detail/reduce.h
index e09652cd9fd..3197b30a97b 100644
--- a/thrust/thrust/system/cpp/detail/reduce.h
+++ b/thrust/thrust/system/cpp/detail/reduce.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits reduce
 #include <thrust/system/detail/sequential/reduce.h>
 
diff --git a/thrust/thrust/system/cpp/detail/reduce_by_key.h b/thrust/thrust/system/cpp/detail/reduce_by_key.h
index a2c7744249b..836628d2fdc 100644
--- a/thrust/thrust/system/cpp/detail/reduce_by_key.h
+++ b/thrust/thrust/system/cpp/detail/reduce_by_key.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits reduce_by_key
 #include <thrust/system/detail/sequential/reduce_by_key.h>
 
diff --git a/thrust/thrust/system/cpp/detail/remove.h b/thrust/thrust/system/cpp/detail/remove.h
index a529f625d6a..f2c673830e2 100644
--- a/thrust/thrust/system/cpp/detail/remove.h
+++ b/thrust/thrust/system/cpp/detail/remove.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits remove
 #include <thrust/system/detail/sequential/remove.h>
 
diff --git a/thrust/thrust/system/cpp/detail/replace.h b/thrust/thrust/system/cpp/detail/replace.h
index c6ae90664ad..c3f89646b4e 100644
--- a/thrust/thrust/system/cpp/detail/replace.h
+++ b/thrust/thrust/system/cpp/detail/replace.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
-// this system has no special version of this algorithm 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
+// this system has no special version of this algorithm
 
diff --git a/thrust/thrust/system/cpp/detail/reverse.h b/thrust/thrust/system/cpp/detail/reverse.h
index c6ae90664ad..c3f89646b4e 100644
--- a/thrust/thrust/system/cpp/detail/reverse.h
+++ b/thrust/thrust/system/cpp/detail/reverse.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
-// this system has no special version of this algorithm 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
+// this system has no special version of this algorithm
 
diff --git a/thrust/thrust/system/cpp/detail/scan.h b/thrust/thrust/system/cpp/detail/scan.h
index 4d38e648437..5c8beb197d9 100644
--- a/thrust/thrust/system/cpp/detail/scan.h
+++ b/thrust/thrust/system/cpp/detail/scan.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits scan
 #include <thrust/system/detail/sequential/scan.h>
 
diff --git a/thrust/thrust/system/cpp/detail/scan_by_key.h b/thrust/thrust/system/cpp/detail/scan_by_key.h
index 1e72e7c38f8..b4aabfd1c1e 100644
--- a/thrust/thrust/system/cpp/detail/scan_by_key.h
+++ b/thrust/thrust/system/cpp/detail/scan_by_key.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits the scan_by_key algorithms
 #include <thrust/system/detail/sequential/scan_by_key.h>
 
diff --git a/thrust/thrust/system/cpp/detail/scatter.h b/thrust/thrust/system/cpp/detail/scatter.h
index c6ae90664ad..c3f89646b4e 100644
--- a/thrust/thrust/system/cpp/detail/scatter.h
+++ b/thrust/thrust/system/cpp/detail/scatter.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
-// this system has no special version of this algorithm 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
+// this system has no special version of this algorithm
 
diff --git a/thrust/thrust/system/cpp/detail/sequence.h b/thrust/thrust/system/cpp/detail/sequence.h
index c6ae90664ad..c3f89646b4e 100644
--- a/thrust/thrust/system/cpp/detail/sequence.h
+++ b/thrust/thrust/system/cpp/detail/sequence.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
-// this system has no special version of this algorithm 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
+// this system has no special version of this algorithm
 
diff --git a/thrust/thrust/system/cpp/detail/set_operations.h b/thrust/thrust/system/cpp/detail/set_operations.h
index bbde20114c2..c5247efee00 100644
--- a/thrust/thrust/system/cpp/detail/set_operations.h
+++ b/thrust/thrust/system/cpp/detail/set_operations.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits the set operations
 #include <thrust/system/detail/sequential/set_operations.h>
 
diff --git a/thrust/thrust/system/cpp/detail/sort.h b/thrust/thrust/system/cpp/detail/sort.h
index ae38b3ba8c7..e25973ea166 100644
--- a/thrust/thrust/system/cpp/detail/sort.h
+++ b/thrust/thrust/system/cpp/detail/sort.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits sort
 #include <thrust/system/detail/sequential/sort.h>
 
diff --git a/thrust/thrust/system/cpp/detail/swap_ranges.h b/thrust/thrust/system/cpp/detail/swap_ranges.h
index 35cd5462207..276381aadff 100644
--- a/thrust/thrust/system/cpp/detail/swap_ranges.h
+++ b/thrust/thrust/system/cpp/detail/swap_ranges.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // cpp has no special swap_ranges
 
diff --git a/thrust/thrust/system/cpp/detail/tabulate.h b/thrust/thrust/system/cpp/detail/tabulate.h
index c6ae90664ad..c3f89646b4e 100644
--- a/thrust/thrust/system/cpp/detail/tabulate.h
+++ b/thrust/thrust/system/cpp/detail/tabulate.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
-// this system has no special version of this algorithm 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
+// this system has no special version of this algorithm
 
diff --git a/thrust/thrust/system/cpp/detail/temporary_buffer.h b/thrust/thrust/system/cpp/detail/temporary_buffer.h
index 2adfaf2810c..d666f267e35 100644
--- a/thrust/thrust/system/cpp/detail/temporary_buffer.h
+++ b/thrust/thrust/system/cpp/detail/temporary_buffer.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system has no special temporary buffer functions
 
diff --git a/thrust/thrust/system/cpp/detail/transform.h b/thrust/thrust/system/cpp/detail/transform.h
index 895164ce5af..434e7a66535 100644
--- a/thrust/thrust/system/cpp/detail/transform.h
+++ b/thrust/thrust/system/cpp/detail/transform.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // cpp has no special transform
 
diff --git a/thrust/thrust/system/cpp/detail/transform_reduce.h b/thrust/thrust/system/cpp/detail/transform_reduce.h
index c6ae90664ad..c3f89646b4e 100644
--- a/thrust/thrust/system/cpp/detail/transform_reduce.h
+++ b/thrust/thrust/system/cpp/detail/transform_reduce.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
-// this system has no special version of this algorithm 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
+// this system has no special version of this algorithm
 
diff --git a/thrust/thrust/system/cpp/detail/transform_scan.h b/thrust/thrust/system/cpp/detail/transform_scan.h
index c6ae90664ad..c3f89646b4e 100644
--- a/thrust/thrust/system/cpp/detail/transform_scan.h
+++ b/thrust/thrust/system/cpp/detail/transform_scan.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
-// this system has no special version of this algorithm 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
+// this system has no special version of this algorithm
 
diff --git a/thrust/thrust/system/cpp/detail/uninitialized_copy.h b/thrust/thrust/system/cpp/detail/uninitialized_copy.h
index c6ae90664ad..c3f89646b4e 100644
--- a/thrust/thrust/system/cpp/detail/uninitialized_copy.h
+++ b/thrust/thrust/system/cpp/detail/uninitialized_copy.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
-// this system has no special version of this algorithm 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
+// this system has no special version of this algorithm
 
diff --git a/thrust/thrust/system/cpp/detail/uninitialized_fill.h b/thrust/thrust/system/cpp/detail/uninitialized_fill.h
index c6ae90664ad..c3f89646b4e 100644
--- a/thrust/thrust/system/cpp/detail/uninitialized_fill.h
+++ b/thrust/thrust/system/cpp/detail/uninitialized_fill.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
-// this system has no special version of this algorithm 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
+// this system has no special version of this algorithm
 
diff --git a/thrust/thrust/system/cpp/detail/unique.h b/thrust/thrust/system/cpp/detail/unique.h
index 2ff23e9d3db..c00503fb282 100644
--- a/thrust/thrust/system/cpp/detail/unique.h
+++ b/thrust/thrust/system/cpp/detail/unique.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits unique
 #include <thrust/system/detail/sequential/unique.h>
 
diff --git a/thrust/thrust/system/cpp/detail/unique_by_key.h b/thrust/thrust/system/cpp/detail/unique_by_key.h
index 1d40011787c..78e73f639e9 100644
--- a/thrust/thrust/system/cpp/detail/unique_by_key.h
+++ b/thrust/thrust/system/cpp/detail/unique_by_key.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits unique_by_key
 #include <thrust/system/detail/sequential/unique_by_key.h>
 
diff --git a/thrust/thrust/system/cpp/detail/vector.inl b/thrust/thrust/system/cpp/detail/vector.inl
index 02980c62a42..96e0f285782 100644
--- a/thrust/thrust/system/cpp/detail/vector.inl
+++ b/thrust/thrust/system/cpp/detail/vector.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/cpp/vector.h>
 #include <utility>
 
@@ -98,13 +104,13 @@ template<typename T, typename Allocator>
     return *this;
   }
 #endif
-  
+
   template<typename T, typename Allocator>
     vector<T,Allocator>
       ::vector(std::initializer_list<T> il)
         : super_t(il)
   {}
-  
+
   template<typename T, typename Allocator>
     vector<T,Allocator>
       ::vector(std::initializer_list<T> il, const Allocator& alloc)
@@ -139,7 +145,7 @@ template<typename T, typename Allocator>
   super_t::operator=(x);
   return *this;
 }
-      
+
 } // end cpp
 } // end system
 THRUST_NAMESPACE_END
diff --git a/thrust/thrust/system/cpp/execution_policy.h b/thrust/thrust/system/cpp/execution_policy.h
index 0d8a9a36790..f7aca1fb22f 100644
--- a/thrust/thrust/system/cpp/execution_policy.h
+++ b/thrust/thrust/system/cpp/execution_policy.h
@@ -22,6 +22,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // get the execution policies definitions first
 #include <thrust/system/cpp/detail/execution_policy.h>
 
diff --git a/thrust/thrust/system/cpp/memory.h b/thrust/thrust/system/cpp/memory.h
index a18abeb8e4f..d6451670c03 100644
--- a/thrust/thrust/system/cpp/memory.h
+++ b/thrust/thrust/system/cpp/memory.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/cpp/memory_resource.h>
 #include <thrust/memory.h>
 #include <thrust/detail/type_traits.h>
diff --git a/thrust/thrust/system/cpp/memory_resource.h b/thrust/thrust/system/cpp/memory_resource.h
index 04b4e3cf8da..1a0206f31e7 100644
--- a/thrust/thrust/system/cpp/memory_resource.h
+++ b/thrust/thrust/system/cpp/memory_resource.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/mr/new.h>
 #include <thrust/mr/fancy_pointer_resource.h>
 
diff --git a/thrust/thrust/system/cpp/pointer.h b/thrust/thrust/system/cpp/pointer.h
index f204fa37509..5e75aafd44c 100644
--- a/thrust/thrust/system/cpp/pointer.h
+++ b/thrust/thrust/system/cpp/pointer.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <type_traits>
 #include <thrust/system/cpp/detail/execution_policy.h>
 #include <thrust/detail/pointer.h>
diff --git a/thrust/thrust/system/cpp/vector.h b/thrust/thrust/system/cpp/vector.h
index 2a418dbc30d..6122dd42bc8 100644
--- a/thrust/thrust/system/cpp/vector.h
+++ b/thrust/thrust/system/cpp/vector.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/cpp/memory.h>
 #include <thrust/detail/vector_base.h>
 #include <vector>
diff --git a/thrust/thrust/system/cuda/config.h b/thrust/thrust/system/cuda/config.h
index f29a72ac863..9ec5137f2af 100644
--- a/thrust/thrust/system/cuda/config.h
+++ b/thrust/thrust/system/cuda/config.h
@@ -34,9 +34,14 @@
 #define THRUST_DEBUG_SYNC_FLAG false
 #endif
 
-
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // We don't directly include <cub/version.cuh> since it doesn't exist in
 // older releases. This header will always pull in version info:
 #include <cub/util_namespace.cuh>
diff --git a/thrust/thrust/system/cuda/detail/adjacent_difference.h b/thrust/thrust/system/cuda/detail/adjacent_difference.h
index 28461123518..85c499b620a 100644
--- a/thrust/thrust/system/cuda/detail/adjacent_difference.h
+++ b/thrust/thrust/system/cuda/detail/adjacent_difference.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 
 #include <thrust/detail/cstdint.h>
diff --git a/thrust/thrust/system/cuda/detail/assign_value.h b/thrust/thrust/system/cuda/detail/assign_value.h
index 8945f1cac37..3e98b37d8e3 100644
--- a/thrust/thrust/system/cuda/detail/assign_value.h
+++ b/thrust/thrust/system/cuda/detail/assign_value.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 #include <thrust/system/cuda/config.h>
 #include <thrust/system/cuda/detail/execution_policy.h>
diff --git a/thrust/thrust/system/cuda/detail/async/copy.h b/thrust/thrust/system/cuda/detail/async/copy.h
index 6f29707594f..af0040f18a2 100644
--- a/thrust/thrust/system/cuda/detail/async/copy.h
+++ b/thrust/thrust/system/cuda/detail/async/copy.h
@@ -30,6 +30,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp14_required.h>
 
 #if THRUST_CPP_DIALECT >= 2014
diff --git a/thrust/thrust/system/cuda/detail/async/customization.h b/thrust/thrust/system/cuda/detail/async/customization.h
index 6f125a6f418..758370cf992 100644
--- a/thrust/thrust/system/cuda/detail/async/customization.h
+++ b/thrust/thrust/system/cuda/detail/async/customization.h
@@ -30,6 +30,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp14_required.h>
 
 #if THRUST_CPP_DIALECT >= 2014
diff --git a/thrust/thrust/system/cuda/detail/async/exclusive_scan.h b/thrust/thrust/system/cuda/detail/async/exclusive_scan.h
index 0b120a434a3..f600def0f85 100644
--- a/thrust/thrust/system/cuda/detail/async/exclusive_scan.h
+++ b/thrust/thrust/system/cuda/detail/async/exclusive_scan.h
@@ -28,6 +28,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp14_required.h>
 
 #if THRUST_CPP_DIALECT >= 2014
diff --git a/thrust/thrust/system/cuda/detail/async/for_each.h b/thrust/thrust/system/cuda/detail/async/for_each.h
index d6809fe0a46..fe500fe8dfa 100644
--- a/thrust/thrust/system/cuda/detail/async/for_each.h
+++ b/thrust/thrust/system/cuda/detail/async/for_each.h
@@ -31,6 +31,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp14_required.h>
 
 #if THRUST_CPP_DIALECT >= 2014
diff --git a/thrust/thrust/system/cuda/detail/async/inclusive_scan.h b/thrust/thrust/system/cuda/detail/async/inclusive_scan.h
index 363347c3531..2c6b3bdb72f 100644
--- a/thrust/thrust/system/cuda/detail/async/inclusive_scan.h
+++ b/thrust/thrust/system/cuda/detail/async/inclusive_scan.h
@@ -28,6 +28,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp14_required.h>
 
 #if THRUST_CPP_DIALECT >= 2014
diff --git a/thrust/thrust/system/cuda/detail/async/reduce.h b/thrust/thrust/system/cuda/detail/async/reduce.h
index 2d0dbfe16f6..73351132df8 100644
--- a/thrust/thrust/system/cuda/detail/async/reduce.h
+++ b/thrust/thrust/system/cuda/detail/async/reduce.h
@@ -32,6 +32,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp14_required.h>
 
 #if THRUST_CPP_DIALECT >= 2014
diff --git a/thrust/thrust/system/cuda/detail/async/scan.h b/thrust/thrust/system/cuda/detail/async/scan.h
index 4a9f31681ec..e946683061c 100644
--- a/thrust/thrust/system/cuda/detail/async/scan.h
+++ b/thrust/thrust/system/cuda/detail/async/scan.h
@@ -27,6 +27,14 @@
 
 #pragma once
 
+#include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/cpp14_required.h>
 
 #include <thrust/system/cuda/detail/async/exclusive_scan.h>
diff --git a/thrust/thrust/system/cuda/detail/async/sort.h b/thrust/thrust/system/cuda/detail/async/sort.h
index f501f19c580..9f1d567cad9 100644
--- a/thrust/thrust/system/cuda/detail/async/sort.h
+++ b/thrust/thrust/system/cuda/detail/async/sort.h
@@ -30,6 +30,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp14_required.h>
 
 #if THRUST_CPP_DIALECT >= 2014
diff --git a/thrust/thrust/system/cuda/detail/async/transform.h b/thrust/thrust/system/cuda/detail/async/transform.h
index a971300f229..4bd8f96bddd 100644
--- a/thrust/thrust/system/cuda/detail/async/transform.h
+++ b/thrust/thrust/system/cuda/detail/async/transform.h
@@ -30,6 +30,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp14_required.h>
 
 #if THRUST_CPP_DIALECT >= 2014
diff --git a/thrust/thrust/system/cuda/detail/binary_search.h b/thrust/thrust/system/cuda/detail/binary_search.h
index fb769a4ac0a..e7c5184af70 100644
--- a/thrust/thrust/system/cuda/detail/binary_search.h
+++ b/thrust/thrust/system/cuda/detail/binary_search.h
@@ -16,4 +16,12 @@
 
 #pragma once
 
+#include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system has no special version of this algorithm
diff --git a/thrust/thrust/system/cuda/detail/cdp_dispatch.h b/thrust/thrust/system/cuda/detail/cdp_dispatch.h
index f94e5dd9265..739d1cb7ecd 100644
--- a/thrust/thrust/system/cuda/detail/cdp_dispatch.h
+++ b/thrust/thrust/system/cuda/detail/cdp_dispatch.h
@@ -21,6 +21,14 @@
 
 #pragma once
 
+#include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/config.cuh>
 #include <cub/detail/detect_cuda_runtime.cuh>
 
diff --git a/thrust/thrust/system/cuda/detail/copy.h b/thrust/thrust/system/cuda/detail/copy.h
index 02a5d2ac1a6..d394131d423 100644
--- a/thrust/thrust/system/cuda/detail/copy.h
+++ b/thrust/thrust/system/cuda/detail/copy.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/advance.h>
 
 #include <thrust/system/cuda/config.h>
diff --git a/thrust/thrust/system/cuda/detail/copy_if.h b/thrust/thrust/system/cuda/detail/copy_if.h
index 5e760c086fb..4d99dda6f35 100644
--- a/thrust/thrust/system/cuda/detail/copy_if.h
+++ b/thrust/thrust/system/cuda/detail/copy_if.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 
 #include <thrust/detail/alignment.h>
diff --git a/thrust/thrust/system/cuda/detail/core/agent_launcher.h b/thrust/thrust/system/cuda/detail/core/agent_launcher.h
index 10df7612cd9..1958e2e20a0 100644
--- a/thrust/thrust/system/cuda/detail/core/agent_launcher.h
+++ b/thrust/thrust/system/cuda/detail/core/agent_launcher.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cub/detail/device_synchronize.cuh>
 
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
@@ -35,30 +41,29 @@
 #include <thrust/system/cuda/detail/core/util.h>
 #include <thrust/system/cuda/detail/guarded_cuda_runtime_api.h>
 
-#include <cuda/std/version> // _LIBCUDACXX_HIDDEN, _LIBCUDACXX_{CLANG,GCC}_DIAGNOSTIC_IGNORED
-
 #include <cassert>
 
 #include <nv/target>
 
-THRUST_NAMESPACE_BEGIN
-namespace cuda_cub {
-namespace core {
-
 /**
  * @def THRUST_DISABLE_KERNEL_VISIBILITY_WARNING_SUPPRESSION
  * If defined, the default suppression of kernel visibility attribute warning is disabled.
  */
 #if !defined(THRUST_DISABLE_KERNEL_VISIBILITY_WARNING_SUPPRESSION)
-_LIBCUDACXX_GCC_DIAGNOSTIC_IGNORED("-Wattributes")
-_LIBCUDACXX_CLANG_DIAGNOSTIC_IGNORED("-Wattributes")                      
-#if !defined(_LIBCUDACXX_COMPILER_NVHPC_CUDA)
-_LIBCUDACXX_NVHPC_DIAGNOSTIC_IGNORED(attribute_requires_external_linkage)
-#endif
-#endif
+_CCCL_DIAG_SUPPRESS_GCC("-Wattributes")
+_CCCL_DIAG_SUPPRESS_CLANG("-Wattributes")
+#if !defined(_CCCL_CUDA_COMPILER_NVHPC)
+_CCCL_DIAG_SUPPRESS_NVHPC(attribute_requires_external_linkage)
+#endif // !_LIBCUDACXX_COMPILER_NVHPC_CUDA
+#endif // !THRUST_DISABLE_KERNEL_VISIBILITY_WARNING_SUPPRESSION
+
+THRUST_NAMESPACE_BEGIN
+
+namespace cuda_cub {
+namespace core {
 
 #ifndef THRUST_DETAIL_KERNEL_ATTRIBUTES
-#define THRUST_DETAIL_KERNEL_ATTRIBUTES __global__ _LIBCUDACXX_HIDDEN
+#define THRUST_DETAIL_KERNEL_ATTRIBUTES CCCL_DETAIL_KERNEL_ATTRIBUTES
 #endif
 
 #if defined(__CUDA_ARCH__) || defined(_NVHPC_CUDA)
@@ -541,7 +546,7 @@ _LIBCUDACXX_NVHPC_DIAGNOSTIC_IGNORED(attribute_requires_external_linkage)
     THRUST_RUNTIME_FUNCTION
     void print_info(K k) const
     {
-      #if THRUST_DEBUG_SYNC_FLAG 
+      #if THRUST_DEBUG_SYNC_FLAG
       cuda_optional<int> occ = max_sm_occupancy(k);
       const int ptx_version = core::get_ptx_version();
       if (count > 0)
@@ -1159,5 +1164,7 @@ _LIBCUDACXX_NVHPC_DIAGNOSTIC_IGNORED(attribute_requires_external_linkage)
 
 } // namespace core
 } // namespace cuda_cub
+
 THRUST_NAMESPACE_END
+
 #endif
diff --git a/thrust/thrust/system/cuda/detail/core/alignment.h b/thrust/thrust/system/cuda/detail/core/alignment.h
index 4b807ebc1ee..8c28547001d 100644
--- a/thrust/thrust/system/cuda/detail/core/alignment.h
+++ b/thrust/thrust/system/cuda/detail/core/alignment.h
@@ -20,6 +20,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/system/cuda/detail/util.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/cuda/detail/core/triple_chevron_launch.h b/thrust/thrust/system/cuda/detail/core/triple_chevron_launch.h
index 28ac1a23053..47af5501934 100644
--- a/thrust/thrust/system/cuda/detail/core/triple_chevron_launch.h
+++ b/thrust/thrust/system/cuda/detail/core/triple_chevron_launch.h
@@ -27,11 +27,15 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/cuda/detail/core/alignment.h>
 #include <thrust/system/cuda/detail/guarded_cuda_runtime_api.h>
 
-#include <cuda/std/version> // _LIBCUDACXX_HIDDEN
-
 #include <cassert>
 
 THRUST_NAMESPACE_BEGIN
@@ -39,7 +43,7 @@ THRUST_NAMESPACE_BEGIN
 namespace cuda_cub {
 namespace launcher {
 
-  struct _LIBCUDACXX_HIDDEN triple_chevron
+  struct _CCCL_ATTRIBUTE_HIDDEN triple_chevron
   {
     typedef size_t Size;
     dim3 const grid;
@@ -127,7 +131,7 @@ namespace launcher {
                               shared_mem,
                               stream);
     }
-    #else 
+    #else
     template<class K, class... Args>
     cudaError_t __device__
     doit_device(K, Args const&... ) const
diff --git a/thrust/thrust/system/cuda/detail/core/util.h b/thrust/thrust/system/cuda/detail/core/util.h
index 3741aec2848..e4dbb8e5195 100644
--- a/thrust/thrust/system/cuda/detail/core/util.h
+++ b/thrust/thrust/system/cuda/detail/core/util.h
@@ -27,6 +27,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/raw_pointer_cast.h>
 #include <thrust/system/cuda/config.h>
 #include <thrust/system/cuda/detail/util.h>
@@ -620,7 +626,7 @@ namespace core {
   inline int get_ptx_version()
   {
     int ptx_version = 0;
-    if (cub::PtxVersion(ptx_version) != cudaSuccess) 
+    if (cub::PtxVersion(ptx_version) != cudaSuccess)
     {
       // Failure might mean that there's no device found
       const int current_device = cub::CurrentDevice();
@@ -629,7 +635,7 @@ namespace core {
         cuda_cub::throw_on_error(cudaErrorNoDevice, "No GPU is available\n");
       }
 
-      // Any subsequent failure means the provided device binary does not match 
+      // Any subsequent failure means the provided device binary does not match
       // the generated function code
       int major = 0, minor = 0;
       cudaError_t attr_status;
@@ -643,7 +649,7 @@ namespace core {
       cuda_cub::throw_on_error(attr_status,
                               "get_ptx_version :"
                               "failed to get minor CUDA device compute capability version.");
-        
+
       // Index from which SM code has to start in the message below
       int code_offset = 37;
       char str[] = "This program was not compiled for SM     \n";
diff --git a/thrust/thrust/system/cuda/detail/count.h b/thrust/thrust/system/cuda/detail/count.h
index b624f39dcca..2fc36c36b61 100644
--- a/thrust/thrust/system/cuda/detail/count.h
+++ b/thrust/thrust/system/cuda/detail/count.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 #include <thrust/system/cuda/config.h>
 
diff --git a/thrust/thrust/system/cuda/detail/cross_system.h b/thrust/thrust/system/cuda/detail/cross_system.h
index 039531d2821..e01721f227e 100644
--- a/thrust/thrust/system/cuda/detail/cross_system.h
+++ b/thrust/thrust/system/cuda/detail/cross_system.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/system/cuda/detail/guarded_cuda_runtime_api.h>
 #include <thrust/system/cpp/detail/execution_policy.h>
 #include <thrust/system/cuda/detail/execution_policy.h>
diff --git a/thrust/thrust/system/cuda/detail/dispatch.h b/thrust/thrust/system/cuda/detail/dispatch.h
index d0e3f94ec82..2e20a19f75f 100644
--- a/thrust/thrust/system/cuda/detail/dispatch.h
+++ b/thrust/thrust/system/cuda/detail/dispatch.h
@@ -16,6 +16,14 @@
 
 #pragma once
 
+#include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/cstdint.h>
 #include <thrust/detail/preprocessor.h>
 #include <thrust/detail/integer_traits.h>
diff --git a/thrust/thrust/system/cuda/detail/equal.h b/thrust/thrust/system/cuda/detail/equal.h
index aec60824526..2adfa868c35 100644
--- a/thrust/thrust/system/cuda/detail/equal.h
+++ b/thrust/thrust/system/cuda/detail/equal.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 #include <thrust/system/cuda/config.h>
 
diff --git a/thrust/thrust/system/cuda/detail/error.inl b/thrust/thrust/system/cuda/detail/error.inl
index e5230521131..2bdd107c77f 100644
--- a/thrust/thrust/system/cuda/detail/error.inl
+++ b/thrust/thrust/system/cuda/detail/error.inl
@@ -19,6 +19,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/system/cuda/error.h>
 #include <thrust/system/cuda/detail/guarded_cuda_runtime_api.h>
 
diff --git a/thrust/thrust/system/cuda/detail/execution_policy.h b/thrust/thrust/system/cuda/detail/execution_policy.h
index 4202424c5e4..318ada1faa1 100644
--- a/thrust/thrust/system/cuda/detail/execution_policy.h
+++ b/thrust/thrust/system/cuda/detail/execution_policy.h
@@ -29,6 +29,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/version.h>
 #include <thrust/detail/execution_policy.h>
 #include <thrust/iterator/detail/any_system_tag.h>
@@ -66,7 +72,7 @@ struct tag : execution_policy<tag>
 template <class Derived>
 struct execution_policy : thrust::execution_policy<Derived>
 {
-  typedef tag tag_type; 
+  typedef tag tag_type;
   operator tag() const { return tag(); }
 };
 
diff --git a/thrust/thrust/system/cuda/detail/extrema.h b/thrust/thrust/system/cuda/detail/extrema.h
index 4fe7ec86b5c..9aea95a1b07 100644
--- a/thrust/thrust/system/cuda/detail/extrema.h
+++ b/thrust/thrust/system/cuda/detail/extrema.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 
 #include <thrust/detail/cstdint.h>
diff --git a/thrust/thrust/system/cuda/detail/fill.h b/thrust/thrust/system/cuda/detail/fill.h
index 80ea68592a2..286937fe10a 100644
--- a/thrust/thrust/system/cuda/detail/fill.h
+++ b/thrust/thrust/system/cuda/detail/fill.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 #include <thrust/system/cuda/detail/util.h>
 #include <thrust/system/cuda/detail/parallel_for.h>
diff --git a/thrust/thrust/system/cuda/detail/find.h b/thrust/thrust/system/cuda/detail/find.h
index b7d2b748f87..adfa93fd484 100644
--- a/thrust/thrust/system/cuda/detail/find.h
+++ b/thrust/thrust/system/cuda/detail/find.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 #include <thrust/system/cuda/config.h>
 
@@ -111,10 +117,10 @@ find_if_n(execution_policy<Derived>& policy,
           Predicate                  predicate)
 {
   typedef typename thrust::tuple<bool,Size> result_type;
-  
+
   // empty sequence
   if(num_items == 0) return first;
-  
+
   // this implementation breaks up the sequence into separate intervals
   // in an attempt to early-out as soon as a value is found
   //
@@ -125,7 +131,7 @@ find_if_n(execution_policy<Derived>& policy,
   // TODO incorporate sizeof(InputType) into interval_threshold and round to multiple of 32
   const Size interval_threshold = 1 << 20;
   const Size interval_size = (thrust::min)(interval_threshold, num_items);
-  
+
   // force transform_iterator output to bool
   typedef transform_input_iterator_t<bool,
                                      InputIt,
@@ -165,7 +171,7 @@ find_if_n(execution_policy<Derived>& policy,
       return first + thrust::get<1>(result);
     }
   }
-  
+
   //nothing was found if we reach here...
   return first + num_items;
 }
diff --git a/thrust/thrust/system/cuda/detail/for_each.h b/thrust/thrust/system/cuda/detail/for_each.h
index 518538ff362..8041de112db 100644
--- a/thrust/thrust/system/cuda/detail/for_each.h
+++ b/thrust/thrust/system/cuda/detail/for_each.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 #include <iterator>
 #include <thrust/system/cuda/config.h>
diff --git a/thrust/thrust/system/cuda/detail/future.inl b/thrust/thrust/system/cuda/detail/future.inl
index f23184aae9d..1ef39fee448 100644
--- a/thrust/thrust/system/cuda/detail/future.inl
+++ b/thrust/thrust/system/cuda/detail/future.inl
@@ -9,6 +9,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp14_required.h>
 
 #if THRUST_CPP_DIALECT >= 2014
@@ -361,7 +367,7 @@ try_acquire_stream(int device, unique_eager_future<X>& parent) noexcept;
 template <typename... Dependencies>
 __host__
 acquired_stream acquire_stream(int device, Dependencies&... deps) noexcept;
-  
+
 template <typename... Dependencies>
 __host__
 unique_eager_event
@@ -520,7 +526,7 @@ public:
 
   // Precondition: `true == valid_content()`.
   __host__
-  pointer data() 
+  pointer data()
   {
     if (!valid_content())
       throw thrust::event_error(event_errc::no_content);
@@ -530,7 +536,7 @@ public:
 
   // Precondition: `true == valid_content()`.
   __host__
-  const_pointer data() const 
+  const_pointer data() const
   {
     if (!valid_content())
       throw thrust::event_error(event_errc::no_content);
@@ -673,7 +679,7 @@ public:
   }
 
   THRUST_NODISCARD __host__ __device__
-  value_type extract() 
+  value_type extract()
   {
     return std::move(value_);
   }
@@ -1331,7 +1337,7 @@ make_dependent_future(ComputeContent&& cc, std::tuple<Dependencies...>&& deps)
   std::unique_ptr<async_signal_type> sig(
     new async_signal_type(std::move(as.stream), std::move(ka), std::move(cc))
   );
- 
+
   // Finally, we create the promise and future objects.
   weak_promise<X, XPointer> child_prom(device_id, sig->data());
   unique_eager_future<X> child_fut(device_id, std::move(sig));
@@ -1350,7 +1356,7 @@ unique_eager_event when_all(Events&&... evs)
 // TODO: Constrain to events, futures, and maybe streams (currently allows keep
 // alives).
 {
-  return detail::make_dependent_event(std::make_tuple(std::move(evs)...)); 
+  return detail::make_dependent_event(std::make_tuple(std::move(evs)...));
 }
 
 // ADL hook for transparent `.after` move support.
diff --git a/thrust/thrust/system/cuda/detail/gather.h b/thrust/thrust/system/cuda/detail/gather.h
index 56ff3aecf16..bdbc7b566dc 100644
--- a/thrust/thrust/system/cuda/detail/gather.h
+++ b/thrust/thrust/system/cuda/detail/gather.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 #include <thrust/system/cuda/detail/transform.h>
 #include <thrust/iterator/permutation_iterator.h>
diff --git a/thrust/thrust/system/cuda/detail/generate.h b/thrust/thrust/system/cuda/detail/generate.h
index ad6340f8313..dd8b7ce9d99 100644
--- a/thrust/thrust/system/cuda/detail/generate.h
+++ b/thrust/thrust/system/cuda/detail/generate.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 #include <iterator>
 #include <thrust/system/cuda/config.h>
diff --git a/thrust/thrust/system/cuda/detail/get_value.h b/thrust/thrust/system/cuda/detail/get_value.h
index 9065f773a80..6364970a243 100644
--- a/thrust/thrust/system/cuda/detail/get_value.h
+++ b/thrust/thrust/system/cuda/detail/get_value.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 #include <thrust/system/cuda/config.h>
 #include <thrust/system/cuda/detail/cross_system.h>
diff --git a/thrust/thrust/system/cuda/detail/guarded_cuda_runtime_api.h b/thrust/thrust/system/cuda/detail/guarded_cuda_runtime_api.h
index 5b0f345a74a..0a61ac2ff2c 100644
--- a/thrust/thrust/system/cuda/detail/guarded_cuda_runtime_api.h
+++ b/thrust/thrust/system/cuda/detail/guarded_cuda_runtime_api.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to check for the existence of macros
 // such as __host__ and __device__, which may already be defined by thrust
 // and to undefine them before entering cuda_runtime_api.h (which will redefine them)
diff --git a/thrust/thrust/system/cuda/detail/guarded_driver_types.h b/thrust/thrust/system/cuda/detail/guarded_driver_types.h
index 076964071cf..7620b08704c 100644
--- a/thrust/thrust/system/cuda/detail/guarded_driver_types.h
+++ b/thrust/thrust/system/cuda/detail/guarded_driver_types.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include <driver_types.h> without causing
 // warnings from redefinitions of __host__ and __device__.
 // carefully save their definitions and restore them
diff --git a/thrust/thrust/system/cuda/detail/inner_product.h b/thrust/thrust/system/cuda/detail/inner_product.h
index 98e9064d286..f65db8df1aa 100644
--- a/thrust/thrust/system/cuda/detail/inner_product.h
+++ b/thrust/thrust/system/cuda/detail/inner_product.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 #include <iterator>
 #include <thrust/system/cuda/detail/reduce.h>
diff --git a/thrust/thrust/system/cuda/detail/internal/copy_cross_system.h b/thrust/thrust/system/cuda/detail/internal/copy_cross_system.h
index a1208c67c9a..468253482e5 100644
--- a/thrust/thrust/system/cuda/detail/internal/copy_cross_system.h
+++ b/thrust/thrust/system/cuda/detail/internal/copy_cross_system.h
@@ -26,12 +26,18 @@
  ******************************************************************************/
 #pragma once
 
