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
35 lines
1.1 KiB
Plaintext
35 lines
1.1 KiB
Plaintext
#include <thrust/copy.h>
|
|
#include <thrust/device_vector.h>
|
|
#include <thrust/functional.h>
|
|
#include <thrust/iterator/counting_iterator.h>
|
|
|
|
#include <iostream>
|
|
#include <iterator>
|
|
|
|
int main()
|
|
{
|
|
// this example computes indices for all the nonzero values in a sequence
|
|
|
|
// sequence of zero and nonzero values
|
|
thrust::device_vector<int> stencil{0, 1, 1, 0, 0, 1, 0, 1};
|
|
|
|
// storage for the nonzero indices
|
|
thrust::device_vector<int> indices(8);
|
|
|
|
// counting iterators define a sequence [0, 8)
|
|
thrust::counting_iterator<int> first(0);
|
|
thrust::counting_iterator<int> last = first + 8;
|
|
|
|
// compute indices of nonzero elements
|
|
using IndexIterator = thrust::device_vector<int>::iterator;
|
|
|
|
IndexIterator indices_end = thrust::copy_if(first, last, stencil.begin(), indices.begin(), cuda::std::identity{});
|
|
// indices now contains [1,2,5,7]
|
|
|
|
// print result
|
|
std::cout << "found " << cuda::std::distance(indices.begin(), indices_end) << " nonzero values at indices:\n";
|
|
thrust::copy(indices.begin(), indices_end, std::ostream_iterator<int>(std::cout, "\n"));
|
|
|
|
return 0;
|
|
}
|