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
85 lines
2.2 KiB
Plaintext
85 lines
2.2 KiB
Plaintext
#include <thrust/copy.h>
|
|
#include <thrust/execution_policy.h>
|
|
|
|
#include <unittest/unittest.h>
|
|
|
|
#ifdef THRUST_TEST_DEVICE_SIDE
|
|
template <typename ExecutionPolicy, typename Iterator1, typename Iterator2>
|
|
__global__ void copy_kernel(ExecutionPolicy exec, Iterator1 first, Iterator1 last, Iterator2 result)
|
|
{
|
|
thrust::copy(exec, first, last, result);
|
|
}
|
|
|
|
template <typename T, typename ExecutionPolicy>
|
|
void TestCopyDevice(ExecutionPolicy exec, size_t n)
|
|
{
|
|
thrust::host_vector<T> h_src = unittest::random_integers<T>(n);
|
|
thrust::host_vector<T> h_dst(n);
|
|
|
|
thrust::device_vector<T> d_src = h_src;
|
|
thrust::device_vector<T> d_dst(n);
|
|
|
|
thrust::copy(h_src.begin(), h_src.end(), h_dst.begin());
|
|
copy_kernel<<<1, 1>>>(exec, d_src.begin(), d_src.end(), d_dst.begin());
|
|
{
|
|
cudaError_t const err = cudaDeviceSynchronize();
|
|
ASSERT_EQUAL(cudaSuccess, err);
|
|
}
|
|
|
|
ASSERT_EQUAL(h_dst, d_dst);
|
|
}
|
|
|
|
template <typename T>
|
|
void TestCopyDeviceSeq(size_t n)
|
|
{
|
|
TestCopyDevice<T>(thrust::seq, n);
|
|
}
|
|
DECLARE_VARIABLE_UNITTEST(TestCopyDeviceSeq);
|
|
|
|
template <typename T>
|
|
void TestCopyDeviceDevice(size_t n)
|
|
{
|
|
TestCopyDevice<T>(thrust::device, n);
|
|
}
|
|
DECLARE_VARIABLE_UNITTEST(TestCopyDeviceDevice);
|
|
|
|
template <typename ExecutionPolicy, typename Iterator1, typename Size, typename Iterator2>
|
|
__global__ void copy_n_kernel(ExecutionPolicy exec, Iterator1 first, Size n, Iterator2 result)
|
|
{
|
|
thrust::copy_n(exec, first, n, result);
|
|
}
|
|
|
|
template <typename T, typename ExecutionPolicy>
|
|
void TestCopyNDevice(ExecutionPolicy exec, size_t n)
|
|
{
|
|
thrust::host_vector<T> h_src = unittest::random_integers<T>(n);
|
|
thrust::host_vector<T> h_dst(n);
|
|
|
|
thrust::device_vector<T> d_src = h_src;
|
|
thrust::device_vector<T> d_dst(n);
|
|
|
|
thrust::copy_n(h_src.begin(), h_src.size(), h_dst.begin());
|
|
copy_n_kernel<<<1, 1>>>(exec, d_src.begin(), d_src.size(), d_dst.begin());
|
|
{
|
|
cudaError_t const err = cudaDeviceSynchronize();
|
|
ASSERT_EQUAL(cudaSuccess, err);
|
|
}
|
|
|
|
ASSERT_EQUAL(h_dst, d_dst);
|
|
}
|
|
|
|
template <typename T>
|
|
void TestCopyNDeviceSeq(size_t n)
|
|
{
|
|
TestCopyNDevice<T>(thrust::seq, n);
|
|
}
|
|
DECLARE_VARIABLE_UNITTEST(TestCopyNDeviceSeq);
|
|
|
|
template <typename T>
|
|
void TestCopyNDeviceDevice(size_t n)
|
|
{
|
|
TestCopyNDevice<T>(thrust::device, n);
|
|
}
|
|
DECLARE_VARIABLE_UNITTEST(TestCopyNDeviceDevice);
|
|
#endif
|