+#include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // XXX
 // this file must not be included on its own, ever,
 // but must be part of include in thrust/system/cuda/detail/copy.h
 
-#include <thrust/detail/config.h>
-
 #include <thrust/system/cuda/config.h>
 
 #include <thrust/distance.h>
@@ -178,7 +184,7 @@ namespace __copy {
     // get type of the input data
     typedef typename thrust::iterator_value<InputIt>::type InputTy;
 
-    // allocate device temp storage 
+    // allocate device temp storage
     thrust::detail::temporary_array<InputTy, D> d_in_ptr(device_s, num_items);
 
     // uninitialize copy into temp device storage
diff --git a/thrust/thrust/system/cuda/detail/internal/copy_device_to_device.h b/thrust/thrust/system/cuda/detail/internal/copy_device_to_device.h
index a0d9d492dac..fccec83922c 100644
--- a/thrust/thrust/system/cuda/detail/internal/copy_device_to_device.h
+++ b/thrust/thrust/system/cuda/detail/internal/copy_device_to_device.h
@@ -29,6 +29,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 #include <thrust/system/cuda/config.h>
 #include <thrust/system/cuda/detail/execution_policy.h>
diff --git a/thrust/thrust/system/cuda/detail/iter_swap.h b/thrust/thrust/system/cuda/detail/iter_swap.h
index c0628610acb..df2a1b450e1 100644
--- a/thrust/thrust/system/cuda/detail/iter_swap.h
+++ b/thrust/thrust/system/cuda/detail/iter_swap.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 
 #include <thrust/system/cuda/config.h>
diff --git a/thrust/thrust/system/cuda/detail/logical.h b/thrust/thrust/system/cuda/detail/logical.h
index c6ae90664ad..c3f89646b4e 100644
--- a/thrust/thrust/system/cuda/detail/logical.h
+++ b/thrust/thrust/system/cuda/detail/logical.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
-// this system has no special version of this algorithm 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
+// this system has no special version of this algorithm
 
diff --git a/thrust/thrust/system/cuda/detail/make_unsigned_special.h b/thrust/thrust/system/cuda/detail/make_unsigned_special.h
index dda73576760..b5e0d5bec73 100644
--- a/thrust/thrust/system/cuda/detail/make_unsigned_special.h
+++ b/thrust/thrust/system/cuda/detail/make_unsigned_special.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 THRUST_NAMESPACE_BEGIN
 namespace cuda_cub {
 
diff --git a/thrust/thrust/system/cuda/detail/malloc_and_free.h b/thrust/thrust/system/cuda/detail/malloc_and_free.h
index 1b12e2cc305..57b5d7e1583 100644
--- a/thrust/thrust/system/cuda/detail/malloc_and_free.h
+++ b/thrust/thrust/system/cuda/detail/malloc_and_free.h
@@ -16,9 +16,16 @@
 
 #pragma once
 
+#include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/system/cuda/detail/guarded_cuda_runtime_api.h>
 
-#include <thrust/detail/config.h>
 #include <thrust/detail/raw_pointer_cast.h>
 #include <thrust/detail/raw_reference_cast.h>
 #include <thrust/detail/seq.h>
diff --git a/thrust/thrust/system/cuda/detail/memory.inl b/thrust/thrust/system/cuda/detail/memory.inl
index f6fc98359af..4ba3180faa0 100644
--- a/thrust/thrust/system/cuda/detail/memory.inl
+++ b/thrust/thrust/system/cuda/detail/memory.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/cuda/memory.h>
 #include <thrust/system/cuda/detail/malloc_and_free.h>
 #include <limits>
diff --git a/thrust/thrust/system/cuda/detail/merge.h b/thrust/thrust/system/cuda/detail/merge.h
index 478e3508d70..2ae9469ad6d 100644
--- a/thrust/thrust/system/cuda/detail/merge.h
+++ b/thrust/thrust/system/cuda/detail/merge.h
@@ -28,6 +28,12 @@ j * Copyright (c) 2016, NVIDIA CORPORATION.  All rights reserved.
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 
 #include <thrust/detail/cstdint.h>
diff --git a/thrust/thrust/system/cuda/detail/mismatch.h b/thrust/thrust/system/cuda/detail/mismatch.h
index b1e2f44d2df..7110c232cbc 100644
--- a/thrust/thrust/system/cuda/detail/mismatch.h
+++ b/thrust/thrust/system/cuda/detail/mismatch.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 #include <thrust/system/cuda/config.h>
 #include <thrust/system/cuda/detail/execution_policy.h>
diff --git a/thrust/thrust/system/cuda/detail/par.h b/thrust/thrust/system/cuda/detail/par.h
index 42c701ca701..56aa3c2414f 100644
--- a/thrust/thrust/system/cuda/detail/par.h
+++ b/thrust/thrust/system/cuda/detail/par.h
@@ -27,6 +27,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/cuda/detail/guarded_cuda_runtime_api.h>
 #include <thrust/system/cuda/detail/execution_policy.h>
 #include <thrust/system/cuda/detail/util.h>
@@ -114,7 +120,7 @@ struct execute_on_stream : execute_on_stream_base<execute_on_stream>
   __host__ __device__
   execute_on_stream() : base_t(){};
   __host__ __device__
-  execute_on_stream(cudaStream_t stream) 
+  execute_on_stream(cudaStream_t stream)
   : base_t(stream){};
 };
 
@@ -125,7 +131,7 @@ struct execute_on_stream_nosync : execute_on_stream_nosync_base<execute_on_strea
   __host__ __device__
   execute_on_stream_nosync() : base_t(){};
   __host__ __device__
-  execute_on_stream_nosync(cudaStream_t stream) 
+  execute_on_stream_nosync(cudaStream_t stream)
   : base_t(stream){};
 };
 
@@ -191,11 +197,11 @@ THRUST_INLINE_CONSTANT par_t par;
 /*! \p thrust::cuda::par_nosync is a parallel execution policy targeting Thrust's CUDA device backend.
  *  Similar to \p thrust::cuda::par it allows execution of Thrust algorithms in a specific CUDA stream.
  *
- *  \p thrust::cuda::par_nosync indicates that an algorithm is free to avoid any synchronization of the 
+ *  \p thrust::cuda::par_nosync indicates that an algorithm is free to avoid any synchronization of the
  *  associated stream that is not strictly required for correctness. Additionally, algorithms may return
  *  before the corresponding kernels are completed, similar to asynchronous kernel launches via <<< >>> syntax.
  *  The user must take care to perform explicit synchronization if necessary.
- *  
+ *
  *  The following code snippet demonstrates how to use \p thrust::cuda::par_nosync :
  *
  *  \code
diff --git a/thrust/thrust/system/cuda/detail/par_to_seq.h b/thrust/thrust/system/cuda/detail/par_to_seq.h
index e710f017b65..a03fa182342 100644
--- a/thrust/thrust/system/cuda/detail/par_to_seq.h
+++ b/thrust/thrust/system/cuda/detail/par_to_seq.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/seq.h>
 #include <thrust/system/cuda/detail/par.h>
 
diff --git a/thrust/thrust/system/cuda/detail/parallel_for.h b/thrust/thrust/system/cuda/detail/parallel_for.h
index 43c3297aaef..8ee1be41bf9 100644
--- a/thrust/thrust/system/cuda/detail/parallel_for.h
+++ b/thrust/thrust/system/cuda/detail/parallel_for.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 
 #include <thrust/detail/type_traits/result_of_adaptable_function.h>
diff --git a/thrust/thrust/system/cuda/detail/partition.h b/thrust/thrust/system/cuda/detail/partition.h
index fad75eb0d83..7ee25157326 100644
--- a/thrust/thrust/system/cuda/detail/partition.h
+++ b/thrust/thrust/system/cuda/detail/partition.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 
 #include <thrust/detail/cstdint.h>
diff --git a/thrust/thrust/system/cuda/detail/per_device_resource.h b/thrust/thrust/system/cuda/detail/per_device_resource.h
index 414ea778813..354fe5d6bce 100644
--- a/thrust/thrust/system/cuda/detail/per_device_resource.h
+++ b/thrust/thrust/system/cuda/detail/per_device_resource.h
@@ -28,6 +28,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp11_required.h>
 
 #if THRUST_CPP_DIALECT >= 2011
diff --git a/thrust/thrust/system/cuda/detail/reduce.h b/thrust/thrust/system/cuda/detail/reduce.h
index 41d9075da75..3258064ddef 100644
--- a/thrust/thrust/system/cuda/detail/reduce.h
+++ b/thrust/thrust/system/cuda/detail/reduce.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 
 #include <thrust/detail/alignment.h>
diff --git a/thrust/thrust/system/cuda/detail/reduce_by_key.h b/thrust/thrust/system/cuda/detail/reduce_by_key.h
index 2933d062acd..bcb5ad60fc8 100644
--- a/thrust/thrust/system/cuda/detail/reduce_by_key.h
+++ b/thrust/thrust/system/cuda/detail/reduce_by_key.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 
 #include <thrust/detail/alignment.h>
diff --git a/thrust/thrust/system/cuda/detail/remove.h b/thrust/thrust/system/cuda/detail/remove.h
index 836d8f5eab6..7d8d85b328d 100644
--- a/thrust/thrust/system/cuda/detail/remove.h
+++ b/thrust/thrust/system/cuda/detail/remove.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 #include <thrust/system/cuda/detail/copy_if.h>
 
@@ -35,7 +41,7 @@ THRUST_NAMESPACE_BEGIN
 namespace cuda_cub {
 
 // in-place
-  
+
 template <class Derived,
           class InputIt,
           class StencilIt,
diff --git a/thrust/thrust/system/cuda/detail/replace.h b/thrust/thrust/system/cuda/detail/replace.h
index af8b8fa9563..31e25e66a33 100644
--- a/thrust/thrust/system/cuda/detail/replace.h
+++ b/thrust/thrust/system/cuda/detail/replace.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 #include <thrust/system/cuda/detail/transform.h>
 #include <thrust/detail/internal_functional.h>
diff --git a/thrust/thrust/system/cuda/detail/reverse.h b/thrust/thrust/system/cuda/detail/reverse.h
index 7c4cb867e32..9f4f831c283 100644
--- a/thrust/thrust/system/cuda/detail/reverse.h
+++ b/thrust/thrust/system/cuda/detail/reverse.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 #include <thrust/system/cuda/detail/execution_policy.h>
 
diff --git a/thrust/thrust/system/cuda/detail/scan.h b/thrust/thrust/system/cuda/detail/scan.h
index fdab8df8447..e2f97a07203 100644
--- a/thrust/thrust/system/cuda/detail/scan.h
+++ b/thrust/thrust/system/cuda/detail/scan.h
@@ -28,9 +28,14 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 
-#include <thrust/detail/config/exec_check_disable.h>
 #include <thrust/detail/cstdint.h>
 #include <thrust/detail/type_traits.h>
 #include <thrust/distance.h>
diff --git a/thrust/thrust/system/cuda/detail/scan_by_key.h b/thrust/thrust/system/cuda/detail/scan_by_key.h
index 0407779c6e0..c358bf2e232 100644
--- a/thrust/thrust/system/cuda/detail/scan_by_key.h
+++ b/thrust/thrust/system/cuda/detail/scan_by_key.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 
 #include <thrust/distance.h>
diff --git a/thrust/thrust/system/cuda/detail/scatter.h b/thrust/thrust/system/cuda/detail/scatter.h
index e297d782d71..84f5afe0bff 100644
--- a/thrust/thrust/system/cuda/detail/scatter.h
+++ b/thrust/thrust/system/cuda/detail/scatter.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 #include <thrust/system/cuda/detail/transform.h>
 #include <thrust/iterator/permutation_iterator.h>
diff --git a/thrust/thrust/system/cuda/detail/sequence.h b/thrust/thrust/system/cuda/detail/sequence.h
index c6ae90664ad..c3f89646b4e 100644
--- a/thrust/thrust/system/cuda/detail/sequence.h
+++ b/thrust/thrust/system/cuda/detail/sequence.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
-// this system has no special version of this algorithm 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
+// this system has no special version of this algorithm
 
diff --git a/thrust/thrust/system/cuda/detail/set_operations.h b/thrust/thrust/system/cuda/detail/set_operations.h
index 98bb4bb5d69..e6b868c6d51 100644
--- a/thrust/thrust/system/cuda/detail/set_operations.h
+++ b/thrust/thrust/system/cuda/detail/set_operations.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 
 #include <thrust/detail/alignment.h>
diff --git a/thrust/thrust/system/cuda/detail/sort.h b/thrust/thrust/system/cuda/detail/sort.h
index db4c211b377..ddb0cf66344 100644
--- a/thrust/thrust/system/cuda/detail/sort.h
+++ b/thrust/thrust/system/cuda/detail/sort.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 #include <thrust/distance.h>
 #include <thrust/extrema.h>
diff --git a/thrust/thrust/system/cuda/detail/swap_ranges.h b/thrust/thrust/system/cuda/detail/swap_ranges.h
index 8f9e4fa8aaf..bbf51789c60 100644
--- a/thrust/thrust/system/cuda/detail/swap_ranges.h
+++ b/thrust/thrust/system/cuda/detail/swap_ranges.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 #include <iterator>
 #include <thrust/system/cuda/detail/transform.h>
diff --git a/thrust/thrust/system/cuda/detail/tabulate.h b/thrust/thrust/system/cuda/detail/tabulate.h
index 67edb8574d4..562f7037706 100644
--- a/thrust/thrust/system/cuda/detail/tabulate.h
+++ b/thrust/thrust/system/cuda/detail/tabulate.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 #include <thrust/distance.h>
 #include <thrust/system/cuda/config.h>
diff --git a/thrust/thrust/system/cuda/detail/temporary_buffer.h b/thrust/thrust/system/cuda/detail/temporary_buffer.h
index 6b527614162..1daf6c04a34 100644
--- a/thrust/thrust/system/cuda/detail/temporary_buffer.h
+++ b/thrust/thrust/system/cuda/detail/temporary_buffer.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system has no special temporary buffer functions
 
diff --git a/thrust/thrust/system/cuda/detail/terminate.h b/thrust/thrust/system/cuda/detail/terminate.h
index 226c9d5ac2a..2a7ebc43153 100644
--- a/thrust/thrust/system/cuda/detail/terminate.h
+++ b/thrust/thrust/system/cuda/detail/terminate.h
@@ -28,6 +28,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/cuda/detail/util.h>
 #include <cstdio>
 
diff --git a/thrust/thrust/system/cuda/detail/transform.h b/thrust/thrust/system/cuda/detail/transform.h
index 3cf171a4724..473a0ba0989 100644
--- a/thrust/thrust/system/cuda/detail/transform.h
+++ b/thrust/thrust/system/cuda/detail/transform.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 #include <thrust/system/cuda/config.h>
 
diff --git a/thrust/thrust/system/cuda/detail/transform_reduce.h b/thrust/thrust/system/cuda/detail/transform_reduce.h
index 60efaae5a55..c0c96b2c594 100644
--- a/thrust/thrust/system/cuda/detail/transform_reduce.h
+++ b/thrust/thrust/system/cuda/detail/transform_reduce.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 #include <iterator>
 #include <thrust/system/cuda/detail/reduce.h>
diff --git a/thrust/thrust/system/cuda/detail/transform_scan.h b/thrust/thrust/system/cuda/detail/transform_scan.h
index 8f14ca8f7f2..fd76a84e795 100644
--- a/thrust/thrust/system/cuda/detail/transform_scan.h
+++ b/thrust/thrust/system/cuda/detail/transform_scan.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 #include <iterator>
 #include <thrust/detail/type_traits.h>
diff --git a/thrust/thrust/system/cuda/detail/uninitialized_copy.h b/thrust/thrust/system/cuda/detail/uninitialized_copy.h
index f21b7c0d62f..3fa8940b627 100644
--- a/thrust/thrust/system/cuda/detail/uninitialized_copy.h
+++ b/thrust/thrust/system/cuda/detail/uninitialized_copy.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 #include <iterator>
 #include <thrust/distance.h>
diff --git a/thrust/thrust/system/cuda/detail/uninitialized_fill.h b/thrust/thrust/system/cuda/detail/uninitialized_fill.h
index 96b9702016b..368349db104 100644
--- a/thrust/thrust/system/cuda/detail/uninitialized_fill.h
+++ b/thrust/thrust/system/cuda/detail/uninitialized_fill.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 #include <iterator>
 #include <thrust/distance.h>
diff --git a/thrust/thrust/system/cuda/detail/unique.h b/thrust/thrust/system/cuda/detail/unique.h
index 653ffa79ab6..aba3a4dc2a9 100644
--- a/thrust/thrust/system/cuda/detail/unique.h
+++ b/thrust/thrust/system/cuda/detail/unique.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 
 #include <thrust/advance.h>
@@ -788,7 +794,7 @@ struct zip_adj_not_predicate {
   bool __host__ __device__ operator()(TupleType&& tuple) {
       return !binary_pred(thrust::get<0>(tuple), thrust::get<1>(tuple));
   }
-  
+
   BinaryPred binary_pred;
 };
 
diff --git a/thrust/thrust/system/cuda/detail/unique_by_key.h b/thrust/thrust/system/cuda/detail/unique_by_key.h
index d5ce8e786f8..f0f7ed93c26 100644
--- a/thrust/thrust/system/cuda/detail/unique_by_key.h
+++ b/thrust/thrust/system/cuda/detail/unique_by_key.h
@@ -28,6 +28,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
 
 #include <thrust/detail/alignment.h>
diff --git a/thrust/thrust/system/cuda/detail/util.h b/thrust/thrust/system/cuda/detail/util.h
index 6d9e3681d8c..c610629af26 100644
--- a/thrust/thrust/system/cuda/detail/util.h
+++ b/thrust/thrust/system/cuda/detail/util.h
@@ -26,16 +26,23 @@
  ******************************************************************************/
 #pragma once
 
+#include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <cstdio>
 #include <exception>
-#include <thrust/detail/config.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/cuda/detail/execution_policy.h>
 #include <thrust/system_error.h>
 #include <thrust/system/cuda/error.h>
 
 #include <cub/detail/device_synchronize.cuh>
-#include <cub/util_arch.cuh>
+#include <cub/config.cuh>
 #include <cub/util_device.cuh>
 
 #include <nv/target>
@@ -318,7 +325,7 @@ struct transform_input_iterator_t
 
   // UnaryOp might not be copy assignable, such as when it is a lambda.  Define
   // an explicit copy assignment operator that doesn't try to assign it.
-  __host__ __device__ 
+  __host__ __device__
   self_t& operator=(const self_t& o)
   {
     input = o.input;
diff --git a/thrust/thrust/system/cuda/error.h b/thrust/thrust/system/cuda/error.h
index b180f8347e0..5b1d764b27f 100644
--- a/thrust/thrust/system/cuda/error.h
+++ b/thrust/thrust/system/cuda/error.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/type_traits.h>
 #include <thrust/system/error_code.h>
 #include <thrust/system/cuda/detail/guarded_driver_types.h>
diff --git a/thrust/thrust/system/cuda/execution_policy.h b/thrust/thrust/system/cuda/execution_policy.h
index c171ac3d969..f1cea2c80b9 100644
--- a/thrust/thrust/system/cuda/execution_policy.h
+++ b/thrust/thrust/system/cuda/execution_policy.h
@@ -27,5 +27,11 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/cuda/detail/execution_policy.h>
 #include <thrust/system/cuda/detail/par.h>
diff --git a/thrust/thrust/system/cuda/future.h b/thrust/thrust/system/cuda/future.h
index 79bfc9134d4..c37f4851c22 100644
--- a/thrust/thrust/system/cuda/future.h
+++ b/thrust/thrust/system/cuda/future.h
@@ -6,6 +6,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp14_required.h>
 
 #if THRUST_CPP_DIALECT >= 2014
@@ -52,14 +58,14 @@ using thrust::system::cuda::when_all;
 } // namespace cuda
 
 template <typename DerivedPolicy>
-__host__ 
+__host__
 thrust::cuda::unique_eager_event
 unique_eager_event_type(
   thrust::cuda::execution_policy<DerivedPolicy> const&
 ) noexcept;
 
 template <typename T, typename DerivedPolicy>
-__host__ 
+__host__
 thrust::cuda::unique_eager_future<T>
 unique_eager_future_type(
   thrust::cuda::execution_policy<DerivedPolicy> const&
diff --git a/thrust/thrust/system/cuda/memory.h b/thrust/thrust/system/cuda/memory.h
index eb8020adb9e..da0708b012f 100644
--- a/thrust/thrust/system/cuda/memory.h
+++ b/thrust/thrust/system/cuda/memory.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/cuda/memory_resource.h>
 #include <thrust/memory.h>
 #include <thrust/detail/type_traits.h>
diff --git a/thrust/thrust/system/cuda/memory_resource.h b/thrust/thrust/system/cuda/memory_resource.h
index 4bf534e401e..68af96b5084 100644
--- a/thrust/thrust/system/cuda/memory_resource.h
+++ b/thrust/thrust/system/cuda/memory_resource.h
@@ -22,6 +22,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/mr/memory_resource.h>
 #include <thrust/system/cuda/detail/guarded_cuda_runtime_api.h>
 #include <thrust/system/cuda/pointer.h>
diff --git a/thrust/thrust/system/cuda/pointer.h b/thrust/thrust/system/cuda/pointer.h
index ace77fbaef4..793751cb040 100644
--- a/thrust/thrust/system/cuda/pointer.h
+++ b/thrust/thrust/system/cuda/pointer.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <type_traits>
 #include <thrust/system/cuda/detail/execution_policy.h>
 #include <thrust/detail/pointer.h>
diff --git a/thrust/thrust/system/cuda/vector.h b/thrust/thrust/system/cuda/vector.h
index fafc7bf17e3..ed6114aadab 100644
--- a/thrust/thrust/system/cuda/vector.h
+++ b/thrust/thrust/system/cuda/vector.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/cuda/memory.h>
 #include <thrust/detail/vector_base.h>
 #include <vector>
diff --git a/thrust/thrust/system/detail/adl/adjacent_difference.h b/thrust/thrust/system/detail/adl/adjacent_difference.h
index c6f6c72820a..980dc4fce48 100644
--- a/thrust/thrust/system/detail/adl/adjacent_difference.h
+++ b/thrust/thrust/system/detail/adl/adjacent_difference.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the adjacent_difference.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch adjacent_difference
diff --git a/thrust/thrust/system/detail/adl/assign_value.h b/thrust/thrust/system/detail/adl/assign_value.h
index d38934affd1..c2a9cda7f78 100644
--- a/thrust/thrust/system/detail/adl/assign_value.h
+++ b/thrust/thrust/system/detail/adl/assign_value.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the assign_value.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch assign_value
diff --git a/thrust/thrust/system/detail/adl/async/copy.h b/thrust/thrust/system/detail/adl/async/copy.h
index 72debb3b667..b9d6a335f91 100644
--- a/thrust/thrust/system/detail/adl/async/copy.h
+++ b/thrust/thrust/system/detail/adl/async/copy.h
@@ -22,6 +22,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 //#include <thrust/system/detail/sequential/async/copy.h>
 
 //#define __THRUST_HOST_SYSTEM_ASYNC_COPY_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/async/copy.h>
diff --git a/thrust/thrust/system/detail/adl/async/for_each.h b/thrust/thrust/system/detail/adl/async/for_each.h
index 08347f65927..e0e71ea8a36 100644
--- a/thrust/thrust/system/detail/adl/async/for_each.h
+++ b/thrust/thrust/system/detail/adl/async/for_each.h
@@ -22,6 +22,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 //#include <thrust/system/detail/sequential/async/for_each.h>
 
 //#define __THRUST_HOST_SYSTEM_ASYNC_FOR_EACH_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/async/for_each.h>
diff --git a/thrust/thrust/system/detail/adl/async/reduce.h b/thrust/thrust/system/detail/adl/async/reduce.h
index f13ab02fdca..0e157cb7970 100644
--- a/thrust/thrust/system/detail/adl/async/reduce.h
+++ b/thrust/thrust/system/detail/adl/async/reduce.h
@@ -22,6 +22,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 //#include <thrust/system/detail/sequential/async/reduce.h>
 
 //#define __THRUST_HOST_SYSTEM_ASYNC_REDUCE_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/async/reduce.h>
diff --git a/thrust/thrust/system/detail/adl/async/scan.h b/thrust/thrust/system/detail/adl/async/scan.h
index a2a90618b48..ac60330e1c7 100644
--- a/thrust/thrust/system/detail/adl/async/scan.h
+++ b/thrust/thrust/system/detail/adl/async/scan.h
@@ -22,6 +22,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 //#include <thrust/system/detail/sequential/async/scan.h>
 
 //#define __THRUST_HOST_SYSTEM_ASYNC_SCAN_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/async/scan.h>
diff --git a/thrust/thrust/system/detail/adl/async/sort.h b/thrust/thrust/system/detail/adl/async/sort.h
index c3a83ad404f..8f07711e6b7 100644
--- a/thrust/thrust/system/detail/adl/async/sort.h
+++ b/thrust/thrust/system/detail/adl/async/sort.h
@@ -22,6 +22,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 //#include <thrust/system/detail/sequential/async/sort.h>
 
 //#define __THRUST_HOST_SYSTEM_ASYNC_SORT_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/async/sort.h>
diff --git a/thrust/thrust/system/detail/adl/async/transform.h b/thrust/thrust/system/detail/adl/async/transform.h
index abb2163ead0..7dc39b0eba0 100644
--- a/thrust/thrust/system/detail/adl/async/transform.h
+++ b/thrust/thrust/system/detail/adl/async/transform.h
@@ -22,6 +22,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 //#include <thrust/system/detail/sequential/async/transform.h>
 
 //#define __THRUST_HOST_SYSTEM_ASYNC_TRANSFORM_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/async/transform.h>
diff --git a/thrust/thrust/system/detail/adl/binary_search.h b/thrust/thrust/system/detail/adl/binary_search.h
index 2f9ac06df9b..a8149753f5c 100644
--- a/thrust/thrust/system/detail/adl/binary_search.h
+++ b/thrust/thrust/system/detail/adl/binary_search.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the binary_search.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch binary_search
diff --git a/thrust/thrust/system/detail/adl/copy.h b/thrust/thrust/system/detail/adl/copy.h
index 0035b83efee..e103a6ed333 100644
--- a/thrust/thrust/system/detail/adl/copy.h
+++ b/thrust/thrust/system/detail/adl/copy.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the copy.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch copy
diff --git a/thrust/thrust/system/detail/adl/copy_if.h b/thrust/thrust/system/detail/adl/copy_if.h
index 31adaf8e1a7..efec0cb9445 100644
--- a/thrust/thrust/system/detail/adl/copy_if.h
+++ b/thrust/thrust/system/detail/adl/copy_if.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the copy_if.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch copy_if
diff --git a/thrust/thrust/system/detail/adl/count.h b/thrust/thrust/system/detail/adl/count.h
index 5d6f1f748ff..60dbab4f29c 100644
--- a/thrust/thrust/system/detail/adl/count.h
+++ b/thrust/thrust/system/detail/adl/count.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the count.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch count
diff --git a/thrust/thrust/system/detail/adl/equal.h b/thrust/thrust/system/detail/adl/equal.h
index 6b02e33b857..3f5bb54b1c0 100644
--- a/thrust/thrust/system/detail/adl/equal.h
+++ b/thrust/thrust/system/detail/adl/equal.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the equal.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch equal
diff --git a/thrust/thrust/system/detail/adl/extrema.h b/thrust/thrust/system/detail/adl/extrema.h
index 62fb39be922..ad4ae3b5a61 100644
--- a/thrust/thrust/system/detail/adl/extrema.h
+++ b/thrust/thrust/system/detail/adl/extrema.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the extrema.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch extrema
diff --git a/thrust/thrust/system/detail/adl/fill.h b/thrust/thrust/system/detail/adl/fill.h
index f76a81b4f34..adfce5a49d5 100644
--- a/thrust/thrust/system/detail/adl/fill.h
+++ b/thrust/thrust/system/detail/adl/fill.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the fill.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch fill
diff --git a/thrust/thrust/system/detail/adl/find.h b/thrust/thrust/system/detail/adl/find.h
index 8d85e09a3ab..1b8543aaaa8 100644
--- a/thrust/thrust/system/detail/adl/find.h
+++ b/thrust/thrust/system/detail/adl/find.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the find.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch find
diff --git a/thrust/thrust/system/detail/adl/for_each.h b/thrust/thrust/system/detail/adl/for_each.h
index 8509edca363..3a9e3adf735 100644
--- a/thrust/thrust/system/detail/adl/for_each.h
+++ b/thrust/thrust/system/detail/adl/for_each.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the for_each.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch for_each
diff --git a/thrust/thrust/system/detail/adl/gather.h b/thrust/thrust/system/detail/adl/gather.h
index 242da3c9095..82ac687ca64 100644
--- a/thrust/thrust/system/detail/adl/gather.h
+++ b/thrust/thrust/system/detail/adl/gather.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the gather.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch gather
diff --git a/thrust/thrust/system/detail/adl/generate.h b/thrust/thrust/system/detail/adl/generate.h
index 5b1d7b4bac5..47dbad27455 100644
--- a/thrust/thrust/system/detail/adl/generate.h
+++ b/thrust/thrust/system/detail/adl/generate.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the generate.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch generate
diff --git a/thrust/thrust/system/detail/adl/get_value.h b/thrust/thrust/system/detail/adl/get_value.h
index 306eb423eb4..4817055baa9 100644
--- a/thrust/thrust/system/detail/adl/get_value.h
+++ b/thrust/thrust/system/detail/adl/get_value.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the get_value.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch get_value
diff --git a/thrust/thrust/system/detail/adl/inner_product.h b/thrust/thrust/system/detail/adl/inner_product.h
index 9423b1bdbf1..ff4756bc5af 100644
--- a/thrust/thrust/system/detail/adl/inner_product.h
+++ b/thrust/thrust/system/detail/adl/inner_product.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the inner_product.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch inner_product
diff --git a/thrust/thrust/system/detail/adl/iter_swap.h b/thrust/thrust/system/detail/adl/iter_swap.h
index d9da52a6274..8bf4f87f123 100644
--- a/thrust/thrust/system/detail/adl/iter_swap.h
+++ b/thrust/thrust/system/detail/adl/iter_swap.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the iter_swap.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch iter_swap
diff --git a/thrust/thrust/system/detail/adl/logical.h b/thrust/thrust/system/detail/adl/logical.h
index bdaad4d293a..9bbdaac40fb 100644
--- a/thrust/thrust/system/detail/adl/logical.h
+++ b/thrust/thrust/system/detail/adl/logical.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the logical.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch logical
diff --git a/thrust/thrust/system/detail/adl/malloc_and_free.h b/thrust/thrust/system/detail/adl/malloc_and_free.h
index c36db0270dd..9234da85e18 100644
--- a/thrust/thrust/system/detail/adl/malloc_and_free.h
+++ b/thrust/thrust/system/detail/adl/malloc_and_free.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the malloc_and_free.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch malloc_and_free
diff --git a/thrust/thrust/system/detail/adl/merge.h b/thrust/thrust/system/detail/adl/merge.h
index 7abca9bcf3f..7974823c9fe 100644
--- a/thrust/thrust/system/detail/adl/merge.h
+++ b/thrust/thrust/system/detail/adl/merge.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the merge.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch merge
diff --git a/thrust/thrust/system/detail/adl/mismatch.h b/thrust/thrust/system/detail/adl/mismatch.h
index 74feb826928..1313425eaf0 100644
--- a/thrust/thrust/system/detail/adl/mismatch.h
+++ b/thrust/thrust/system/detail/adl/mismatch.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the mismatch.h header
 // of the host and device systems. It should be #included in any
 // code which uses adl to dispatch mismatch
diff --git a/thrust/thrust/system/detail/adl/partition.h b/thrust/thrust/system/detail/adl/partition.h
index a45f845a5c6..60d2d8de2f3 100644
--- a/thrust/thrust/system/detail/adl/partition.h
+++ b/thrust/thrust/system/detail/adl/partition.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the partition.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch partition
diff --git a/thrust/thrust/system/detail/adl/per_device_resource.h b/thrust/thrust/system/detail/adl/per_device_resource.h
index 721f49e03fd..afc5df0031b 100644
--- a/thrust/thrust/system/detail/adl/per_device_resource.h
+++ b/thrust/thrust/system/detail/adl/per_device_resource.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the per_device_resource.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch per_device_resource
diff --git a/thrust/thrust/system/detail/adl/reduce.h b/thrust/thrust/system/detail/adl/reduce.h
index 8a9673b3f95..ce86610ebbc 100644
--- a/thrust/thrust/system/detail/adl/reduce.h
+++ b/thrust/thrust/system/detail/adl/reduce.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the reduce.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch reduce
diff --git a/thrust/thrust/system/detail/adl/reduce_by_key.h b/thrust/thrust/system/detail/adl/reduce_by_key.h
index 0605f9befaf..49afe209f44 100644
--- a/thrust/thrust/system/detail/adl/reduce_by_key.h
+++ b/thrust/thrust/system/detail/adl/reduce_by_key.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the reduce_by_key.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch reduce_by_key
diff --git a/thrust/thrust/system/detail/adl/remove.h b/thrust/thrust/system/detail/adl/remove.h
index c281379d54e..ba14b96bd4f 100644
--- a/thrust/thrust/system/detail/adl/remove.h
+++ b/thrust/thrust/system/detail/adl/remove.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the remove.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch remove
diff --git a/thrust/thrust/system/detail/adl/replace.h b/thrust/thrust/system/detail/adl/replace.h
index d8fb5746f1c..c2efd40e80a 100644
--- a/thrust/thrust/system/detail/adl/replace.h
+++ b/thrust/thrust/system/detail/adl/replace.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the replace.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch replace
diff --git a/thrust/thrust/system/detail/adl/reverse.h b/thrust/thrust/system/detail/adl/reverse.h
index f6bd8947ee4..f0a4972c648 100644
--- a/thrust/thrust/system/detail/adl/reverse.h
+++ b/thrust/thrust/system/detail/adl/reverse.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the reverse.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch reverse
diff --git a/thrust/thrust/system/detail/adl/scan.h b/thrust/thrust/system/detail/adl/scan.h
index a2491041058..904e9c4f9ce 100644
--- a/thrust/thrust/system/detail/adl/scan.h
+++ b/thrust/thrust/system/detail/adl/scan.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the scan.h header
 // of the host and device systems. It should be #included in any
 // code which uses adl to dispatch scan
diff --git a/thrust/thrust/system/detail/adl/scan_by_key.h b/thrust/thrust/system/detail/adl/scan_by_key.h
index 94f73503cac..bb4f9d4af76 100644
--- a/thrust/thrust/system/detail/adl/scan_by_key.h
+++ b/thrust/thrust/system/detail/adl/scan_by_key.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the scan_by_key.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch scan_by_key
diff --git a/thrust/thrust/system/detail/adl/scatter.h b/thrust/thrust/system/detail/adl/scatter.h
index d9f42b28b13..d562fbec037 100644
--- a/thrust/thrust/system/detail/adl/scatter.h
+++ b/thrust/thrust/system/detail/adl/scatter.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the scatter.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch scatter
diff --git a/thrust/thrust/system/detail/adl/sequence.h b/thrust/thrust/system/detail/adl/sequence.h
index d3c2a20f47e..4e364a84db2 100644
--- a/thrust/thrust/system/detail/adl/sequence.h
+++ b/thrust/thrust/system/detail/adl/sequence.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the sequence.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch sequence
diff --git a/thrust/thrust/system/detail/adl/set_operations.h b/thrust/thrust/system/detail/adl/set_operations.h
index 7d09355e12a..dae16bd3714 100644
--- a/thrust/thrust/system/detail/adl/set_operations.h
+++ b/thrust/thrust/system/detail/adl/set_operations.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the set_operations.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch set_operations
diff --git a/thrust/thrust/system/detail/adl/sort.h b/thrust/thrust/system/detail/adl/sort.h
index 1f6118c90bf..d0687f311b6 100644
--- a/thrust/thrust/system/detail/adl/sort.h
+++ b/thrust/thrust/system/detail/adl/sort.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the sort.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch sort
diff --git a/thrust/thrust/system/detail/adl/swap_ranges.h b/thrust/thrust/system/detail/adl/swap_ranges.h
index 1ca3719d96c..af2393b8125 100644
--- a/thrust/thrust/system/detail/adl/swap_ranges.h
+++ b/thrust/thrust/system/detail/adl/swap_ranges.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the swap_ranges.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch swap_ranges
diff --git a/thrust/thrust/system/detail/adl/tabulate.h b/thrust/thrust/system/detail/adl/tabulate.h
index 6ae2b22a5cb..ba9a013ac7a 100644
--- a/thrust/thrust/system/detail/adl/tabulate.h
+++ b/thrust/thrust/system/detail/adl/tabulate.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the tabulate.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch tabulate
diff --git a/thrust/thrust/system/detail/adl/temporary_buffer.h b/thrust/thrust/system/detail/adl/temporary_buffer.h
index 0cada5ee4b1..e916fe63d8f 100644
--- a/thrust/thrust/system/detail/adl/temporary_buffer.h
+++ b/thrust/thrust/system/detail/adl/temporary_buffer.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the temporary_buffer.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch get_temporary_buffer or return_temporary_buffer
diff --git a/thrust/thrust/system/detail/adl/transform.h b/thrust/thrust/system/detail/adl/transform.h
index b70333093fd..a3e0757cc7c 100644
--- a/thrust/thrust/system/detail/adl/transform.h
+++ b/thrust/thrust/system/detail/adl/transform.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the transform.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch transform
diff --git a/thrust/thrust/system/detail/adl/transform_reduce.h b/thrust/thrust/system/detail/adl/transform_reduce.h
index e3f9494dfa6..45483fea753 100644
--- a/thrust/thrust/system/detail/adl/transform_reduce.h
+++ b/thrust/thrust/system/detail/adl/transform_reduce.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the transform_reduce.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch transform_reduce
diff --git a/thrust/thrust/system/detail/adl/transform_scan.h b/thrust/thrust/system/detail/adl/transform_scan.h
index 3a05c7eeed9..972d46c723a 100644
--- a/thrust/thrust/system/detail/adl/transform_scan.h
+++ b/thrust/thrust/system/detail/adl/transform_scan.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the transform_scan.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch transform_scan
diff --git a/thrust/thrust/system/detail/adl/uninitialized_copy.h b/thrust/thrust/system/detail/adl/uninitialized_copy.h
index a13b18aa8d7..f31e4b2c66d 100644
--- a/thrust/thrust/system/detail/adl/uninitialized_copy.h
+++ b/thrust/thrust/system/detail/adl/uninitialized_copy.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the uninitialized_copy.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch uninitialized_copy
diff --git a/thrust/thrust/system/detail/adl/uninitialized_fill.h b/thrust/thrust/system/detail/adl/uninitialized_fill.h
index 98b57836e12..bd98e450f46 100644
--- a/thrust/thrust/system/detail/adl/uninitialized_fill.h
+++ b/thrust/thrust/system/detail/adl/uninitialized_fill.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the uninitialized_fill.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch uninitialized_fill
diff --git a/thrust/thrust/system/detail/adl/unique.h b/thrust/thrust/system/detail/adl/unique.h
index 4082f529926..d866ad04a49 100644
--- a/thrust/thrust/system/detail/adl/unique.h
+++ b/thrust/thrust/system/detail/adl/unique.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the unique.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch unique
diff --git a/thrust/thrust/system/detail/adl/unique_by_key.h b/thrust/thrust/system/detail/adl/unique_by_key.h
index dcf9acd42cd..d6e6020550a 100644
--- a/thrust/thrust/system/detail/adl/unique_by_key.h
+++ b/thrust/thrust/system/detail/adl/unique_by_key.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // the purpose of this header is to #include the unique_by_key.h header
 // of the sequential, host, and device systems. It should be #included in any
 // code which uses adl to dispatch unique_by_key
diff --git a/thrust/thrust/system/detail/bad_alloc.h b/thrust/thrust/system/detail/bad_alloc.h
index ae5dd5994ef..32b1bf0a75f 100644
--- a/thrust/thrust/system/detail/bad_alloc.h
+++ b/thrust/thrust/system/detail/bad_alloc.h
@@ -17,11 +17,17 @@
 
