Files
project_6/cccl_upstream/thrust/testing/partition_point.cu
EngineX CI 56fd68e7dd [INFRA] Import NVIDIA/CCCL upstream as optimization reference library
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
2026-07-30 09:35:51 +00:00

118 lines
3.0 KiB
Plaintext

#include <thrust/functional.h>
#include <thrust/iterator/retag.h>
#include <thrust/partition.h>
#include <unittest/unittest.h>
template <typename T>
struct is_even
{
_CCCL_HOST_DEVICE bool operator()(T x) const
{
return ((int) x % 2) == 0;
}
};
template <typename Vector>
void TestPartitionPointSimple()
{
using Iterator = typename Vector::iterator;
Vector v{1, 1, 1, 0};
Iterator first = v.begin();
Iterator last = v.begin() + 4;
Iterator ref = first + 3;
ASSERT_EQUAL_QUIET(ref, thrust::partition_point(first, last, ::cuda::std::identity{}));
last = v.begin() + 3;
ref = last;
ASSERT_EQUAL_QUIET(ref, thrust::partition_point(first, last, ::cuda::std::identity{}));
}
DECLARE_VECTOR_UNITTEST(TestPartitionPointSimple);
template <class Vector>
void TestPartitionPoint()
{
using T = typename Vector::value_type;
using Iterator = typename Vector::iterator;
const size_t n = (1 << 16) + 13;
Vector v = unittest::random_integers<T>(n);
Iterator ref = thrust::stable_partition(v.begin(), v.end(), is_even<T>());
ASSERT_EQUAL(ref - v.begin(), thrust::partition_point(v.begin(), v.end(), is_even<T>()) - v.begin());
}
DECLARE_INTEGRAL_VECTOR_UNITTEST(TestPartitionPoint);
template <typename ForwardIterator, typename Predicate>
ForwardIterator partition_point(my_system& system, ForwardIterator first, ForwardIterator, Predicate)
{
system.validate_dispatch();
return first;
}
void TestPartitionPointDispatchExplicit()
{
thrust::device_vector<int> vec(1);
my_system sys(0);
thrust::partition_point(sys, vec.begin(), vec.begin(), 0);
ASSERT_EQUAL(true, sys.is_valid());
}
DECLARE_UNITTEST(TestPartitionPointDispatchExplicit);
template <typename ForwardIterator, typename Predicate>
ForwardIterator partition_point(my_tag, ForwardIterator first, ForwardIterator, Predicate)
{
*first = 13;
return first;
}
void TestPartitionPointDispatchImplicit()
{
thrust::device_vector<int> vec(1);
thrust::partition_point(thrust::retag<my_tag>(vec.begin()), thrust::retag<my_tag>(vec.begin()), 0);
ASSERT_EQUAL(13, vec.front());
}
DECLARE_UNITTEST(TestPartitionPointDispatchImplicit);
struct test_less_than
{
long long expected;
_CCCL_DEVICE bool operator()(long long y)
{
return y < expected;
}
};
void TestPartitionPointWithBigIndexesHelper(int magnitude)
{
thrust::counting_iterator<long long> begin(0);
thrust::counting_iterator<long long> end = begin + (1ll << magnitude);
ASSERT_EQUAL(::cuda::std::distance(begin, end), 1ll << magnitude);
test_less_than fn = {(1ll << magnitude) - 17};
ASSERT_EQUAL(::cuda::std::distance(begin, thrust::partition_point(thrust::device, begin, end, fn)),
(1ll << magnitude) - 17);
}
#ifndef THRUST_FORCE_32_BIT_OFFSET_TYPE
void TestPartitionPointWithBigIndexes()
{
TestPartitionPointWithBigIndexesHelper(30);
TestPartitionPointWithBigIndexesHelper(31);
TestPartitionPointWithBigIndexesHelper(32);
TestPartitionPointWithBigIndexesHelper(33);
}
DECLARE_UNITTEST(TestPartitionPointWithBigIndexes);
#endif // THRUST_FORCE_32_BIT_OFFSET_TYPE