[CCCL] 瘦身 + 补全: 移除 cudax/python/libcudacxx-tests 冗余文件, 新增 c2h 测试助手 + cmake 构建系统 + 8 个 CUDA thrust examples
变更摘要:
- 删除: cudax/ (783 files, 7.2M) — 实验性组件，竞赛不需要
- 删除: python/ (226 files, 2.0M) — Python 绑定，竞赛不需要
- 删除: libcudacxx/{test,benchmarks,codegen,cmake,share} (4432 files, 31M)
保留: libcudacxx/include/ (1463 headers, cuda::std 编译依赖)
- 新增: c2h/ (27 files) — CUB Catch2 测试辅助头文件，编译 243 个测试必需
- 新增: cmake/ (29 files) — CCCL 原生 CMake 构建系统
- 新增: thrust/examples/cuda/ (7 files) + cpp_integration/ (1 file)
async_reduce, custom_temporary_allocation, explicit_cuda_stream,
global_device_vector, range_view, unwrap_pointer, wrap_pointer, device
结果: cccl_upstream 从 74M→35M (瘦身 53%), 核心内容 100% 保留:
27/27 tuning headers, 78 benchmarks, 243 tests,
60 thrust examples, 18 CUB examples, 全部编译头文件
This commit is contained in:
muh-bot
@@ -1,184 +0,0 @@
|
||||
//===----------------------------------------------------------------------===//
|
||||
//
|
||||
// Part of CUDA Experimental in CUDA C++ Core Libraries,
|
||||
// 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) 2024 NVIDIA CORPORATION & AFFILIATES.
|
||||
//
|
||||
//===----------------------------------------------------------------------===//
|
||||
|
||||
#include <cuda/std/type_traits>
|
||||
#include <cuda/std/utility>
|
||||
|
||||
#include <cuda/experimental/launch.cuh>
|
||||
#include <cuda/experimental/stream.cuh>
|
||||
|
||||
#include <testing.cuh>
|
||||
#include <utility.cuh>
|
||||
|
||||
C2H_CCCLRT_TEST("Can create a stream and launch work into it", "[stream]")
|
||||
{
|
||||
cudax::stream str{cuda::device_ref{0}};
|
||||
::test::pinned<int> i(0);
|
||||
cudax::launch(str, ::test::one_thread_dims, ::test::assign_42{}, i.get());
|
||||
str.sync();
|
||||
REQUIRE(*i == 42);
|
||||
}
|
||||
|
||||
C2H_CCCLRT_TEST("From native handle", "[stream]")
|
||||
{
|
||||
cuda::__ensure_current_context guard(cuda::device_ref{0});
|
||||
cudaStream_t handle;
|
||||
REQUIRE_CUDART(cudaStreamCreate(&handle));
|
||||
{
|
||||
auto stream = cudax::stream::from_native_handle(handle);
|
||||
|
||||
::test::pinned<int> i(0);
|
||||
cudax::launch(stream, ::test::one_thread_dims, ::test::assign_42{}, i.get());
|
||||
stream.sync();
|
||||
REQUIRE(*i == 42);
|
||||
(void) stream.release();
|
||||
}
|
||||
REQUIRE_CUDART(cudaStreamDestroy(handle));
|
||||
}
|
||||
|
||||
template <typename StreamType>
|
||||
void add_dependency_test(const StreamType& waiter, const StreamType& waitee)
|
||||
{
|
||||
REQUIRE(waiter != waitee);
|
||||
|
||||
auto verify_dependency = [&](const auto& insert_dependency) {
|
||||
::test::pinned<int> i(0);
|
||||
::cuda::atomic_ref atomic_i(*i);
|
||||
|
||||
cudax::launch(waitee, ::test::one_thread_dims, ::test::spin_until_80{}, i.get());
|
||||
cudax::launch(waitee, ::test::one_thread_dims, ::test::assign_42{}, i.get());
|
||||
insert_dependency();
|
||||
cudax::launch(waiter, ::test::one_thread_dims, ::test::verify_42{}, i.get());
|
||||
REQUIRE(atomic_i.load() != 42);
|
||||
REQUIRE(!waiter.is_done());
|
||||
atomic_i.store(80);
|
||||
waiter.sync();
|
||||
waitee.sync();
|
||||
};
|
||||
|
||||
SECTION("Stream wait declared event")
|
||||
{
|
||||
verify_dependency([&]() {
|
||||
cuda::event ev(waitee);
|
||||
waiter.wait(ev);
|
||||
});
|
||||
}
|
||||
|
||||
SECTION("Stream wait returned event")
|
||||
{
|
||||
verify_dependency([&]() {
|
||||
auto ev = waitee.record_event();
|
||||
waiter.wait(ev);
|
||||
});
|
||||
}
|
||||
|
||||
SECTION("Stream wait returned timed event")
|
||||
{
|
||||
verify_dependency([&]() {
|
||||
auto ev = waitee.record_timed_event();
|
||||
waiter.wait(ev);
|
||||
});
|
||||
}
|
||||
|
||||
SECTION("Stream wait stream")
|
||||
{
|
||||
verify_dependency([&]() {
|
||||
waiter.wait(waitee);
|
||||
});
|
||||
}
|
||||
}
|
||||
|
||||
C2H_CCCLRT_TEST("Can add dependency into a stream", "[stream]")
|
||||
{
|
||||
