Expose PSTL algorithms through <cuda/std/algorithm> and <cuda/std/numeric>

libcudacxx/benchmarks/bench/adjacent_difference/basic.cu

@@ -12,7 +12,7 @@
 #include <thrust/device_vector.h>
 
 #include <cuda/memory_pool>
-#include <cuda/std/__pstl_algorithm>
+#include <cuda/std/execution>
 #include <cuda/stream>
 
 #include "nvbench_helper.cuh"

libcudacxx/benchmarks/bench/adjacent_find/basic.cu

@@ -12,8 +12,7 @@
 #include <thrust/sequence.h>
 
 #include <cuda/memory_pool>
-#include <cuda/std/__pstl_algorithm>
-#include <cuda/std/algorithm>
+#include <cuda/std/execution>
 #include <cuda/stream>
 
 #include "nvbench_helper.cuh"

libcudacxx/benchmarks/bench/all_of/basic.cu

@@ -11,7 +11,7 @@
 #include <thrust/device_vector.h>
 
 #include <cuda/memory_pool>
-#include <cuda/std/__pstl_algorithm>
+#include <cuda/std/execution>
 #include <cuda/stream>
 
 #include "nvbench_helper.cuh"
@@ -41,8 +41,8 @@ static void basic(nvbench::state& state, nvbench::type_list<T>)
   const auto mismatch_point = static_cast<std::size_t>(elements * common_prefix);
 
   thrust::device_vector<T> dinput(elements, thrust::no_init);
-  cuda::std::fill(cuda::execution::__cub_par_unseq, dinput.begin(), dinput.begin() + mismatch_point, T{0});
-  cuda::std::fill(cuda::execution::__cub_par_unseq, dinput.begin() + mismatch_point, dinput.end(), val);
+  cuda::std::fill(cuda::execution::gpu, dinput.begin(), dinput.begin() + mismatch_point, T{0});
+  cuda::std::fill(cuda::execution::gpu, dinput.begin() + mismatch_point, dinput.end(), val);
 
   state.add_global_memory_reads<T>(mismatch_point + 1);
   state.add_global_memory_writes<size_t>(1);

libcudacxx/benchmarks/bench/any_of/basic.cu

@@ -11,7 +11,7 @@
 #include <thrust/device_vector.h>
 
 #include <cuda/memory_pool>
-#include <cuda/std/__pstl_algorithm>
+#include <cuda/std/execution>
 #include <cuda/stream>
 
 #include "nvbench_helper.cuh"
@@ -41,8 +41,8 @@ static void basic(nvbench::state& state, nvbench::type_list<T>)
   const auto mismatch_point = static_cast<std::size_t>(elements * common_prefix);
 
   thrust::device_vector<T> dinput(elements, thrust::no_init);
-  cuda::std::fill(cuda::execution::__cub_par_unseq, dinput.begin(), dinput.begin() + mismatch_point, T{0});
-  cuda::std::fill(cuda::execution::__cub_par_unseq, dinput.begin() + mismatch_point, dinput.end(), val);
+  cuda::std::fill(cuda::execution::gpu, dinput.begin(), dinput.begin() + mismatch_point, T{0});
+  cuda::std::fill(cuda::execution::gpu, dinput.begin() + mismatch_point, dinput.end(), val);
 
   state.add_global_memory_reads<T>(mismatch_point + 1);
   state.add_global_memory_writes<size_t>(1);

libcudacxx/benchmarks/bench/copy/basic.cu

@@ -11,7 +11,7 @@
 #include <thrust/device_vector.h>
 
 #include <cuda/memory_pool>
-#include <cuda/std/__pstl_algorithm>
+#include <cuda/std/execution>
 #include <cuda/stream>
 
 #include "nvbench_helper.cuh"

libcudacxx/benchmarks/bench/copy_if/basic.cu

@@ -11,8 +11,8 @@
 #include <thrust/device_vector.h>
 
 #include <cuda/memory_pool>
-#include <cuda/std/__pstl_algorithm>
 #include <cuda/std/complex>
+#include <cuda/std/execution>
 #include <cuda/stream>
 
