Work around issues with cccl main (#15552)

This gets cuDF build cccl main on 12.3. There is one issue with the cuco tuple helpers but that will be fixed on the cuco side --------- Co-authored-by: Bernhard Manfred Gruber <[email protected]> Co-authored-by: Bradley Dice <[email protected]> Co-authored-by: Nghia Truong <[email protected]> Co-authored-by: ptaylor <[email protected]> Co-authored-by: Paul Taylor <[email protected]> Co-authored-by: Yunsong Wang <[email protected]>
rapidsai · May 28, 2024 · 29429f7 · 29429f7
1 parent 8a40567
commit 29429f7
Show file tree

Hide file tree

Showing 15 changed files with 209 additions and 27 deletions.
diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml
@@ -7,11 +7,13 @@ repos:
       - id: trailing-whitespace
         exclude: |
           (?x)^(
+            ^cpp/cmake/thirdparty/patches/.*|
             ^python/cudf/cudf/tests/data/subword_tokenizer_data/.*
           )
       - id: end-of-file-fixer
         exclude: |
           (?x)^(
+            ^cpp/cmake/thirdparty/patches/.*|
             ^python/cudf/cudf/tests/data/subword_tokenizer_data/.*
           )
   - repo: https://github.com/PyCQA/isort

diff --git a/cpp/cmake/thirdparty/patches/cccl_override.json b/cpp/cmake/thirdparty/patches/cccl_override.json
@@ -18,6 +18,11 @@
           "issue" : "thrust::copy introduced a change in behavior that causes failures with cudaErrorInvalidValue.",
           "fixed_in" : ""
         },
+        {
+          "file" : "${current_json_dir}/revert_pr_211_cccl_2.5.0.diff",
+          "issue" : "thrust::copy introduced a change in behavior that causes failures with cudaErrorInvalidValue.",
+          "fixed_in" : ""
+        },
         {
           "file": "cccl/kernel_pointer_hiding.diff",
           "issue": "Hide APIs that accept kernel pointers [https://github.com/NVIDIA/cccl/pull/1395]",
@@ -28,15 +33,30 @@
           "issue" : "Remove 64bit dispatching as not needed by libcudf and results in compiling twice as many kernels [https://github.com/rapidsai/cudf/pull/11437]",
           "fixed_in" : ""
         },
+        {
+          "file" : "${current_json_dir}/thrust_disable_64bit_dispatching_cccl_2.5.0.diff",
+          "issue" : "Remove 64bit dispatching as not needed by libcudf and results in compiling twice as many kernels [https://github.com/rapidsai/cudf/pull/11437]",
+          "fixed_in" : ""
+        },
         {
           "file" : "${current_json_dir}/thrust_faster_sort_compile_times.diff",
           "issue" : "Improve Thrust sort compile times by not unrolling loops for inlined comparators [https://github.com/rapidsai/cudf/pull/10577]",
           "fixed_in" : ""
         },
+        {
+          "file" : "${current_json_dir}/thrust_faster_sort_compile_times_cccl_2.5.0.diff",
+          "issue" : "Improve Thrust sort compile times by not unrolling loops for inlined comparators [https://github.com/rapidsai/cudf/pull/10577]",
+          "fixed_in" : ""
+        },
         {
           "file" : "${current_json_dir}/thrust_faster_scan_compile_times.diff",
           "issue" : "Improve Thrust scan compile times by reducing the number of kernels generated [https://github.com/rapidsai/cudf/pull/8183]",
           "fixed_in" : ""
+        },
+        {
+          "file" : "${current_json_dir}/thrust_faster_scan_compile_times_cccl_2.5.0.diff",
+          "issue" : "Improve Thrust scan compile times by reducing the number of kernels generated [https://github.com/rapidsai/cudf/pull/8183]",
+          "fixed_in" : ""
         }
       ]
     }

