56fd68e7dd
CCCL (CUDA C++ Core Libraries) provides: - CUB: device/block/warp-level GPU primitives (reduce, scan, sort, topk) - Thrust: high-level parallel algorithms (transform_reduce, sort, scan) - libcudacxx: CUDA C++ standard library (atomics, barriers, memory) - cudax: experimental features (memory resources, allocators) - Tuning policies: per-SM hardware-specific algorithm parameters Competition optimization vectors mapped to CCCL: - Output TPS (83% weight): warp_reduce, block_reduce, device_topk - Input TPS (14% weight): device_scan, block_load, prefetch - Cache TPS (3% weight): prefix caching strategy patterns - Memory (0.9 util): pooled/cached/buddy allocators Source: https://github.com/NVIDIA/cccl (shallow clone, HEAD only) License: Apache-2.0
152 lines
2.8 KiB
Plaintext
152 lines
2.8 KiB
Plaintext
//===----------------------------------------------------------------------===//
|
|
//
|
|
// 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.
|
|
//
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
#ifndef __COMMON_UTILITY_H__
|
|
#define __COMMON_UTILITY_H__
|
|
|
|
#include <cuda_runtime_api.h>
|
|
// cuda_runtime_api needs to come first
|
|
|
|
#include <cuda/__runtime/api_wrapper.h>
|
|
#include <cuda/__stream/stream_ref.h>
|
|
#include <cuda/atomic>
|
|
#include <cuda/std/utility>
|
|
|
|
#include <cuda/experimental/launch.cuh>
|
|
|
|
#include <new> // IWYU pragma: keep (needed for placement new)
|
|
|
|
#include "testing.cuh"
|
|
|
|
namespace
|
|
{
|
|
namespace test
|
|
{
|
|
constexpr auto one_thread_dims = cuda::make_config(cuda::block_dims<1>(), cuda::grid_dims<1>());
|
|
|
|
struct _malloc_pinned
|
|
{
|
|
private:
|
|
void* pv = nullptr;
|
|
|
|
public:
|
|
explicit _malloc_pinned(std::size_t size)
|
|
{
|
|
cuda::__ensure_current_context guard(cuda::device_ref{0});
|
|
_CCCL_TRY_CUDA_API(::cudaMallocHost, "failed to allocate pinned memory", &pv, size);
|
|
}
|
|
|
|
~_malloc_pinned()
|
|
{
|
|
cuda::__ensure_current_context guard(cuda::device_ref{0});
|
|
[[maybe_unused]] auto status = ::cudaFreeHost(pv);
|
|
}
|
|
|
|
template <class T>
|
|
T* get_as() const noexcept
|
|
{
|
|
return static_cast<T*>(pv);
|
|
}
|
|
};
|
|
|
|
template <class T>
|
|
struct pinned
|
|
{
|
|
private:
|
|
_malloc_pinned _mem;
|
|
|
|
public:
|
|
explicit pinned(T t)
|
|
: _mem(sizeof(T))
|
|
{
|
|
::new (_mem.get_as<void>()) T(std::move(t));
|
|
}
|
|
|
|
~pinned()
|
|
{
|
|
get()->~T();
|
|
}
|
|
|
|
T* get() noexcept
|
|
{
|
|
return _mem.get_as<T>();
|
|
}
|
|
const T* get() const noexcept
|
|
{
|
|
return _mem.get_as<T>();
|
|
}
|
|
|
|
T& operator*() noexcept
|
|
{
|
|
return *get();
|
|
}
|
|
const T& operator*() const noexcept
|
|
{
|
|
return *get();
|
|
}
|
|
};
|
|
|
|
template <int N>
|
|
struct assign_n
|
|
{
|
|
__device__ constexpr void operator()(int* pi) const noexcept
|
|
{
|
|
*pi = N;
|
|
}
|
|
};
|
|
|
|
template <int N>
|
|
struct verify_n
|
|
{
|
|
__device__ void operator()(int* pi) const noexcept
|
|
{
|
|
REQUIRE_DEVICE(*pi == N);
|
|
}
|
|
};
|
|
|
|
using assign_42 = assign_n<42>;
|
|
using verify_42 = verify_n<42>;
|
|
|
|
struct atomic_add_one
|
|
{
|
|
__device__ void operator()(int* pi) const noexcept
|
|
{
|
|
cuda::atomic_ref atomic_pi(*pi);
|
|
atomic_pi.fetch_add(1);
|
|
}
|
|
};
|
|
|
|
struct atomic_sub_one
|
|
{
|
|
__device__ void operator()(int* pi) const noexcept
|
|
{
|
|
cuda::atomic_ref atomic_pi(*pi);
|
|
atomic_pi.fetch_sub(1);
|
|
}
|
|
};
|
|
|
|
struct spin_until_80
|
|
{
|
|
__device__ void operator()(int* pi) const noexcept
|
|
{
|
|
cuda::atomic_ref atomic_pi(*pi);
|
|
while (atomic_pi.load() != 80)
|
|
;
|
|
}
|
|
};
|
|
|
|
struct empty_kernel
|
|
{
|
|
__device__ void operator()() const noexcept {}
|
|
};
|
|
} // namespace test
|
|
} // namespace
|
|
#endif // __COMMON_UTILITY_H__
|