 #include "nvbench_helper.cuh"

libcudacxx/benchmarks/bench/copy_n/basic.cu

@@ -11,7 +11,7 @@
 #include <thrust/device_vector.h>
 
 #include <cuda/memory_pool>
-#include <cuda/std/__pstl_algorithm>
+#include <cuda/std/execution>
 #include <cuda/stream>
 
 #include "nvbench_helper.cuh"

libcudacxx/benchmarks/bench/count/basic.cu

@@ -11,7 +11,7 @@
 #include <thrust/device_vector.h>
 
 #include <cuda/memory_pool>
-#include <cuda/std/__pstl_algorithm>
+#include <cuda/std/execution>
 #include <cuda/stream>
 
 #include "nvbench_helper.cuh"

libcudacxx/benchmarks/bench/count_if/basic.cu

@@ -11,7 +11,7 @@
 #include <thrust/device_vector.h>
 
 #include <cuda/memory_pool>
-#include <cuda/std/__pstl_algorithm>
+#include <cuda/std/execution>
 #include <cuda/stream>
 
 #include "nvbench_helper.cuh"

libcudacxx/benchmarks/bench/equal/basic.cu

@@ -11,7 +11,7 @@
 #include <thrust/device_vector.h>
 
 #include <cuda/memory_pool>
-#include <cuda/std/__pstl_algorithm>
+#include <cuda/std/execution>
 #include <cuda/stream>
 
 #include "nvbench_helper.cuh"
@@ -26,8 +26,8 @@ static void range_iter(nvbench::state& state, nvbench::type_list<T>)
   const auto mismatch_point = static_cast<std::size_t>(elements * common_prefix);
 
   thrust::device_vector<T> dinput(elements, thrust::no_init);
-  cuda::std::fill(cuda::execution::__cub_par_unseq, dinput.begin(), dinput.begin() + mismatch_point, T{0});
-  cuda::std::fill(cuda::execution::__cub_par_unseq, dinput.begin() + mismatch_point, dinput.end(), val);
+  cuda::std::fill(cuda::execution::gpu, dinput.begin(), dinput.begin() + mismatch_point, T{0});
+  cuda::std::fill(cuda::execution::gpu, dinput.begin() + mismatch_point, dinput.end(), val);
 
   state.add_global_memory_reads<T>(mismatch_point + 1);
   state.add_global_memory_writes<size_t>(1);
@@ -56,8 +56,8 @@ static void range_range(nvbench::state& state, nvbench::type_list<T>)
   const auto mismatch_point = static_cast<std::size_t>(elements * common_prefix);
 
   thrust::device_vector<T> dinput(elements, thrust::no_init);
-  cuda::std::fill(cuda::execution::__cub_par_unseq, dinput.begin(), dinput.begin() + mismatch_point, T{0});
-  cuda::std::fill(cuda::execution::__cub_par_unseq, dinput.begin() + mismatch_point, dinput.end(), val);
+  cuda::std::fill(cuda::execution::gpu, dinput.begin(), dinput.begin() + mismatch_point, T{0});
+  cuda::std::fill(cuda::execution::gpu, dinput.begin() + mismatch_point, dinput.end(), val);
 
   state.add_global_memory_reads<T>(mismatch_point + 1);
   state.add_global_memory_writes<size_t>(1);

libcudacxx/benchmarks/bench/exclusive_scan/basic.cu

@@ -11,7 +11,7 @@
 #include <thrust/device_vector.h>
 
 #include <cuda/memory_pool>
-#include <cuda/std/__pstl_algorithm>
+#include <cuda/std/execution>
 #include <cuda/stream>
 
 #include "nvbench_helper.cuh"

libcudacxx/benchmarks/bench/exclusive_scan/max.cu

@@ -11,7 +11,7 @@
 #include <thrust/device_vector.h>
 
 #include <cuda/memory_pool>
-#include <cuda/std/__pstl_algorithm>
+#include <cuda/std/execution>
 #include <cuda/stream>
 
 #include "nvbench_helper.cuh"

libcudacxx/benchmarks/bench/fill/basic.cu

@@ -11,7 +11,7 @@
 #include <thrust/device_vector.h>
 
 #include <cuda/memory_pool>
-#include <cuda/std/__pstl_algorithm>
+#include <cuda/std/execution>
 #include <cuda/stream>
 