diff --git a/cpp/cmake/thirdparty/patches/revert_pr_211_cccl_2.5.0.diff b/cpp/cmake/thirdparty/patches/revert_pr_211_cccl_2.5.0.diff
@@ -0,0 +1,47 @@
+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 046eb83c0..8047c9701 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
+@@ -53,41 +53,15 @@ namespace cuda_cub
+
+ namespace __copy
+ {
+-template <class Derived, class InputIt, class OutputIt>
+-OutputIt THRUST_RUNTIME_FUNCTION device_to_device(
+-  execution_policy<Derived>& policy, InputIt first, InputIt last, OutputIt result, thrust::detail::true_type)
+-{
+-  typedef typename thrust::iterator_traits<InputIt>::value_type InputTy;
+-  const auto n = thrust::distance(first, last);
+-  if (n > 0)
+-  {
+-    cudaError status;
+-    status = trivial_copy_device_to_device(
+-      policy,
+-      reinterpret_cast<InputTy*>(thrust::raw_pointer_cast(&*result)),
+-      reinterpret_cast<InputTy const*>(thrust::raw_pointer_cast(&*first)),
+-      n);
+-    cuda_cub::throw_on_error(status, "__copy:: D->D: failed");
+-  }
+-
+-  return result + n;
+-}
+
+ template <class Derived, class InputIt, class OutputIt>
+ OutputIt THRUST_RUNTIME_FUNCTION device_to_device(
+-  execution_policy<Derived>& policy, InputIt first, InputIt last, OutputIt result, thrust::detail::false_type)
++  execution_policy<Derived>& policy, InputIt first, InputIt last, OutputIt result)
+ {
+   typedef typename thrust::iterator_traits<InputIt>::value_type InputTy;
+   return cuda_cub::transform(policy, first, last, result, thrust::identity<InputTy>());
+ }
+
+-template <class Derived, class InputIt, class OutputIt>
+-OutputIt THRUST_RUNTIME_FUNCTION
+-device_to_device(execution_policy<Derived>& policy, InputIt first, InputIt last, OutputIt result)
+-{
+-  return device_to_device(
+-    policy, first, last, result, typename is_indirectly_trivially_relocatable_to<InputIt, OutputIt>::type());
+-}
+ } // namespace __copy
+
+ } // namespace cuda_cub
diff --git a/cpp/cmake/thirdparty/patches/thrust_disable_64bit_dispatching_cccl_2.5.0.diff b/cpp/cmake/thirdparty/patches/thrust_disable_64bit_dispatching_cccl_2.5.0.diff
@@ -0,0 +1,25 @@
+diff --git a/thrust/thrust/system/cuda/detail/dispatch.h b/thrust/thrust/system/cuda/detail/dispatch.h
+index 2a3cc4e33..8fb337b26 100644
+--- a/thrust/thrust/system/cuda/detail/dispatch.h
++++ b/thrust/thrust/system/cuda/detail/dispatch.h
+@@ -44,8 +44,7 @@
+   }                                                                                   \
+   else                                                                                \
+   {                                                                                   \
+-    auto THRUST_PP_CAT2(count, _fixed) = static_cast<thrust::detail::int64_t>(count); \
+-    status                             = call arguments;                              \
++    throw std::runtime_error("THRUST_INDEX_TYPE_DISPATCH 64-bit count is unsupported in libcudf"); \
+   }
+
+ /**
+@@ -66,9 +65,7 @@
+   }                                                                                          \
+   else                                                                                       \
+   {                                                                                          \
+-    auto THRUST_PP_CAT2(count1, _fixed) = static_cast<thrust::detail::int64_t>(count1);      \
+-    auto THRUST_PP_CAT2(count2, _fixed) = static_cast<thrust::detail::int64_t>(count2);      \
+-    status                              = call arguments;                                    \