 #pragma once
 
+#include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <new>
 #include <string>
 
-#include <thrust/detail/config.h>
-
 THRUST_NAMESPACE_BEGIN
 namespace system
 {
@@ -51,7 +57,7 @@ class bad_alloc
   private:
     std::string m_what;
 }; // end bad_alloc
-  
+
 } // end detail
 } // end system
 THRUST_NAMESPACE_END
diff --git a/thrust/thrust/system/detail/errno.h b/thrust/thrust/system/detail/errno.h
index 69cb2bd9806..8e29bf7cb6b 100644
--- a/thrust/thrust/system/detail/errno.h
+++ b/thrust/thrust/system/detail/errno.h
@@ -19,6 +19,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // The rationale for the existence of these apparently redundant definitions is
 // to provide them portably and to avoid bringing in system headers which might
 // pollute the global namespace. These identifiers are in lowercase to avoid
diff --git a/thrust/thrust/system/detail/error_category.inl b/thrust/thrust/system/detail/error_category.inl
index 45fd15a3fd5..68332dd3644 100644
--- a/thrust/thrust/system/detail/error_category.inl
+++ b/thrust/thrust/system/detail/error_category.inl
@@ -19,6 +19,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/system/error_code.h>
 #include <thrust/system/detail/errno.h>
 #include <thrust/functional.h>
diff --git a/thrust/thrust/system/detail/error_code.inl b/thrust/thrust/system/detail/error_code.inl
index 2b819c0485e..16611bf1b5b 100644
--- a/thrust/thrust/system/detail/error_code.inl
+++ b/thrust/thrust/system/detail/error_code.inl
@@ -19,6 +19,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/system/error_code.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/error_condition.inl b/thrust/thrust/system/detail/error_condition.inl
index 0daf1f2938b..903a4fc20b6 100644
--- a/thrust/thrust/system/detail/error_condition.inl
+++ b/thrust/thrust/system/detail/error_condition.inl
@@ -19,6 +19,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/system/detail/error_condition.inl>
 #include <thrust/functional.h>
 
diff --git a/thrust/thrust/system/detail/generic/adjacent_difference.h b/thrust/thrust/system/detail/generic/adjacent_difference.h
index 43592e15b03..0e0b13d841c 100644
--- a/thrust/thrust/system/detail/generic/adjacent_difference.h
+++ b/thrust/thrust/system/detail/generic/adjacent_difference.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/generic/adjacent_difference.inl b/thrust/thrust/system/detail/generic/adjacent_difference.inl
index 50412932888..1a6128cfa41 100644
--- a/thrust/thrust/system/detail/generic/adjacent_difference.inl
+++ b/thrust/thrust/system/detail/generic/adjacent_difference.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/adjacent_difference.h>
 #include <thrust/adjacent_difference.h>
 #include <thrust/functional.h>
diff --git a/thrust/thrust/system/detail/generic/advance.h b/thrust/thrust/system/detail/generic/advance.h
index 4d6562e00fa..cc298954174 100644
--- a/thrust/thrust/system/detail/generic/advance.h
+++ b/thrust/thrust/system/detail/generic/advance.h
@@ -19,6 +19,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 THRUST_NAMESPACE_BEGIN
 namespace system
 {
diff --git a/thrust/thrust/system/detail/generic/advance.inl b/thrust/thrust/system/detail/generic/advance.inl
index 21555ebb03f..c2600cc56cc 100644
--- a/thrust/thrust/system/detail/generic/advance.inl
+++ b/thrust/thrust/system/detail/generic/advance.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/advance.h>
 #include <thrust/iterator/iterator_traits.h>
 
diff --git a/thrust/thrust/system/detail/generic/binary_search.h b/thrust/thrust/system/detail/generic/binary_search.h
index 6603f6c302f..77a4ce73ea6 100644
--- a/thrust/thrust/system/detail/generic/binary_search.h
+++ b/thrust/thrust/system/detail/generic/binary_search.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 
 THRUST_NAMESPACE_BEGIN
@@ -35,7 +41,7 @@ namespace generic
 
 template <typename DerivedPolicy, typename ForwardIterator, typename T>
 __host__ __device__
-ForwardIterator lower_bound(thrust::execution_policy<DerivedPolicy> &exec, 
+ForwardIterator lower_bound(thrust::execution_policy<DerivedPolicy> &exec,
                             ForwardIterator begin,
                             ForwardIterator end,
                             const T& value);
@@ -45,7 +51,7 @@ __host__ __device__
 ForwardIterator lower_bound(thrust::execution_policy<DerivedPolicy> &exec,
                             ForwardIterator begin,
                             ForwardIterator end,
-                            const T& value, 
+                            const T& value,
                             StrictWeakOrdering comp);
 
 
@@ -58,10 +64,10 @@ ForwardIterator upper_bound(thrust::execution_policy<DerivedPolicy> &exec,
 
 template <typename DerivedPolicy, typename ForwardIterator, typename T, typename StrictWeakOrdering>
 __host__ __device__
-ForwardIterator upper_bound(thrust::execution_policy<DerivedPolicy> &exec, 
+ForwardIterator upper_bound(thrust::execution_policy<DerivedPolicy> &exec,
                             ForwardIterator begin,
                             ForwardIterator end,
-                            const T& value, 
+                            const T& value,
                             StrictWeakOrdering comp);
 
 
@@ -77,16 +83,16 @@ __host__ __device__
 bool binary_search(thrust::execution_policy<DerivedPolicy> &exec,
                    ForwardIterator begin,
                    ForwardIterator end,
-                   const T& value, 
+                   const T& value,
                    StrictWeakOrdering comp);
 
 
 template <typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator>
 __host__ __device__
 OutputIterator lower_bound(thrust::execution_policy<DerivedPolicy> &exec,
-                           ForwardIterator begin, 
+                           ForwardIterator begin,
                            ForwardIterator end,
-                           InputIterator values_begin, 
+                           InputIterator values_begin,
                            InputIterator values_end,
                            OutputIterator output);
 
@@ -94,9 +100,9 @@ OutputIterator lower_bound(thrust::execution_policy<DerivedPolicy> &exec,
 template <typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator, typename StrictWeakOrdering>
 __host__ __device__
 OutputIterator lower_bound(thrust::execution_policy<DerivedPolicy> &exec,
-                           ForwardIterator begin, 
+                           ForwardIterator begin,
                            ForwardIterator end,
-                           InputIterator values_begin, 
+                           InputIterator values_begin,
                            InputIterator values_end,
                            OutputIterator output,
                            StrictWeakOrdering comp);
@@ -105,9 +111,9 @@ OutputIterator lower_bound(thrust::execution_policy<DerivedPolicy> &exec,
 template <typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator>
 __host__ __device__
 OutputIterator upper_bound(thrust::execution_policy<DerivedPolicy> &exec,
-                           ForwardIterator begin, 
+                           ForwardIterator begin,
                            ForwardIterator end,
-                           InputIterator values_begin, 
+                           InputIterator values_begin,
                            InputIterator values_end,
                            OutputIterator output);
 
@@ -115,9 +121,9 @@ OutputIterator upper_bound(thrust::execution_policy<DerivedPolicy> &exec,
 template <typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator, typename StrictWeakOrdering>
 __host__ __device__
 OutputIterator upper_bound(thrust::execution_policy<DerivedPolicy> &exec,
-                           ForwardIterator begin, 
+                           ForwardIterator begin,
                            ForwardIterator end,
-                           InputIterator values_begin, 
+                           InputIterator values_begin,
                            InputIterator values_end,
                            OutputIterator output,
                            StrictWeakOrdering comp);
@@ -126,9 +132,9 @@ OutputIterator upper_bound(thrust::execution_policy<DerivedPolicy> &exec,
 template <typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator>
 __host__ __device__
 OutputIterator binary_search(thrust::execution_policy<DerivedPolicy> &exec,
-                             ForwardIterator begin, 
+                             ForwardIterator begin,
                              ForwardIterator end,
-                             InputIterator values_begin, 
+                             InputIterator values_begin,
                              InputIterator values_end,
                              OutputIterator output);
 
@@ -136,9 +142,9 @@ OutputIterator binary_search(thrust::execution_policy<DerivedPolicy> &exec,
 template <typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator, typename StrictWeakOrdering>
 __host__ __device__
 OutputIterator binary_search(thrust::execution_policy<DerivedPolicy> &exec,
-                             ForwardIterator begin, 
+                             ForwardIterator begin,
                              ForwardIterator end,
-                             InputIterator values_begin, 
+                             InputIterator values_begin,
                              InputIterator values_end,
                              OutputIterator output,
                              StrictWeakOrdering comp);
diff --git a/thrust/thrust/system/detail/generic/binary_search.inl b/thrust/thrust/system/detail/generic/binary_search.inl
index bc60bb8e502..391263eea08 100644
--- a/thrust/thrust/system/detail/generic/binary_search.inl
+++ b/thrust/thrust/system/detail/generic/binary_search.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/distance.h>
 #include <thrust/binary_search.h>
 #include <thrust/iterator/zip_iterator.h>
diff --git a/thrust/thrust/system/detail/generic/copy.h b/thrust/thrust/system/detail/generic/copy.h
index 36ac71899de..201a33489a3 100644
--- a/thrust/thrust/system/detail/generic/copy.h
+++ b/thrust/thrust/system/detail/generic/copy.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/generic/copy.inl b/thrust/thrust/system/detail/generic/copy.inl
index 34d66baa6b2..57d040385e5 100644
--- a/thrust/thrust/system/detail/generic/copy.inl
+++ b/thrust/thrust/system/detail/generic/copy.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/copy.h>
 #include <thrust/functional.h>
 #include <thrust/detail/internal_functional.h>
diff --git a/thrust/thrust/system/detail/generic/copy_if.h b/thrust/thrust/system/detail/generic/copy_if.h
index 6a13edfdaf3..44105eb0140 100644
--- a/thrust/thrust/system/detail/generic/copy_if.h
+++ b/thrust/thrust/system/detail/generic/copy_if.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/generic/copy_if.inl b/thrust/thrust/system/detail/generic/copy_if.inl
index 5a6edd72e18..5dbff9fff55 100644
--- a/thrust/thrust/system/detail/generic/copy_if.inl
+++ b/thrust/thrust/system/detail/generic/copy_if.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/copy_if.h>
 #include <thrust/detail/copy_if.h>
 #include <thrust/iterator/iterator_traits.h>
@@ -58,7 +64,7 @@ OutputIterator copy_if(thrust::execution_policy<DerivedPolicy> &exec,
                        Predicate pred)
 {
   THRUST_DISABLE_MSVC_POSSIBLE_LOSS_OF_DATA_WARNING(IndexType n = thrust::distance(first, last));
-  
+
   // compute {0,1} predicates
   thrust::detail::temporary_array<IndexType, DerivedPolicy> predicates(exec, n);
   thrust::transform(exec,
@@ -66,7 +72,7 @@ OutputIterator copy_if(thrust::execution_policy<DerivedPolicy> &exec,
                     stencil + n,
                     predicates.begin(),
                     thrust::detail::predicate_to_integral<Predicate,IndexType>(pred));
-  
+
   // scan {0,1} predicates
   thrust::detail::temporary_array<IndexType, DerivedPolicy> scatter_indices(exec, n);
   thrust::exclusive_scan(exec,
@@ -75,7 +81,7 @@ OutputIterator copy_if(thrust::execution_policy<DerivedPolicy> &exec,
                          scatter_indices.begin(),
                          static_cast<IndexType>(0),
                          thrust::plus<IndexType>());
-  
+
   // scatter the true elements
   thrust::scatter_if(exec,
                      first,
@@ -84,10 +90,10 @@ OutputIterator copy_if(thrust::execution_policy<DerivedPolicy> &exec,
                      predicates.begin(),
                      result,
                      thrust::identity<IndexType>());
-  
+
   // find the end of the new sequence
   IndexType output_size = scatter_indices[n - 1] + predicates[n - 1];
-  
+
   return result + output_size;
 }
 
@@ -128,17 +134,17 @@ __host__ __device__
                           Predicate pred)
 {
   typedef typename thrust::iterator_traits<InputIterator1>::difference_type difference_type;
-  
+
   // empty sequence
   if(first == last)
     return result;
-  
+
   difference_type n = thrust::distance(first, last);
-  
+
   // create an unsigned version of n (we know n is positive from the comparison above)
   // to avoid a warning in the compare below
   typename thrust::detail::make_unsigned<difference_type>::type unsigned_n(n);
-  
+
   // use 32-bit indices when possible (almost always)
   if(sizeof(difference_type) > sizeof(unsigned int) && unsigned_n > thrust::detail::integer_traits<unsigned int>::const_max)
   {
diff --git a/thrust/thrust/system/detail/generic/count.h b/thrust/thrust/system/detail/generic/count.h
index 295d36e6b9d..80523a7dc1f 100644
--- a/thrust/thrust/system/detail/generic/count.h
+++ b/thrust/thrust/system/detail/generic/count.h
@@ -18,6 +18,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/generic/count.inl b/thrust/thrust/system/detail/generic/count.inl
index dafc1c1df2a..d1187a8d667 100644
--- a/thrust/thrust/system/detail/generic/count.inl
+++ b/thrust/thrust/system/detail/generic/count.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/count.h>
 #include <thrust/transform_reduce.h>
 #include <thrust/detail/internal_functional.h>
diff --git a/thrust/thrust/system/detail/generic/distance.h b/thrust/thrust/system/detail/generic/distance.h
index 4627376b575..2d48211af10 100644
--- a/thrust/thrust/system/detail/generic/distance.h
+++ b/thrust/thrust/system/detail/generic/distance.h
@@ -18,6 +18,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/iterator/iterator_traits.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/generic/distance.inl b/thrust/thrust/system/detail/generic/distance.inl
index 46bad7ba74e..7356946a033 100644
--- a/thrust/thrust/system/detail/generic/distance.inl
+++ b/thrust/thrust/system/detail/generic/distance.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/distance.h>
 #include <thrust/iterator/iterator_traits.h>
 
diff --git a/thrust/thrust/system/detail/generic/equal.h b/thrust/thrust/system/detail/generic/equal.h
index 4afd88d001f..cbbd6ffb2e3 100644
--- a/thrust/thrust/system/detail/generic/equal.h
+++ b/thrust/thrust/system/detail/generic/equal.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/generic/equal.inl b/thrust/thrust/system/detail/generic/equal.inl
index c023070cdcb..a9e318c5751 100644
--- a/thrust/thrust/system/detail/generic/equal.inl
+++ b/thrust/thrust/system/detail/generic/equal.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/equal.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/detail/internal_functional.h>
diff --git a/thrust/thrust/system/detail/generic/extrema.h b/thrust/thrust/system/detail/generic/extrema.h
index e3b447958e0..88d1a859b0c 100644
--- a/thrust/thrust/system/detail/generic/extrema.h
+++ b/thrust/thrust/system/detail/generic/extrema.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/pair.h>
 #include <thrust/system/detail/generic/tag.h>
 
@@ -67,14 +73,14 @@ ForwardIterator min_element(thrust::execution_policy<DerivedPolicy> &exec,
 template <typename DerivedPolicy, typename ForwardIterator>
 __host__ __device__
 thrust::pair<ForwardIterator,ForwardIterator> minmax_element(thrust::execution_policy<DerivedPolicy> &exec,
-                                                             ForwardIterator first, 
+                                                             ForwardIterator first,
                                                              ForwardIterator last);
 
 
 template <typename DerivedPolicy, typename ForwardIterator, typename BinaryPredicate>
 __host__ __device__
 thrust::pair<ForwardIterator,ForwardIterator> minmax_element(thrust::execution_policy<DerivedPolicy> &exec,
-                                                             ForwardIterator first, 
+                                                             ForwardIterator first,
                                                              ForwardIterator last,
                                                              BinaryPredicate comp);
 
diff --git a/thrust/thrust/system/detail/generic/extrema.inl b/thrust/thrust/system/detail/generic/extrema.inl
index 744d137de1d..a98ce903ce2 100644
--- a/thrust/thrust/system/detail/generic/extrema.inl
+++ b/thrust/thrust/system/detail/generic/extrema.inl
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/get_iterator_value.h>
 #include <thrust/extrema.h>
 #include <thrust/functional.h>
@@ -49,19 +55,19 @@ namespace detail
 //////////////
 //
 
-// return the smaller/larger element making sure to prefer the 
+// return the smaller/larger element making sure to prefer the
 // first occurance of the minimum/maximum element
 template <typename InputType, typename IndexType, typename BinaryPredicate>
 struct min_element_reduction
 {
   BinaryPredicate comp;
 
-  __host__ __device__ 
+  __host__ __device__
   min_element_reduction(BinaryPredicate comp) : comp(comp){}
 
-  __host__ __device__ 
+  __host__ __device__
   thrust::tuple<InputType, IndexType>
-  operator()(const thrust::tuple<InputType, IndexType>& lhs, 
+  operator()(const thrust::tuple<InputType, IndexType>& lhs,
              const thrust::tuple<InputType, IndexType>& rhs )
   {
     if(comp(thrust::get<0>(lhs), thrust::get<0>(rhs)))
@@ -83,12 +89,12 @@ struct max_element_reduction
 {
   BinaryPredicate comp;
 
-  __host__ __device__ 
+  __host__ __device__
   max_element_reduction(BinaryPredicate comp) : comp(comp){}
 
-  __host__ __device__ 
+  __host__ __device__
   thrust::tuple<InputType, IndexType>
-  operator()(const thrust::tuple<InputType, IndexType>& lhs, 
+  operator()(const thrust::tuple<InputType, IndexType>& lhs,
              const thrust::tuple<InputType, IndexType>& rhs )
   {
     if(comp(thrust::get<0>(lhs), thrust::get<0>(rhs)))
@@ -105,7 +111,7 @@ struct max_element_reduction
 }; // end max_element_reduction
 
 
-// return the smaller & larger element making sure to prefer the 
+// return the smaller & larger element making sure to prefer the
 // first occurance of the minimum/maximum element
 template <typename InputType, typename IndexType, typename BinaryPredicate>
 struct minmax_element_reduction
@@ -115,9 +121,9 @@ struct minmax_element_reduction
   __host__ __device__
   minmax_element_reduction(BinaryPredicate comp) : comp(comp){}
 
-  __host__ __device__ 
+  __host__ __device__
   thrust::tuple< thrust::tuple<InputType,IndexType>, thrust::tuple<InputType,IndexType> >
-  operator()(const thrust::tuple< thrust::tuple<InputType,IndexType>, thrust::tuple<InputType,IndexType> >& lhs, 
+  operator()(const thrust::tuple< thrust::tuple<InputType,IndexType>, thrust::tuple<InputType,IndexType> >& lhs,
              const thrust::tuple< thrust::tuple<InputType,IndexType>, thrust::tuple<InputType,IndexType> >& rhs )
   {
 
@@ -130,7 +136,7 @@ struct minmax_element_reduction
 template <typename InputType, typename IndexType>
 struct duplicate_tuple
 {
-  __host__ __device__ 
+  __host__ __device__
   thrust::tuple< thrust::tuple<InputType,IndexType>, thrust::tuple<InputType,IndexType> >
   operator()(const thrust::tuple<InputType,IndexType>& t)
   {
@@ -219,7 +225,7 @@ ForwardIterator max_element(thrust::execution_policy<DerivedPolicy> &exec,
 template <typename DerivedPolicy, typename ForwardIterator>
 __host__ __device__
 thrust::pair<ForwardIterator,ForwardIterator> minmax_element(thrust::execution_policy<DerivedPolicy> &exec,
-                                                             ForwardIterator first, 
+                                                             ForwardIterator first,
                                                              ForwardIterator last)
 {
   typedef typename thrust::iterator_value<ForwardIterator>::type value_type;
@@ -231,7 +237,7 @@ thrust::pair<ForwardIterator,ForwardIterator> minmax_element(thrust::execution_p
 template <typename DerivedPolicy, typename ForwardIterator, typename BinaryPredicate>
 __host__ __device__
 thrust::pair<ForwardIterator,ForwardIterator> minmax_element(thrust::execution_policy<DerivedPolicy> &exec,
-                                                             ForwardIterator first, 
+                                                             ForwardIterator first,
                                                              ForwardIterator last,
                                                              BinaryPredicate comp)
 {
@@ -241,7 +247,7 @@ thrust::pair<ForwardIterator,ForwardIterator> minmax_element(thrust::execution_p
   typedef typename thrust::iterator_traits<ForwardIterator>::value_type      InputType;
   typedef typename thrust::iterator_traits<ForwardIterator>::difference_type IndexType;
 
-  thrust::tuple< thrust::tuple<InputType,IndexType>, thrust::tuple<InputType,IndexType> > result = 
+  thrust::tuple< thrust::tuple<InputType,IndexType>, thrust::tuple<InputType,IndexType> > result =
     thrust::transform_reduce
       (exec,
        thrust::make_zip_iterator(thrust::make_tuple(first, thrust::counting_iterator<IndexType>(0))),
diff --git a/thrust/thrust/system/detail/generic/fill.h b/thrust/thrust/system/detail/generic/fill.h
index 5a881359b56..95179d287e1 100644
--- a/thrust/thrust/system/detail/generic/fill.h
+++ b/thrust/thrust/system/detail/generic/fill.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/internal_functional.h>
 #include <thrust/generate.h>
 #include <thrust/system/detail/generic/tag.h>
diff --git a/thrust/thrust/system/detail/generic/find.h b/thrust/thrust/system/detail/generic/find.h
index 6db441d02f7..5825fe11530 100644
--- a/thrust/thrust/system/detail/generic/find.h
+++ b/thrust/thrust/system/detail/generic/find.h
@@ -18,6 +18,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/generic/find.inl b/thrust/thrust/system/detail/generic/find.inl
index 8bd619561e9..c0077e9c70a 100644
--- a/thrust/thrust/system/detail/generic/find.inl
+++ b/thrust/thrust/system/detail/generic/find.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/find.h>
 #include <thrust/reduce.h>
 
diff --git a/thrust/thrust/system/detail/generic/for_each.h b/thrust/thrust/system/detail/generic/for_each.h
index 3c2ec12cd21..d650ee20b00 100644
--- a/thrust/thrust/system/detail/generic/for_each.h
+++ b/thrust/thrust/system/detail/generic/for_each.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 #include <thrust/system/detail/generic/tag.h>
 #include <thrust/detail/static_assert.h>
diff --git a/thrust/thrust/system/detail/generic/gather.h b/thrust/thrust/system/detail/generic/gather.h
index 5b6b4183113..c8d94c9514f 100644
--- a/thrust/thrust/system/detail/generic/gather.h
+++ b/thrust/thrust/system/detail/generic/gather.h
@@ -18,6 +18,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/generic/gather.inl b/thrust/thrust/system/detail/generic/gather.inl
index 7ab550edf14..d8d897302fe 100644
--- a/thrust/thrust/system/detail/generic/gather.inl
+++ b/thrust/thrust/system/detail/generic/gather.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/gather.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/functional.h>
diff --git a/thrust/thrust/system/detail/generic/generate.h b/thrust/thrust/system/detail/generic/generate.h
index a9846c5bed6..8806f4c4bc8 100644
--- a/thrust/thrust/system/detail/generic/generate.h
+++ b/thrust/thrust/system/detail/generic/generate.h
@@ -18,6 +18,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/generic/generate.inl b/thrust/thrust/system/detail/generic/generate.inl
index 869e0f32bf4..7c6a949e913 100644
--- a/thrust/thrust/system/detail/generic/generate.inl
+++ b/thrust/thrust/system/detail/generic/generate.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/generate.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/detail/internal_functional.h>
diff --git a/thrust/thrust/system/detail/generic/inner_product.h b/thrust/thrust/system/detail/generic/inner_product.h
index 62d10d31f58..92c4915a548 100644
--- a/thrust/thrust/system/detail/generic/inner_product.h
+++ b/thrust/thrust/system/detail/generic/inner_product.h
@@ -18,6 +18,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 
 THRUST_NAMESPACE_BEGIN
@@ -44,7 +50,7 @@ OutputType inner_product(thrust::execution_policy<DerivedPolicy> &exec,
                          InputIterator1 first1,
                          InputIterator1 last1,
                          InputIterator2 first2,
-                         OutputType init, 
+                         OutputType init,
                          BinaryFunction1 binary_op1,
                          BinaryFunction2 binary_op2);
 
diff --git a/thrust/thrust/system/detail/generic/inner_product.inl b/thrust/thrust/system/detail/generic/inner_product.inl
index 5055ec10f66..92246ae46e9 100644
--- a/thrust/thrust/system/detail/generic/inner_product.inl
+++ b/thrust/thrust/system/detail/generic/inner_product.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/inner_product.h>
 #include <thrust/functional.h>
 #include <thrust/detail/internal_functional.h>
diff --git a/thrust/thrust/system/detail/generic/logical.h b/thrust/thrust/system/detail/generic/logical.h
index e261154e248..ab74f97126b 100644
--- a/thrust/thrust/system/detail/generic/logical.h
+++ b/thrust/thrust/system/detail/generic/logical.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 #include <thrust/detail/internal_functional.h>
 #include <thrust/find.h>
diff --git a/thrust/thrust/system/detail/generic/memory.h b/thrust/thrust/system/detail/generic/memory.h
index 675cc73023c..69a0598e764 100644
--- a/thrust/thrust/system/detail/generic/memory.h
+++ b/thrust/thrust/system/detail/generic/memory.h
@@ -23,6 +23,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 #include <thrust/system/detail/generic/tag.h>
 #include <thrust/detail/type_traits.h>
diff --git a/thrust/thrust/system/detail/generic/memory.inl b/thrust/thrust/system/detail/generic/memory.inl
index b857290985c..8b4b4a120e8 100644
--- a/thrust/thrust/system/detail/generic/memory.inl
+++ b/thrust/thrust/system/detail/generic/memory.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/type_traits/pointer_traits.h>
 #include <thrust/system/detail/generic/memory.h>
 #include <thrust/system/detail/adl/malloc_and_free.h>
diff --git a/thrust/thrust/system/detail/generic/merge.h b/thrust/thrust/system/detail/generic/merge.h
index 6e82464074e..83a64187ebc 100644
--- a/thrust/thrust/system/detail/generic/merge.h
+++ b/thrust/thrust/system/detail/generic/merge.h
@@ -18,6 +18,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/generic/merge.inl b/thrust/thrust/system/detail/generic/merge.inl
index 03b77e6235b..18dd9f0323e 100644
--- a/thrust/thrust/system/detail/generic/merge.inl
+++ b/thrust/thrust/system/detail/generic/merge.inl
@@ -18,6 +18,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/static_assert.h>
 #include <thrust/system/detail/generic/merge.h>
 #include <thrust/merge.h>
diff --git a/thrust/thrust/system/detail/generic/mismatch.h b/thrust/thrust/system/detail/generic/mismatch.h
index 4a71cd34414..aff424d0d04 100644
--- a/thrust/thrust/system/detail/generic/mismatch.h
+++ b/thrust/thrust/system/detail/generic/mismatch.h
@@ -18,6 +18,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/generic/mismatch.inl b/thrust/thrust/system/detail/generic/mismatch.inl
index f6b9674a1cc..717d576e2cd 100644
--- a/thrust/thrust/system/detail/generic/mismatch.inl
+++ b/thrust/thrust/system/detail/generic/mismatch.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/mismatch.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/detail/internal_functional.h>
diff --git a/thrust/thrust/system/detail/generic/partition.h b/thrust/thrust/system/detail/generic/partition.h
index 113d6ecbc3d..6b216dac154 100644
--- a/thrust/thrust/system/detail/generic/partition.h
+++ b/thrust/thrust/system/detail/generic/partition.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/generic/partition.inl b/thrust/thrust/system/detail/generic/partition.inl
index ab56fdd5751..0a835fcedbd 100644
--- a/thrust/thrust/system/detail/generic/partition.inl
+++ b/thrust/thrust/system/detail/generic/partition.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/partition.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/pair.h>
diff --git a/thrust/thrust/system/detail/generic/per_device_resource.h b/thrust/thrust/system/detail/generic/per_device_resource.h
index 606f91f3658..1565fee9c1f 100644
--- a/thrust/thrust/system/detail/generic/per_device_resource.h
+++ b/thrust/thrust/system/detail/generic/per_device_resource.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 #include <thrust/system/detail/generic/tag.h>
 #include <thrust/mr/memory_resource.h>
diff --git a/thrust/thrust/system/detail/generic/reduce.h b/thrust/thrust/system/detail/generic/reduce.h
index f28b11a87bb..54c415aa265 100644
--- a/thrust/thrust/system/detail/generic/reduce.h
+++ b/thrust/thrust/system/detail/generic/reduce.h
@@ -18,6 +18,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 #include <thrust/iterator/iterator_traits.h>
 
diff --git a/thrust/thrust/system/detail/generic/reduce.inl b/thrust/thrust/system/detail/generic/reduce.inl
index d673d0cf882..bfc478ed55a 100644
--- a/thrust/thrust/system/detail/generic/reduce.inl
+++ b/thrust/thrust/system/detail/generic/reduce.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/reduce.h>
 #include <thrust/system/detail/generic/reduce.h>
 #include <thrust/iterator/iterator_traits.h>
diff --git a/thrust/thrust/system/detail/generic/reduce_by_key.h b/thrust/thrust/system/detail/generic/reduce_by_key.h
index 8ba47e11f0a..00abf0c85bf 100644
--- a/thrust/thrust/system/detail/generic/reduce_by_key.h
+++ b/thrust/thrust/system/detail/generic/reduce_by_key.h
@@ -18,6 +18,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 #include <thrust/iterator/iterator_traits.h>
 