 #include "nvbench_helper.cuh"

libcudacxx/benchmarks/bench/fill_n/basic.cu

@@ -11,7 +11,7 @@
 #include <thrust/device_vector.h>
 
 #include <cuda/memory_pool>
-#include <cuda/std/__pstl_algorithm>
+#include <cuda/std/execution>
 #include <cuda/stream>
 
 #include "nvbench_helper.cuh"

libcudacxx/benchmarks/bench/find/basic.cu

@@ -11,7 +11,7 @@
 #include <thrust/device_vector.h>
 
 #include <cuda/memory_pool>
-#include <cuda/std/__pstl_algorithm>
+#include <cuda/std/execution>
 #include <cuda/stream>
 
 #include "nvbench_helper.cuh"
@@ -26,8 +26,8 @@ static void basic(nvbench::state& state, nvbench::type_list<T>)
   const auto mismatch_point = static_cast<std::size_t>(elements * common_prefix);
 
   thrust::device_vector<T> dinput(elements, thrust::no_init);
-  cuda::std::fill(cuda::execution::__cub_par_unseq, dinput.begin(), dinput.begin() + mismatch_point, T{0});
-  cuda::std::fill(cuda::execution::__cub_par_unseq, dinput.begin() + mismatch_point, dinput.end(), val);
+  cuda::std::fill(cuda::execution::gpu, dinput.begin(), dinput.begin() + mismatch_point, T{0});
+  cuda::std::fill(cuda::execution::gpu, dinput.begin() + mismatch_point, dinput.end(), val);
 
   state.add_global_memory_reads<T>(mismatch_point + 1);
   state.add_global_memory_writes<size_t>(1);

libcudacxx/benchmarks/bench/find_if/basic.cu

@@ -12,7 +12,7 @@
 
 #include <cuda/functional>
 #include <cuda/memory_pool>
-#include <cuda/std/__pstl_algorithm>
+#include <cuda/std/execution>
 #include <cuda/stream>
 
 #include "nvbench_helper.cuh"
@@ -27,8 +27,8 @@ static void basic(nvbench::state& state, nvbench::type_list<T>)
   const auto mismatch_point = static_cast<std::size_t>(elements * common_prefix);
 
   thrust::device_vector<T> dinput(elements, thrust::no_init);
-  cuda::std::fill(cuda::execution::__cub_par_unseq, dinput.begin(), dinput.begin() + mismatch_point, T{0});
-  cuda::std::fill(cuda::execution::__cub_par_unseq, dinput.begin() + mismatch_point, dinput.end(), val);
+  cuda::std::fill(cuda::execution::gpu, dinput.begin(), dinput.begin() + mismatch_point, T{0});
+  cuda::std::fill(cuda::execution::gpu, dinput.begin() + mismatch_point, dinput.end(), val);
 
   state.add_global_memory_reads<T>(mismatch_point + 1);
   state.add_global_memory_writes<size_t>(1);

libcudacxx/benchmarks/bench/find_if_not/basic.cu

@@ -11,7 +11,7 @@
 #include <thrust/device_vector.h>
 
 #include <cuda/memory_pool>
-#include <cuda/std/__pstl_algorithm>
+#include <cuda/std/execution>
 #include <cuda/stream>
 
 #include "nvbench_helper.cuh"
@@ -41,8 +41,8 @@ static void basic(nvbench::state& state, nvbench::type_list<T>)
   const auto mismatch_point = static_cast<std::size_t>(elements * common_prefix);
 
   thrust::device_vector<T> dinput(elements, thrust::no_init);
-  cuda::std::fill(cuda::execution::__cub_par_unseq, dinput.begin(), dinput.begin() + mismatch_point, T{0});
-  cuda::std::fill(cuda::execution::__cub_par_unseq, dinput.begin() + mismatch_point, dinput.end(), val);
+  cuda::std::fill(cuda::execution::gpu, dinput.begin(), dinput.begin() + mismatch_point, T{0});
+  cuda::std::fill(cuda::execution::gpu, dinput.begin() + mismatch_point, dinput.end(), val);
 
   state.add_global_memory_reads<T>(mismatch_point + 1);
   state.add_global_memory_writes<size_t>(1);

libcudacxx/benchmarks/bench/for_each/basic.cu

@@ -11,7 +11,7 @@
 #include <thrust/device_vector.h>
 
 #include <cuda/memory_pool>
-#include <cuda/std/__pstl_algorithm>
+#include <cuda/std/execution>
 #include <cuda/stream>
 