++    throw std::runtime_error("THRUST_DOUBLE_INDEX_TYPE_DISPATCH 64-bit count is unsupported in libcudf"); \
+   }
+ /**
+  * Dispatch between 32-bit and 64-bit index based versions of the same algorithm
diff --git a/cpp/cmake/thirdparty/patches/thrust_faster_scan_compile_times_cccl_2.5.0.diff b/cpp/cmake/thirdparty/patches/thrust_faster_scan_compile_times_cccl_2.5.0.diff
@@ -0,0 +1,39 @@
+diff --git a/cub/cub/device/dispatch/dispatch_radix_sort.cuh b/cub/cub/device/dispatch/dispatch_radix_sort.cuh
+index 0606485bb..dbb99ff13 100644
+--- a/cub/cub/device/dispatch/dispatch_radix_sort.cuh
++++ b/cub/cub/device/dispatch/dispatch_radix_sort.cuh
+@@ -1085,7 +1085,7 @@ struct DeviceRadixSortPolicy
+   };
+
+   /// SM60 (GP100)
+-  struct Policy600 : ChainedPolicy<600, Policy600, Policy500>
++  struct Policy600 : ChainedPolicy<600, Policy600, Policy600>
+   {
+     enum
+     {
+diff --git a/cub/cub/device/dispatch/dispatch_reduce.cuh b/cub/cub/device/dispatch/dispatch_reduce.cuh
+index f39613adb..75bd16ff9 100644
+--- a/cub/cub/device/dispatch/dispatch_reduce.cuh
++++ b/cub/cub/device/dispatch/dispatch_reduce.cuh
+@@ -488,7 +488,7 @@ struct DeviceReducePolicy
+   };
+
+   /// SM60
+-  struct Policy600 : ChainedPolicy<600, Policy600, Policy350>
++  struct Policy600 : ChainedPolicy<600, Policy600, Policy600>
+   {
+     static constexpr int threads_per_block  = 256;
+     static constexpr int items_per_thread   = 16;
+diff --git a/cub/cub/device/dispatch/tuning/tuning_scan.cuh b/cub/cub/device/dispatch/tuning/tuning_scan.cuh
+index 419908c4e..6ab0840e1 100644
+--- a/cub/cub/device/dispatch/tuning/tuning_scan.cuh
++++ b/cub/cub/device/dispatch/tuning/tuning_scan.cuh
+@@ -339,7 +339,7 @@ struct DeviceScanPolicy
+   /// SM600
+   struct Policy600
+       : DefaultTuning
+-      , ChainedPolicy<600, Policy600, Policy520>
++      , ChainedPolicy<600, Policy600, Policy600>
+   {};
+
+   /// SM800
diff --git a/cpp/cmake/thirdparty/patches/thrust_faster_sort_compile_times_cccl_2.5.0.diff b/cpp/cmake/thirdparty/patches/thrust_faster_sort_compile_times_cccl_2.5.0.diff
@@ -0,0 +1,39 @@
+diff --git a/cub/cub/block/block_merge_sort.cuh b/cub/cub/block/block_merge_sort.cuh
+index eb76ebb0b..c6c529a50 100644
+--- a/cub/cub/block/block_merge_sort.cuh
++++ b/cub/cub/block/block_merge_sort.cuh
+@@ -95,7 +95,7 @@ _CCCL_DEVICE _CCCL_FORCEINLINE void SerialMerge(
+   KeyT key1 = keys_shared[keys1_beg];
+   KeyT key2 = keys_shared[keys2_beg];
+
+-#pragma unroll
++#pragma unroll 1
+   for (int item = 0; item < ITEMS_PER_THREAD; ++item)
+   {
+     bool p = (keys2_beg < keys2_end) && ((keys1_beg >= keys1_end) || compare_op(key2, key1));
+@@ -376,7 +376,7 @@ public:
+       //
+       KeyT max_key = oob_default;
+
+-#pragma unroll
++#pragma unroll 1
+       for (int item = 1; item < ITEMS_PER_THREAD; ++item)
+       {
+         if (ITEMS_PER_THREAD * linear_tid + item < valid_items)
+diff --git a/cub/cub/thread/thread_sort.cuh b/cub/cub/thread/thread_sort.cuh
+index 7d9e8622f..da5627306 100644
+--- a/cub/cub/thread/thread_sort.cuh
++++ b/cub/cub/thread/thread_sort.cuh
+@@ -87,10 +87,10 @@ StableOddEvenSort(KeyT (&keys)[ITEMS_PER_THREAD], ValueT (&items)[ITEMS_PER_THRE
+ {
+   constexpr bool KEYS_ONLY = ::cuda::std::is_same<ValueT, NullType>::value;
+
+-#pragma unroll
++#pragma unroll 1
+   for (int i = 0; i < ITEMS_PER_THREAD; ++i)
+   {
+-#pragma unroll
++#pragma unroll 1
+     for (int j = 1 & i; j < ITEMS_PER_THREAD - 1; j += 2)
+     {
+       if (compare_op(keys[j + 1], keys[j]))
diff --git a/cpp/src/io/comp/statistics.cu b/cpp/src/io/comp/statistics.cu
@@ -18,6 +18,7 @@
 