@@ -38,7 +44,7 @@ template<typename DerivedPolicy,
 __host__ __device__
   thrust::pair<OutputIterator1,OutputIterator2>
     reduce_by_key(thrust::execution_policy<DerivedPolicy> &exec,
-                  InputIterator1 keys_first, 
+                  InputIterator1 keys_first,
                   InputIterator1 keys_last,
                   InputIterator2 values_first,
                   OutputIterator1 keys_output,
@@ -53,7 +59,7 @@ template<typename DerivedPolicy,
 __host__ __device__
   thrust::pair<OutputIterator1,OutputIterator2>
     reduce_by_key(thrust::execution_policy<DerivedPolicy> &exec,
-                  InputIterator1 keys_first, 
+                  InputIterator1 keys_first,
                   InputIterator1 keys_last,
                   InputIterator2 values_first,
                   OutputIterator1 keys_output,
@@ -70,7 +76,7 @@ template<typename DerivedPolicy,
 __host__ __device__
   thrust::pair<OutputIterator1,OutputIterator2>
     reduce_by_key(thrust::execution_policy<DerivedPolicy> &exec,
-                  InputIterator1 keys_first, 
+                  InputIterator1 keys_first,
                   InputIterator1 keys_last,
                   InputIterator2 values_first,
                   OutputIterator1 keys_output,
diff --git a/thrust/thrust/system/detail/generic/reduce_by_key.inl b/thrust/thrust/system/detail/generic/reduce_by_key.inl
index 2ea73fedaa1..9ade5eaaec1 100644
--- a/thrust/thrust/system/detail/generic/reduce_by_key.inl
+++ b/thrust/thrust/system/detail/generic/reduce_by_key.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/iterator/detail/minimum_system.h>
 #include <thrust/detail/type_traits.h>
diff --git a/thrust/thrust/system/detail/generic/remove.h b/thrust/thrust/system/detail/generic/remove.h
index 37354ef805c..2c9a8f9abae 100644
--- a/thrust/thrust/system/detail/generic/remove.h
+++ b/thrust/thrust/system/detail/generic/remove.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/generic/remove.inl b/thrust/thrust/system/detail/generic/remove.inl
index e51a3caeee1..d71a2e2e0fb 100644
--- a/thrust/thrust/system/detail/generic/remove.inl
+++ b/thrust/thrust/system/detail/generic/remove.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/remove.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/detail/copy_if.h>
diff --git a/thrust/thrust/system/detail/generic/replace.h b/thrust/thrust/system/detail/generic/replace.h
index 0821d6c07db..a294f108a8b 100644
--- a/thrust/thrust/system/detail/generic/replace.h
+++ b/thrust/thrust/system/detail/generic/replace.h
@@ -18,6 +18,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/generic/replace.inl b/thrust/thrust/system/detail/generic/replace.inl
index ed845dd458c..d49f85cc078 100644
--- a/thrust/thrust/system/detail/generic/replace.inl
+++ b/thrust/thrust/system/detail/generic/replace.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/functional.h>
 #include <thrust/system/detail/generic/replace.h>
 #include <thrust/transform.h>
diff --git a/thrust/thrust/system/detail/generic/reverse.h b/thrust/thrust/system/detail/generic/reverse.h
index 65c77ae7502..3e8451a6cd1 100644
--- a/thrust/thrust/system/detail/generic/reverse.h
+++ b/thrust/thrust/system/detail/generic/reverse.h
@@ -18,6 +18,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/generic/reverse.inl b/thrust/thrust/system/detail/generic/reverse.inl
index 1ce6db38b98..2ee86fcba11 100644
--- a/thrust/thrust/system/detail/generic/reverse.inl
+++ b/thrust/thrust/system/detail/generic/reverse.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/reverse.h>
 #include <thrust/advance.h>
 #include <thrust/distance.h>
diff --git a/thrust/thrust/system/detail/generic/scalar/binary_search.h b/thrust/thrust/system/detail/generic/scalar/binary_search.h
index 3e019c2233e..51842eb5103 100644
--- a/thrust/thrust/system/detail/generic/scalar/binary_search.h
+++ b/thrust/thrust/system/detail/generic/scalar/binary_search.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/pair.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/generic/scalar/binary_search.inl b/thrust/thrust/system/detail/generic/scalar/binary_search.inl
index 61c71fba4df..095a8a13a51 100644
--- a/thrust/thrust/system/detail/generic/scalar/binary_search.inl
+++ b/thrust/thrust/system/detail/generic/scalar/binary_search.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/pair.h>
 #include <thrust/detail/function.h>
 #include <thrust/iterator/iterator_traits.h>
diff --git a/thrust/thrust/system/detail/generic/scan.h b/thrust/thrust/system/detail/generic/scan.h
index 476441ab6f6..b96ab3462cb 100644
--- a/thrust/thrust/system/detail/generic/scan.h
+++ b/thrust/thrust/system/detail/generic/scan.h
@@ -18,6 +18,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 
 THRUST_NAMESPACE_BEGIN
@@ -39,7 +45,7 @@ __host__ __device__
                                 OutputIterator result);
 
 
-// XXX it is an error to call this function; it has no implementation 
+// XXX it is an error to call this function; it has no implementation
 template<typename ExecutionPolicy,
          typename InputIterator,
          typename OutputIterator,
@@ -74,7 +80,7 @@ __host__ __device__
                                 T init);
 
 
-// XXX it is an error to call this function; it has no implementation 
+// XXX it is an error to call this function; it has no implementation
 template<typename ExecutionPolicy,
          typename InputIterator,
          typename OutputIterator,
diff --git a/thrust/thrust/system/detail/generic/scan.inl b/thrust/thrust/system/detail/generic/scan.inl
index 45a2aadd087..788761487e2 100644
--- a/thrust/thrust/system/detail/generic/scan.inl
+++ b/thrust/thrust/system/detail/generic/scan.inl
@@ -18,6 +18,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/static_assert.h>
 #include <thrust/system/detail/generic/scan.h>
 #include <thrust/iterator/iterator_traits.h>
diff --git a/thrust/thrust/system/detail/generic/scan_by_key.h b/thrust/thrust/system/detail/generic/scan_by_key.h
index 9e38ac93392..eba28de7feb 100644
--- a/thrust/thrust/system/detail/generic/scan_by_key.h
+++ b/thrust/thrust/system/detail/generic/scan_by_key.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/generic/scan_by_key.inl b/thrust/thrust/system/detail/generic/scan_by_key.inl
index 0e3100224a4..242c6852c9f 100644
--- a/thrust/thrust/system/detail/generic/scan_by_key.inl
+++ b/thrust/thrust/system/detail/generic/scan_by_key.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cstdint.h>
 #include <thrust/system/detail/generic/scan_by_key.h>
 #include <thrust/functional.h>
diff --git a/thrust/thrust/system/detail/generic/scatter.h b/thrust/thrust/system/detail/generic/scatter.h
index 6bb7949efa4..00b89287b78 100644
--- a/thrust/thrust/system/detail/generic/scatter.h
+++ b/thrust/thrust/system/detail/generic/scatter.h
@@ -18,6 +18,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/generic/scatter.inl b/thrust/thrust/system/detail/generic/scatter.inl
index 5b479870897..0880a495d2b 100644
--- a/thrust/thrust/system/detail/generic/scatter.inl
+++ b/thrust/thrust/system/detail/generic/scatter.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/scatter.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/functional.h>
diff --git a/thrust/thrust/system/detail/generic/select_system.h b/thrust/thrust/system/detail/generic/select_system.h
index 7619b80e5ad..4d7657cf356 100644
--- a/thrust/thrust/system/detail/generic/select_system.h
+++ b/thrust/thrust/system/detail/generic/select_system.h
@@ -18,6 +18,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 #include <thrust/detail/type_traits.h>
 #include <thrust/iterator/detail/minimum_system.h>
diff --git a/thrust/thrust/system/detail/generic/select_system.inl b/thrust/thrust/system/detail/generic/select_system.inl
index b69d17c4507..b793fca5e4e 100644
--- a/thrust/thrust/system/detail/generic/select_system.inl
+++ b/thrust/thrust/system/detail/generic/select_system.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 #include <thrust/detail/type_traits.h>
 #include <thrust/system/detail/generic/select_system_exists.h>
diff --git a/thrust/thrust/system/detail/generic/select_system_exists.h b/thrust/thrust/system/detail/generic/select_system_exists.h
index 29d05781d9e..1a6943f0730 100644
--- a/thrust/thrust/system/detail/generic/select_system_exists.h
+++ b/thrust/thrust/system/detail/generic/select_system_exists.h
@@ -23,6 +23,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 THRUST_NAMESPACE_BEGIN
 
 // forward declaration of any_system_tag for any_conversion below
diff --git a/thrust/thrust/system/detail/generic/sequence.h b/thrust/thrust/system/detail/generic/sequence.h
index 26bf17bb8ba..2fa79d9256b 100644
--- a/thrust/thrust/system/detail/generic/sequence.h
+++ b/thrust/thrust/system/detail/generic/sequence.h
@@ -18,6 +18,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/generic/sequence.inl b/thrust/thrust/system/detail/generic/sequence.inl
index 7d0c12de3c2..55f8e9e333c 100644
--- a/thrust/thrust/system/detail/generic/sequence.inl
+++ b/thrust/thrust/system/detail/generic/sequence.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/sequence.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/tabulate.h>
diff --git a/thrust/thrust/system/detail/generic/set_operations.h b/thrust/thrust/system/detail/generic/set_operations.h
index 37665d78dd9..7285876b54e 100644
--- a/thrust/thrust/system/detail/generic/set_operations.h
+++ b/thrust/thrust/system/detail/generic/set_operations.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 #include <thrust/pair.h>
 
diff --git a/thrust/thrust/system/detail/generic/set_operations.inl b/thrust/thrust/system/detail/generic/set_operations.inl
index 4363be5c0d8..61f9c0156b8 100644
--- a/thrust/thrust/system/detail/generic/set_operations.inl
+++ b/thrust/thrust/system/detail/generic/set_operations.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/static_assert.h>
 #include <thrust/system/detail/generic/set_operations.h>
 #include <thrust/functional.h>
diff --git a/thrust/thrust/system/detail/generic/shuffle.h b/thrust/thrust/system/detail/generic/shuffle.h
index 8f8e21afdf9..34aea081fb0 100644
--- a/thrust/thrust/system/detail/generic/shuffle.h
+++ b/thrust/thrust/system/detail/generic/shuffle.h
@@ -20,8 +20,14 @@
 
 #pragma once
 
-
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/cpp11_required.h>
 
 #if THRUST_CPP_DIALECT >= 2011
diff --git a/thrust/thrust/system/detail/generic/sort.h b/thrust/thrust/system/detail/generic/sort.h
index cd8d455624c..87da3bc4cb6 100644
--- a/thrust/thrust/system/detail/generic/sort.h
+++ b/thrust/thrust/system/detail/generic/sort.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/generic/sort.inl b/thrust/thrust/system/detail/generic/sort.inl
index 632cab435ea..e2f3c31184e 100644
--- a/thrust/thrust/system/detail/generic/sort.inl
+++ b/thrust/thrust/system/detail/generic/sort.inl
@@ -18,6 +18,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/sort.h>
 #include <thrust/functional.h>
 #include <thrust/iterator/iterator_traits.h>
@@ -44,7 +50,7 @@ __host__ __device__
             RandomAccessIterator first,
             RandomAccessIterator last)
 {
-  typedef typename thrust::iterator_value<RandomAccessIterator>::type value_type; 
+  typedef typename thrust::iterator_value<RandomAccessIterator>::type value_type;
   thrust::sort(exec, first, last, thrust::less<value_type>());
 } // end sort()
 
diff --git a/thrust/thrust/system/detail/generic/swap_ranges.h b/thrust/thrust/system/detail/generic/swap_ranges.h
index edb5acf31fb..b7270530247 100644
--- a/thrust/thrust/system/detail/generic/swap_ranges.h
+++ b/thrust/thrust/system/detail/generic/swap_ranges.h
@@ -18,6 +18,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/generic/swap_ranges.inl b/thrust/thrust/system/detail/generic/swap_ranges.inl
index ea42df35bcd..0f68ebd80a7 100644
--- a/thrust/thrust/system/detail/generic/swap_ranges.inl
+++ b/thrust/thrust/system/detail/generic/swap_ranges.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/swap_ranges.h>
 #include <thrust/tuple.h>
 #include <thrust/iterator/zip_iterator.h>
diff --git a/thrust/thrust/system/detail/generic/tabulate.h b/thrust/thrust/system/detail/generic/tabulate.h
index 041093e82e6..4c6df103a8b 100644
--- a/thrust/thrust/system/detail/generic/tabulate.h
+++ b/thrust/thrust/system/detail/generic/tabulate.h
@@ -18,6 +18,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/generic/tabulate.inl b/thrust/thrust/system/detail/generic/tabulate.inl
index 0fd2121c149..da2a6b82c9d 100644
--- a/thrust/thrust/system/detail/generic/tabulate.inl
+++ b/thrust/thrust/system/detail/generic/tabulate.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tabulate.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/transform.h>
diff --git a/thrust/thrust/system/detail/generic/tag.h b/thrust/thrust/system/detail/generic/tag.h
index 48f0947974f..1a02f1af0bd 100644
--- a/thrust/thrust/system/detail/generic/tag.h
+++ b/thrust/thrust/system/detail/generic/tag.h
@@ -23,6 +23,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 THRUST_NAMESPACE_BEGIN
 namespace system
 {
diff --git a/thrust/thrust/system/detail/generic/temporary_buffer.h b/thrust/thrust/system/detail/generic/temporary_buffer.h
index 6b7e01ff24b..94d8cb64be6 100644
--- a/thrust/thrust/system/detail/generic/temporary_buffer.h
+++ b/thrust/thrust/system/detail/generic/temporary_buffer.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 #include <thrust/pair.h>
 #include <thrust/detail/pointer.h>
diff --git a/thrust/thrust/system/detail/generic/temporary_buffer.inl b/thrust/thrust/system/detail/generic/temporary_buffer.inl
index 254c48cb915..7a792e5041b 100644
--- a/thrust/thrust/system/detail/generic/temporary_buffer.inl
+++ b/thrust/thrust/system/detail/generic/temporary_buffer.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/temporary_buffer.h>
 #include <thrust/detail/pointer.h>
 #include <thrust/detail/malloc_and_free.h>
diff --git a/thrust/thrust/system/detail/generic/transform.h b/thrust/thrust/system/detail/generic/transform.h
index 30e032696bc..175fcb672b8 100644
--- a/thrust/thrust/system/detail/generic/transform.h
+++ b/thrust/thrust/system/detail/generic/transform.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/generic/transform.inl b/thrust/thrust/system/detail/generic/transform.inl
index 122c42580b0..6bc3292117e 100644
--- a/thrust/thrust/system/detail/generic/transform.inl
+++ b/thrust/thrust/system/detail/generic/transform.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/transform.h>
 #include <thrust/for_each.h>
 #include <thrust/iterator/iterator_traits.h>
diff --git a/thrust/thrust/system/detail/generic/transform_reduce.h b/thrust/thrust/system/detail/generic/transform_reduce.h
index af510296e1c..e5c07e2d415 100644
--- a/thrust/thrust/system/detail/generic/transform_reduce.h
+++ b/thrust/thrust/system/detail/generic/transform_reduce.h
@@ -18,6 +18,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 
 THRUST_NAMESPACE_BEGIN
@@ -30,8 +36,8 @@ namespace generic
 
 
 template<typename ExecutionPolicy,
-         typename InputIterator, 
-         typename UnaryFunction, 
+         typename InputIterator,
+         typename UnaryFunction,
          typename OutputType,
          typename BinaryFunction>
 __host__ __device__
diff --git a/thrust/thrust/system/detail/generic/transform_reduce.inl b/thrust/thrust/system/detail/generic/transform_reduce.inl
index 539c3b22c04..580da0c1e06 100644
--- a/thrust/thrust/system/detail/generic/transform_reduce.inl
+++ b/thrust/thrust/system/detail/generic/transform_reduce.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/transform_reduce.h>
 #include <thrust/reduce.h>
 #include <thrust/iterator/transform_iterator.h>
diff --git a/thrust/thrust/system/detail/generic/transform_scan.h b/thrust/thrust/system/detail/generic/transform_scan.h
index 05054c965a0..9226615a0b1 100644
--- a/thrust/thrust/system/detail/generic/transform_scan.h
+++ b/thrust/thrust/system/detail/generic/transform_scan.h
@@ -18,6 +18,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/generic/transform_scan.inl b/thrust/thrust/system/detail/generic/transform_scan.inl
index c9c97668731..23e2582b6f1 100644
--- a/thrust/thrust/system/detail/generic/transform_scan.inl
+++ b/thrust/thrust/system/detail/generic/transform_scan.inl
@@ -18,6 +18,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/type_traits.h>
 #include <thrust/detail/type_traits/function_traits.h>
 #include <thrust/detail/type_traits/iterator/is_output_iterator.h>
diff --git a/thrust/thrust/system/detail/generic/uninitialized_copy.h b/thrust/thrust/system/detail/generic/uninitialized_copy.h
index bac5bcf9609..022f016055e 100644
--- a/thrust/thrust/system/detail/generic/uninitialized_copy.h
+++ b/thrust/thrust/system/detail/generic/uninitialized_copy.h
@@ -18,6 +18,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/generic/uninitialized_copy.inl b/thrust/thrust/system/detail/generic/uninitialized_copy.inl
index 679d1f6bae8..ba7a5a27dc1 100644
--- a/thrust/thrust/system/detail/generic/uninitialized_copy.inl
+++ b/thrust/thrust/system/detail/generic/uninitialized_copy.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/uninitialized_copy.h>
 #include <thrust/copy.h>
 #include <thrust/for_each.h>
diff --git a/thrust/thrust/system/detail/generic/uninitialized_fill.h b/thrust/thrust/system/detail/generic/uninitialized_fill.h
index 4f5404508c0..dc88adc0e9c 100644
--- a/thrust/thrust/system/detail/generic/uninitialized_fill.h
+++ b/thrust/thrust/system/detail/generic/uninitialized_fill.h
@@ -18,6 +18,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/generic/uninitialized_fill.inl b/thrust/thrust/system/detail/generic/uninitialized_fill.inl
index 06241494548..169abebb309 100644
--- a/thrust/thrust/system/detail/generic/uninitialized_fill.inl
+++ b/thrust/thrust/system/detail/generic/uninitialized_fill.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/uninitialized_fill.h>
 #include <thrust/fill.h>
 #include <thrust/detail/internal_functional.h>
diff --git a/thrust/thrust/system/detail/generic/unique.h b/thrust/thrust/system/detail/generic/unique.h
index ce3bff884b3..786da30caa9 100644
--- a/thrust/thrust/system/detail/generic/unique.h
+++ b/thrust/thrust/system/detail/generic/unique.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/generic/unique.inl b/thrust/thrust/system/detail/generic/unique.inl
index bb66e3585d2..e85768cd3df 100644
--- a/thrust/thrust/system/detail/generic/unique.inl
+++ b/thrust/thrust/system/detail/generic/unique.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/unique.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/transform.h>
@@ -112,9 +118,9 @@ __host__ __device__
                  BinaryPredicate binary_pred)
 {
   thrust::detail::head_flags<ForwardIterator, BinaryPredicate> stencil(first, last, binary_pred);
-  
+
   using namespace thrust::placeholders;
-  
+
   return thrust::count_if(exec, stencil.begin(), stencil.end(), _1);
 } // end unique_copy()
 
diff --git a/thrust/thrust/system/detail/generic/unique_by_key.h b/thrust/thrust/system/detail/generic/unique_by_key.h
index 0ea9e7cc845..c070e738cc0 100644
--- a/thrust/thrust/system/detail/generic/unique_by_key.h
+++ b/thrust/thrust/system/detail/generic/unique_by_key.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/tag.h>
 #include <thrust/pair.h>
 
@@ -35,7 +41,7 @@ template<typename ExecutionPolicy,
 __host__ __device__
   thrust::pair<ForwardIterator1,ForwardIterator2>
     unique_by_key(thrust::execution_policy<ExecutionPolicy> &exec,
-                  ForwardIterator1 keys_first, 
+                  ForwardIterator1 keys_first,
                   ForwardIterator1 keys_last,
                   ForwardIterator2 values_first);
 
@@ -47,7 +53,7 @@ template<typename ExecutionPolicy,
 __host__ __device__
   thrust::pair<ForwardIterator1,ForwardIterator2>
     unique_by_key(thrust::execution_policy<ExecutionPolicy> &exec,
-                  ForwardIterator1 keys_first, 
+                  ForwardIterator1 keys_first,
                   ForwardIterator1 keys_last,
                   ForwardIterator2 values_first,
                   BinaryPredicate binary_pred);
@@ -61,7 +67,7 @@ template<typename ExecutionPolicy,
 __host__ __device__
   thrust::pair<OutputIterator1,OutputIterator2>
     unique_by_key_copy(thrust::execution_policy<ExecutionPolicy> &exec,
-                       InputIterator1 keys_first, 
+                       InputIterator1 keys_first,
                        InputIterator1 keys_last,
                        InputIterator2 values_first,
                        OutputIterator1 keys_output,
@@ -77,7 +83,7 @@ template<typename ExecutionPolicy,
 __host__ __device__
   thrust::pair<OutputIterator1,OutputIterator2>
     unique_by_key_copy(thrust::execution_policy<ExecutionPolicy> &exec,
-                       InputIterator1 keys_first, 
+                       InputIterator1 keys_first,
                        InputIterator1 keys_last,
                        InputIterator2 values_first,
                        OutputIterator1 keys_output,
diff --git a/thrust/thrust/system/detail/generic/unique_by_key.inl b/thrust/thrust/system/detail/generic/unique_by_key.inl
index ffcf1dd0c74..801bdea5abf 100644
--- a/thrust/thrust/system/detail/generic/unique_by_key.inl
+++ b/thrust/thrust/system/detail/generic/unique_by_key.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/unique_by_key.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/iterator/detail/minimum_system.h>
diff --git a/thrust/thrust/system/detail/internal/decompose.h b/thrust/thrust/system/detail/internal/decompose.h
index 58af7c55144..131fb258f52 100644
--- a/thrust/thrust/system/detail/internal/decompose.h
+++ b/thrust/thrust/system/detail/internal/decompose.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 THRUST_NAMESPACE_BEGIN
 namespace system
 {
diff --git a/thrust/thrust/system/detail/sequential/adjacent_difference.h b/thrust/thrust/system/detail/sequential/adjacent_difference.h
index 4a9dad82c93..414470442b2 100644
--- a/thrust/thrust/system/detail/sequential/adjacent_difference.h
+++ b/thrust/thrust/system/detail/sequential/adjacent_difference.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/sequential/execution_policy.h>
 
diff --git a/thrust/thrust/system/detail/sequential/assign_value.h b/thrust/thrust/system/detail/sequential/assign_value.h
index 0eb145d135a..511a8507f8b 100644
--- a/thrust/thrust/system/detail/sequential/assign_value.h
+++ b/thrust/thrust/system/detail/sequential/assign_value.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/sequential/execution_policy.h>
 #include <thrust/detail/raw_pointer_cast.h>
 
diff --git a/thrust/thrust/system/detail/sequential/binary_search.h b/thrust/thrust/system/detail/sequential/binary_search.h
index 2da5080f4f8..37ffb195b49 100644
--- a/thrust/thrust/system/detail/sequential/binary_search.h
+++ b/thrust/thrust/system/detail/sequential/binary_search.h
@@ -23,6 +23,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/advance.h>
 #include <thrust/distance.h>
 #include <thrust/iterator/iterator_traits.h>
@@ -91,7 +97,7 @@ __host__ __device__
 ForwardIterator upper_bound(sequential::execution_policy<DerivedPolicy> &,
                             ForwardIterator first,
                             ForwardIterator last,
-                            const T& val, 
+                            const T& val,
                             StrictWeakOrdering comp)
 {
   // wrap comp
@@ -136,7 +142,7 @@ __host__ __device__
 bool binary_search(sequential::execution_policy<DerivedPolicy> &exec,
                    ForwardIterator first,
                    ForwardIterator last,
-                   const T& val, 
+                   const T& val,
                    StrictWeakOrdering comp)
 {
   ForwardIterator iter = sequential::lower_bound(exec, first, last, val, comp);
diff --git a/thrust/thrust/system/detail/sequential/copy.h b/thrust/thrust/system/detail/sequential/copy.h
index 0dd2cdad5c3..a657297698c 100644
--- a/thrust/thrust/system/detail/sequential/copy.h
+++ b/thrust/thrust/system/detail/sequential/copy.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/sequential/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/sequential/copy.inl b/thrust/thrust/system/detail/sequential/copy.inl
index 850f20f1eaa..50007f948fc 100644
--- a/thrust/thrust/system/detail/sequential/copy.inl
+++ b/thrust/thrust/system/detail/sequential/copy.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/sequential/copy.h>
 #include <thrust/detail/type_traits.h>
 #include <thrust/system/detail/sequential/general_copy.h>
diff --git a/thrust/thrust/system/detail/sequential/copy_backward.h b/thrust/thrust/system/detail/sequential/copy_backward.h
index d127ac80db2..b85834a601f 100644
--- a/thrust/thrust/system/detail/sequential/copy_backward.h
+++ b/thrust/thrust/system/detail/sequential/copy_backward.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 THRUST_NAMESPACE_BEGIN
 namespace system
 {
@@ -31,8 +37,8 @@ __thrust_exec_check_disable__
 template<typename BidirectionalIterator1,
          typename BidirectionalIterator2>
 __host__ __device__
-BidirectionalIterator2 copy_backward(BidirectionalIterator1 first, 
-                                     BidirectionalIterator1 last, 
+BidirectionalIterator2 copy_backward(BidirectionalIterator1 first,
+                                     BidirectionalIterator1 last,
                                      BidirectionalIterator2 result)
 {
   while (first != last)
diff --git a/thrust/thrust/system/detail/sequential/copy_if.h b/thrust/thrust/system/detail/sequential/copy_if.h
index 3c00956deb3..169654517ac 100644
--- a/thrust/thrust/system/detail/sequential/copy_if.h
+++ b/thrust/thrust/system/detail/sequential/copy_if.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/function.h>
 #include <thrust/system/detail/sequential/execution_policy.h>
 
diff --git a/thrust/thrust/system/detail/sequential/count.h b/thrust/thrust/system/detail/sequential/count.h
index ed3cadffe36..aa47b74f536 100644
--- a/thrust/thrust/system/detail/sequential/count.h
+++ b/thrust/thrust/system/detail/sequential/count.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system has no special count functions
 
diff --git a/thrust/thrust/system/detail/sequential/equal.h b/thrust/thrust/system/detail/sequential/equal.h
index 9d31e70f6fd..be6983aab32 100644
--- a/thrust/thrust/system/detail/sequential/equal.h
+++ b/thrust/thrust/system/detail/sequential/equal.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system has no special equal functions
 
diff --git a/thrust/thrust/system/detail/sequential/execution_policy.h b/thrust/thrust/system/detail/sequential/execution_policy.h
index 99d78fc27f0..dd547d77f76 100644
--- a/thrust/thrust/system/detail/sequential/execution_policy.h
+++ b/thrust/thrust/system/detail/sequential/execution_policy.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/sequential/extrema.h b/thrust/thrust/system/detail/sequential/extrema.h
index 5e5c62da694..0859adbc3de 100644
--- a/thrust/thrust/system/detail/sequential/extrema.h
+++ b/thrust/thrust/system/detail/sequential/extrema.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/pair.h>
 #include <thrust/detail/function.h>
 #include <thrust/system/detail/sequential/execution_policy.h>
@@ -41,7 +47,7 @@ template<typename DerivedPolicy,
          typename BinaryPredicate>
 __host__ __device__
 ForwardIterator min_element(sequential::execution_policy<DerivedPolicy> &,
-                            ForwardIterator first, 
+                            ForwardIterator first,
                             ForwardIterator last,
                             BinaryPredicate comp)
 {
@@ -71,7 +77,7 @@ template<typename DerivedPolicy,
          typename BinaryPredicate>
 __host__ __device__
 ForwardIterator max_element(sequential::execution_policy<DerivedPolicy> &,
-                            ForwardIterator first, 
+                            ForwardIterator first,
                             ForwardIterator last,
                             BinaryPredicate comp)
 {
@@ -101,7 +107,7 @@ template<typename DerivedPolicy,
          typename BinaryPredicate>
 __host__ __device__
 thrust::pair<ForwardIterator,ForwardIterator> minmax_element(sequential::execution_policy<DerivedPolicy> &,
-                                                             ForwardIterator first, 
+                                                             ForwardIterator first,
                                                              ForwardIterator last,
                                                              BinaryPredicate comp)
 {
@@ -110,7 +116,7 @@ thrust::pair<ForwardIterator,ForwardIterator> minmax_element(sequential::executi
     BinaryPredicate,
     bool
   > wrapped_comp(comp);
-  
+
   ForwardIterator imin = first;
   ForwardIterator imax = first;
 
diff --git a/thrust/thrust/system/detail/sequential/fill.h b/thrust/thrust/system/detail/sequential/fill.h
index 20c636096c2..85313ac6548 100644
--- a/thrust/thrust/system/detail/sequential/fill.h
+++ b/thrust/thrust/system/detail/sequential/fill.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system has no special fill functions
 
diff --git a/thrust/thrust/system/detail/sequential/find.h b/thrust/thrust/system/detail/sequential/find.h
index 54c238c7158..ea19b6957c8 100644
--- a/thrust/thrust/system/detail/sequential/find.h
+++ b/thrust/thrust/system/detail/sequential/find.h
@@ -16,12 +16,18 @@
 
 
 /*! \file find.h
- *  \brief Sequential implementation of find_if. 
+ *  \brief Sequential implementation of find_if.
  */
 
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/function.h>
 #include <thrust/system/detail/sequential/execution_policy.h>
 
diff --git a/thrust/thrust/system/detail/sequential/for_each.h b/thrust/thrust/system/detail/sequential/for_each.h
index 7058c56f28e..c115b901cf9 100644
--- a/thrust/thrust/system/detail/sequential/for_each.h
+++ b/thrust/thrust/system/detail/sequential/for_each.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/function.h>
 #include <thrust/system/detail/sequential/execution_policy.h>
 
diff --git a/thrust/thrust/system/detail/sequential/gather.h b/thrust/thrust/system/detail/sequential/gather.h
index a5657585eae..76eb47da643 100644
--- a/thrust/thrust/system/detail/sequential/gather.h
+++ b/thrust/thrust/system/detail/sequential/gather.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system has no special gather functions
 
diff --git a/thrust/thrust/system/detail/sequential/general_copy.h b/thrust/thrust/system/detail/sequential/general_copy.h
index 6ea87bbacbe..a1579d5d1b4 100644
--- a/thrust/thrust/system/detail/sequential/general_copy.h
+++ b/thrust/thrust/system/detail/sequential/general_copy.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/raw_reference_cast.h>
 #include <thrust/detail/type_traits.h>
 
diff --git a/thrust/thrust/system/detail/sequential/generate.h b/thrust/thrust/system/detail/sequential/generate.h
index ac38be51617..eb534e994c4 100644
--- a/thrust/thrust/system/detail/sequential/generate.h
+++ b/thrust/thrust/system/detail/sequential/generate.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system has no special generate functions
 
diff --git a/thrust/thrust/system/detail/sequential/get_value.h b/thrust/thrust/system/detail/sequential/get_value.h
index 90752d8671a..f06567334eb 100644
--- a/thrust/thrust/system/detail/sequential/get_value.h
+++ b/thrust/thrust/system/detail/sequential/get_value.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/sequential/execution_policy.h>
 #include <thrust/detail/raw_pointer_cast.h>
 
diff --git a/thrust/thrust/system/detail/sequential/inner_product.h b/thrust/thrust/system/detail/sequential/inner_product.h
index e6b3c0ae174..28d3f13a3ba 100644
--- a/thrust/thrust/system/detail/sequential/inner_product.h
+++ b/thrust/thrust/system/detail/sequential/inner_product.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system has no special inner_product functions
 
diff --git a/thrust/thrust/system/detail/sequential/insertion_sort.h b/thrust/thrust/system/detail/sequential/insertion_sort.h
index 9acccd8e92e..4b9d283573c 100644
--- a/thrust/thrust/system/detail/sequential/insertion_sort.h
+++ b/thrust/thrust/system/detail/sequential/insertion_sort.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/detail/function.h>
 #include <thrust/system/detail/sequential/copy_backward.h>
diff --git a/thrust/thrust/system/detail/sequential/iter_swap.h b/thrust/thrust/system/detail/sequential/iter_swap.h
index 7a5c481fc03..bcb98dd2acd 100644
--- a/thrust/thrust/system/detail/sequential/iter_swap.h
+++ b/thrust/thrust/system/detail/sequential/iter_swap.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/sequential/execution_policy.h>
 #include <thrust/detail/raw_pointer_cast.h>
 #include <thrust/detail/swap.h>
diff --git a/thrust/thrust/system/detail/sequential/logical.h b/thrust/thrust/system/detail/sequential/logical.h
index ee458627356..9bed89ef06e 100644
--- a/thrust/thrust/system/detail/sequential/logical.h
+++ b/thrust/thrust/system/detail/sequential/logical.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system has no special logical functions
 
diff --git a/thrust/thrust/system/detail/sequential/malloc_and_free.h b/thrust/thrust/system/detail/sequential/malloc_and_free.h
index b250140e083..452ff12beda 100644
--- a/thrust/thrust/system/detail/sequential/malloc_and_free.h
+++ b/thrust/thrust/system/detail/sequential/malloc_and_free.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/sequential/execution_policy.h>
 #include <cstdlib> // for malloc & free
 #include <thrust/detail/raw_pointer_cast.h>
diff --git a/thrust/thrust/system/detail/sequential/merge.h b/thrust/thrust/system/detail/sequential/merge.h
index a45e18004d4..3cb0041e269 100644
--- a/thrust/thrust/system/detail/sequential/merge.h
+++ b/thrust/thrust/system/detail/sequential/merge.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/sequential/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/sequential/merge.inl b/thrust/thrust/system/detail/sequential/merge.inl
index 08d7c0b0d4e..ee11e602a3c 100644
--- a/thrust/thrust/system/detail/sequential/merge.inl
+++ b/thrust/thrust/system/detail/sequential/merge.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/sequential/merge.h>
 #include <thrust/detail/copy.h>
 #include <thrust/iterator/iterator_traits.h>
diff --git a/thrust/thrust/system/detail/sequential/mismatch.h b/thrust/thrust/system/detail/sequential/mismatch.h
index e6094d261a0..5d077bbdabd 100644
--- a/thrust/thrust/system/detail/sequential/mismatch.h
+++ b/thrust/thrust/system/detail/sequential/mismatch.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system has no special mismatch functions
 
diff --git a/thrust/thrust/system/detail/sequential/partition.h b/thrust/thrust/system/detail/sequential/partition.h
index 43d5b0e2378..3878a3b2bdc 100644
--- a/thrust/thrust/system/detail/sequential/partition.h
+++ b/thrust/thrust/system/detail/sequential/partition.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/pair.h>
 #include <thrust/detail/temporary_array.h>
 #include <thrust/detail/function.h>
diff --git a/thrust/thrust/system/detail/sequential/per_device_resource.h b/thrust/thrust/system/detail/sequential/per_device_resource.h
index 1b8d61f9216..64c721210a7 100644
--- a/thrust/thrust/system/detail/sequential/per_device_resource.h
+++ b/thrust/thrust/system/detail/sequential/per_device_resource.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system has no special per device resource functions
 
diff --git a/thrust/thrust/system/detail/sequential/reduce.h b/thrust/thrust/system/detail/sequential/reduce.h
index a532f71b246..e14fb4755f0 100644
--- a/thrust/thrust/system/detail/sequential/reduce.h
+++ b/thrust/thrust/system/detail/sequential/reduce.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/function.h>
 #include <thrust/system/detail/sequential/execution_policy.h>
 
