ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-04-19 22:39:03 +00:00
Code Issues Packages Projects Releases Wiki Activity

[CK_TILE] Adjust kBlockSize of reduce example for better perf (#1779)

Browse Source 
* Observed a 2x perf improvement with kBlockSize = 256
* Using 512 threads may lead to redundant computations
This commit is contained in:
ClementLinCF
2025-01-13 12:50:32 +08:00
committed by GitHub
parent 3d50f57f43
commit 0b8f117f1a
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
example/ck_tile/05_reduce/reduce.cpp
View File

@@ -52,7 +52,7 @@ bool run(const ck_tile::ArgParser& arg_parser)
// using WarpTile = ck_tile::sequence<1, 512>;
// using Vector = ck_tile::sequence<1, 8>;
constexpr ck_tile::index_t kBlockSize = 512;
constexpr ck_tile::index_t kBlockSize = 256;
constexpr ck_tile::index_t kBlockPerCu = 1;
ck_tile::index_t kGridSize = (m / BlockTile::at(ck_tile::number<0>{}));
std::cout << "grid size " << kGridSize << std::endl;