 #include <rmm/exec_policy.hpp>
 
+#include <cuda/functional>
 #include <thrust/transform_reduce.h>
 
 namespace cudf::io {
@@ -32,9 +33,9 @@ writer_compression_statistics collect_compression_statistics(
     rmm::exec_policy(stream),
     results.begin(),
     results.end(),
-    [] __device__(auto& res) {
+    cuda::proclaim_return_type<size_t>([] __device__(compression_result const& res) {
       return res.status == compression_status::SUCCESS ? res.bytes_written : 0;
-    },
+    }),
     0ul,
     thrust::plus<size_t>());
 
@@ -47,9 +48,9 @@ writer_compression_statistics collect_compression_statistics(
       rmm::exec_policy(stream),
       zipped_begin,
       zipped_end,
-      [status] __device__(auto tup) {
+      cuda::proclaim_return_type<size_t>([status] __device__(auto tup) {
         return thrust::get<1>(tup).status == status ? thrust::get<0>(tup).size() : 0;
-      },
+      }),
       0ul,
       thrust::plus<size_t>());
   };

diff --git a/cpp/src/io/orc/reader_impl_decode.cu b/cpp/src/io/orc/reader_impl_decode.cu
@@ -692,8 +692,7 @@ std::vector<range> find_table_splits(table_view const& input,
      d_sizes  = d_segmented_sizes->view().begin<size_type>()] __device__(auto const segment_idx) {
       // Since the number of rows may not divisible by segment_length,
       // the last segment may be shorter than the others.
-      auto const current_length =
-        cuda::std::min(segment_length, num_rows - segment_length * segment_idx);
+      auto const current_length = min(segment_length, num_rows - segment_length * segment_idx);
       auto const size = d_sizes[segment_idx] / CHAR_BIT;  // divide by CHAR_BIT to get size in bytes
       return cumulative_size{static_cast<std::size_t>(current_length),
                              static_cast<std::size_t>(size)};

diff --git a/cpp/src/io/orc/stripe_init.cu b/cpp/src/io/orc/stripe_init.cu
@@ -561,20 +561,26 @@ void __host__ ParseCompressedStripeData(CompressedStreamInfo* strm_info,
                                         uint32_t log2maxcr,
                                         rmm::cuda_stream_view stream)
 {
-  dim3 dim_block(128, 1);
-  dim3 dim_grid((num_streams + 3) >> 2, 1);  // 1 stream per warp, 4 warps per block
-  gpuParseCompressedStripeData<<<dim_grid, dim_block, 0, stream.value()>>>(
-    strm_info, num_streams, compression_block_size, log2maxcr);
+  auto const num_blocks = (num_streams + 3) >> 2;  // 1 stream per warp, 4 warps per block
+  if (num_blocks > 0) {
+    dim3 dim_block(128, 1);
+    dim3 dim_grid(num_blocks, 1);
+    gpuParseCompressedStripeData<<<dim_grid, dim_block, 0, stream.value()>>>(
+      strm_info, num_streams, compression_block_size, log2maxcr);
+  }
 }
 
 void __host__ PostDecompressionReassemble(CompressedStreamInfo* strm_info,
                                           int32_t num_streams,
                                           rmm::cuda_stream_view stream)
 {
-  dim3 dim_block(128, 1);
-  dim3 dim_grid((num_streams + 3) >> 2, 1);  // 1 stream per warp, 4 warps per block
-  gpuPostDecompressionReassemble<<<dim_grid, dim_block, 0, stream.value()>>>(strm_info,
-                                                                             num_streams);
+  auto const num_blocks = (num_streams + 3) >> 2;  // 1 stream per warp, 4 warps per block
+  if (num_blocks > 0) {
+    dim3 dim_block(128, 1);
+    dim3 dim_grid(num_blocks, 1);
+    gpuPostDecompressionReassemble<<<dim_grid, dim_block, 0, stream.value()>>>(strm_info,
+                                                                               num_streams);
+  }
 }
 
 void __host__ ParseRowGroupIndex(RowGroup* row_groups,

diff --git a/cpp/src/io/parquet/page_string_decode.cu b/cpp/src/io/parquet/page_string_decode.cu
@@ -1197,14 +1197,17 @@ void ComputePageStringSizes(cudf::detail::hostdevice_span<PageInfo> pages,
   cudf::detail::join_streams(streams, stream);
 
   // check for needed temp space for DELTA_BYTE_ARRAY
-  auto const need_sizes = thrust::any_of(
-    rmm::exec_policy(stream), pages.device_begin(), pages.device_end(), [] __device__(auto& page) {
-      return page.temp_string_size != 0;
-    });
+  auto const need_sizes =
+    thrust::any_of(rmm::exec_policy(stream),
+                   pages.device_begin(),
+                   pages.device_end(),
+                   cuda::proclaim_return_type<bool>(
+                     [] __device__(auto& page) { return page.temp_string_size != 0; }));
 
   if (need_sizes) {
     // sum up all of the temp_string_sizes
-    auto const page_sizes = [] __device__(PageInfo const& page) { return page.temp_string_size; };
+    auto const page_sizes = cuda::proclaim_return_type<int64_t>(
+      [] __device__(PageInfo const& page) { return page.temp_string_size; });
     auto const total_size = thrust::transform_reduce(rmm::exec_policy(stream),
                                                      pages.device_begin(),
                                                      pages.device_end(),