@@ -36,7 +42,7 @@ namespace sequential
 
 __thrust_exec_check_disable__
 template<typename DerivedPolicy,
-         typename InputIterator, 
+         typename InputIterator,
          typename OutputType,
          typename BinaryFunction>
 __host__ __device__
diff --git a/thrust/thrust/system/detail/sequential/reduce_by_key.h b/thrust/thrust/system/detail/sequential/reduce_by_key.h
index ef17ac5b0ea..cad30bb417c 100644
--- a/thrust/thrust/system/detail/sequential/reduce_by_key.h
+++ b/thrust/thrust/system/detail/sequential/reduce_by_key.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/pair.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/detail/sequential/execution_policy.h>
@@ -41,7 +47,7 @@ template<typename DerivedPolicy,
 __host__ __device__
   thrust::pair<OutputIterator1,OutputIterator2>
     reduce_by_key(sequential::execution_policy<DerivedPolicy> &,
-                  InputIterator1 keys_first, 
+                  InputIterator1 keys_first,
                   InputIterator1 keys_last,
                   InputIterator2 values_first,
                   OutputIterator1 keys_output,
diff --git a/thrust/thrust/system/detail/sequential/remove.h b/thrust/thrust/system/detail/sequential/remove.h
index df564f15b03..1444db22fc9 100644
--- a/thrust/thrust/system/detail/sequential/remove.h
+++ b/thrust/thrust/system/detail/sequential/remove.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/function.h>
 #include <thrust/system/detail/sequential/execution_policy.h>
 
@@ -57,7 +63,7 @@ __host__ __device__
   if(first == last)
     return first;
 
-  // result always trails first 
+  // result always trails first
   ForwardIterator result = first;
 
   ++first;
@@ -104,7 +110,7 @@ __host__ __device__
   if(first == last)
     return first;
 
-  // result always trails first 
+  // result always trails first
   ForwardIterator result = first;
 
   ++first;
diff --git a/thrust/thrust/system/detail/sequential/replace.h b/thrust/thrust/system/detail/sequential/replace.h
index f5c8e838571..8a2fb0e0463 100644
--- a/thrust/thrust/system/detail/sequential/replace.h
+++ b/thrust/thrust/system/detail/sequential/replace.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system has no special replace functions
 
diff --git a/thrust/thrust/system/detail/sequential/reverse.h b/thrust/thrust/system/detail/sequential/reverse.h
index f80974e8a8d..9cd83565804 100644
--- a/thrust/thrust/system/detail/sequential/reverse.h
+++ b/thrust/thrust/system/detail/sequential/reverse.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system has no special reverse functions
 
diff --git a/thrust/thrust/system/detail/sequential/scan.h b/thrust/thrust/system/detail/sequential/scan.h
index c5fce247582..fee6f11571f 100644
--- a/thrust/thrust/system/detail/sequential/scan.h
+++ b/thrust/thrust/system/detail/sequential/scan.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/sequential/execution_policy.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/detail/type_traits.h>
@@ -111,7 +117,7 @@ __host__ __device__
   }
 
   return result;
-} 
+}
 
 
 } // end namespace sequential
diff --git a/thrust/thrust/system/detail/sequential/scan_by_key.h b/thrust/thrust/system/detail/sequential/scan_by_key.h
index c428c1050e8..11e43969304 100644
--- a/thrust/thrust/system/detail/sequential/scan_by_key.h
+++ b/thrust/thrust/system/detail/sequential/scan_by_key.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/detail/function.h>
 #include <thrust/system/detail/sequential/execution_policy.h>
@@ -131,7 +137,7 @@ __host__ __device__
       if (!binary_pred(temp_key, key))
         next = init;  // reset sum
 
-      *result = next;  
+      *result = next;
       next = binary_op(next, temp_value);
 
       temp_key = key;
diff --git a/thrust/thrust/system/detail/sequential/scatter.h b/thrust/thrust/system/detail/sequential/scatter.h
index d6817b4cb26..166b455c1ae 100644
--- a/thrust/thrust/system/detail/sequential/scatter.h
+++ b/thrust/thrust/system/detail/sequential/scatter.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system has no special scatter functions
 
diff --git a/thrust/thrust/system/detail/sequential/sequence.h b/thrust/thrust/system/detail/sequential/sequence.h
index 0c090f96f34..a65392eb799 100644
--- a/thrust/thrust/system/detail/sequential/sequence.h
+++ b/thrust/thrust/system/detail/sequential/sequence.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system has no special sequence functions
 
diff --git a/thrust/thrust/system/detail/sequential/set_operations.h b/thrust/thrust/system/detail/sequential/set_operations.h
index 678754b45dd..028fe9bcb53 100644
--- a/thrust/thrust/system/detail/sequential/set_operations.h
+++ b/thrust/thrust/system/detail/sequential/set_operations.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/sequential/execution_policy.h>
 #include <thrust/detail/copy.h>
 #include <thrust/detail/function.h>
diff --git a/thrust/thrust/system/detail/sequential/sort.h b/thrust/thrust/system/detail/sequential/sort.h
index 34cc7a8ba89..6e635deb1ea 100644
--- a/thrust/thrust/system/detail/sequential/sort.h
+++ b/thrust/thrust/system/detail/sequential/sort.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/sequential/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/sequential/sort.inl b/thrust/thrust/system/detail/sequential/sort.inl
index 241a860aff5..30a4715b15d 100644
--- a/thrust/thrust/system/detail/sequential/sort.inl
+++ b/thrust/thrust/system/detail/sequential/sort.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/reverse.h>
 #include <thrust/detail/type_traits.h>
 #include <thrust/iterator/iterator_traits.h>
diff --git a/thrust/thrust/system/detail/sequential/stable_merge_sort.h b/thrust/thrust/system/detail/sequential/stable_merge_sort.h
index 64aa2bf96ce..e9a8bedc288 100644
--- a/thrust/thrust/system/detail/sequential/stable_merge_sort.h
+++ b/thrust/thrust/system/detail/sequential/stable_merge_sort.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/sequential/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/sequential/stable_merge_sort.inl b/thrust/thrust/system/detail/sequential/stable_merge_sort.inl
index 02f384afba1..a98e10541f9 100644
--- a/thrust/thrust/system/detail/sequential/stable_merge_sort.inl
+++ b/thrust/thrust/system/detail/sequential/stable_merge_sort.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/detail/temporary_array.h>
 #include <thrust/merge.h>
diff --git a/thrust/thrust/system/detail/sequential/stable_primitive_sort.h b/thrust/thrust/system/detail/sequential/stable_primitive_sort.h
index acbb81217bd..d4e16a7d18d 100644
--- a/thrust/thrust/system/detail/sequential/stable_primitive_sort.h
+++ b/thrust/thrust/system/detail/sequential/stable_primitive_sort.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/sequential/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/sequential/stable_primitive_sort.inl b/thrust/thrust/system/detail/sequential/stable_primitive_sort.inl
index 9897d6798da..94ba396b08d 100644
--- a/thrust/thrust/system/detail/sequential/stable_primitive_sort.inl
+++ b/thrust/thrust/system/detail/sequential/stable_primitive_sort.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/sequential/stable_primitive_sort.h>
 #include <thrust/system/detail/sequential/stable_radix_sort.h>
 #include <thrust/functional.h>
diff --git a/thrust/thrust/system/detail/sequential/stable_radix_sort.h b/thrust/thrust/system/detail/sequential/stable_radix_sort.h
index 1e9713a2cc9..bb7b941d744 100644
--- a/thrust/thrust/system/detail/sequential/stable_radix_sort.h
+++ b/thrust/thrust/system/detail/sequential/stable_radix_sort.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/sequential/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/detail/sequential/stable_radix_sort.inl b/thrust/thrust/system/detail/sequential/stable_radix_sort.inl
index 83d95ebfd49..463d380aee0 100644
--- a/thrust/thrust/system/detail/sequential/stable_radix_sort.inl
+++ b/thrust/thrust/system/detail/sequential/stable_radix_sort.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/copy.h>
 #include <thrust/functional.h>
 #include <thrust/iterator/iterator_traits.h>
diff --git a/thrust/thrust/system/detail/sequential/swap_ranges.h b/thrust/thrust/system/detail/sequential/swap_ranges.h
index 497497d6a15..c41998fc52d 100644
--- a/thrust/thrust/system/detail/sequential/swap_ranges.h
+++ b/thrust/thrust/system/detail/sequential/swap_ranges.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system has no special swap_ranges functions
 
diff --git a/thrust/thrust/system/detail/sequential/tabulate.h b/thrust/thrust/system/detail/sequential/tabulate.h
index 926cdec15a0..8a556172336 100644
--- a/thrust/thrust/system/detail/sequential/tabulate.h
+++ b/thrust/thrust/system/detail/sequential/tabulate.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system has no special tabulate functions
 
diff --git a/thrust/thrust/system/detail/sequential/temporary_buffer.h b/thrust/thrust/system/detail/sequential/temporary_buffer.h
index 2adfaf2810c..d666f267e35 100644
--- a/thrust/thrust/system/detail/sequential/temporary_buffer.h
+++ b/thrust/thrust/system/detail/sequential/temporary_buffer.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system has no special temporary buffer functions
 
diff --git a/thrust/thrust/system/detail/sequential/transform.h b/thrust/thrust/system/detail/sequential/transform.h
index 30305c15287..0cd9491b4a4 100644
--- a/thrust/thrust/system/detail/sequential/transform.h
+++ b/thrust/thrust/system/detail/sequential/transform.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system has no special transform functions
 
diff --git a/thrust/thrust/system/detail/sequential/transform_reduce.h b/thrust/thrust/system/detail/sequential/transform_reduce.h
index 8d2a1b3850d..15e8ed52d08 100644
--- a/thrust/thrust/system/detail/sequential/transform_reduce.h
+++ b/thrust/thrust/system/detail/sequential/transform_reduce.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system has no special transform_reduce functions
 
diff --git a/thrust/thrust/system/detail/sequential/transform_scan.h b/thrust/thrust/system/detail/sequential/transform_scan.h
index c6ae90664ad..c3f89646b4e 100644
--- a/thrust/thrust/system/detail/sequential/transform_scan.h
+++ b/thrust/thrust/system/detail/sequential/transform_scan.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
-// this system has no special version of this algorithm 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
+// this system has no special version of this algorithm
 
diff --git a/thrust/thrust/system/detail/sequential/trivial_copy.h b/thrust/thrust/system/detail/sequential/trivial_copy.h
index ea55c8fd201..44047db2b9b 100644
--- a/thrust/thrust/system/detail/sequential/trivial_copy.h
+++ b/thrust/thrust/system/detail/sequential/trivial_copy.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <cstring>
 #include <thrust/system/detail/sequential/general_copy.h>
 
diff --git a/thrust/thrust/system/detail/sequential/uninitialized_copy.h b/thrust/thrust/system/detail/sequential/uninitialized_copy.h
index c6ae90664ad..c3f89646b4e 100644
--- a/thrust/thrust/system/detail/sequential/uninitialized_copy.h
+++ b/thrust/thrust/system/detail/sequential/uninitialized_copy.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
-// this system has no special version of this algorithm 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
+// this system has no special version of this algorithm
 
diff --git a/thrust/thrust/system/detail/sequential/uninitialized_fill.h b/thrust/thrust/system/detail/sequential/uninitialized_fill.h
index 65e59fae5dc..cdf362d9d9d 100644
--- a/thrust/thrust/system/detail/sequential/uninitialized_fill.h
+++ b/thrust/thrust/system/detail/sequential/uninitialized_fill.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system has no special unintialized_fill functions
 
diff --git a/thrust/thrust/system/detail/sequential/unique.h b/thrust/thrust/system/detail/sequential/unique.h
index c4fe5268a33..e3897ba21dc 100644
--- a/thrust/thrust/system/detail/sequential/unique.h
+++ b/thrust/thrust/system/detail/sequential/unique.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/sequential/execution_policy.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/pair.h>
diff --git a/thrust/thrust/system/detail/sequential/unique_by_key.h b/thrust/thrust/system/detail/sequential/unique_by_key.h
index d30cc7c7153..1cd0cedcfc4 100644
--- a/thrust/thrust/system/detail/sequential/unique_by_key.h
+++ b/thrust/thrust/system/detail/sequential/unique_by_key.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/sequential/execution_policy.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/pair.h>
@@ -45,7 +51,7 @@ template<typename DerivedPolicy,
 __host__ __device__
   thrust::pair<OutputIterator1,OutputIterator2>
     unique_by_key_copy(sequential::execution_policy<DerivedPolicy> &,
-                       InputIterator1 keys_first, 
+                       InputIterator1 keys_first,
                        InputIterator1 keys_last,
                        InputIterator2 values_first,
                        OutputIterator1 keys_output,
@@ -98,7 +104,7 @@ template<typename DerivedPolicy,
 __host__ __device__
   thrust::pair<ForwardIterator1,ForwardIterator2>
     unique_by_key(sequential::execution_policy<DerivedPolicy> &exec,
-                  ForwardIterator1 keys_first, 
+                  ForwardIterator1 keys_first,
                   ForwardIterator1 keys_last,
                   ForwardIterator2 values_first,
                   BinaryPredicate binary_pred)
diff --git a/thrust/thrust/system/detail/system_error.inl b/thrust/thrust/system/detail/system_error.inl
index 075fe88e463..dbdf7d82243 100644
--- a/thrust/thrust/system/detail/system_error.inl
+++ b/thrust/thrust/system/detail/system_error.inl
@@ -19,6 +19,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/system/system_error.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/error_code.h b/thrust/thrust/system/error_code.h
index d460a315b1d..23e818a93ee 100644
--- a/thrust/thrust/system/error_code.h
+++ b/thrust/thrust/system/error_code.h
@@ -23,6 +23,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/type_traits.h>
 #include <thrust/system/detail/errno.h>
 #include <iostream>
@@ -419,7 +425,7 @@ class error_condition
 
     // XXX replace below with this upon c++0x
     //explicit operator bool (void) const;
-    
+
     /*! \return <tt>value() != 0</tt>.
      */
     inline operator bool (void) const;
diff --git a/thrust/thrust/system/omp/detail/adjacent_difference.h b/thrust/thrust/system/omp/detail/adjacent_difference.h
index 622ee61bafd..6706826e6ca 100644
--- a/thrust/thrust/system/omp/detail/adjacent_difference.h
+++ b/thrust/thrust/system/omp/detail/adjacent_difference.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/omp/detail/execution_policy.h>
 #include <thrust/system/detail/generic/adjacent_difference.h>
 
diff --git a/thrust/thrust/system/omp/detail/assign_value.h b/thrust/thrust/system/omp/detail/assign_value.h
index cf244a02193..5924261c7ca 100644
--- a/thrust/thrust/system/omp/detail/assign_value.h
+++ b/thrust/thrust/system/omp/detail/assign_value.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits assign_value
 #include <thrust/system/cpp/detail/assign_value.h>
 
diff --git a/thrust/thrust/system/omp/detail/binary_search.h b/thrust/thrust/system/omp/detail/binary_search.h
index 1ed700bd8c8..15de5974ee9 100644
--- a/thrust/thrust/system/omp/detail/binary_search.h
+++ b/thrust/thrust/system/omp/detail/binary_search.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/omp/detail/execution_policy.h>
 #include <thrust/system/detail/generic/binary_search.h>
 
@@ -33,7 +39,7 @@ template <typename DerivedPolicy, typename ForwardIterator, typename T, typename
 ForwardIterator lower_bound(execution_policy<DerivedPolicy> &exec,
                             ForwardIterator begin,
                             ForwardIterator end,
-                            const T& value, 
+                            const T& value,
                             StrictWeakOrdering comp)
 {
     // omp prefers generic::lower_bound to cpp::lower_bound
@@ -45,7 +51,7 @@ template <typename DerivedPolicy, typename ForwardIterator, typename T, typename
 ForwardIterator upper_bound(execution_policy<DerivedPolicy> &exec,
                             ForwardIterator begin,
                             ForwardIterator end,
-                            const T& value, 
+                            const T& value,
                             StrictWeakOrdering comp)
 {
     // omp prefers generic::upper_bound to cpp::upper_bound
@@ -57,7 +63,7 @@ template <typename DerivedPolicy, typename ForwardIterator, typename T, typename
 bool binary_search(execution_policy<DerivedPolicy> &exec,
                    ForwardIterator begin,
                    ForwardIterator end,
-                   const T& value, 
+                   const T& value,
                    StrictWeakOrdering comp)
 {
     // omp prefers generic::binary_search to cpp::binary_search
diff --git a/thrust/thrust/system/omp/detail/copy.h b/thrust/thrust/system/omp/detail/copy.h
index ae7b1eed777..9a735ca4b94 100644
--- a/thrust/thrust/system/omp/detail/copy.h
+++ b/thrust/thrust/system/omp/detail/copy.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/omp/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/omp/detail/copy.inl b/thrust/thrust/system/omp/detail/copy.inl
index 47f606dda63..de6ad0c851b 100644
--- a/thrust/thrust/system/omp/detail/copy.inl
+++ b/thrust/thrust/system/omp/detail/copy.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/omp/detail/copy.h>
 #include <thrust/system/detail/generic/copy.h>
 #include <thrust/system/detail/sequential/copy.h>
@@ -101,7 +107,7 @@ OutputIterator copy(execution_policy<DerivedPolicy> &exec,
 {
   typedef typename thrust::iterator_traversal<InputIterator>::type  traversal1;
   typedef typename thrust::iterator_traversal<OutputIterator>::type traversal2;
-  
+
   typedef typename thrust::detail::minimum_type<traversal1,traversal2>::type traversal;
 
   // dispatch on minimum traversal
@@ -121,7 +127,7 @@ OutputIterator copy_n(execution_policy<DerivedPolicy> &exec,
 {
   typedef typename thrust::iterator_traversal<InputIterator>::type  traversal1;
   typedef typename thrust::iterator_traversal<OutputIterator>::type traversal2;
-  
+
   typedef typename thrust::detail::minimum_type<traversal1,traversal2>::type traversal;
 
   // dispatch on minimum traversal
diff --git a/thrust/thrust/system/omp/detail/copy_if.h b/thrust/thrust/system/omp/detail/copy_if.h
index b33fd96df12..67be8febb13 100644
--- a/thrust/thrust/system/omp/detail/copy_if.h
+++ b/thrust/thrust/system/omp/detail/copy_if.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/omp/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/omp/detail/copy_if.inl b/thrust/thrust/system/omp/detail/copy_if.inl
index 8e597d4fcc5..43117e74d21 100644
--- a/thrust/thrust/system/omp/detail/copy_if.inl
+++ b/thrust/thrust/system/omp/detail/copy_if.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/omp/detail/copy_if.h>
 #include <thrust/system/detail/generic/copy_if.h>
 
diff --git a/thrust/thrust/system/omp/detail/count.h b/thrust/thrust/system/omp/detail/count.h
index fde1728b772..2b57cde7e4f 100644
--- a/thrust/thrust/system/omp/detail/count.h
+++ b/thrust/thrust/system/omp/detail/count.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits count
 #include <thrust/system/cpp/detail/count.h>
 
diff --git a/thrust/thrust/system/omp/detail/default_decomposition.h b/thrust/thrust/system/omp/detail/default_decomposition.h
index 2fe0a24fd74..232ab0422be 100644
--- a/thrust/thrust/system/omp/detail/default_decomposition.h
+++ b/thrust/thrust/system/omp/detail/default_decomposition.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/internal/decompose.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/omp/detail/default_decomposition.inl b/thrust/thrust/system/omp/detail/default_decomposition.inl
index 0698d53fbdc..ac7e36245eb 100644
--- a/thrust/thrust/system/omp/detail/default_decomposition.inl
+++ b/thrust/thrust/system/omp/detail/default_decomposition.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/omp/detail/default_decomposition.h>
 
 // don't attempt to #include this file without omp support
diff --git a/thrust/thrust/system/omp/detail/equal.h b/thrust/thrust/system/omp/detail/equal.h
index 13398fc9db5..b0dc474862c 100644
--- a/thrust/thrust/system/omp/detail/equal.h
+++ b/thrust/thrust/system/omp/detail/equal.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits equal
 #include <thrust/system/cpp/detail/equal.h>
 
diff --git a/thrust/thrust/system/omp/detail/execution_policy.h b/thrust/thrust/system/omp/detail/execution_policy.h
index f9b45312be8..850084cab9a 100644
--- a/thrust/thrust/system/omp/detail/execution_policy.h
+++ b/thrust/thrust/system/omp/detail/execution_policy.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/cpp/detail/execution_policy.h>
 #include <thrust/system/tbb/detail/execution_policy.h>
 #include <thrust/iterator/detail/any_system_tag.h>
@@ -58,7 +64,7 @@ template<typename Derived>
   struct execution_policy
     : thrust::system::cpp::detail::execution_policy<Derived>
 {
-  typedef tag tag_type; 
+  typedef tag tag_type;
   operator tag() const { return tag(); }
 };
 
diff --git a/thrust/thrust/system/omp/detail/extrema.h b/thrust/thrust/system/omp/detail/extrema.h
index bde4e5f80cf..fd9cbe89b66 100644
--- a/thrust/thrust/system/omp/detail/extrema.h
+++ b/thrust/thrust/system/omp/detail/extrema.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/omp/detail/execution_policy.h>
 #include <thrust/system/detail/generic/extrema.h>
 
@@ -30,7 +36,7 @@ namespace detail
 
 template <typename DerivedPolicy, typename ForwardIterator, typename BinaryPredicate>
 ForwardIterator max_element(execution_policy<DerivedPolicy> &exec,
-                            ForwardIterator first, 
+                            ForwardIterator first,
                             ForwardIterator last,
                             BinaryPredicate comp)
 {
@@ -40,7 +46,7 @@ ForwardIterator max_element(execution_policy<DerivedPolicy> &exec,
 
 template <typename DerivedPolicy, typename ForwardIterator, typename BinaryPredicate>
 ForwardIterator min_element(execution_policy<DerivedPolicy> &exec,
-                            ForwardIterator first, 
+                            ForwardIterator first,
                             ForwardIterator last,
                             BinaryPredicate comp)
 {
@@ -50,7 +56,7 @@ ForwardIterator min_element(execution_policy<DerivedPolicy> &exec,
 
 template <typename DerivedPolicy, typename ForwardIterator, typename BinaryPredicate>
 thrust::pair<ForwardIterator,ForwardIterator> minmax_element(execution_policy<DerivedPolicy> &exec,
-                                                             ForwardIterator first, 
+                                                             ForwardIterator first,
                                                              ForwardIterator last,
                                                              BinaryPredicate comp)
 {
diff --git a/thrust/thrust/system/omp/detail/fill.h b/thrust/thrust/system/omp/detail/fill.h
index 6665a264873..3bda479c892 100644
--- a/thrust/thrust/system/omp/detail/fill.h
+++ b/thrust/thrust/system/omp/detail/fill.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits fill
 #include <thrust/system/cpp/detail/fill.h>
 
diff --git a/thrust/thrust/system/omp/detail/find.h b/thrust/thrust/system/omp/detail/find.h
index d2abac95e30..6a8209c13c8 100644
--- a/thrust/thrust/system/omp/detail/find.h
+++ b/thrust/thrust/system/omp/detail/find.h
@@ -16,12 +16,18 @@
 
 
 /*! \file find.h
- *  \brief OpenMP implementation of find_if. 
+ *  \brief OpenMP implementation of find_if.
  */
 
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/find.h>
 #include <thrust/system/omp/detail/execution_policy.h>
 
diff --git a/thrust/thrust/system/omp/detail/for_each.h b/thrust/thrust/system/omp/detail/for_each.h
index a2030f37464..f138622e1bf 100644
--- a/thrust/thrust/system/omp/detail/for_each.h
+++ b/thrust/thrust/system/omp/detail/for_each.h
@@ -16,13 +16,19 @@
 
 
 /*! \file for_each.h
- *  \brief Defines the interface for a function that executes a 
+ *  \brief Defines the interface for a function that executes a
  *  function or functional for each value in a given range.
  */
 
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/omp/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/omp/detail/for_each.inl b/thrust/thrust/system/omp/detail/for_each.inl
index 4246d538070..80b47d4d173 100644
--- a/thrust/thrust/system/omp/detail/for_each.inl
+++ b/thrust/thrust/system/omp/detail/for_each.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/function.h>
 #include <thrust/detail/static_assert.h>
 #include <thrust/distance.h>
diff --git a/thrust/thrust/system/omp/detail/gather.h b/thrust/thrust/system/omp/detail/gather.h
index 098e0f4fbad..6630a1dd322 100644
--- a/thrust/thrust/system/omp/detail/gather.h
+++ b/thrust/thrust/system/omp/detail/gather.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits gather
 #include <thrust/system/cpp/detail/gather.h>
 
diff --git a/thrust/thrust/system/omp/detail/generate.h b/thrust/thrust/system/omp/detail/generate.h
index f907b6acc07..847f2b151f5 100644
--- a/thrust/thrust/system/omp/detail/generate.h
+++ b/thrust/thrust/system/omp/detail/generate.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits generate
 #include <thrust/system/cpp/detail/generate.h>
 
diff --git a/thrust/thrust/system/omp/detail/get_value.h b/thrust/thrust/system/omp/detail/get_value.h
index 23a11a8574f..cc3c0652d23 100644
--- a/thrust/thrust/system/omp/detail/get_value.h
+++ b/thrust/thrust/system/omp/detail/get_value.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits get_value
 #include <thrust/system/cpp/detail/get_value.h>
 
diff --git a/thrust/thrust/system/omp/detail/inner_product.h b/thrust/thrust/system/omp/detail/inner_product.h
index e8cf941a1dc..4a11b03869f 100644
--- a/thrust/thrust/system/omp/detail/inner_product.h
+++ b/thrust/thrust/system/omp/detail/inner_product.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits inner_product
 #include <thrust/system/cpp/detail/inner_product.h>
 
diff --git a/thrust/thrust/system/omp/detail/iter_swap.h b/thrust/thrust/system/omp/detail/iter_swap.h
index a096739947c..8e4fe18ba0e 100644
--- a/thrust/thrust/system/omp/detail/iter_swap.h
+++ b/thrust/thrust/system/omp/detail/iter_swap.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits iter_swap
 #include <thrust/system/cpp/detail/iter_swap.h>
 
diff --git a/thrust/thrust/system/omp/detail/logical.h b/thrust/thrust/system/omp/detail/logical.h
index 4199063183d..755655faecd 100644
--- a/thrust/thrust/system/omp/detail/logical.h
+++ b/thrust/thrust/system/omp/detail/logical.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits logical
 #include <thrust/system/cpp/detail/logical.h>
 
diff --git a/thrust/thrust/system/omp/detail/malloc_and_free.h b/thrust/thrust/system/omp/detail/malloc_and_free.h
index 01ab1e6dbe1..d68883f4561 100644
--- a/thrust/thrust/system/omp/detail/malloc_and_free.h
+++ b/thrust/thrust/system/omp/detail/malloc_and_free.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits malloc and free
 #include <thrust/system/cpp/detail/malloc_and_free.h>
 
diff --git a/thrust/thrust/system/omp/detail/memory.inl b/thrust/thrust/system/omp/detail/memory.inl
index db9b4f07b80..d66050d7ac2 100644
--- a/thrust/thrust/system/omp/detail/memory.inl
+++ b/thrust/thrust/system/omp/detail/memory.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/cpp/detail/execution_policy.h>
 #include <thrust/system/omp/memory.h>
 #include <thrust/system/cpp/memory.h>
diff --git a/thrust/thrust/system/omp/detail/merge.h b/thrust/thrust/system/omp/detail/merge.h
index 6a7bf6a7f80..1938e2e3940 100644
--- a/thrust/thrust/system/omp/detail/merge.h
+++ b/thrust/thrust/system/omp/detail/merge.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits merge
 #include <thrust/system/cpp/detail/merge.h>
 
diff --git a/thrust/thrust/system/omp/detail/mismatch.h b/thrust/thrust/system/omp/detail/mismatch.h
index b5c6b2c4bdd..2b9f3a3645c 100644
--- a/thrust/thrust/system/omp/detail/mismatch.h
+++ b/thrust/thrust/system/omp/detail/mismatch.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits mismatch
 #include <thrust/system/cpp/detail/mismatch.h>
 
diff --git a/thrust/thrust/system/omp/detail/par.h b/thrust/thrust/system/omp/detail/par.h
index b81a5d4894d..e1a07606453 100644
--- a/thrust/thrust/system/omp/detail/par.h
+++ b/thrust/thrust/system/omp/detail/par.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/allocator_aware_execution_policy.h>
 #include <thrust/system/omp/detail/execution_policy.h>
 
diff --git a/thrust/thrust/system/omp/detail/partition.h b/thrust/thrust/system/omp/detail/partition.h
index 7a6f4a93447..66f94abe693 100644
--- a/thrust/thrust/system/omp/detail/partition.h
+++ b/thrust/thrust/system/omp/detail/partition.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/omp/detail/execution_policy.h>
 #include <thrust/pair.h>
 
diff --git a/thrust/thrust/system/omp/detail/partition.inl b/thrust/thrust/system/omp/detail/partition.inl
index ba0a09eafb0..5737762ec40 100644
--- a/thrust/thrust/system/omp/detail/partition.inl
+++ b/thrust/thrust/system/omp/detail/partition.inl
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/omp/detail/partition.h>
 #include <thrust/system/detail/generic/partition.h>
 
diff --git a/thrust/thrust/system/omp/detail/per_device_resource.h b/thrust/thrust/system/omp/detail/per_device_resource.h
index 1b8d61f9216..64c721210a7 100644
--- a/thrust/thrust/system/omp/detail/per_device_resource.h
+++ b/thrust/thrust/system/omp/detail/per_device_resource.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system has no special per device resource functions
 
diff --git a/thrust/thrust/system/omp/detail/pragma_omp.h b/thrust/thrust/system/omp/detail/pragma_omp.h
index a8eeae23449..1b78d526651 100644
--- a/thrust/thrust/system/omp/detail/pragma_omp.h
+++ b/thrust/thrust/system/omp/detail/pragma_omp.h
@@ -29,6 +29,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_MSVC
 // MSVC ICEs when using the standard C++11 `_Pragma` operator with OpenMP
 // directives.
diff --git a/thrust/thrust/system/omp/detail/reduce.h b/thrust/thrust/system/omp/detail/reduce.h
index 5e5f2106ef7..39e4057847c 100644
--- a/thrust/thrust/system/omp/detail/reduce.h
+++ b/thrust/thrust/system/omp/detail/reduce.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/omp/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
@@ -34,7 +40,7 @@ namespace detail
 