 #include "nvbench_helper.cuh"

libcudacxx/benchmarks/bench/for_each_n/basic.cu

@@ -11,7 +11,7 @@
 #include <thrust/device_vector.h>
 
 #include <cuda/memory_pool>
-#include <cuda/std/__pstl_algorithm>
+#include <cuda/std/execution>
 #include <cuda/stream>
 
 #include "nvbench_helper.cuh"

libcudacxx/benchmarks/bench/generate/basic.cu

@@ -4,7 +4,7 @@
 #include <thrust/device_vector.h>
 
 #include <cuda/memory_pool>
-#include <cuda/std/__pstl_algorithm>
+#include <cuda/std/execution>
 #include <cuda/stream>
 
 #include "nvbench_helper.cuh"

libcudacxx/benchmarks/bench/generate_n/basic.cu

@@ -4,7 +4,7 @@
 #include <thrust/device_vector.h>
 
 #include <cuda/memory_pool>
-#include <cuda/std/__pstl_algorithm>
+#include <cuda/std/execution>
 #include <cuda/stream>
 
 #include "nvbench_helper.cuh"

libcudacxx/benchmarks/bench/inclusive_scan/basic.cu

@@ -11,7 +11,7 @@
 #include <thrust/device_vector.h>
 
 #include <cuda/memory_pool>
-#include <cuda/std/__pstl_algorithm>
+#include <cuda/std/execution>
 #include <cuda/stream>
 
 #include "nvbench_helper.cuh"

libcudacxx/benchmarks/bench/inclusive_scan/max.cu

@@ -11,7 +11,7 @@
 #include <thrust/device_vector.h>
 
 #include <cuda/memory_pool>
-#include <cuda/std/__pstl_algorithm>
+#include <cuda/std/execution>
 #include <cuda/stream>
 
 #include "nvbench_helper.cuh"

libcudacxx/benchmarks/bench/merge/basic.cu

@@ -14,7 +14,7 @@
 #include <thrust/sort.h>
 
 #include <cuda/memory_pool>
-#include <cuda/std/__pstl_algorithm>
+#include <cuda/std/execution>
 #include <cuda/stream>
 
 #include "nvbench_helper.cuh"

libcudacxx/benchmarks/bench/mismatch/basic.cu

@@ -11,7 +11,7 @@
 #include <thrust/device_vector.h>
 
 #include <cuda/memory_pool>
-#include <cuda/std/__pstl_algorithm>
+#include <cuda/std/execution>
 #include <cuda/stream>
 
 #include "nvbench_helper.cuh"
@@ -26,8 +26,8 @@ static void range_iter(nvbench::state& state, nvbench::type_list<T>)
   const auto mismatch_point = static_cast<std::size_t>(elements * common_prefix);
 
   thrust::device_vector<T> dinput(elements, thrust::no_init);
-  cuda::std::fill(cuda::execution::__cub_par_unseq, dinput.begin(), dinput.begin() + mismatch_point, T{0});
-  cuda::std::fill(cuda::execution::__cub_par_unseq, dinput.begin() + mismatch_point, dinput.end(), val);
+  cuda::std::fill(cuda::execution::gpu, dinput.begin(), dinput.begin() + mismatch_point, T{0});
+  cuda::std::fill(cuda::execution::gpu, dinput.begin() + mismatch_point, dinput.end(), val);
 
   state.add_global_memory_reads<T>(mismatch_point + 1);
   state.add_global_memory_writes<size_t>(1);
@@ -56,8 +56,8 @@ static void range_range(nvbench::state& state, nvbench::type_list<T>)
   const auto mismatch_point = static_cast<std::size_t>(elements * common_prefix);
 