diff --git a/cpp/src/io/parquet/reader_impl_preprocess.cu b/cpp/src/io/parquet/reader_impl_preprocess.cu
@@ -452,9 +452,9 @@ std::string encoding_to_string(Encoding encoding)
 [[nodiscard]] std::string list_unsupported_encodings(device_span<PageInfo const> pages,
                                                      rmm::cuda_stream_view stream)
 {
-  auto const to_mask = [] __device__(auto const& page) {
+  auto const to_mask = cuda::proclaim_return_type<uint32_t>([] __device__(auto const& page) {
     return is_supported_encoding(page.encoding) ? 0U : encoding_to_mask(page.encoding);
-  };
+  });
   uint32_t const unsupported = thrust::transform_reduce(
     rmm::exec_policy(stream), pages.begin(), pages.end(), to_mask, 0U, thrust::bit_or<uint32_t>());
   return encoding_bitmask_to_str(unsupported);

diff --git a/cpp/src/io/utilities/data_casting.cu b/cpp/src/io/utilities/data_casting.cu
@@ -34,6 +34,7 @@
 #include <rmm/resource_ref.hpp>
 
 #include <cub/cub.cuh>
+#include <cuda/functional>
 #include <thrust/copy.h>
 #include <thrust/functional.h>
 #include <thrust/transform_reduce.h>
@@ -783,7 +784,8 @@ template <typename SymbolT>
 struct to_string_view_pair {
   SymbolT const* data;
   to_string_view_pair(SymbolT const* _data) : data(_data) {}
-  __device__ auto operator()(thrust::tuple<size_type, size_type> ip)
+  __device__ thrust::pair<char const*, std::size_t> operator()(
+    thrust::tuple<size_type, size_type> ip)
   {
     return thrust::pair<char const*, std::size_t>{data + thrust::get<0>(ip),
                                                   static_cast<std::size_t>(thrust::get<1>(ip))};
@@ -805,7 +807,7 @@ static std::unique_ptr<column> parse_string(string_view_pair_it str_tuples,
     rmm::exec_policy(stream),
     str_tuples,
     str_tuples + col_size,
-    [] __device__(auto t) { return t.second; },
+    cuda::proclaim_return_type<std::size_t>([] __device__(auto t) { return t.second; }),
     size_type{0},
     thrust::maximum<size_type>{});
 

diff --git a/cpp/src/join/distinct_hash_join.cu b/cpp/src/join/distinct_hash_join.cu
@@ -182,7 +182,7 @@ distinct_hash_join<HasNested>::inner_join(rmm::cuda_stream_view stream,
     thrust::make_transform_output_iterator(probe_indices->begin(), output_fn{});
 
   auto const [probe_indices_end, _] = this->_hash_table.retrieve(
-    iter, iter + probe_table_num_rows, probe_indices_begin, build_indices_begin, stream.value());
+    iter, iter + probe_table_num_rows, probe_indices_begin, build_indices_begin, {stream.value()});
 
   auto const actual_size = std::distance(probe_indices_begin, probe_indices_end);
   build_indices->resize(actual_size, stream);

diff --git a/cpp/src/strings/split/split_re.cu b/cpp/src/strings/split/split_re.cu
@@ -219,9 +219,9 @@ std::unique_ptr<table> split_re(strings_column_view const& input,
     rmm::exec_policy(stream),
     thrust::make_counting_iterator<size_type>(0),
     thrust::make_counting_iterator<size_type>(strings_count),
-    [d_offsets] __device__(auto const idx) -> size_type {
+    cuda::proclaim_return_type<size_type>([d_offsets] __device__(auto const idx) -> size_type {
       return static_cast<size_type>(d_offsets[idx + 1] - d_offsets[idx]);
-    },
+    }),
     0,
     thrust::maximum<size_type>{});
 

diff --git a/cpp/tests/hash_map/map_test.cu b/cpp/tests/hash_map/map_test.cu
@@ -69,7 +69,6 @@ struct InsertTest : public cudf::test::BaseFixture {
 
 using TestTypes = ::testing::Types<key_value_types<int32_t, int32_t>,
                                    key_value_types<int64_t, int64_t>,
-                                   key_value_types<int16_t, int16_t>,
                                    key_value_types<int32_t, float>,
                                    key_value_types<int64_t, double>>;