cudax::stream waiter{cuda::device_ref{0}}, waitee{cuda::device_ref{0}};
|
||||
|
||||
add_dependency_test<cudax::stream>(waiter, waitee);
|
||||
add_dependency_test<cudax::stream_ref>(waiter, waitee);
|
||||
}
|
||||
|
||||
C2H_CCCLRT_TEST("Stream priority", "[stream]")
|
||||
{
|
||||
cudax::stream stream_default_prio{cuda::device_ref{0}};
|
||||
REQUIRE(stream_default_prio.priority() == cudax::stream::default_priority);
|
||||
|
||||
auto priority = cudax::stream::default_priority - 1;
|
||||
cudax::stream stream{cuda::device_ref{0}, priority};
|
||||
REQUIRE(stream.priority() == priority);
|
||||
}
|
||||
|
||||
C2H_CCCLRT_TEST("Stream get device", "[stream]")
|
||||
{
|
||||
cudax::stream dev0_stream(cuda::device_ref{0});
|
||||
REQUIRE(dev0_stream.device() == 0);
|
||||
|
||||
cudax::__ensure_current_device guard(cuda::device_ref{*std::prev(cuda::devices.end())});
|
||||
cudaStream_t stream_handle;
|
||||
REQUIRE_CUDART(cudaStreamCreate(&stream_handle));
|
||||
auto stream_cudart = cudax::stream::from_native_handle(stream_handle);
|
||||
REQUIRE(stream_cudart.device() == *std::prev(cuda::devices.end()));
|
||||
auto stream_ref_cudart = cudax::stream_ref(stream_handle);
|
||||
REQUIRE(stream_ref_cudart.device() == *std::prev(cuda::devices.end()));
|
||||
|
||||
INFO("Can create a side stream using logical device");
|
||||
{
|
||||
if (test::cuda_driver_version() >= 12050)
|
||||
{
|
||||
auto ldev = dev0_stream.logical_device();
|
||||
REQUIRE(ldev.kind() == cudax::logical_device::kinds::device);
|
||||
cudax::stream side_stream(ldev);
|
||||
REQUIRE(side_stream.device() == dev0_stream.device());
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
C2H_CCCLRT_TEST("Stream ID", "[stream]")
|
||||
{
|
||||
STATIC_REQUIRE(cuda::std::is_same_v<unsigned long long, cuda::std::underlying_type_t<cuda::stream_id>>);
|
||||
STATIC_REQUIRE(cuda::std::is_same_v<cuda::stream_id, decltype(cuda::std::declval<cudax::stream_ref>().id())>);
|
||||
|
||||
cudax::stream stream1{cuda::device_ref{0}};
|
||||
cudax::stream stream2{cuda::device_ref{0}};
|
||||
|
||||
// Test that id() returns a valid ID
|
||||
auto id1 = stream1.id();
|
||||
auto id2 = stream2.id();
|
||||
|
||||
// Test that different streams have different IDs
|
||||
#if _CCCL_COMPILER(NVHPC, <, 25, 11)
|
||||
REQUIRE(cuda::std::to_underlying(id1) != cuda::std::to_underlying(id2));
|
||||
#else // ^^^ _CCCL_COMPILER(NVHPC, <, 25, 11) ^^^ / vvv !_CCCL_COMPILER(NVHPC, <, 25, 11) vvv
|
||||
REQUIRE(id1 != id2);
|
||||
#endif // ^^^ !_CCCL_COMPILER(NVHPC, <, 25, 11) ^^^
|
||||
|
||||
// Test that the same stream returns the same ID when called multiple times
|
||||
#if _CCCL_COMPILER(NVHPC, <, 25, 11)
|
||||
REQUIRE(cuda::std::to_underlying(stream1.id()) == cuda::std::to_underlying(id1));
|
||||
REQUIRE(cuda::std::to_underlying(stream2.id()) == cuda::std::to_underlying(id2));
|
||||
#else // ^^^ _CCCL_COMPILER(NVHPC, <, 25, 11) ^^^ / vvv !_CCCL_COMPILER(NVHPC, <, 25, 11) vvv
|
||||
REQUIRE(stream1.id() == id1);
|
||||
REQUIRE(stream2.id() == id2);
|
||||
#endif // ^^^ !_CCCL_COMPILER(NVHPC, <, 25, 11) ^^^
|
||||
|
||||
{
|
||||
// Test that stream_ref also supports id()
|
||||
// NULL stream needs a device to be set
|
||||
cuda::__ensure_current_context guard(cuda::device_ref{0});
|
||||
cuda::stream_ref ref1(::cudaStream_t{});
|
||||
cuda::stream_ref ref2(stream1);
|
||||
|
||||
#if _CCCL_COMPILER(NVHPC, <, 25, 11)
|
||||
REQUIRE(cuda::std::to_underlying(ref1.id()) != cuda::std::to_underlying(ref2.id()));
|
||||
REQUIRE(cuda::std::to_underlying(ref2.id()) == cuda::std::to_underlying(id1));
|
||||
#else // ^^^ _CCCL_COMPILER(NVHPC, <, 25, 11) ^^^ / vvv !_CCCL_COMPILER(NVHPC, <, 25, 11) vvv
|
||||
REQUIRE(ref1.id() != ref2.id());
|
||||
REQUIRE(ref2.id() == id1);
|
||||
#endif // ^^^ !_CCCL_COMPILER(NVHPC, <, 25, 11) ^^^
|
||||
}
|
||||
}
|
||||
Reference in New Issue
Block a user