   thrust::device_vector<T> dinput(elements, thrust::no_init);
-  cuda::std::fill(cuda::execution::__cub_par_unseq, dinput.begin(), dinput.begin() + mismatch_point, T{0});
-  cuda::std::fill(cuda::execution::__cub_par_unseq, dinput.begin() + mismatch_point, dinput.end(), val);
+  cuda::std::fill(cuda::execution::gpu, dinput.begin(), dinput.begin() + mismatch_point, T{0});
+  cuda::std::fill(cuda::execution::gpu, dinput.begin() + mismatch_point, dinput.end(), val);
 
   state.add_global_memory_reads<T>(mismatch_point + 1);
   state.add_global_memory_writes<size_t>(1);

libcudacxx/benchmarks/bench/none_of/basic.cu

@@ -11,7 +11,7 @@
 #include <thrust/device_vector.h>
 
 #include <cuda/memory_pool>
-#include <cuda/std/__pstl_algorithm>
+#include <cuda/std/execution>
 #include <cuda/stream>
 
 #include "nvbench_helper.cuh"
@@ -41,8 +41,8 @@ static void basic(nvbench::state& state, nvbench::type_list<T>)
   const auto mismatch_point = static_cast<std::size_t>(elements * common_prefix);
 
   thrust::device_vector<T> dinput(elements, thrust::no_init);
-  cuda::std::fill(cuda::execution::__cub_par_unseq, dinput.begin(), dinput.begin() + mismatch_point, T{0});
-  cuda::std::fill(cuda::execution::__cub_par_unseq, dinput.begin() + mismatch_point, dinput.end(), val);
+  cuda::std::fill(cuda::execution::gpu, dinput.begin(), dinput.begin() + mismatch_point, T{0});
+  cuda::std::fill(cuda::execution::gpu, dinput.begin() + mismatch_point, dinput.end(), val);
 
   state.add_global_memory_reads<T>(mismatch_point + 1);
   state.add_global_memory_writes<size_t>(1);

libcudacxx/benchmarks/bench/reduce/basic.cu

@@ -11,7 +11,7 @@
 #include <thrust/device_vector.h>
 
 #include <cuda/memory_pool>
-#include <cuda/std/__pstl_algorithm>
+#include <cuda/std/execution>
 #include <cuda/stream>
 
 #include "nvbench_helper.cuh"

libcudacxx/benchmarks/bench/remove/basic.cu

@@ -11,8 +11,8 @@
 #include <thrust/device_vector.h>
 
 #include <cuda/memory_pool>
-#include <cuda/std/__pstl_algorithm>
 #include <cuda/std/complex>
+#include <cuda/std/execution>
 #include <cuda/stream>
 
 #include "nvbench_helper.cuh"
@@ -23,7 +23,7 @@ static void basic(nvbench::state& state, nvbench::type_list<T>)
   const auto elements = static_cast<std::size_t>(state.get_int64("Elements"));
 
   thrust::device_vector<T> in = generate(elements, bit_entropy::_1_000, T{0}, T{42});
-  const auto count            = cuda::std::count(cuda::execution::__cub_par_unseq, in.begin(), in.end(), T{42});
+  const auto count            = cuda::std::count(cuda::execution::gpu, in.begin(), in.end(), T{42});
 
   state.add_element_count(elements);
   state.add_global_memory_reads<T>(elements);

libcudacxx/benchmarks/bench/remove_copy/basic.cu

@@ -11,8 +11,8 @@
 #include <thrust/device_vector.h>
 
 #include <cuda/memory_pool>
-#include <cuda/std/__pstl_algorithm>
 #include <cuda/std/complex>
+#include <cuda/std/execution>
 #include <cuda/stream>
 
 #include "nvbench_helper.cuh"
@@ -24,7 +24,7 @@ static void basic(nvbench::state& state, nvbench::type_list<T>)
 
   thrust::device_vector<T> in = generate(elements, bit_entropy::_1_000, T{0}, T{42});
   thrust::device_vector<T> out(elements, thrust::no_init);
-  const auto count = cuda::std::count(cuda::execution::__cub_par_unseq, in.begin(), in.end(), T{42});
+  const auto count = cuda::std::count(cuda::execution::gpu, in.begin(), in.end(), T{42});
 
   state.add_element_count(elements);
   state.add_global_memory_reads<T>(elements);