cub::WarpMergeSort
Defined in cub/warp/warp_merge_sort.cuh
-
template<typename KeyT, int ITEMS_PER_THREAD, int LOGICAL_WARP_THREADS = detail::warp_threads, typename ValueT = NullType>
class WarpMergeSort : public cub::BlockMergeSortStrategy<KeyT, NullType, detail::warp_threads, ITEMS_PER_THREAD, WarpMergeSort<KeyT, ITEMS_PER_THREAD, detail::warp_threads, NullType>> The WarpMergeSort class provides methods for sorting items partitioned across a CUDA warp using a merge sorting method.
Overview
WarpMergeSort arranges items into ascending order using a comparison functor with less-than semantics. Merge sort can handle arbitrary types and comparison functors.
A Simple Example
The code snippet below illustrates a sort of 64 integer keys that are partitioned across 16 threads where each thread owns 4 consecutive items.
#include <cub/cub.cuh> // or equivalently <cub/warp/warp_merge_sort.cuh> struct CustomLess { template <typename DataType> __device__ bool operator()(const DataType &lhs, const DataType &rhs) { return lhs < rhs; } }; __global__ void ExampleKernel(...) { constexpr int warp_threads = 16; constexpr int block_threads = 256; constexpr int items_per_thread = 4; constexpr int warps_per_block = block_threads / warp_threads; const int warp_id = static_cast<int>(threadIdx.x) / warp_threads; // Specialize WarpMergeSort for a virtual warp of 16 threads // owning 4 integer items each using WarpMergeSortT = cub::WarpMergeSort<int, items_per_thread, warp_threads>; // Allocate shared memory for WarpMergeSort __shared__ typename WarpMergeSortT::TempStorage temp_storage[warps_per_block]; // Obtain a segment of consecutive items that are blocked across threads int thread_keys[items_per_thread]; // ... WarpMergeSortT(temp_storage[warp_id]).Sort(thread_keys, CustomLess()); // ... }
Suppose the set of input
thread_keysacross a warp of threads is{ [0,64,1,63], [2,62,3,61], [4,60,5,59], ..., [31,34,32,33] }. The corresponding outputthread_keysin those threads will be{ [0,1,2,3], [4,5,6,7], [8,9,10,11], ..., [31,32,33,34] }.- Template Parameters
KeyT – Key type
ITEMS_PER_THREAD – The number of items per thread
LOGICAL_WARP_THREADS – [optional] The number of threads per “logical” warp (may be less than the number of hardware warp threads). Default is the warp size of the targeted CUDA compute-capability (e.g., 32 threads for SM86). Must be a power of two.
ValueT – [optional] Value type (default: cub::NullType, which indicates a keys-only sort)