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
123 lines
3.9 KiB
Plaintext
123 lines
3.9 KiB
Plaintext
#include <thrust/extrema.h>
|
|
#include <thrust/functional.h>
|
|
#include <thrust/iterator/retag.h>
|
|
#include <thrust/iterator/transform_iterator.h>
|
|
|
|
#include <cuda/iterator>
|
|
|
|
#include <unittest/unittest.h>
|
|
|
|
template <class Vector>
|
|
void TestMaxElementSimple()
|
|
{
|
|
using T = typename Vector::value_type;
|
|
|
|
Vector data{3, 5, 1, 2, 5, 1};
|
|
|
|
ASSERT_EQUAL(*thrust::max_element(data.begin(), data.end()), 5);
|
|
ASSERT_EQUAL(thrust::max_element(data.begin(), data.end()) - data.begin(), 1);
|
|
|
|
ASSERT_EQUAL(*thrust::max_element(data.begin(), data.end(), ::cuda::std::greater<T>()), 1);
|
|
ASSERT_EQUAL(thrust::max_element(data.begin(), data.end(), ::cuda::std::greater<T>()) - data.begin(), 2);
|
|
}
|
|
DECLARE_VECTOR_UNITTEST(TestMaxElementSimple);
|
|
|
|
template <class Vector>
|
|
void TestMaxElementWithTransform()
|
|
{
|
|
using T = typename Vector::value_type;
|
|
|
|
Vector data{3, 5, 1, 2, 5, 1};
|
|
|
|
ASSERT_EQUAL(*thrust::max_element(thrust::make_transform_iterator(data.begin(), ::cuda::std::negate<T>()),
|
|
thrust::make_transform_iterator(data.end(), ::cuda::std::negate<T>())),
|
|
-1);
|
|
ASSERT_EQUAL(*thrust::max_element(thrust::make_transform_iterator(data.begin(), ::cuda::std::negate<T>()),
|
|
thrust::make_transform_iterator(data.end(), ::cuda::std::negate<T>()),
|
|
::cuda::std::greater<T>()),
|
|
-5);
|
|
}
|
|
DECLARE_VECTOR_UNITTEST(TestMaxElementWithTransform);
|
|
|
|
template <typename T>
|
|
void TestMaxElement(const size_t n)
|
|
{
|
|
thrust::host_vector<T> h_data = unittest::random_samples<T>(n);
|
|
thrust::device_vector<T> d_data = h_data;
|
|
|
|
typename thrust::host_vector<T>::iterator h_max = thrust::max_element(h_data.begin(), h_data.end());
|
|
typename thrust::device_vector<T>::iterator d_max = thrust::max_element(d_data.begin(), d_data.end());
|
|
|
|
ASSERT_EQUAL(h_max - h_data.begin(), d_max - d_data.begin());
|
|
|
|
typename thrust::host_vector<T>::iterator h_min =
|
|
thrust::max_element(h_data.begin(), h_data.end(), ::cuda::std::greater<T>());
|
|
typename thrust::device_vector<T>::iterator d_min =
|
|
thrust::max_element(d_data.begin(), d_data.end(), ::cuda::std::greater<T>());
|
|
|
|
ASSERT_EQUAL(h_min - h_data.begin(), d_min - d_data.begin());
|
|
}
|
|
DECLARE_VARIABLE_UNITTEST(TestMaxElement);
|
|
|
|
template <typename ForwardIterator>
|
|
ForwardIterator max_element(my_system& system, ForwardIterator first, ForwardIterator)
|
|
{
|
|
system.validate_dispatch();
|
|
return first;
|
|
}
|
|
|
|
void TestMaxElementDispatchExplicit()
|
|
{
|
|
thrust::device_vector<int> vec(1);
|
|
|
|
my_system sys(0);
|
|
thrust::max_element(sys, vec.begin(), vec.end());
|
|
|
|
ASSERT_EQUAL(true, sys.is_valid());
|
|
}
|
|
DECLARE_UNITTEST(TestMaxElementDispatchExplicit);
|
|
|
|
template <typename ForwardIterator>
|
|
ForwardIterator max_element(my_tag, ForwardIterator first, ForwardIterator)
|
|
{
|
|
*first = 13;
|
|
return first;
|
|
}
|
|
|
|
void TestMaxElementDispatchImplicit()
|
|
{
|
|
thrust::device_vector<int> vec(1);
|
|
|
|
thrust::max_element(thrust::retag<my_tag>(vec.begin()), thrust::retag<my_tag>(vec.end()));
|
|
|
|
ASSERT_EQUAL(13, vec.front());
|
|
}
|
|
DECLARE_UNITTEST(TestMaxElementDispatchImplicit);
|
|
|
|
void TestMaxElementWithBigIndexesHelper(int magnitude)
|
|
{
|
|
thrust::counting_iterator<long long> begin(1);
|
|
thrust::counting_iterator<long long> end = begin + (1ll << magnitude);
|
|
ASSERT_EQUAL(::cuda::std::distance(begin, end), 1ll << magnitude);
|
|
|
|
ASSERT_EQUAL(*thrust::max_element(thrust::device, begin, end), (1ll << magnitude));
|
|
}
|
|
|
|
void TestMaxElementWithBigIndexes()
|
|
{
|
|
TestMaxElementWithBigIndexesHelper(30);
|
|
#ifndef THRUST_FORCE_32_BIT_OFFSET_TYPE
|
|
TestMaxElementWithBigIndexesHelper(31);
|
|
TestMaxElementWithBigIndexesHelper(32);
|
|
TestMaxElementWithBigIndexesHelper(33);
|
|
#endif
|
|
}
|
|
DECLARE_UNITTEST(TestMaxElementWithBigIndexes);
|
|
|
|
void TestMaxElementCudaIterator()
|
|
{
|
|
auto pos = thrust::max_element(thrust::device, cuda::counting_iterator{0}, cuda::counting_iterator{0} + 100);
|
|
ASSERT_EQUAL(*pos, 99);
|
|
}
|
|
DECLARE_UNITTEST(TestMaxElementCudaIterator);
|