 
 template<typename DerivedPolicy,
-         typename InputIterator, 
+         typename InputIterator,
          typename OutputType,
          typename BinaryFunction>
   OutputType reduce(execution_policy<DerivedPolicy> &exec,
diff --git a/thrust/thrust/system/omp/detail/reduce.inl b/thrust/thrust/system/omp/detail/reduce.inl
index 6a5723780b4..473dfe0c8be 100644
--- a/thrust/thrust/system/omp/detail/reduce.inl
+++ b/thrust/thrust/system/omp/detail/reduce.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/system/omp/detail/reduce.h>
 #include <thrust/system/omp/detail/default_decomposition.h>
diff --git a/thrust/thrust/system/omp/detail/reduce_by_key.h b/thrust/thrust/system/omp/detail/reduce_by_key.h
index 005616de5f1..997c2c88951 100644
--- a/thrust/thrust/system/omp/detail/reduce_by_key.h
+++ b/thrust/thrust/system/omp/detail/reduce_by_key.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/omp/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
@@ -42,7 +48,7 @@ template <typename DerivedPolicy,
           typename BinaryFunction>
   thrust::pair<OutputIterator1,OutputIterator2>
     reduce_by_key(execution_policy<DerivedPolicy> &exec,
-                  InputIterator1 keys_first, 
+                  InputIterator1 keys_first,
                   InputIterator1 keys_last,
                   InputIterator2 values_first,
                   OutputIterator1 keys_output,
diff --git a/thrust/thrust/system/omp/detail/reduce_by_key.inl b/thrust/thrust/system/omp/detail/reduce_by_key.inl
index 4088d06345e..3bc9e0c0ff6 100644
--- a/thrust/thrust/system/omp/detail/reduce_by_key.inl
+++ b/thrust/thrust/system/omp/detail/reduce_by_key.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/omp/detail/reduce_by_key.h>
 #include <thrust/system/detail/generic/reduce_by_key.h>
 #include <thrust/distance.h>
diff --git a/thrust/thrust/system/omp/detail/reduce_intervals.h b/thrust/thrust/system/omp/detail/reduce_intervals.h
index 1c69fc621ef..58938c53607 100644
--- a/thrust/thrust/system/omp/detail/reduce_intervals.h
+++ b/thrust/thrust/system/omp/detail/reduce_intervals.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/omp/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/omp/detail/reduce_intervals.inl b/thrust/thrust/system/omp/detail/reduce_intervals.inl
index 2668a7b6059..7254b1cbab2 100644
--- a/thrust/thrust/system/omp/detail/reduce_intervals.inl
+++ b/thrust/thrust/system/omp/detail/reduce_intervals.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/omp/detail/reduce_intervals.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/detail/function.h>
diff --git a/thrust/thrust/system/omp/detail/remove.h b/thrust/thrust/system/omp/detail/remove.h
index 9b2d46e75c6..793fa98f78e 100644
--- a/thrust/thrust/system/omp/detail/remove.h
+++ b/thrust/thrust/system/omp/detail/remove.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/omp/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/omp/detail/remove.inl b/thrust/thrust/system/omp/detail/remove.inl
index 5330f140725..82772cc1d79 100644
--- a/thrust/thrust/system/omp/detail/remove.inl
+++ b/thrust/thrust/system/omp/detail/remove.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/omp/detail/remove.h>
 #include <thrust/system/detail/generic/remove.h>
 
diff --git a/thrust/thrust/system/omp/detail/replace.h b/thrust/thrust/system/omp/detail/replace.h
index 95c5a14ba3d..12b1603a6a8 100644
--- a/thrust/thrust/system/omp/detail/replace.h
+++ b/thrust/thrust/system/omp/detail/replace.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits this algorithm
 #include <thrust/system/cpp/detail/scatter.h>
 
diff --git a/thrust/thrust/system/omp/detail/reverse.h b/thrust/thrust/system/omp/detail/reverse.h
index 1f3e0325e25..aedea33aa82 100644
--- a/thrust/thrust/system/omp/detail/reverse.h
+++ b/thrust/thrust/system/omp/detail/reverse.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits reverse
 #include <thrust/system/cpp/detail/reverse.h>
 
diff --git a/thrust/thrust/system/omp/detail/scan.h b/thrust/thrust/system/omp/detail/scan.h
index f47dbbc3087..f1f901dcab5 100644
--- a/thrust/thrust/system/omp/detail/scan.h
+++ b/thrust/thrust/system/omp/detail/scan.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits scan
 #include <thrust/system/cpp/detail/scan.h>
 
diff --git a/thrust/thrust/system/omp/detail/scan_by_key.h b/thrust/thrust/system/omp/detail/scan_by_key.h
index 7f6b42d5441..5c7e3d9942a 100644
--- a/thrust/thrust/system/omp/detail/scan_by_key.h
+++ b/thrust/thrust/system/omp/detail/scan_by_key.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits this algorithm
 #include <thrust/system/cpp/detail/scan_by_key.h>
 
diff --git a/thrust/thrust/system/omp/detail/scatter.h b/thrust/thrust/system/omp/detail/scatter.h
index 95c5a14ba3d..12b1603a6a8 100644
--- a/thrust/thrust/system/omp/detail/scatter.h
+++ b/thrust/thrust/system/omp/detail/scatter.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits this algorithm
 #include <thrust/system/cpp/detail/scatter.h>
 
diff --git a/thrust/thrust/system/omp/detail/sequence.h b/thrust/thrust/system/omp/detail/sequence.h
index c33b2d4333c..8d6e3c34f65 100644
--- a/thrust/thrust/system/omp/detail/sequence.h
+++ b/thrust/thrust/system/omp/detail/sequence.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits sequence
 #include <thrust/system/cpp/detail/sequence.h>
 
diff --git a/thrust/thrust/system/omp/detail/set_operations.h b/thrust/thrust/system/omp/detail/set_operations.h
index 421fa8a4bd9..890da986216 100644
--- a/thrust/thrust/system/omp/detail/set_operations.h
+++ b/thrust/thrust/system/omp/detail/set_operations.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits set_operations
 #include <thrust/system/cpp/detail/set_operations.h>
 
diff --git a/thrust/thrust/system/omp/detail/sort.h b/thrust/thrust/system/omp/detail/sort.h
index cf0b8c6d636..b673c572189 100644
--- a/thrust/thrust/system/omp/detail/sort.h
+++ b/thrust/thrust/system/omp/detail/sort.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/omp/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
@@ -34,7 +40,7 @@ void stable_sort(execution_policy<DerivedPolicy> &exec,
                  RandomAccessIterator first,
                  RandomAccessIterator last,
                  StrictWeakOrdering comp);
-    
+
 template<typename DerivedPolicy,
          typename RandomAccessIterator1,
          typename RandomAccessIterator2,
diff --git a/thrust/thrust/system/omp/detail/sort.inl b/thrust/thrust/system/omp/detail/sort.inl
index a0867ca4d3e..6a78af10616 100644
--- a/thrust/thrust/system/omp/detail/sort.inl
+++ b/thrust/thrust/system/omp/detail/sort.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // don't attempt to #include this file without omp support
 #if (THRUST_DEVICE_COMPILER_IS_OMP_CAPABLE == THRUST_TRUE)
 #include <omp.h>
diff --git a/thrust/thrust/system/omp/detail/swap_ranges.h b/thrust/thrust/system/omp/detail/swap_ranges.h
index 8c3338b1baf..2540b86e17b 100644
--- a/thrust/thrust/system/omp/detail/swap_ranges.h
+++ b/thrust/thrust/system/omp/detail/swap_ranges.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // omp inherits swap_ranges
 #include <thrust/system/cpp/detail/swap_ranges.h>
 
diff --git a/thrust/thrust/system/omp/detail/tabulate.h b/thrust/thrust/system/omp/detail/tabulate.h
index ea135c70706..930d9efd15f 100644
--- a/thrust/thrust/system/omp/detail/tabulate.h
+++ b/thrust/thrust/system/omp/detail/tabulate.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits tabulate
 #include <thrust/system/cpp/detail/tabulate.h>
 
diff --git a/thrust/thrust/system/omp/detail/temporary_buffer.h b/thrust/thrust/system/omp/detail/temporary_buffer.h
index 2adfaf2810c..d666f267e35 100644
--- a/thrust/thrust/system/omp/detail/temporary_buffer.h
+++ b/thrust/thrust/system/omp/detail/temporary_buffer.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system has no special temporary buffer functions
 
diff --git a/thrust/thrust/system/omp/detail/transform.h b/thrust/thrust/system/omp/detail/transform.h
index 20d606dfbee..8a572e02bc3 100644
--- a/thrust/thrust/system/omp/detail/transform.h
+++ b/thrust/thrust/system/omp/detail/transform.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // omp inherits transform
 #include <thrust/system/cpp/detail/transform.h>
 
diff --git a/thrust/thrust/system/omp/detail/transform_reduce.h b/thrust/thrust/system/omp/detail/transform_reduce.h
index a8736bd75d0..5fb51bf8a07 100644
--- a/thrust/thrust/system/omp/detail/transform_reduce.h
+++ b/thrust/thrust/system/omp/detail/transform_reduce.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits transform_reduce
 #include <thrust/system/cpp/detail/transform_reduce.h>
 
diff --git a/thrust/thrust/system/omp/detail/transform_scan.h b/thrust/thrust/system/omp/detail/transform_scan.h
index 75b075b6b16..02371a3fc09 100644
--- a/thrust/thrust/system/omp/detail/transform_scan.h
+++ b/thrust/thrust/system/omp/detail/transform_scan.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits transform_scan
 #include <thrust/system/cpp/detail/transform_scan.h>
 
diff --git a/thrust/thrust/system/omp/detail/uninitialized_copy.h b/thrust/thrust/system/omp/detail/uninitialized_copy.h
index bda06ac13e9..5ba0d83cde8 100644
--- a/thrust/thrust/system/omp/detail/uninitialized_copy.h
+++ b/thrust/thrust/system/omp/detail/uninitialized_copy.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits uninitialized_copy
 #include <thrust/system/cpp/detail/uninitialized_copy.h>
 
diff --git a/thrust/thrust/system/omp/detail/uninitialized_fill.h b/thrust/thrust/system/omp/detail/uninitialized_fill.h
index 764de876233..f1b4a302c6d 100644
--- a/thrust/thrust/system/omp/detail/uninitialized_fill.h
+++ b/thrust/thrust/system/omp/detail/uninitialized_fill.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits uninitialized_fill
 #include <thrust/system/cpp/detail/uninitialized_fill.h>
 
diff --git a/thrust/thrust/system/omp/detail/unique.h b/thrust/thrust/system/omp/detail/unique.h
index cf8025665ac..a95aaad5037 100644
--- a/thrust/thrust/system/omp/detail/unique.h
+++ b/thrust/thrust/system/omp/detail/unique.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/omp/detail/execution_policy.h>
 #include <thrust/pair.h>
 
@@ -60,7 +66,7 @@ template<typename DerivedPolicy,
 
 
 } // end namespace detail
-} // end namespace omp 
+} // end namespace omp
 } // end namespace system
 THRUST_NAMESPACE_END
 
diff --git a/thrust/thrust/system/omp/detail/unique.inl b/thrust/thrust/system/omp/detail/unique.inl
index 9a93fb13584..90d10b548bc 100644
--- a/thrust/thrust/system/omp/detail/unique.inl
+++ b/thrust/thrust/system/omp/detail/unique.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/omp/detail/unique.h>
 #include <thrust/system/detail/generic/unique.h>
 #include <thrust/pair.h>
@@ -73,7 +79,7 @@ template<typename DerivedPolicy,
 
 
 } // end namespace detail
-} // end namespace omp 
+} // end namespace omp
 } // end namespace system
 THRUST_NAMESPACE_END
 
diff --git a/thrust/thrust/system/omp/detail/unique_by_key.h b/thrust/thrust/system/omp/detail/unique_by_key.h
index 43859b64e89..84d721e5e97 100644
--- a/thrust/thrust/system/omp/detail/unique_by_key.h
+++ b/thrust/thrust/system/omp/detail/unique_by_key.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/omp/detail/execution_policy.h>
 #include <thrust/pair.h>
 
@@ -35,7 +41,7 @@ template<typename DerivedPolicy,
          typename BinaryPredicate>
   thrust::pair<ForwardIterator1,ForwardIterator2>
     unique_by_key(execution_policy<DerivedPolicy> &exec,
-                  ForwardIterator1 keys_first, 
+                  ForwardIterator1 keys_first,
                   ForwardIterator1 keys_last,
                   ForwardIterator2 values_first,
                   BinaryPredicate binary_pred);
@@ -49,7 +55,7 @@ template<typename DerivedPolicy,
          typename BinaryPredicate>
   thrust::pair<OutputIterator1,OutputIterator2>
     unique_by_key_copy(execution_policy<DerivedPolicy> &exec,
-                       InputIterator1 keys_first, 
+                       InputIterator1 keys_first,
                        InputIterator1 keys_last,
                        InputIterator2 values_first,
                        OutputIterator1 keys_output,
@@ -58,7 +64,7 @@ template<typename DerivedPolicy,
 
 
 } // end namespace detail
-} // end namespace omp 
+} // end namespace omp
 } // end namespace system
 THRUST_NAMESPACE_END
 
diff --git a/thrust/thrust/system/omp/detail/unique_by_key.inl b/thrust/thrust/system/omp/detail/unique_by_key.inl
index 6610c8a0080..ca183c527bc 100644
--- a/thrust/thrust/system/omp/detail/unique_by_key.inl
+++ b/thrust/thrust/system/omp/detail/unique_by_key.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/omp/detail/unique_by_key.h>
 #include <thrust/system/detail/generic/unique_by_key.h>
 #include <thrust/pair.h>
@@ -36,7 +42,7 @@ template<typename DerivedPolicy,
          typename BinaryPredicate>
   thrust::pair<ForwardIterator1,ForwardIterator2>
     unique_by_key(execution_policy<DerivedPolicy> &exec,
-                  ForwardIterator1 keys_first, 
+                  ForwardIterator1 keys_first,
                   ForwardIterator1 keys_last,
                   ForwardIterator2 values_first,
                   BinaryPredicate binary_pred)
@@ -54,7 +60,7 @@ template<typename DerivedPolicy,
          typename BinaryPredicate>
   thrust::pair<OutputIterator1,OutputIterator2>
     unique_by_key_copy(execution_policy<DerivedPolicy> &exec,
-                       InputIterator1 keys_first, 
+                       InputIterator1 keys_first,
                        InputIterator1 keys_last,
                        InputIterator2 values_first,
                        OutputIterator1 keys_output,
diff --git a/thrust/thrust/system/omp/execution_policy.h b/thrust/thrust/system/omp/execution_policy.h
index c027d6be67d..f499eaed51b 100644
--- a/thrust/thrust/system/omp/execution_policy.h
+++ b/thrust/thrust/system/omp/execution_policy.h
@@ -16,12 +16,18 @@
 
 #pragma once
 
+#include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 /*! \file thrust/system/omp/execution_policy.h
  *  \brief Execution policies for Thrust's OpenMP system.
  */
 
-#include <thrust/detail/config.h>
-
 // get the execution policies definitions first
 #include <thrust/system/omp/detail/execution_policy.h>
 
diff --git a/thrust/thrust/system/omp/memory.h b/thrust/thrust/system/omp/memory.h
index 31f407c4c6f..b568b77079c 100644
--- a/thrust/thrust/system/omp/memory.h
+++ b/thrust/thrust/system/omp/memory.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/omp/memory_resource.h>
 #include <thrust/memory.h>
 #include <thrust/detail/type_traits.h>
diff --git a/thrust/thrust/system/omp/memory_resource.h b/thrust/thrust/system/omp/memory_resource.h
index d8eed0c0f9e..c2d3c9e650c 100644
--- a/thrust/thrust/system/omp/memory_resource.h
+++ b/thrust/thrust/system/omp/memory_resource.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/mr/new.h>
 #include <thrust/mr/fancy_pointer_resource.h>
 
diff --git a/thrust/thrust/system/omp/pointer.h b/thrust/thrust/system/omp/pointer.h
index 2be42e4fcf6..3cf07954160 100644
--- a/thrust/thrust/system/omp/pointer.h
+++ b/thrust/thrust/system/omp/pointer.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <type_traits>
 #include <thrust/system/omp/detail/execution_policy.h>
 #include <thrust/detail/pointer.h>
diff --git a/thrust/thrust/system/omp/vector.h b/thrust/thrust/system/omp/vector.h
index 179b5207dd6..98d8765155f 100644
--- a/thrust/thrust/system/omp/vector.h
+++ b/thrust/thrust/system/omp/vector.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/omp/memory.h>
 #include <thrust/detail/vector_base.h>
 #include <vector>
diff --git a/thrust/thrust/system/system_error.h b/thrust/thrust/system/system_error.h
index fb31a2da805..438518e72ea 100644
--- a/thrust/thrust/system/system_error.h
+++ b/thrust/thrust/system/system_error.h
@@ -23,6 +23,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <stdexcept>
 #include <string>
 
@@ -90,7 +96,7 @@ class system_error
 {
   public:
     // [19.5.5.2] Class system_error members
-    
+
     /*! Constructs an object of class \p system_error.
      *  \param ec The value returned by \p code().
      *  \param what_arg A string to include in the result returned by \p what().
@@ -141,7 +147,7 @@ class system_error
     /*! Destructor does not throw.
      */
     inline virtual ~system_error(void) noexcept {};
-    
+
     /*! Returns an object encoding the error.
      *  \return <tt>ec</tt> or <tt>error_code(ev, ecat)</tt>, from the
      *          constructor, as appropriate.
diff --git a/thrust/thrust/system/tbb/detail/adjacent_difference.h b/thrust/thrust/system/tbb/detail/adjacent_difference.h
index ab519d11eb4..70d93b46286 100644
--- a/thrust/thrust/system/tbb/detail/adjacent_difference.h
+++ b/thrust/thrust/system/tbb/detail/adjacent_difference.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/tbb/detail/execution_policy.h>
 #include <thrust/system/detail/generic/adjacent_difference.h>
 
diff --git a/thrust/thrust/system/tbb/detail/assign_value.h b/thrust/thrust/system/tbb/detail/assign_value.h
index cf244a02193..5924261c7ca 100644
--- a/thrust/thrust/system/tbb/detail/assign_value.h
+++ b/thrust/thrust/system/tbb/detail/assign_value.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits assign_value
 #include <thrust/system/cpp/detail/assign_value.h>
 
diff --git a/thrust/thrust/system/tbb/detail/binary_search.h b/thrust/thrust/system/tbb/detail/binary_search.h
index 0847e5d1fdb..eed753e11c0 100644
--- a/thrust/thrust/system/tbb/detail/binary_search.h
+++ b/thrust/thrust/system/tbb/detail/binary_search.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits binary_search
 #include <thrust/system/cpp/detail/binary_search.h>
 
diff --git a/thrust/thrust/system/tbb/detail/copy.h b/thrust/thrust/system/tbb/detail/copy.h
index 30e95a98ce2..f4d737b0494 100644
--- a/thrust/thrust/system/tbb/detail/copy.h
+++ b/thrust/thrust/system/tbb/detail/copy.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/tbb/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/tbb/detail/copy.inl b/thrust/thrust/system/tbb/detail/copy.inl
index 1016f40d4da..bb20559b5fb 100644
--- a/thrust/thrust/system/tbb/detail/copy.inl
+++ b/thrust/thrust/system/tbb/detail/copy.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/tbb/detail/copy.h>
 #include <thrust/system/detail/generic/copy.h>
 #include <thrust/system/detail/sequential/copy.h>
@@ -101,7 +107,7 @@ OutputIterator copy(execution_policy<DerivedPolicy> &exec,
 {
   typedef typename thrust::iterator_traversal<InputIterator>::type  traversal1;
   typedef typename thrust::iterator_traversal<OutputIterator>::type traversal2;
-  
+
   typedef typename thrust::detail::minimum_type<traversal1,traversal2>::type traversal;
 
   // dispatch on minimum traversal
@@ -121,7 +127,7 @@ OutputIterator copy_n(execution_policy<DerivedPolicy> &exec,
 {
   typedef typename thrust::iterator_traversal<InputIterator>::type  traversal1;
   typedef typename thrust::iterator_traversal<OutputIterator>::type traversal2;
-  
+
   typedef typename thrust::detail::minimum_type<traversal1,traversal2>::type traversal;
 
   // dispatch on minimum traversal
diff --git a/thrust/thrust/system/tbb/detail/copy_if.h b/thrust/thrust/system/tbb/detail/copy_if.h
index db860f37758..87b60db56b5 100644
--- a/thrust/thrust/system/tbb/detail/copy_if.h
+++ b/thrust/thrust/system/tbb/detail/copy_if.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/tbb/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/tbb/detail/copy_if.inl b/thrust/thrust/system/tbb/detail/copy_if.inl
index aa2379b8d03..f566020efe4 100644
--- a/thrust/thrust/system/tbb/detail/copy_if.inl
+++ b/thrust/thrust/system/tbb/detail/copy_if.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/function.h>
 #include <thrust/system/tbb/detail/copy_if.h>
 #include <thrust/iterator/iterator_traits.h>
@@ -66,13 +72,13 @@ struct body
         ++sum;
     }
   }
-  
+
   void operator()(const ::tbb::blocked_range<Size>& r, ::tbb::final_scan_tag)
   {
     InputIterator1  iter1 = first   + r.begin();
     InputIterator2  iter2 = stencil + r.begin();
     OutputIterator  iter3 = result  + sum;
-      
+
     for (Size i = r.begin(); i != r.end(); ++i, ++iter1, ++iter2)
     {
       if (pred(*iter2))
@@ -87,12 +93,12 @@ struct body
   void reverse_join(body& b)
   {
     sum = b.sum + sum;
-  } 
+  }
 
   void assign(body& b)
   {
     sum = b.sum;
-  } 
+  }
 }; // end body
 
 } // end copy_if_detail
@@ -108,9 +114,9 @@ template<typename InputIterator1,
                          OutputIterator result,
                          Predicate pred)
 {
-  typedef typename thrust::iterator_difference<InputIterator1>::type Size; 
+  typedef typename thrust::iterator_difference<InputIterator1>::type Size;
   typedef typename copy_if_detail::body<InputIterator1,InputIterator2,OutputIterator,Predicate,Size> Body;
-  
+
   Size n = thrust::distance(first, last);
 
   if (n != 0)
diff --git a/thrust/thrust/system/tbb/detail/count.h b/thrust/thrust/system/tbb/detail/count.h
index fde1728b772..2b57cde7e4f 100644
--- a/thrust/thrust/system/tbb/detail/count.h
+++ b/thrust/thrust/system/tbb/detail/count.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits count
 #include <thrust/system/cpp/detail/count.h>
 
diff --git a/thrust/thrust/system/tbb/detail/equal.h b/thrust/thrust/system/tbb/detail/equal.h
index 13398fc9db5..b0dc474862c 100644
--- a/thrust/thrust/system/tbb/detail/equal.h
+++ b/thrust/thrust/system/tbb/detail/equal.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits equal
 #include <thrust/system/cpp/detail/equal.h>
 
diff --git a/thrust/thrust/system/tbb/detail/execution_policy.h b/thrust/thrust/system/tbb/detail/execution_policy.h
index ac4a788e760..a67b7db379d 100644
--- a/thrust/thrust/system/tbb/detail/execution_policy.h
+++ b/thrust/thrust/system/tbb/detail/execution_policy.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/cpp/detail/execution_policy.h>
 #include <thrust/iterator/detail/any_system_tag.h>
 #include <thrust/detail/type_traits.h>
@@ -57,7 +63,7 @@ template<typename Derived>
   struct execution_policy
     : thrust::system::cpp::detail::execution_policy<Derived>
 {
-  typedef tag tag_type; 
+  typedef tag tag_type;
   operator tag() const { return tag(); }
 };
 
diff --git a/thrust/thrust/system/tbb/detail/extrema.h b/thrust/thrust/system/tbb/detail/extrema.h
index c6c747f426b..2cfdac87f8a 100644
--- a/thrust/thrust/system/tbb/detail/extrema.h
+++ b/thrust/thrust/system/tbb/detail/extrema.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/tbb/detail/execution_policy.h>
 #include <thrust/system/detail/generic/extrema.h>
 
@@ -30,7 +36,7 @@ namespace detail
 
 template <typename DerivedPolicy, typename ForwardIterator, typename BinaryPredicate>
 ForwardIterator max_element(execution_policy<DerivedPolicy> &exec,
-                            ForwardIterator first, 
+                            ForwardIterator first,
                             ForwardIterator last,
                             BinaryPredicate comp)
 {
@@ -40,7 +46,7 @@ ForwardIterator max_element(execution_policy<DerivedPolicy> &exec,
 
 template <typename DerivedPolicy, typename ForwardIterator, typename BinaryPredicate>
 ForwardIterator min_element(execution_policy<DerivedPolicy> &exec,
-                            ForwardIterator first, 
+                            ForwardIterator first,
                             ForwardIterator last,
                             BinaryPredicate comp)
 {
@@ -50,7 +56,7 @@ ForwardIterator min_element(execution_policy<DerivedPolicy> &exec,
 
 template <typename DerivedPolicy, typename ForwardIterator, typename BinaryPredicate>
 thrust::pair<ForwardIterator,ForwardIterator> minmax_element(execution_policy<DerivedPolicy> &exec,
-                                                             ForwardIterator first, 
+                                                             ForwardIterator first,
                                                              ForwardIterator last,
                                                              BinaryPredicate comp)
 {
diff --git a/thrust/thrust/system/tbb/detail/fill.h b/thrust/thrust/system/tbb/detail/fill.h
index 6665a264873..3bda479c892 100644
--- a/thrust/thrust/system/tbb/detail/fill.h
+++ b/thrust/thrust/system/tbb/detail/fill.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits fill
 #include <thrust/system/cpp/detail/fill.h>
 
diff --git a/thrust/thrust/system/tbb/detail/find.h b/thrust/thrust/system/tbb/detail/find.h
index e5dea8e779d..ecd6e11f7cf 100644
--- a/thrust/thrust/system/tbb/detail/find.h
+++ b/thrust/thrust/system/tbb/detail/find.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/detail/generic/find.h>
 #include <thrust/system/tbb/detail/execution_policy.h>
 
diff --git a/thrust/thrust/system/tbb/detail/for_each.h b/thrust/thrust/system/tbb/detail/for_each.h
index 26c4b539b05..94b7803101e 100644
--- a/thrust/thrust/system/tbb/detail/for_each.h
+++ b/thrust/thrust/system/tbb/detail/for_each.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/tbb/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/tbb/detail/for_each.inl b/thrust/thrust/system/tbb/detail/for_each.inl
index 21dfce9aebf..8e95e221cdf 100644
--- a/thrust/thrust/system/tbb/detail/for_each.inl
+++ b/thrust/thrust/system/tbb/detail/for_each.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/static_assert.h>
 #include <thrust/distance.h>
 #include <thrust/iterator/iterator_traits.h>
diff --git a/thrust/thrust/system/tbb/detail/gather.h b/thrust/thrust/system/tbb/detail/gather.h
index 098e0f4fbad..6630a1dd322 100644
--- a/thrust/thrust/system/tbb/detail/gather.h
+++ b/thrust/thrust/system/tbb/detail/gather.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits gather
 #include <thrust/system/cpp/detail/gather.h>
 
diff --git a/thrust/thrust/system/tbb/detail/generate.h b/thrust/thrust/system/tbb/detail/generate.h
index f907b6acc07..847f2b151f5 100644
--- a/thrust/thrust/system/tbb/detail/generate.h
+++ b/thrust/thrust/system/tbb/detail/generate.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits generate
 #include <thrust/system/cpp/detail/generate.h>
 
diff --git a/thrust/thrust/system/tbb/detail/get_value.h b/thrust/thrust/system/tbb/detail/get_value.h
index 23a11a8574f..cc3c0652d23 100644
--- a/thrust/thrust/system/tbb/detail/get_value.h
+++ b/thrust/thrust/system/tbb/detail/get_value.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits get_value
 #include <thrust/system/cpp/detail/get_value.h>
 
diff --git a/thrust/thrust/system/tbb/detail/inner_product.h b/thrust/thrust/system/tbb/detail/inner_product.h
index e8cf941a1dc..4a11b03869f 100644
--- a/thrust/thrust/system/tbb/detail/inner_product.h
+++ b/thrust/thrust/system/tbb/detail/inner_product.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits inner_product
 #include <thrust/system/cpp/detail/inner_product.h>
 
diff --git a/thrust/thrust/system/tbb/detail/iter_swap.h b/thrust/thrust/system/tbb/detail/iter_swap.h
index a096739947c..8e4fe18ba0e 100644
--- a/thrust/thrust/system/tbb/detail/iter_swap.h
+++ b/thrust/thrust/system/tbb/detail/iter_swap.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits iter_swap
 #include <thrust/system/cpp/detail/iter_swap.h>
 
diff --git a/thrust/thrust/system/tbb/detail/logical.h b/thrust/thrust/system/tbb/detail/logical.h
index 4199063183d..755655faecd 100644
--- a/thrust/thrust/system/tbb/detail/logical.h
+++ b/thrust/thrust/system/tbb/detail/logical.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits logical
 #include <thrust/system/cpp/detail/logical.h>
 
diff --git a/thrust/thrust/system/tbb/detail/malloc_and_free.h b/thrust/thrust/system/tbb/detail/malloc_and_free.h
index 01ab1e6dbe1..d68883f4561 100644
--- a/thrust/thrust/system/tbb/detail/malloc_and_free.h
+++ b/thrust/thrust/system/tbb/detail/malloc_and_free.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits malloc and free
 #include <thrust/system/cpp/detail/malloc_and_free.h>
 
diff --git a/thrust/thrust/system/tbb/detail/memory.inl b/thrust/thrust/system/tbb/detail/memory.inl
index 32e28300a41..d0985ef568c 100644
--- a/thrust/thrust/system/tbb/detail/memory.inl
+++ b/thrust/thrust/system/tbb/detail/memory.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/cpp/detail/execution_policy.h>
 #include <thrust/system/tbb/memory.h>
 #include <thrust/system/cpp/memory.h>
diff --git a/thrust/thrust/system/tbb/detail/merge.h b/thrust/thrust/system/tbb/detail/merge.h
index 014e2eb8bce..a64b03eb637 100644
--- a/thrust/thrust/system/tbb/detail/merge.h
+++ b/thrust/thrust/system/tbb/detail/merge.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/tbb/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/tbb/detail/merge.inl b/thrust/thrust/system/tbb/detail/merge.inl
index 89a01aebfc8..4b61e16bc63 100644
--- a/thrust/thrust/system/tbb/detail/merge.inl
+++ b/thrust/thrust/system/tbb/detail/merge.inl
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/detail/temporary_array.h>
 #include <thrust/system/tbb/detail/execution_policy.h>
diff --git a/thrust/thrust/system/tbb/detail/mismatch.h b/thrust/thrust/system/tbb/detail/mismatch.h
index b5c6b2c4bdd..2b9f3a3645c 100644
--- a/thrust/thrust/system/tbb/detail/mismatch.h
+++ b/thrust/thrust/system/tbb/detail/mismatch.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits mismatch
 #include <thrust/system/cpp/detail/mismatch.h>
 
diff --git a/thrust/thrust/system/tbb/detail/par.h b/thrust/thrust/system/tbb/detail/par.h
index 308d41e1366..74a5f6ef6d0 100644
--- a/thrust/thrust/system/tbb/detail/par.h
+++ b/thrust/thrust/system/tbb/detail/par.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/allocator_aware_execution_policy.h>
 #include <thrust/system/tbb/detail/execution_policy.h>
 
diff --git a/thrust/thrust/system/tbb/detail/partition.h b/thrust/thrust/system/tbb/detail/partition.h
index f9c56b92b52..ec60d3d456c 100644
--- a/thrust/thrust/system/tbb/detail/partition.h
+++ b/thrust/thrust/system/tbb/detail/partition.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/tbb/detail/execution_policy.h>
 #include <thrust/pair.h>
 
diff --git a/thrust/thrust/system/tbb/detail/partition.inl b/thrust/thrust/system/tbb/detail/partition.inl
index 74ad809da4e..4cdb0492247 100644
--- a/thrust/thrust/system/tbb/detail/partition.inl
+++ b/thrust/thrust/system/tbb/detail/partition.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/tbb/detail/partition.h>
 #include <thrust/system/detail/generic/partition.h>
 
diff --git a/thrust/thrust/system/tbb/detail/per_device_resource.h b/thrust/thrust/system/tbb/detail/per_device_resource.h
index 1b8d61f9216..64c721210a7 100644
--- a/thrust/thrust/system/tbb/detail/per_device_resource.h
+++ b/thrust/thrust/system/tbb/detail/per_device_resource.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system has no special per device resource functions
 
diff --git a/thrust/thrust/system/tbb/detail/reduce.h b/thrust/thrust/system/tbb/detail/reduce.h
index 81e8d1f6f30..b12c7f43927 100644
--- a/thrust/thrust/system/tbb/detail/reduce.h
+++ b/thrust/thrust/system/tbb/detail/reduce.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/tbb/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
@@ -34,7 +40,7 @@ namespace detail
 
 
 template<typename DerivedPolicy,
-         typename InputIterator, 
+         typename InputIterator,
          typename OutputType,
          typename BinaryFunction>
   OutputType reduce(execution_policy<DerivedPolicy> &exec,
diff --git a/thrust/thrust/system/tbb/detail/reduce.inl b/thrust/thrust/system/tbb/detail/reduce.inl
index 47fe6616d6b..96621d9bf63 100644
--- a/thrust/thrust/system/tbb/detail/reduce.inl
+++ b/thrust/thrust/system/tbb/detail/reduce.inl
@@ -18,6 +18,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/function.h>
 #include <thrust/detail/static_assert.h>
 #include <thrust/iterator/iterator_traits.h>
@@ -60,7 +66,7 @@ struct body
   void operator()(const ::tbb::blocked_range<Size> &r)
   {
     // we assume that blocked_range specifies a contiguous range of integers
-    
+
     if (r.empty()) return; // nothing to do
 
     RandomAccessIterator iter = first + r.begin();
@@ -85,7 +91,7 @@ struct body
       sum = binary_op(sum, temp);
     }
   } // end operator()()
-  
+
   void join(body& b)
   {
     sum = binary_op(sum, b.sum);
@@ -96,7 +102,7 @@ struct body
 
 
 template<typename DerivedPolicy,
-         typename InputIterator, 
+         typename InputIterator,
          typename OutputType,
          typename BinaryFunction>
   OutputType reduce(execution_policy<DerivedPolicy> &,
@@ -105,7 +111,7 @@ template<typename DerivedPolicy,
                     OutputType init,
                     BinaryFunction binary_op)
 {
-  typedef typename thrust::iterator_difference<InputIterator>::type Size; 
+  typedef typename thrust::iterator_difference<InputIterator>::type Size;
 
   Size n = thrust::distance(begin, end);
 
diff --git a/thrust/thrust/system/tbb/detail/reduce_by_key.h b/thrust/thrust/system/tbb/detail/reduce_by_key.h
index 04d46e7c0fa..0b14ca2787e 100644
--- a/thrust/thrust/system/tbb/detail/reduce_by_key.h
+++ b/thrust/thrust/system/tbb/detail/reduce_by_key.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/tbb/detail/execution_policy.h>
 #include <thrust/pair.h>
 
@@ -38,7 +44,7 @@ template<typename DerivedPolicy,
          typename BinaryFunction>
   thrust::pair<OutputIterator1,OutputIterator2>
     reduce_by_key(execution_policy<DerivedPolicy> &exec,
-                  InputIterator1 keys_first, 
+                  InputIterator1 keys_first,
                   InputIterator1 keys_last,
                   InputIterator2 values_first,
                   OutputIterator1 keys_output,
diff --git a/thrust/thrust/system/tbb/detail/reduce_by_key.inl b/thrust/thrust/system/tbb/detail/reduce_by_key.inl
index 693abb2e739..4053f0faf1e 100644
--- a/thrust/thrust/system/tbb/detail/reduce_by_key.inl
+++ b/thrust/thrust/system/tbb/detail/reduce_by_key.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/tbb/detail/reduce_by_key.h>
 #include <thrust/iterator/reverse_iterator.h>
 #include <thrust/detail/seq.h>
@@ -125,7 +131,7 @@ template<typename InputIterator1,
     typename thrust::iterator_value<InputIterator1>::type,
     typename partial_sum_type<InputIterator2,BinaryFunction>::type
   >
-    reduce_by_key_with_carry(InputIterator1 keys_first, 
+    reduce_by_key_with_carry(InputIterator1 keys_first,
                              InputIterator1 keys_last,
                              InputIterator2 values_first,
                              OutputIterator1 keys_output,
@@ -145,7 +151,7 @@ template<typename InputIterator1,
   // finish with sequential reduce_by_key
   thrust::tie(keys_output, values_output) =
     thrust::reduce_by_key(thrust::seq, keys_first, keys_last, values_first, keys_output, values_output, binary_pred, binary_op);
-  
+
   return thrust::make_tuple(keys_output, values_output, carry.first, carry.second);
 }
 
