diff --git a/include/cuda/std/detail/__atomic b/include/cuda/std/detail/__atomic
index e5fe05034d..14f49434c2 100644
--- a/include/cuda/std/detail/__atomic
+++ b/include/cuda/std/detail/__atomic
@@ -44,7 +44,6 @@
 #define __ATOMIC_THREAD 10
 #endif //__ATOMIC_BLOCK
 
-
 _LIBCUDACXX_BEGIN_NAMESPACE_CUDA
 
 namespace detail {
@@ -102,7 +101,8 @@ namespace detail {
 
 _LIBCUDACXX_END_NAMESPACE_CUDA
 
-#ifdef _MSC_VER
+#if defined(_LIBCUDACXX_COMPILER_MSVC)
+    // Inject atomic intrinsics built from MSVC compiler intrinsics
     #include "libcxx/include/support/win32/atomic_msvc.h"
 #endif
 
diff --git a/libcxx/include/atomic b/libcxx/include/atomic
index c34cec008b..be235a64dc 100644
--- a/libcxx/include/atomic
+++ b/libcxx/include/atomic
@@ -555,6 +555,11 @@ void atomic_signal_fence(memory_order m) noexcept;
 #include <type_traits>
 #include <version>
 #include <__pragma_push>
+
+#if defined(_LIBCUDACXX_COMPILER_MSVC)
+#include "support/win32/atomic_msvc.h"
+#endif
+
 #endif //__cuda_std__
 
 #if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
diff --git a/libcxx/include/support/win32/atomic_msvc.h b/libcxx/include/support/win32/atomic_msvc.h
index 11cb4f6f01..6b1f2da88e 100644
--- a/libcxx/include/support/win32/atomic_msvc.h
+++ b/libcxx/include/support/win32/atomic_msvc.h
@@ -66,7 +66,7 @@ void __atomic_load_relaxed(const volatile _Type *__ptr, _Type *__ret) {
 template<class _Type>
 void __atomic_load(const volatile _Type *__ptr, _Type *__ret, int __memorder) {
     switch (__memorder) {
-    case __ATOMIC_SEQ_CST: _Memory_barrier();
+    case __ATOMIC_SEQ_CST: _Memory_barrier(); _LIBCUDACXX_FALLTHROUGH();
     case __ATOMIC_CONSUME:
     case __ATOMIC_ACQUIRE: __atomic_load_relaxed(__ptr, __ret); _Compiler_or_memory_barrier(); break;
     case __ATOMIC_RELAXED: __atomic_load_relaxed(__ptr, __ret); break;
@@ -98,7 +98,7 @@ template<class _Type>
 void __atomic_store(volatile _Type *__ptr, _Type *__val, int __memorder) {
     switch (__memorder) {
     case __ATOMIC_RELEASE: _Compiler_or_memory_barrier(); __atomic_store_relaxed(__ptr, __val); break;
-    case __ATOMIC_SEQ_CST: _Memory_barrier();
+    case __ATOMIC_SEQ_CST: _Memory_barrier(); _LIBCUDACXX_FALLTHROUGH();
     case __ATOMIC_RELAXED: __atomic_store_relaxed(__ptr, __val); break;
     default: assert(0);
     }
@@ -149,7 +149,8 @@ bool __atomic_compare_exchange(_Type volatile *__ptr, _Type *__expected, const _
     bool success = false;
     switch (detail::__stronger_order_cuda(__success_memorder, __failure_memorder)) {
     case __ATOMIC_RELEASE: _Compiler_or_memory_barrier(); success = __atomic_compare_exchange_relaxed(__ptr, __expected, __desired); break;
-    case __ATOMIC_ACQ_REL: _Compiler_or_memory_barrier();
+    case __ATOMIC_ACQ_REL: _Compiler_or_memory_barrier(); _LIBCUDACXX_FALLTHROUGH();
+    case __ATOMIC_CONSUME:
     case __ATOMIC_ACQUIRE: success = __atomic_compare_exchange_relaxed(__ptr, __expected, __desired); _Compiler_or_memory_barrier(); break;
     case __ATOMIC_SEQ_CST: _Memory_barrier(); success = __atomic_compare_exchange_relaxed(__ptr, __expected, __desired); _Compiler_or_memory_barrier(); break;
     case __ATOMIC_RELAXED: success = __atomic_compare_exchange_relaxed(__ptr, __expected, __desired); break;
@@ -182,7 +183,8 @@ template<class _Type>
 void __atomic_exchange(_Type volatile *__ptr, const _Type *__val, _Type *__ret, int __memorder) {
     switch (__memorder) {
     case __ATOMIC_RELEASE: _Compiler_or_memory_barrier(); __atomic_exchange_relaxed(__ptr, __val, __ret);break;
-    case __ATOMIC_ACQ_REL: _Compiler_or_memory_barrier();
+    case __ATOMIC_ACQ_REL: _Compiler_or_memory_barrier(); _LIBCUDACXX_FALLTHROUGH();
+    case __ATOMIC_CONSUME:
     case __ATOMIC_ACQUIRE: __atomic_exchange_relaxed(__ptr, __val, __ret); _Compiler_or_memory_barrier(); break;
     case __ATOMIC_SEQ_CST: _Memory_barrier(); __atomic_exchange_relaxed(__ptr, __val, __ret); _Compiler_or_memory_barrier(); break;
     case __ATOMIC_RELAXED: __atomic_exchange_relaxed(__ptr, __val, __ret); break;
@@ -217,7 +219,8 @@ _Type __atomic_fetch_add(_Type volatile *__ptr, _Delta __val, int __memorder) {
 