@@ -153,7 +159,7 @@ template<typename InputIterator1,
 template<typename Iterator>
   bool interval_has_carry(size_t interval_idx, size_t interval_size, size_t num_intervals, Iterator tail_flags)
 {
-  // to discover whether the interval has a carry, look at the tail_flag corresponding to its last element 
+  // to discover whether the interval has a carry, look at the tail_flag corresponding to its last element
   // the final interval never has a carry by definition
   return (interval_idx + 1 < num_intervals) ? !tail_flags[(interval_idx + 1) * interval_size - 1] : false;
 }
@@ -226,7 +232,7 @@ template<typename Iterator1, typename Iterator2, typename Iterator3, typename It
 
     // store to carry only when we actually have a carry
     // store to my_keys_result & my_values_result otherwise
-    
+
     // create tail_flags so we can check for a carry
     thrust::detail::tail_flags<Iterator1,BinaryPredicate> flags = thrust::detail::make_tail_flags(keys_first, keys_first + n, binary_pred);
 
@@ -259,7 +265,7 @@ template<typename Iterator1, typename Iterator2, typename Iterator3, typename It
 template<typename DerivedPolicy, typename Iterator1, typename Iterator2, typename Iterator3, typename Iterator4, typename BinaryPredicate, typename BinaryFunction>
   thrust::pair<Iterator3,Iterator4>
     reduce_by_key(thrust::tbb::execution_policy<DerivedPolicy> &exec,
-                  Iterator1 keys_first, Iterator1 keys_last, 
+                  Iterator1 keys_first, Iterator1 keys_last,
                   Iterator2 values_first,
                   Iterator3 keys_result,
                   Iterator4 values_result,
@@ -300,7 +306,7 @@ template<typename DerivedPolicy, typename Iterator1, typename Iterator2, typenam
 
   // scan the counts to get each body's output offset
   thrust::inclusive_scan(thrust::seq,
-                         interval_output_offsets.begin() + 1, interval_output_offsets.end(), 
+                         interval_output_offsets.begin() + 1, interval_output_offsets.end(),
                          interval_output_offsets.begin() + 1);
 
   // do a reduce_by_key serially in each thread
diff --git a/thrust/thrust/system/tbb/detail/reduce_intervals.h b/thrust/thrust/system/tbb/detail/reduce_intervals.h
index 7164c3f97c5..dbb198ae7de 100644
--- a/thrust/thrust/system/tbb/detail/reduce_intervals.h
+++ b/thrust/thrust/system/tbb/detail/reduce_intervals.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/tbb/detail/execution_policy.h>
 #include <thrust/detail/seq.h>
 
diff --git a/thrust/thrust/system/tbb/detail/remove.h b/thrust/thrust/system/tbb/detail/remove.h
index 34cd917994a..f13bdc9ccf6 100644
--- a/thrust/thrust/system/tbb/detail/remove.h
+++ b/thrust/thrust/system/tbb/detail/remove.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/omp/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/tbb/detail/remove.inl b/thrust/thrust/system/tbb/detail/remove.inl
index 76d77e64ba5..f2a35f1e534 100644
--- a/thrust/thrust/system/tbb/detail/remove.inl
+++ b/thrust/thrust/system/tbb/detail/remove.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/tbb/detail/remove.h>
 #include <thrust/system/detail/generic/remove.h>
 
diff --git a/thrust/thrust/system/tbb/detail/replace.h b/thrust/thrust/system/tbb/detail/replace.h
index 95c5a14ba3d..12b1603a6a8 100644
--- a/thrust/thrust/system/tbb/detail/replace.h
+++ b/thrust/thrust/system/tbb/detail/replace.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits this algorithm
 #include <thrust/system/cpp/detail/scatter.h>
 
diff --git a/thrust/thrust/system/tbb/detail/reverse.h b/thrust/thrust/system/tbb/detail/reverse.h
index 1f3e0325e25..aedea33aa82 100644
--- a/thrust/thrust/system/tbb/detail/reverse.h
+++ b/thrust/thrust/system/tbb/detail/reverse.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits reverse
 #include <thrust/system/cpp/detail/reverse.h>
 
diff --git a/thrust/thrust/system/tbb/detail/scan.h b/thrust/thrust/system/tbb/detail/scan.h
index b31b46317cf..769eb449e5a 100644
--- a/thrust/thrust/system/tbb/detail/scan.h
+++ b/thrust/thrust/system/tbb/detail/scan.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/tbb/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/tbb/detail/scan.inl b/thrust/thrust/system/tbb/detail/scan.inl
index d6e89498390..86857f5bc81 100644
--- a/thrust/thrust/system/tbb/detail/scan.inl
+++ b/thrust/thrust/system/tbb/detail/scan.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/tbb/detail/scan.h>
 #include <thrust/distance.h>
 #include <thrust/advance.h>
@@ -53,16 +59,16 @@ struct inclusive_body
   inclusive_body(InputIterator input, OutputIterator output, BinaryFunction binary_op, ValueType dummy)
     : input(input), output(output), binary_op(binary_op), sum(dummy), first_call(true)
   {}
-    
+
   inclusive_body(inclusive_body& b, ::tbb::split)
     : input(b.input), output(b.output), binary_op(b.binary_op), sum(b.sum), first_call(true)
   {}
 
-  template<typename Size> 
+  template<typename Size>
   void operator()(const ::tbb::blocked_range<Size>& r, ::tbb::pre_scan_tag)
   {
     InputIterator iter = input + r.begin();
- 
+
     ValueType temp = *iter;
 
     ++iter;
@@ -74,11 +80,11 @@ struct inclusive_body
       sum = temp;
     else
       sum = binary_op(sum, temp);
-      
+
     first_call = false;
   }
-  
-  template<typename Size> 
+
+  template<typename Size>
   void operator()(const ::tbb::blocked_range<Size>& r, ::tbb::final_scan_tag)
   {
     InputIterator  iter1 = input  + r.begin();
@@ -109,12 +115,12 @@ struct inclusive_body
     {
       sum = binary_op(b.sum, sum);
     }
-  } 
+  }
 
   void assign(inclusive_body& b)
   {
     sum = b.sum;
-  } 
+  }
 };
 
 
@@ -133,16 +139,16 @@ struct exclusive_body
   exclusive_body(InputIterator input, OutputIterator output, BinaryFunction binary_op, ValueType init)
     : input(input), output(output), binary_op(binary_op), sum(init), first_call(true)
   {}
-    
+
   exclusive_body(exclusive_body& b, ::tbb::split)
     : input(b.input), output(b.output), binary_op(b.binary_op), sum(b.sum), first_call(true)
   {}
 
-  template<typename Size> 
+  template<typename Size>
   void operator()(const ::tbb::blocked_range<Size>& r, ::tbb::pre_scan_tag)
   {
     InputIterator iter = input + r.begin();
- 
+
     ValueType temp = *iter;
 
     ++iter;
@@ -154,11 +160,11 @@ struct exclusive_body
       sum = temp;
     else
       sum = binary_op(sum, temp);
-      
+
     first_call = false;
   }
-  
-  template<typename Size> 
+
+  template<typename Size>
   void operator()(const ::tbb::blocked_range<Size>& r, ::tbb::final_scan_tag)
   {
     InputIterator  iter1 = input  + r.begin();
@@ -170,7 +176,7 @@ struct exclusive_body
       *iter2 = sum;
       sum = temp;
     }
-    
+
     first_call = false;
   }
 
@@ -187,7 +193,7 @@ struct exclusive_body
   void assign(exclusive_body& b)
   {
     sum = b.sum;
-  } 
+  }
 };
 
 } // end scan_detail
@@ -250,7 +256,7 @@ template<typename InputIterator,
   thrust::advance(result, n);
 
   return result;
-} 
+}
 
 } // end namespace detail
 } // end namespace tbb
diff --git a/thrust/thrust/system/tbb/detail/scan_by_key.h b/thrust/thrust/system/tbb/detail/scan_by_key.h
index 2b5fa36483c..737df3473df 100644
--- a/thrust/thrust/system/tbb/detail/scan_by_key.h
+++ b/thrust/thrust/system/tbb/detail/scan_by_key.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits scan_by_key
 #include <thrust/system/cpp/detail/scan_by_key.h>
 
diff --git a/thrust/thrust/system/tbb/detail/scatter.h b/thrust/thrust/system/tbb/detail/scatter.h
index 95c5a14ba3d..12b1603a6a8 100644
--- a/thrust/thrust/system/tbb/detail/scatter.h
+++ b/thrust/thrust/system/tbb/detail/scatter.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits this algorithm
 #include <thrust/system/cpp/detail/scatter.h>
 
diff --git a/thrust/thrust/system/tbb/detail/sequence.h b/thrust/thrust/system/tbb/detail/sequence.h
index c33b2d4333c..8d6e3c34f65 100644
--- a/thrust/thrust/system/tbb/detail/sequence.h
+++ b/thrust/thrust/system/tbb/detail/sequence.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits sequence
 #include <thrust/system/cpp/detail/sequence.h>
 
diff --git a/thrust/thrust/system/tbb/detail/set_operations.h b/thrust/thrust/system/tbb/detail/set_operations.h
index 421fa8a4bd9..890da986216 100644
--- a/thrust/thrust/system/tbb/detail/set_operations.h
+++ b/thrust/thrust/system/tbb/detail/set_operations.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits set_operations
 #include <thrust/system/cpp/detail/set_operations.h>
 
diff --git a/thrust/thrust/system/tbb/detail/sort.h b/thrust/thrust/system/tbb/detail/sort.h
index 9c58bf6d49f..b89693a2da8 100644
--- a/thrust/thrust/system/tbb/detail/sort.h
+++ b/thrust/thrust/system/tbb/detail/sort.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/tbb/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/system/tbb/detail/sort.inl b/thrust/thrust/system/tbb/detail/sort.inl
index 103710fba54..5b72a6f6380 100644
--- a/thrust/thrust/system/tbb/detail/sort.inl
+++ b/thrust/thrust/system/tbb/detail/sort.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/temporary_array.h>
 #include <thrust/detail/copy.h>
 #include <thrust/iterator/iterator_traits.h>
diff --git a/thrust/thrust/system/tbb/detail/swap_ranges.h b/thrust/thrust/system/tbb/detail/swap_ranges.h
index 4a0b06cbe9e..a57487f9fa1 100644
--- a/thrust/thrust/system/tbb/detail/swap_ranges.h
+++ b/thrust/thrust/system/tbb/detail/swap_ranges.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // tbb inherits swap_ranges
 #include <thrust/system/cpp/detail/swap_ranges.h>
 
diff --git a/thrust/thrust/system/tbb/detail/tabulate.h b/thrust/thrust/system/tbb/detail/tabulate.h
index ea135c70706..930d9efd15f 100644
--- a/thrust/thrust/system/tbb/detail/tabulate.h
+++ b/thrust/thrust/system/tbb/detail/tabulate.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits tabulate
 #include <thrust/system/cpp/detail/tabulate.h>
 
diff --git a/thrust/thrust/system/tbb/detail/temporary_buffer.h b/thrust/thrust/system/tbb/detail/temporary_buffer.h
index 2adfaf2810c..d666f267e35 100644
--- a/thrust/thrust/system/tbb/detail/temporary_buffer.h
+++ b/thrust/thrust/system/tbb/detail/temporary_buffer.h
@@ -18,5 +18,11 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system has no special temporary buffer functions
 
diff --git a/thrust/thrust/system/tbb/detail/transform.h b/thrust/thrust/system/tbb/detail/transform.h
index 20d606dfbee..8a572e02bc3 100644
--- a/thrust/thrust/system/tbb/detail/transform.h
+++ b/thrust/thrust/system/tbb/detail/transform.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // omp inherits transform
 #include <thrust/system/cpp/detail/transform.h>
 
diff --git a/thrust/thrust/system/tbb/detail/transform_reduce.h b/thrust/thrust/system/tbb/detail/transform_reduce.h
index a8736bd75d0..5fb51bf8a07 100644
--- a/thrust/thrust/system/tbb/detail/transform_reduce.h
+++ b/thrust/thrust/system/tbb/detail/transform_reduce.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits transform_reduce
 #include <thrust/system/cpp/detail/transform_reduce.h>
 
diff --git a/thrust/thrust/system/tbb/detail/transform_scan.h b/thrust/thrust/system/tbb/detail/transform_scan.h
index 75b075b6b16..02371a3fc09 100644
--- a/thrust/thrust/system/tbb/detail/transform_scan.h
+++ b/thrust/thrust/system/tbb/detail/transform_scan.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits transform_scan
 #include <thrust/system/cpp/detail/transform_scan.h>
 
diff --git a/thrust/thrust/system/tbb/detail/uninitialized_copy.h b/thrust/thrust/system/tbb/detail/uninitialized_copy.h
index bda06ac13e9..5ba0d83cde8 100644
--- a/thrust/thrust/system/tbb/detail/uninitialized_copy.h
+++ b/thrust/thrust/system/tbb/detail/uninitialized_copy.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits uninitialized_copy
 #include <thrust/system/cpp/detail/uninitialized_copy.h>
 
diff --git a/thrust/thrust/system/tbb/detail/uninitialized_fill.h b/thrust/thrust/system/tbb/detail/uninitialized_fill.h
index 764de876233..f1b4a302c6d 100644
--- a/thrust/thrust/system/tbb/detail/uninitialized_fill.h
+++ b/thrust/thrust/system/tbb/detail/uninitialized_fill.h
@@ -18,6 +18,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // this system inherits uninitialized_fill
 #include <thrust/system/cpp/detail/uninitialized_fill.h>
 
diff --git a/thrust/thrust/system/tbb/detail/unique.h b/thrust/thrust/system/tbb/detail/unique.h
index 843e6406e3a..66ecebafe2d 100644
--- a/thrust/thrust/system/tbb/detail/unique.h
+++ b/thrust/thrust/system/tbb/detail/unique.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/tbb/detail/execution_policy.h>
 #include <thrust/pair.h>
 
@@ -60,7 +66,7 @@ template<typename DerivedPolicy,
 
 
 } // end namespace detail
-} // end namespace tbb 
+} // end namespace tbb
 } // end namespace system
 THRUST_NAMESPACE_END
 
diff --git a/thrust/thrust/system/tbb/detail/unique.inl b/thrust/thrust/system/tbb/detail/unique.inl
index 136af897c69..9345cc08e69 100644
--- a/thrust/thrust/system/tbb/detail/unique.inl
+++ b/thrust/thrust/system/tbb/detail/unique.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/tbb/detail/unique.h>
 #include <thrust/system/detail/generic/unique.h>
 #include <thrust/pair.h>
@@ -73,7 +79,7 @@ template<typename DerivedPolicy,
 
 
 } // end namespace detail
-} // end namespace tbb 
+} // end namespace tbb
 } // end namespace system
 THRUST_NAMESPACE_END
 
diff --git a/thrust/thrust/system/tbb/detail/unique_by_key.h b/thrust/thrust/system/tbb/detail/unique_by_key.h
index 513bb386e23..9adbc07f501 100644
--- a/thrust/thrust/system/tbb/detail/unique_by_key.h
+++ b/thrust/thrust/system/tbb/detail/unique_by_key.h
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/tbb/detail/execution_policy.h>
 #include <thrust/pair.h>
 
@@ -35,7 +41,7 @@ template<typename DerivedPolicy,
          typename BinaryPredicate>
   thrust::pair<ForwardIterator1,ForwardIterator2>
     unique_by_key(execution_policy<DerivedPolicy> &exec,
-                  ForwardIterator1 keys_first, 
+                  ForwardIterator1 keys_first,
                   ForwardIterator1 keys_last,
                   ForwardIterator2 values_first,
                   BinaryPredicate binary_pred);
@@ -49,7 +55,7 @@ template<typename DerivedPolicy,
          typename BinaryPredicate>
   thrust::pair<OutputIterator1,OutputIterator2>
     unique_by_key_copy(execution_policy<DerivedPolicy> &exec,
-                       InputIterator1 keys_first, 
+                       InputIterator1 keys_first,
                        InputIterator1 keys_last,
                        InputIterator2 values_first,
                        OutputIterator1 keys_output,
@@ -58,7 +64,7 @@ template<typename DerivedPolicy,
 
 
 } // end namespace detail
-} // end namespace tbb 
+} // end namespace tbb
 } // end namespace system
 THRUST_NAMESPACE_END
 
diff --git a/thrust/thrust/system/tbb/detail/unique_by_key.inl b/thrust/thrust/system/tbb/detail/unique_by_key.inl
index dbd5922b072..d4cc4d23c50 100644
--- a/thrust/thrust/system/tbb/detail/unique_by_key.inl
+++ b/thrust/thrust/system/tbb/detail/unique_by_key.inl
@@ -17,6 +17,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/tbb/detail/unique_by_key.h>
 #include <thrust/system/detail/generic/unique_by_key.h>
 #include <thrust/pair.h>
@@ -36,7 +42,7 @@ template<typename DerivedPolicy,
          typename BinaryPredicate>
   thrust::pair<ForwardIterator1,ForwardIterator2>
     unique_by_key(execution_policy<DerivedPolicy> &exec,
-                  ForwardIterator1 keys_first, 
+                  ForwardIterator1 keys_first,
                   ForwardIterator1 keys_last,
                   ForwardIterator2 values_first,
                   BinaryPredicate binary_pred)
@@ -54,7 +60,7 @@ template<typename DerivedPolicy,
          typename BinaryPredicate>
   thrust::pair<OutputIterator1,OutputIterator2>
     unique_by_key_copy(execution_policy<DerivedPolicy> &exec,
-                       InputIterator1 keys_first, 
+                       InputIterator1 keys_first,
                        InputIterator1 keys_last,
                        InputIterator2 values_first,
                        OutputIterator1 keys_output,
diff --git a/thrust/thrust/system/tbb/execution_policy.h b/thrust/thrust/system/tbb/execution_policy.h
index bfa6b7893b1..b305519f423 100644
--- a/thrust/thrust/system/tbb/execution_policy.h
+++ b/thrust/thrust/system/tbb/execution_policy.h
@@ -16,12 +16,18 @@
 
 #pragma once
 
+#include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 /*! \file thrust/system/tbb/execution_policy.h
  *  \brief Execution policies for Thrust's TBB system.
  */
 
-#include <thrust/detail/config.h>
-
 // get the execution policies definitions first
 #include <thrust/system/tbb/detail/execution_policy.h>
 
diff --git a/thrust/thrust/system/tbb/memory.h b/thrust/thrust/system/tbb/memory.h
index 3bd4422320e..49453b3755f 100644
--- a/thrust/thrust/system/tbb/memory.h
+++ b/thrust/thrust/system/tbb/memory.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/tbb/memory_resource.h>
 #include <thrust/memory.h>
 #include <thrust/detail/type_traits.h>
diff --git a/thrust/thrust/system/tbb/memory_resource.h b/thrust/thrust/system/tbb/memory_resource.h
index a698b924271..1ba3a020ccc 100644
--- a/thrust/thrust/system/tbb/memory_resource.h
+++ b/thrust/thrust/system/tbb/memory_resource.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/mr/new.h>
 #include <thrust/mr/fancy_pointer_resource.h>
 
diff --git a/thrust/thrust/system/tbb/pointer.h b/thrust/thrust/system/tbb/pointer.h
index 065e1a548f0..794429d4d80 100644
--- a/thrust/thrust/system/tbb/pointer.h
+++ b/thrust/thrust/system/tbb/pointer.h
@@ -21,6 +21,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <type_traits>
 #include <thrust/system/tbb/detail/execution_policy.h>
 #include <thrust/detail/pointer.h>
diff --git a/thrust/thrust/system/tbb/vector.h b/thrust/thrust/system/tbb/vector.h
index 8cbbabbd242..d6615e500fe 100644
--- a/thrust/thrust/system/tbb/vector.h
+++ b/thrust/thrust/system/tbb/vector.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/system/tbb/memory.h>
 #include <thrust/detail/vector_base.h>
 #include <vector>
diff --git a/thrust/thrust/system_error.h b/thrust/thrust/system_error.h
index 6bf240e5123..3d657b24833 100644
--- a/thrust/thrust/system_error.h
+++ b/thrust/thrust/system_error.h
@@ -22,6 +22,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 THRUST_NAMESPACE_BEGIN
 
 /*! \addtogroup system
diff --git a/thrust/thrust/tabulate.h b/thrust/thrust/tabulate.h
index 7cb7945508d..725ed981fcb 100644
--- a/thrust/thrust/tabulate.h
+++ b/thrust/thrust/tabulate.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
diff --git a/thrust/thrust/transform.h b/thrust/thrust/transform.h
index 2d064c13b3e..06886523a31 100644
--- a/thrust/thrust/transform.h
+++ b/thrust/thrust/transform.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
@@ -36,16 +42,16 @@ THRUST_NAMESPACE_BEGIN
 
 
 /*! This version of \p transform applies a unary function to each element
- *  of an input sequence and stores the result in the corresponding 
- *  position in an output sequence.  Specifically, for each iterator 
- *  <tt>i</tt> in the range [\p first, \p last) the operation 
+ *  of an input sequence and stores the result in the corresponding
+ *  position in an output sequence.  Specifically, for each iterator
+ *  <tt>i</tt> in the range [\p first, \p last) the operation
  *  <tt>op(*i)</tt> is performed and the result is assigned to <tt>*o</tt>,
  *  where <tt>o</tt> is the corresponding output iterator in the range
  *  [\p result, \p result + (\p last - \p first) ).  The input and
  *  output sequences may coincide, resulting in an in-place transformation.
  *
  *  The algorithm's execution is parallelized as determined by \p exec.
- *    
+ *
  *  \param exec The execution policy to use for parallelization.
  *  \param first The beginning of the input sequence.
  *  \param last The end of the input sequence.
@@ -70,9 +76,9 @@ THRUST_NAMESPACE_BEGIN
  *  #include <thrust/functional.h>
  *  #include <thrust/execution_policy.h>
  *  ...
- *  
+ *
  *  int data[10] = {-5, 0, 2, -3, 2, 4, 0, -1, 2, 8};
- * 
+ *
  *  thrust::negate<int> op;
  *
  *  thrust::transform(thrust::host, data, data + 10, data, op); // in-place transformation
@@ -92,16 +98,16 @@ __host__ __device__
                            OutputIterator result,
                            UnaryFunction op);
 
-	
+
 /*! This version of \p transform applies a unary function to each element
- *  of an input sequence and stores the result in the corresponding 
- *  position in an output sequence.  Specifically, for each iterator 
- *  <tt>i</tt> in the range [\p first, \p last) the operation 
+ *  of an input sequence and stores the result in the corresponding
+ *  position in an output sequence.  Specifically, for each iterator
+ *  <tt>i</tt> in the range [\p first, \p last) the operation
  *  <tt>op(*i)</tt> is performed and the result is assigned to <tt>*o</tt>,
  *  where <tt>o</tt> is the corresponding output iterator in the range
  *  [\p result, \p result + (\p last - \p first) ).  The input and
  *  output sequences may coincide, resulting in an in-place transformation.
- *    
+ *
  *  \param first The beginning of the input sequence.
  *  \param last The end of the input sequence.
  *  \param result The beginning of the output sequence.
@@ -121,9 +127,9 @@ __host__ __device__
  *  \code
  *  #include <thrust/transform.h>
  *  #include <thrust/functional.h>
- *  
+ *
  *  int data[10] = {-5, 0, 2, -3, 2, 4, 0, -1, 2, 8};
- * 
+ *
  *  thrust::negate<int> op;
  *
  *  thrust::transform(data, data + 10, data, op); // in-place transformation
@@ -144,16 +150,16 @@ template<typename InputIterator,
 /*! This version of \p transform applies a binary function to each pair
  *  of elements from two input sequences and stores the result in the
  *  corresponding position in an output sequence.  Specifically, for
- *  each iterator <tt>i</tt> in the range [\p first1, \p last1) and 
+ *  each iterator <tt>i</tt> in the range [\p first1, \p last1) and
  *  <tt>j = first + (i - first1)</tt> in the range [\p first2, \p last2)
- *  the operation <tt>op(*i,*j)</tt> is performed and the result is 
+ *  the operation <tt>op(*i,*j)</tt> is performed and the result is
  *  assigned to <tt>*o</tt>,  where <tt>o</tt> is the corresponding
  *  output iterator in the range [\p result, \p result + (\p last - \p first) ).
- *  The input and output sequences may coincide, resulting in an 
+ *  The input and output sequences may coincide, resulting in an
  *  in-place transformation.
  *
  *  The algorithm's execution is parallelized as determined by \p exec.
- *    
+ *
  *  \param exec The execution policy to use for parallelization.
  *  \param first1 The beginning of the first input sequence.
  *  \param last1 The end of the first input sequence.
@@ -182,11 +188,11 @@ template<typename InputIterator,
  *  #include <thrust/functional.h>
  *  #include <thrust/execution_policy.h>
  *  ...
- *  
+ *
  *  int input1[6] = {-5,  0,  2,  3,  2,  4};
  *  int input2[6] = { 3,  6, -2,  1,  2,  3};
  *  int output[6];
- * 
+ *
  *  thrust::plus<int> op;
  *
  *  thrust::transform(thrust::host, input1, input1 + 6, input2, output, op);
@@ -212,14 +218,14 @@ __host__ __device__
 /*! This version of \p transform applies a binary function to each pair
  *  of elements from two input sequences and stores the result in the
  *  corresponding position in an output sequence.  Specifically, for
- *  each iterator <tt>i</tt> in the range [\p first1, \p last1) and 
+ *  each iterator <tt>i</tt> in the range [\p first1, \p last1) and
  *  <tt>j = first + (i - first1)</tt> in the range [\p first2, \p last2)
- *  the operation <tt>op(*i,*j)</tt> is performed and the result is 
+ *  the operation <tt>op(*i,*j)</tt> is performed and the result is
  *  assigned to <tt>*o</tt>,  where <tt>o</tt> is the corresponding
  *  output iterator in the range [\p result, \p result + (\p last - \p first) ).
- *  The input and output sequences may coincide, resulting in an 
+ *  The input and output sequences may coincide, resulting in an
  *  in-place transformation.
- *    
+ *
  *  \param first1 The beginning of the first input sequence.
  *  \param last1 The end of the first input sequence.
  *  \param first2 The beginning of the second input sequence.
@@ -243,11 +249,11 @@ __host__ __device__
  *  \code
  *  #include <thrust/transform.h>
  *  #include <thrust/functional.h>
- *  
+ *
  *  int input1[6] = {-5,  0,  2,  3,  2,  4};
  *  int input2[6] = { 3,  6, -2,  1,  2,  3};
  *  int output[6];
- * 
+ *
  *  thrust::plus<int> op;
  *
  *  thrust::transform(input1, input1 + 6, input2, output, op);
@@ -268,7 +274,7 @@ template<typename InputIterator1,
 
 
 /*! This version of \p transform_if conditionally applies a unary function
- *  to each element of an input sequence and stores the result in the corresponding 
+ *  to each element of an input sequence and stores the result in the corresponding
  *  position in an output sequence if the corresponding position in the input sequence
  *  satifies a predicate. Otherwise, the corresponding position in the
  *  output sequence is not modified.
@@ -282,7 +288,7 @@ template<typename InputIterator1,
  *  resulting in an in-place transformation.
  *
  *  The algorithm's execution is parallelized as determined by \p exec.
- *    
+ *
  *  \param exec The execution policy to use for parallelization.
  *  \param first The beginning of the input sequence.
  *  \param last The end of the input sequence.
@@ -310,7 +316,7 @@ template<typename InputIterator1,
  *  #include <thrust/functional.h>
  *  #include <thrust/execution_policy.h>
  *  ...
- *  
+ *
  *  int data[10]    = {-5, 0, 2, -3, 2, 4, 0, -1, 2, 8};
  *
  *  struct is_odd
@@ -321,7 +327,7 @@ template<typename InputIterator1,
  *      return x % 2;
  *    }
  *  };
- * 
+ *
  *  thrust::negate<int> op;
  *  thrust::identity<int> identity;
  *
@@ -347,7 +353,7 @@ __host__ __device__
 
 
 /*! This version of \p transform_if conditionally applies a unary function
- *  to each element of an input sequence and stores the result in the corresponding 
+ *  to each element of an input sequence and stores the result in the corresponding
  *  position in an output sequence if the corresponding position in the input sequence
  *  satifies a predicate. Otherwise, the corresponding position in the
  *  output sequence is not modified.
@@ -359,7 +365,7 @@ __host__ __device__
  *  <tt>[result, result + (last - first) )</tt>. Otherwise, <tt>op(*i)</tt> is
  *  not evaluated and no assignment occurs. The input and output sequences may coincide,
  *  resulting in an in-place transformation.
- *    
+ *
  *  \param first The beginning of the input sequence.
  *  \param last The end of the input sequence.
  *  \param result The beginning of the output sequence.
@@ -382,7 +388,7 @@ __host__ __device__
  *  \code
  *  #include <thrust/transform.h>
  *  #include <thrust/functional.h>
- *  
+ *
  *  int data[10]    = {-5, 0, 2, -3, 2, 4, 0, -1, 2, 8};
  *
  *  struct is_odd
@@ -393,7 +399,7 @@ __host__ __device__
  *      return x % 2;
  *    }
  *  };
- * 
+ *
  *  thrust::negate<int> op;
  *  thrust::identity<int> identity;
  *
@@ -416,7 +422,7 @@ template<typename InputIterator,
 
 
 /*! This version of \p transform_if conditionally applies a unary function
- *  to each element of an input sequence and stores the result in the corresponding 
+ *  to each element of an input sequence and stores the result in the corresponding
  *  position in an output sequence if the corresponding position in a stencil sequence
  *  satisfies a predicate. Otherwise, the corresponding position in the
  *  output sequence is not modified.
@@ -431,7 +437,7 @@ template<typename InputIterator,
  *  resulting in an in-place transformation.
  *
  *  The algorithm's execution is parallelized as determined by \p exec.
- *    
+ *
  *  \param exec The execution policy to use for parallelization.
  *  \param first The beginning of the input sequence.
  *  \param last The end of the input sequence.
@@ -462,10 +468,10 @@ template<typename InputIterator,
  *  #include <thrust/functional.h>
  *  #include <thrust/execution_policy.h>
  *  ...
- *  
+ *
  *  int data[10]    = {-5, 0, 2, -3, 2, 4, 0, -1, 2, 8};
  *  int stencil[10] = { 1, 0, 1,  0, 1, 0, 1,  0, 1, 0};
- * 
+ *
  *  thrust::negate<int> op;
  *  thrust::identity<int> identity;
  *
@@ -492,7 +498,7 @@ __host__ __device__
 
 
 /*! This version of \p transform_if conditionally applies a unary function
- *  to each element of an input sequence and stores the result in the corresponding 
+ *  to each element of an input sequence and stores the result in the corresponding
  *  position in an output sequence if the corresponding position in a stencil sequence
  *  satisfies a predicate. Otherwise, the corresponding position in the
  *  output sequence is not modified.
@@ -505,7 +511,7 @@ __host__ __device__
  *  <tt>[result, result + (last - first) )</tt>. Otherwise, <tt>op(*i)</tt> is
  *  not evaluated and no assignment occurs. The input and output sequences may coincide,
  *  resulting in an in-place transformation.
- *    
+ *
  *  \param first The beginning of the input sequence.
  *  \param last The end of the input sequence.
  *  \param stencil The beginning of the stencil sequence.
@@ -531,10 +537,10 @@ __host__ __device__
  *  \code
  *  #include <thrust/transform.h>
  *  #include <thrust/functional.h>
- *  
+ *
  *  int data[10]    = {-5, 0, 2, -3, 2, 4, 0, -1, 2, 8};
  *  int stencil[10] = { 1, 0, 1,  0, 1, 0, 1,  0, 1, 0};
- * 
+ *
  *  thrust::negate<int> op;
  *  thrust::identity<int> identity;
  *
@@ -558,12 +564,12 @@ template<typename InputIterator1,
 
 
 /*! This version of \p transform_if conditionally applies a binary function
- *  to each pair of elements from two input sequences and stores the result in the corresponding 
+ *  to each pair of elements from two input sequences and stores the result in the corresponding
  *  position in an output sequence if the corresponding position in a stencil sequence
  *  satifies a predicate. Otherwise, the corresponding position in the
  *  output sequence is not modified.
  *
- *  Specifically, for each iterator <tt>i</tt> in the range <tt>[first1, last1)</tt> and 
+ *  Specifically, for each iterator <tt>i</tt> in the range <tt>[first1, last1)</tt> and
  *  <tt>j = first2 + (i - first1)</tt> in the range <tt>[first2, first2 + (last1 - first1) )</tt>,
  *  the predicate <tt>pred(*s)</tt> is evaluated, where <tt>s</tt> is the corresponding input
  *  iterator in the range <tt>[stencil, stencil + (last1 - first1) )</tt>. If this predicate
@@ -574,7 +580,7 @@ template<typename InputIterator1,
  *  resulting in an in-place transformation.
  *
  *  The algorithm's execution is parallelized as determined by \p exec.
- *    
+ *
  *  \param exec The execution policy to use for parallelization.
  *  \param first1 The beginning of the first input sequence.
  *  \param last1 The end of the first input sequence.
@@ -607,12 +613,12 @@ template<typename InputIterator1,
  *  #include <thrust/functional.h>
  *  #include <thrust/execution_policy.h>
  *  ...
- *  
+ *
  *  int input1[6]  = {-5,  0,  2,  3,  2,  4};
  *  int input2[6]  = { 3,  6, -2,  1,  2,  3};
  *  int stencil[8] = { 1,  0,  1,  0,  1,  0};
  *  int output[6];
- * 
+ *
  *  thrust::plus<int> op;
  *  thrust::identity<int> identity;
  *
@@ -641,12 +647,12 @@ __host__ __device__
 
 
 /*! This version of \p transform_if conditionally applies a binary function
- *  to each pair of elements from two input sequences and stores the result in the corresponding 
+ *  to each pair of elements from two input sequences and stores the result in the corresponding
  *  position in an output sequence if the corresponding position in a stencil sequence
  *  satifies a predicate. Otherwise, the corresponding position in the
  *  output sequence is not modified.
  *
- *  Specifically, for each iterator <tt>i</tt> in the range <tt>[first1, last1)</tt> and 
+ *  Specifically, for each iterator <tt>i</tt> in the range <tt>[first1, last1)</tt> and
  *  <tt>j = first2 + (i - first1)</tt> in the range <tt>[first2, first2 + (last1 - first1) )</tt>,
  *  the predicate <tt>pred(*s)</tt> is evaluated, where <tt>s</tt> is the corresponding input
  *  iterator in the range <tt>[stencil, stencil + (last1 - first1) )</tt>. If this predicate
@@ -655,7 +661,7 @@ __host__ __device__
  *  <tt>[result, result + (last1 - first1) )</tt>. Otherwise, <tt>binary_op(*i,*j)</tt> is
  *  not evaluated and no assignment occurs. The input and output sequences may coincide,
  *  resulting in an in-place transformation.
- *    
+ *
  *  \param first1 The beginning of the first input sequence.
  *  \param last1 The end of the first input sequence.
  *  \param first2 The beginning of the second input sequence.
@@ -683,12 +689,12 @@ __host__ __device__
  *  \code
  *  #include <thrust/transform.h>
  *  #include <thrust/functional.h>
- *  
+ *
  *  int input1[6]  = {-5,  0,  2,  3,  2,  4};
  *  int input2[6]  = { 3,  6, -2,  1,  2,  3};
  *  int stencil[8] = { 1,  0,  1,  0,  1,  0};
  *  int output[6];
- * 
+ *
  *  thrust::plus<int> op;
  *  thrust::identity<int> identity;
  *
diff --git a/thrust/thrust/transform_reduce.h b/thrust/thrust/transform_reduce.h
index 11d6b84c3eb..17ae56af125 100644
--- a/thrust/thrust/transform_reduce.h
+++ b/thrust/thrust/transform_reduce.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
@@ -43,10 +49,10 @@ THRUST_NAMESPACE_BEGIN
  *  \p transform_reduce performs a reduction on the transformation of the
  *  sequence <tt>[first, last)</tt> according to \p unary_op. Specifically,
  *  \p unary_op is applied to each element of the sequence and then the result
- *  is reduced to a single value with \p binary_op using the initial value 
- *  \p init.  Note that the transformation \p unary_op is not applied to 
- *  the initial value \p init.  The order of reduction is not specified, 
- *  so \p binary_op must be both commutative and associative. 
+ *  is reduced to a single value with \p binary_op using the initial value
+ *  \p init.  Note that the transformation \p unary_op is not applied to
+ *  the initial value \p init.  The order of reduction is not specified,
+ *  so \p binary_op must be both commutative and associative.
  *
  *  The algorithm's execution is parallelized as determined by \p exec.
  *
@@ -101,8 +107,8 @@ THRUST_NAMESPACE_BEGIN
  *  \see \c reduce
  */
 template<typename DerivedPolicy,
-         typename InputIterator, 
-         typename UnaryFunction, 
+         typename InputIterator,
+         typename UnaryFunction,
          typename OutputType,
          typename BinaryFunction>
 __host__ __device__
@@ -123,10 +129,10 @@ __host__ __device__
  *  \p transform_reduce performs a reduction on the transformation of the
  *  sequence <tt>[first, last)</tt> according to \p unary_op. Specifically,
  *  \p unary_op is applied to each element of the sequence and then the result
- *  is reduced to a single value with \p binary_op using the initial value 
- *  \p init.  Note that the transformation \p unary_op is not applied to 
- *  the initial value \p init.  The order of reduction is not specified, 
- *  so \p binary_op must be both commutative and associative. 
+ *  is reduced to a single value with \p binary_op using the initial value
+ *  \p init.  Note that the transformation \p unary_op is not applied to
+ *  the initial value \p init.  The order of reduction is not specified,
+ *  so \p binary_op must be both commutative and associative.
  *
  *  \param first The beginning of the sequence.
  *  \param last The end of the sequence.
@@ -174,8 +180,8 @@ __host__ __device__
  *  \see \c transform
  *  \see \c reduce
  */
-template<typename InputIterator, 
-         typename UnaryFunction, 
+template<typename InputIterator,
+         typename UnaryFunction,
          typename OutputType,
          typename BinaryFunction>
   OutputType transform_reduce(InputIterator first,
diff --git a/thrust/thrust/transform_scan.h b/thrust/thrust/transform_scan.h
index 6c0fe811668..1831c7d84be 100644
--- a/thrust/thrust/transform_scan.h
+++ b/thrust/thrust/transform_scan.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
@@ -33,7 +39,7 @@ THRUST_NAMESPACE_BEGIN
  *  \ingroup algorithms
  *  \{
  */
-	
+
 /*! \addtogroup transformed_prefixsums Transformed Prefix Sums
  *  \ingroup prefixsums
  *  \{
@@ -81,7 +87,7 @@ THRUST_NAMESPACE_BEGIN
  *  #include <thrust/transform_scan.h>
  *  #include <thrust/execution_policy.h>
  *  ...
- *  
+ *
  *  int data[6] = {1, 0, 2, 2, 1, 3};
  *
  *  thrust::negate<int> unary_op;
@@ -144,7 +150,7 @@ __host__ __device__
  *
  *  \code
  *  #include <thrust/transform_scan.h>
- *  
+ *
  *  int data[6] = {1, 0, 2, 2, 1, 3};
  *
  *  thrust::negate<int> unary_op;
@@ -175,10 +181,10 @@ template<typename InputIterator,
  *  tranformation defined by \p unary_op into a temporary sequence and then
  *  performing an \p exclusive_scan on the tranformed sequence.  In most
  *  cases, fusing these two operations together is more efficient, since
- *  fewer memory reads and writes are required. In 
- *  \p transform_exclusive_scan, \p init is assigned to <tt>\*result</tt> 
+ *  fewer memory reads and writes are required. In
+ *  \p transform_exclusive_scan, \p init is assigned to <tt>\*result</tt>
  *  and the result of <tt>binary_op(init, unary_op(\*first))</tt> is assigned
- *  to <tt>\*(result + 1)</tt>, and so on.  The transform scan operation is 
+ *  to <tt>\*(result + 1)</tt>, and so on.  The transform scan operation is
  *  permitted to be in-place.
  *
  *  The algorithm's execution is parallelized as determined by \p exec.
@@ -213,7 +219,7 @@ template<typename InputIterator,
  *  #include <thrust/transform_scan.h>
  *  #include <thrust/execution_policy.h>
  *  ...
- *  
+ *
  *  int data[6] = {1, 0, 2, 2, 1, 3};
  *
  *  thrust::negate<int> unary_op;
@@ -249,10 +255,10 @@ __host__ __device__
  *  tranformation defined by \p unary_op into a temporary sequence and then
  *  performing an \p exclusive_scan on the tranformed sequence.  In most
  *  cases, fusing these two operations together is more efficient, since
- *  fewer memory reads and writes are required. In 
- *  \p transform_exclusive_scan, \p init is assigned to <tt>\*result</tt> 
+ *  fewer memory reads and writes are required. In
+ *  \p transform_exclusive_scan, \p init is assigned to <tt>\*result</tt>
  *  and the result of <tt>binary_op(init, unary_op(\*first))</tt> is assigned
- *  to <tt>\*(result + 1)</tt>, and so on.  The transform scan operation is 
+ *  to <tt>\*(result + 1)</tt>, and so on.  The transform scan operation is
  *  permitted to be in-place.
  *
  *  \param first The beginning of the input sequence.
@@ -280,7 +286,7 @@ __host__ __device__
  *
  *  \code
  *  #include <thrust/transform_scan.h>
- *  
+ *
  *  int data[6] = {1, 0, 2, 2, 1, 3};
  *
  *  thrust::negate<int> unary_op;
diff --git a/thrust/thrust/tuple.h b/thrust/thrust/tuple.h
index 20b9fe5a2b6..75922592b6c 100644
--- a/thrust/thrust/tuple.h
+++ b/thrust/thrust/tuple.h
@@ -31,11 +31,17 @@
 
 #include <thrust/detail/config.h>
 
-#include <tuple>
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 
 #include <cuda/std/tuple>
 #include <cuda/std/utility>
 
+#include <tuple>
+
 THRUST_NAMESPACE_BEGIN
 
 
diff --git a/thrust/thrust/type_traits/integer_sequence.h b/thrust/thrust/type_traits/integer_sequence.h
index 26ea542134a..427a5cf8e14 100644
--- a/thrust/thrust/type_traits/integer_sequence.h
+++ b/thrust/thrust/type_traits/integer_sequence.h
@@ -23,6 +23,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp11_required.h>
 
 #if THRUST_CPP_DIALECT >= 2011
diff --git a/thrust/thrust/type_traits/is_contiguous_iterator.h b/thrust/thrust/type_traits/is_contiguous_iterator.h
index eaa088978c4..3eb2575c244 100644
--- a/thrust/thrust/type_traits/is_contiguous_iterator.h
+++ b/thrust/thrust/type_traits/is_contiguous_iterator.h
@@ -23,6 +23,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/raw_pointer_cast.h>
 #include <thrust/detail/type_traits.h>
 #include <thrust/detail/type_traits/pointer_traits.h>
diff --git a/thrust/thrust/type_traits/is_execution_policy.h b/thrust/thrust/type_traits/is_execution_policy.h
index f83751ea2da..18c7269d13f 100644
--- a/thrust/thrust/type_traits/is_execution_policy.h
+++ b/thrust/thrust/type_traits/is_execution_policy.h
@@ -22,6 +22,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/detail/type_traits.h>
 #include <thrust/detail/execution_policy.h>
 
diff --git a/thrust/thrust/type_traits/is_operator_less_or_greater_function_object.h b/thrust/thrust/type_traits/is_operator_less_or_greater_function_object.h
index ef5a19f691e..0f9c21ee1c9 100644
--- a/thrust/thrust/type_traits/is_operator_less_or_greater_function_object.h
+++ b/thrust/thrust/type_traits/is_operator_less_or_greater_function_object.h
@@ -23,6 +23,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/functional.h>
 #include <thrust/detail/type_traits.h>
 #include <thrust/detail/type_traits/pointer_traits.h>
diff --git a/thrust/thrust/type_traits/is_operator_plus_function_object.h b/thrust/thrust/type_traits/is_operator_plus_function_object.h
index 800847532cf..f7372c1d7d6 100644
--- a/thrust/thrust/type_traits/is_operator_plus_function_object.h
+++ b/thrust/thrust/type_traits/is_operator_plus_function_object.h
@@ -23,6 +23,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/functional.h>
 #include <thrust/detail/type_traits.h>
 #include <thrust/detail/type_traits/pointer_traits.h>
diff --git a/thrust/thrust/type_traits/is_trivially_relocatable.h b/thrust/thrust/type_traits/is_trivially_relocatable.h
index 21d1f09d8d7..4ab9f04fa77 100644
--- a/thrust/thrust/type_traits/is_trivially_relocatable.h
+++ b/thrust/thrust/type_traits/is_trivially_relocatable.h
@@ -24,6 +24,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/static_assert.h>
 #include <thrust/detail/type_traits.h>
 #include <thrust/type_traits/is_contiguous_iterator.h>
diff --git a/thrust/thrust/type_traits/logical_metafunctions.h b/thrust/thrust/type_traits/logical_metafunctions.h
index 914b477e8ef..5cddc88722a 100644
--- a/thrust/thrust/type_traits/logical_metafunctions.h
+++ b/thrust/thrust/type_traits/logical_metafunctions.h
@@ -25,6 +25,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp11_required.h>
 
 #if THRUST_CPP_DIALECT >= 2011
diff --git a/thrust/thrust/type_traits/remove_cvref.h b/thrust/thrust/type_traits/remove_cvref.h
index 1da2e0de35c..5189d7ebd1c 100644
--- a/thrust/thrust/type_traits/remove_cvref.h
+++ b/thrust/thrust/type_traits/remove_cvref.h
@@ -23,6 +23,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if  THRUST_CPP_DIALECT >= 2017
 #if __has_include(<version>)
 #  include <version>
diff --git a/thrust/thrust/type_traits/void_t.h b/thrust/thrust/type_traits/void_t.h
index ed12d861dcb..d10d96f2643 100644
--- a/thrust/thrust/type_traits/void_t.h
+++ b/thrust/thrust/type_traits/void_t.h
@@ -22,6 +22,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #if THRUST_CPP_DIALECT >= 2017
 #  include <type_traits>
 #endif
diff --git a/thrust/thrust/uninitialized_copy.h b/thrust/thrust/uninitialized_copy.h
index 94c2763e3e4..459af1e1251 100644
--- a/thrust/thrust/uninitialized_copy.h
+++ b/thrust/thrust/uninitialized_copy.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
@@ -66,13 +72,13 @@ THRUST_NAMESPACE_BEGIN
  *  #include <thrust/device_malloc.h>
  *  #include <thrust/device_vector.h>
  *  #include <thrust/execution_policy.h>
- *  
+ *
  *  struct Int
  *  {
  *    __host__ __device__
  *    Int(int x) : val(x) {}
  *    int val;
- *  };  
+ *  };
  *  ...
  *  const int N = 137;
  *
@@ -128,13 +134,13 @@ __host__ __device__
  *  #include <thrust/uninitialized_copy.h>
  *  #include <thrust/device_malloc.h>
  *  #include <thrust/device_vector.h>
- *  
+ *
  *  struct Int
  *  {
  *    __host__ __device__
  *    Int(int x) : val(x) {}
  *    int val;
- *  };  
+ *  };
  *  ...
  *  const int N = 137;
  *
@@ -195,13 +201,13 @@ template<typename InputIterator, typename ForwardIterator>
  *  #include <thrust/device_malloc.h>
  *  #include <thrust/device_vector.h>
  *  #include <thrust/execution_policy.h>
- *  
+ *
  *  struct Int
  *  {
  *    __host__ __device__
  *    Int(int x) : val(x) {}
  *    int val;
- *  };  
+ *  };
  *  ...
  *  const int N = 137;
  *
@@ -259,13 +265,13 @@ __host__ __device__
  *  #include <thrust/uninitialized_copy.h>
  *  #include <thrust/device_malloc.h>
  *  #include <thrust/device_vector.h>
- *  
+ *
  *  struct Int
  *  {
  *    __host__ __device__
  *    Int(int x) : val(x) {}
  *    int val;
- *  };  
+ *  };
  *  ...
  *  const int N = 137;
  *
diff --git a/thrust/thrust/uninitialized_fill.h b/thrust/thrust/uninitialized_fill.h
index b46758a3c8c..a46c78bd405 100644
--- a/thrust/thrust/uninitialized_fill.h
+++ b/thrust/thrust/uninitialized_fill.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 
 THRUST_NAMESPACE_BEGIN
@@ -42,7 +48,7 @@ THRUST_NAMESPACE_BEGIN
  *  calling \p ForwardIterator's \c value_type's copy constructor.
  *
  *  The algorithm's execution is parallelized as determined by \p exec.
- *  
+ *
  *  \param exec The execution policy to use for parallelization.
  *  \param first The first element of the range of interest.
  *  \param last The last element of the range of interest.
@@ -60,13 +66,13 @@ THRUST_NAMESPACE_BEGIN
  *  #include <thrust/uninitialized_fill.h>
  *  #include <thrust/device_malloc.h>
  *  #include <thrust/execution_policy.h>
- *  
+ *
  *  struct Int
  *  {
  *    __host__ __device__
  *    Int(int x) : val(x) {}
  *    int val;
- *  };  
+ *  };
  *  ...
  *  const int N = 137;
  *
@@ -101,7 +107,7 @@ __host__ __device__
  *  in that range. That is, for each iterator \c i in the range <tt>[first, last)</tt>,
  *  \p uninitialized_fill creates a copy of \c x in the location pointed to \c i by
  *  calling \p ForwardIterator's \c value_type's copy constructor.
- *  
+ *
  *  \param first The first element of the range of interest.
  *  \param last The last element of the range of interest.
  *  \param x The value to use as the exemplar of the copy constructor.
@@ -116,13 +122,13 @@ __host__ __device__
  *  \code
  *  #include <thrust/uninitialized_fill.h>
  *  #include <thrust/device_malloc.h>
- *  
+ *
  *  struct Int
  *  {
  *    __host__ __device__
  *    Int(int x) : val(x) {}
  *    int val;
- *  };  
+ *  };
  *  ...
  *  const int N = 137;
  *
@@ -157,7 +163,7 @@ template<typename ForwardIterator, typename T>
  *  calling \p ForwardIterator's \c value_type's copy constructor.
  *
  *  The algorithm's execution is parallelized as determined by \p exec.
- *  
+ *
  *  \param exec The execution policy to use for parallelization.
  *  \param first The first element of the range of interest.
  *  \param n The size of the range of interest.
@@ -176,13 +182,13 @@ template<typename ForwardIterator, typename T>
  *  #include <thrust/uninitialized_fill.h>
  *  #include <thrust/device_malloc.h>
  *  #include <thrust/execution_policy.h>
- *  
+ *
  *  struct Int
  *  {
  *    __host__ __device__
  *    Int(int x) : val(x) {}
  *    int val;
- *  };  
+ *  };
  *  ...
  *  const int N = 137;
  *
@@ -217,7 +223,7 @@ __host__ __device__
  *  in that range. That is, for each iterator \c i in the range <tt>[first, first+n)</tt>,
  *  \p uninitialized_fill creates a copy of \c x in the location pointed to \c i by
  *  calling \p ForwardIterator's \c value_type's copy constructor.
- *  
+ *
  *  \param first The first element of the range of interest.
  *  \param n The size of the range of interest.
  *  \param x The value to use as the exemplar of the copy constructor.
@@ -233,13 +239,13 @@ __host__ __device__
  *  \code
  *  #include <thrust/uninitialized_fill.h>
  *  #include <thrust/device_malloc.h>
- *  
+ *
  *  struct Int
  *  {
  *    __host__ __device__
  *    Int(int x) : val(x) {}
  *    int val;
- *  };  
+ *  };
  *  ...
  *  const int N = 137;
  *
diff --git a/thrust/thrust/unique.h b/thrust/thrust/unique.h
index 234cd493569..a3a6a7fdbe2 100644
--- a/thrust/thrust/unique.h
+++ b/thrust/thrust/unique.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/execution_policy.h>
 #include <thrust/iterator/iterator_traits.h>
 #include <thrust/pair.h>
@@ -34,8 +40,8 @@ THRUST_NAMESPACE_BEGIN
 
 
 /*! For each group of consecutive elements in the range <tt>[first, last)</tt>
- *  with the same value, \p unique removes all but the first element of 
- *  the group. The return value is an iterator \c new_last such that 
+ *  with the same value, \p unique removes all but the first element of
+ *  the group. The return value is an iterator \c new_last such that
  *  no two consecutive elements in the range <tt>[first, new_last)</tt> are
  *  equal. The iterators in the range <tt>[new_last, last)</tt> are all still
  *  dereferenceable, but the elements that they point to are unspecified.
@@ -83,8 +89,8 @@ ForwardIterator unique(const thrust::detail::execution_policy_base<DerivedPolicy
 
 
 /*! For each group of consecutive elements in the range <tt>[first, last)</tt>
- *  with the same value, \p unique removes all but the first element of 
- *  the group. The return value is an iterator \c new_last such that 
+ *  with the same value, \p unique removes all but the first element of
+ *  the group. The return value is an iterator \c new_last such that
  *  no two consecutive elements in the range <tt>[first, new_last)</tt> are
  *  equal. The iterators in the range <tt>[new_last, last)</tt> are all still
  *  dereferenceable, but the elements that they point to are unspecified.
@@ -123,8 +129,8 @@ ForwardIterator unique(ForwardIterator first,
 
 
 /*! For each group of consecutive elements in the range <tt>[first, last)</tt>
- *  with the same value, \p unique removes all but the first element of 
- *  the group. The return value is an iterator \c new_last such that 
+ *  with the same value, \p unique removes all but the first element of
+ *  the group. The return value is an iterator \c new_last such that
  *  no two consecutive elements in the range <tt>[first, new_last)</tt> are
  *  equal. The iterators in the range <tt>[new_last, last)</tt> are all still
  *  dereferenceable, but the elements that they point to are unspecified.
@@ -177,8 +183,8 @@ ForwardIterator unique(const thrust::detail::execution_policy_base<DerivedPolicy
 
 
 /*! For each group of consecutive elements in the range <tt>[first, last)</tt>
- *  with the same value, \p unique removes all but the first element of 
- *  the group. The return value is an iterator \c new_last such that 
+ *  with the same value, \p unique removes all but the first element of
+ *  the group. The return value is an iterator \c new_last such that
  *  no two consecutive elements in the range <tt>[first, new_last)</tt> are
  *  equal. The iterators in the range <tt>[new_last, last)</tt> are all still
  *  dereferenceable, but the elements that they point to are unspecified.
@@ -224,14 +230,14 @@ ForwardIterator unique(ForwardIterator first,
 /*! \p unique_copy copies elements from the range <tt>[first, last)</tt>
  * to a range beginning with \p result, except that in a consecutive group
  * of duplicate elements only the first one is copied. The return value
- * is the end of the range to which the elements are copied. 
+ * is the end of the range to which the elements are copied.
  *
  * The reason there are two different versions of unique_copy is that there
  * are two different definitions of what it means for a consecutive group of
  * elements to be duplicates. In the first version, the test is simple
  * equality: the elements in a range <tt>[f, l)</tt> are duplicates if,
- * for every iterator \p i in the range, either <tt>i == f</tt> or else 
- * <tt>*i == *(i-1)</tt>. In the second, the test is an arbitrary 
+ * for every iterator \p i in the range, either <tt>i == f</tt> or else
+ * <tt>*i == *(i-1)</tt>. In the second, the test is an arbitrary
  * \p BinaryPredicate \p binary_pred: the elements in <tt>[f, l)</tt> are
  * duplicates if, for every iterator \p i in the range, either <tt>i == f</tt>
  * or else <tt>binary_pred(*i, *(i-1))</tt> is \p true.
@@ -286,14 +292,14 @@ OutputIterator unique_copy(const thrust::detail::execution_policy_base<DerivedPo
 /*! \p unique_copy copies elements from the range <tt>[first, last)</tt>
  * to a range beginning with \p result, except that in a consecutive group
  * of duplicate elements only the first one is copied. The return value
- * is the end of the range to which the elements are copied. 
+ * is the end of the range to which the elements are copied.
  *
  * The reason there are two different versions of unique_copy is that there
  * are two different definitions of what it means for a consecutive group of
  * elements to be duplicates. In the first version, the test is simple
  * equality: the elements in a range <tt>[f, l)</tt> are duplicates if,
- * for every iterator \p i in the range, either <tt>i == f</tt> or else 
- * <tt>*i == *(i-1)</tt>. In the second, the test is an arbitrary 
+ * for every iterator \p i in the range, either <tt>i == f</tt> or else
+ * <tt>*i == *(i-1)</tt>. In the second, the test is an arbitrary
  * \p BinaryPredicate \p binary_pred: the elements in <tt>[f, l)</tt> are
  * duplicates if, for every iterator \p i in the range, either <tt>i == f</tt>
  * or else <tt>binary_pred(*i, *(i-1))</tt> is \p true.
@@ -339,9 +345,9 @@ OutputIterator unique_copy(InputIterator first,
 /*! \p unique_copy copies elements from the range <tt>[first, last)</tt>
  * to a range beginning with \p result, except that in a consecutive group
  * of duplicate elements only the first one is copied. The return value
- * is the end of the range to which the elements are copied. 
+ * is the end of the range to which the elements are copied.
  *
- * This version of \p unique_copy uses the function object \c binary_pred 
+ * This version of \p unique_copy uses the function object \c binary_pred
  * to test for equality.
  *
  * The algorithm's execution is parallelized as determined by \p exec.
@@ -391,14 +397,14 @@ OutputIterator unique_copy(const thrust::detail::execution_policy_base<DerivedPo
                            InputIterator last,
                            OutputIterator result,
                            BinaryPredicate binary_pred);
-                       
+
 
 /*! \p unique_copy copies elements from the range <tt>[first, last)</tt>
  * to a range beginning with \p result, except that in a consecutive group
  * of duplicate elements only the first one is copied. The return value
- * is the end of the range to which the elements are copied. 
+ * is the end of the range to which the elements are copied.
  *
- * This version of \p unique_copy uses the function object \c binary_pred 
+ * This version of \p unique_copy uses the function object \c binary_pred
  * to test for equality.
  *
  *  \param first The beginning of the input range.
@@ -443,16 +449,16 @@ OutputIterator unique_copy(InputIterator first,
 
 /*! \p unique_by_key is a generalization of \p unique to key-value pairs.
  *  For each group of consecutive keys in the range <tt>[keys_first, keys_last)</tt>
- *  that are equal, \p unique_by_key removes all but the first element of 
+ *  that are equal, \p unique_by_key removes all but the first element of
  *  the group.  Similarly, the corresponding values in the range
- *  <tt>[values_first, values_first + (keys_last - keys_first))</tt> 
+ *  <tt>[values_first, values_first + (keys_last - keys_first))</tt>
  *  are also removed.
  *
  *  The return value is a \p pair of iterators <tt>(new_keys_last,new_values_last)</tt>
  *  such that no two consecutive elements in the range <tt>[keys_first, new_keys_last)</tt>
  *  are equal.
  *
- *  This version of \p unique_by_key uses \c operator== to test for equality and 
+ *  This version of \p unique_by_key uses \c operator== to test for equality and
  *  \c project1st to reduce values with equal keys.
  *
  *  The algorithm's execution is parallelized as determined by \p exec.
@@ -501,23 +507,23 @@ template<typename DerivedPolicy,
 __host__ __device__
   thrust::pair<ForwardIterator1,ForwardIterator2>
   unique_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
-                ForwardIterator1 keys_first, 
+                ForwardIterator1 keys_first,
                 ForwardIterator1 keys_last,
                 ForwardIterator2 values_first);
 
 
 /*! \p unique_by_key is a generalization of \p unique to key-value pairs.
  *  For each group of consecutive keys in the range <tt>[keys_first, keys_last)</tt>
- *  that are equal, \p unique_by_key removes all but the first element of 
+ *  that are equal, \p unique_by_key removes all but the first element of
  *  the group.  Similarly, the corresponding values in the range
- *  <tt>[values_first, values_first + (keys_last - keys_first))</tt> 
+ *  <tt>[values_first, values_first + (keys_last - keys_first))</tt>
  *  are also removed.
  *
  *  The return value is a \p pair of iterators <tt>(new_keys_last,new_values_last)</tt>
  *  such that no two consecutive elements in the range <tt>[keys_first, new_keys_last)</tt>
  *  are equal.
  *
- *  This version of \p unique_by_key uses \c operator== to test for equality and 
+ *  This version of \p unique_by_key uses \c operator== to test for equality and
  *  \c project1st to reduce values with equal keys.
  *
  *  \param keys_first The beginning of the key range.
@@ -557,16 +563,16 @@ __host__ __device__
 template<typename ForwardIterator1,
          typename ForwardIterator2>
   thrust::pair<ForwardIterator1,ForwardIterator2>
-  unique_by_key(ForwardIterator1 keys_first, 
+  unique_by_key(ForwardIterator1 keys_first,
                 ForwardIterator1 keys_last,
                 ForwardIterator2 values_first);
 
 
 /*! \p unique_by_key is a generalization of \p unique to key-value pairs.
  *  For each group of consecutive keys in the range <tt>[keys_first, keys_last)</tt>
- *  that are equal, \p unique_by_key removes all but the first element of 
+ *  that are equal, \p unique_by_key removes all but the first element of
  *  the group.  Similarly, the corresponding values in the range
- *  <tt>[values_first, values_first + (keys_last - keys_first))</tt> 
+ *  <tt>[values_first, values_first + (keys_last - keys_first))</tt>
  *  are also removed.
  *
  *  This version of \p unique_by_key uses the function object \c binary_pred
@@ -622,7 +628,7 @@ template<typename DerivedPolicy,
 __host__ __device__
   thrust::pair<ForwardIterator1,ForwardIterator2>
     unique_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
-                  ForwardIterator1 keys_first, 
+                  ForwardIterator1 keys_first,
                   ForwardIterator1 keys_last,
                   ForwardIterator2 values_first,
                   BinaryPredicate binary_pred);
@@ -630,9 +636,9 @@ __host__ __device__
 
 /*! \p unique_by_key is a generalization of \p unique to key-value pairs.
  *  For each group of consecutive keys in the range <tt>[keys_first, keys_last)</tt>
- *  that are equal, \p unique_by_key removes all but the first element of 
+ *  that are equal, \p unique_by_key removes all but the first element of
  *  the group.  Similarly, the corresponding values in the range
- *  <tt>[values_first, values_first + (keys_last - keys_first))</tt> 
+ *  <tt>[values_first, values_first + (keys_last - keys_first))</tt>
  *  are also removed.
  *
  *  This version of \p unique_by_key uses the function object \c binary_pred
@@ -679,7 +685,7 @@ template<typename ForwardIterator1,
          typename ForwardIterator2,
          typename BinaryPredicate>
   thrust::pair<ForwardIterator1,ForwardIterator2>
-  unique_by_key(ForwardIterator1 keys_first, 
+  unique_by_key(ForwardIterator1 keys_first,
                 ForwardIterator1 keys_last,
                 ForwardIterator2 values_first,
                 BinaryPredicate binary_pred);
@@ -748,7 +754,7 @@ template<typename DerivedPolicy,
 __host__ __device__
   thrust::pair<OutputIterator1,OutputIterator2>
     unique_by_key_copy(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
-                       InputIterator1 keys_first, 
+                       InputIterator1 keys_first,
                        InputIterator1 keys_last,
                        InputIterator2 values_first,
                        OutputIterator1 keys_result,
@@ -809,7 +815,7 @@ template<typename InputIterator1,
          typename OutputIterator1,
          typename OutputIterator2>
   thrust::pair<OutputIterator1,OutputIterator2>
-  unique_by_key_copy(InputIterator1 keys_first, 
+  unique_by_key_copy(InputIterator1 keys_first,
                      InputIterator1 keys_last,
                      InputIterator2 values_first,
                      OutputIterator1 keys_result,
@@ -883,7 +889,7 @@ template<typename DerivedPolicy,
 __host__ __device__
   thrust::pair<OutputIterator1,OutputIterator2>
     unique_by_key_copy(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
-                       InputIterator1 keys_first, 
+                       InputIterator1 keys_first,
                        InputIterator1 keys_last,
                        InputIterator2 values_first,
                        OutputIterator1 keys_result,
@@ -949,7 +955,7 @@ template<typename InputIterator1,
          typename OutputIterator2,
          typename BinaryPredicate>
   thrust::pair<OutputIterator1,OutputIterator2>
-  unique_by_key_copy(InputIterator1 keys_first, 
+  unique_by_key_copy(InputIterator1 keys_first,
                      InputIterator1 keys_last,
                      InputIterator2 values_first,
                      OutputIterator1 keys_result,
@@ -958,7 +964,7 @@ template<typename InputIterator1,
 
 
 /*! \p unique_count counts runs of equal elements in the range <tt>[first, last)</tt>
- *  with the same value, 
+ *  with the same value,
  *
  *  This version of \p unique_count uses the function object \p binary_pred to test for equality.
  *
@@ -1005,7 +1011,7 @@ __host__ __device__
 
 
 /*! \p unique_count counts runs of equal elements in the range <tt>[first, last)</tt>
- *  with the same value, 
+ *  with the same value,
  *
  *  This version of \p unique_count uses \c operator== to test for equality.
  *
@@ -1050,7 +1056,7 @@ __host__ __device__
 
 
 /*! \p unique_count counts runs of equal elements in the range <tt>[first, last)</tt>
- *  with the same value, 
+ *  with the same value,
  *
  *  This version of \p unique_count uses the function object \p binary_pred to test for equality.
  *
@@ -1092,7 +1098,7 @@ __host__ __device__
 
 
 /*! \p unique_count counts runs of equal elements in the range <tt>[first, last)</tt>
- *  with the same value, 
+ *  with the same value,
  *
  *  This version of \p unique_count uses \c operator== to test for equality.
  *
diff --git a/thrust/thrust/universal_allocator.h b/thrust/thrust/universal_allocator.h
index 8d85cd20de0..742bb559a3d 100644
--- a/thrust/thrust/universal_allocator.h
+++ b/thrust/thrust/universal_allocator.h
@@ -24,6 +24,12 @@
 
 #include <thrust/detail/config.h>
 
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 // #include the device system's vector header
 #define __THRUST_DEVICE_SYSTEM_MEMORY_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/memory.h>
 #include __THRUST_DEVICE_SYSTEM_MEMORY_HEADER
diff --git a/thrust/thrust/universal_ptr.h b/thrust/thrust/universal_ptr.h
index 9d1de19d5f1..122ade0ed11 100644
--- a/thrust/thrust/universal_ptr.h
+++ b/thrust/thrust/universal_ptr.h
@@ -22,5 +22,13 @@
 
 #pragma once
 
+#include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 #include <thrust/universal_allocator.h>
 
diff --git a/thrust/thrust/universal_vector.h b/thrust/thrust/universal_vector.h
index 0ce38fd868c..311f31473f7 100644
--- a/thrust/thrust/universal_vector.h
+++ b/thrust/thrust/universal_vector.h
@@ -22,6 +22,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/universal_allocator.h>
 
 // #include the device system's vector header
diff --git a/thrust/thrust/version.h b/thrust/thrust/version.h
index 71f1adb6951..8c3fa09acaf 100644
--- a/thrust/thrust/version.h
+++ b/thrust/thrust/version.h
@@ -28,6 +28,14 @@
 
 #pragma once
 
+#include <thrust/detail/config/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
+
 //  This is the only Thrust header that is guaranteed to
 //  change with every Thrust release.
 //
diff --git a/thrust/thrust/zip_function.h b/thrust/thrust/zip_function.h
index 7653f9b7f3e..b5b725b61dd 100644
--- a/thrust/thrust/zip_function.h
+++ b/thrust/thrust/zip_function.h
@@ -8,6 +8,12 @@
 #pragma once
 
 #include <thrust/detail/config.h>
+
+#if defined(_CCCL_COMPILER_NVHPC) && defined(_CCCL_USE_IMPLICIT_SYSTEM_DEADER)
+#pragma GCC system_header
+#else // ^^^ _CCCL_COMPILER_NVHPC ^^^ / vvv !_CCCL_COMPILER_NVHPC vvv
+_CCCL_IMPLICIT_SYSTEM_HEADER
+#endif // !_CCCL_COMPILER_NVHPC
 #include <thrust/detail/cpp11_required.h>
 #include <thrust/detail/modern_gcc_required.h>