     switch (__memorder) {
     case __ATOMIC_RELEASE: _Compiler_or_memory_barrier(); __atomic_fetch_add_relaxed(__ptr, &__val, __dest);break;
-    case __ATOMIC_ACQ_REL: _Compiler_or_memory_barrier();
+    case __ATOMIC_ACQ_REL: _Compiler_or_memory_barrier(); _LIBCUDACXX_FALLTHROUGH();
+    case __ATOMIC_CONSUME:
     case __ATOMIC_ACQUIRE: __atomic_fetch_add_relaxed(__ptr, &__val, __dest); _Compiler_or_memory_barrier(); break;
     case __ATOMIC_SEQ_CST: _Memory_barrier(); __atomic_fetch_add_relaxed(__ptr, &__val, __dest); _Compiler_or_memory_barrier(); break;
     case __ATOMIC_RELAXED: __atomic_fetch_add_relaxed(__ptr, &__val, __dest); break;
@@ -258,7 +261,8 @@ _Type __atomic_fetch_and(_Type volatile *__ptr, _Delta __val, int __memorder) {
 
     switch (__memorder) {
     case __ATOMIC_RELEASE: _Compiler_or_memory_barrier(); __atomic_fetch_and_relaxed(__ptr, &__val, __dest);break;
-    case __ATOMIC_ACQ_REL: _Compiler_or_memory_barrier();
+    case __ATOMIC_ACQ_REL: _Compiler_or_memory_barrier(); _LIBCUDACXX_FALLTHROUGH();
+    case __ATOMIC_CONSUME:
     case __ATOMIC_ACQUIRE: __atomic_fetch_and_relaxed(__ptr, &__val, __dest); _Compiler_or_memory_barrier(); break;
     case __ATOMIC_SEQ_CST: _Memory_barrier(); __atomic_fetch_and_relaxed(__ptr, &__val, __dest); _Compiler_or_memory_barrier(); break;
     case __ATOMIC_RELAXED: __atomic_fetch_and_relaxed(__ptr, &__val, __dest); break;
@@ -294,7 +298,8 @@ _Type __atomic_fetch_xor(_Type volatile *__ptr, _Delta __val, int __memorder) {
 
     switch (__memorder) {
     case __ATOMIC_RELEASE: _Compiler_or_memory_barrier(); __atomic_fetch_xor_relaxed(__ptr, &__val, __dest);break;
-    case __ATOMIC_ACQ_REL: _Compiler_or_memory_barrier();
+    case __ATOMIC_ACQ_REL: _Compiler_or_memory_barrier(); _LIBCUDACXX_FALLTHROUGH();
+    case __ATOMIC_CONSUME:
     case __ATOMIC_ACQUIRE: __atomic_fetch_xor_relaxed(__ptr, &__val, __dest); _Compiler_or_memory_barrier(); break;
     case __ATOMIC_SEQ_CST: _Memory_barrier(); __atomic_fetch_xor_relaxed(__ptr, &__val, __dest); _Compiler_or_memory_barrier(); break;
     case __ATOMIC_RELAXED: __atomic_fetch_xor_relaxed(__ptr, &__val, __dest); break;
@@ -330,7 +335,8 @@ _Type __atomic_fetch_or(_Type volatile *__ptr, _Delta __val, int __memorder) {
 
     switch (__memorder) {
     case __ATOMIC_RELEASE: _Compiler_or_memory_barrier(); __atomic_fetch_or_relaxed(__ptr, &__val, __dest);break;
-    case __ATOMIC_ACQ_REL: _Compiler_or_memory_barrier();
+    case __ATOMIC_ACQ_REL: _Compiler_or_memory_barrier(); _LIBCUDACXX_FALLTHROUGH();
+    case __ATOMIC_CONSUME:
     case __ATOMIC_ACQUIRE: __atomic_fetch_or_relaxed(__ptr, &__val, __dest); _Compiler_or_memory_barrier(); break;
     case __ATOMIC_SEQ_CST: _Memory_barrier(); __atomic_fetch_or_relaxed(__ptr, &__val, __dest); _Compiler_or_memory_barrier(); break;
     case __ATOMIC_RELAXED: __atomic_fetch_or_relaxed(__ptr, &__val, __dest); break;