ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-16 19:09:59 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge commit '753b6227c507c45f4838971067f3cf158a392d55' into develop

Browse Source
This commit is contained in:
assistant-librarian[bot]
2025-08-13 23:12:01 +00:00
parent ffae627353
commit 9f8adac9f9
17 changed files with 239 additions and 257 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
example/ck_tile/39_copy/copy_basic.cpp
View File

@@ -54,7 +54,7 @@ bool run(const ck_tile::ArgParser& arg_parser)
x_buf.ToDevice(x_host.data());
// Define tile configuration
using Vector = ck_tile::sequence<1, 4>; // vector size along M and N dimension
using ThreadTile = ck_tile::sequence<1, 4>; // per-thread tile size along M and N
using WaveTile = ck_tile::sequence<64, 4>; // wave size along M and N dimension
using BlockWaves = ck_tile::sequence<4, 1>; // number of waves along M dimension
using BlockTile = ck_tile::sequence<512, 4>; // block size along M and N dimension
@@ -65,7 +65,7 @@ bool run(const ck_tile::ArgParser& arg_parser)
std::cout << "grid size (number of blocks per grid) " << kGridSize << std::endl;
// Define kernel types
using Shape = ck_tile::TileCopyShape<BlockWaves, BlockTile, WaveTile, Vector>;
using Shape = ck_tile::TileCopyShape<BlockWaves, BlockTile, WaveTile, ThreadTile>;
using Problem = ck_tile::TileCopyProblem<XDataType, Shape>;
using Policy = ck_tile::TileCopyPolicy<Problem>;
using Kernel = ck_tile::ElementWiseTileCopyKernel<Problem, Policy>;
@@ -88,8 +88,9 @@ bool run(const ck_tile::ArgParser& arg_parser)
<< " " << BlockTile::at(ck_tile::number<1>{}) << std::endl;
std::cout << "wave tile (number of elements per wave) " << WaveTile::at(ck_tile::number<0>{})
<< " " << WaveTile::at(ck_tile::number<1>{}) << std::endl;
std::cout << "vector (number of elements per thread) " << Vector::at(ck_tile::number<0>{})
<< " " << Vector::at(ck_tile::number<1>{}) << std::endl;
std::cout << "thread tile (number of elements per thread) "
<< ThreadTile::at(ck_tile::number<0>{}) << " " << ThreadTile::at(ck_tile::number<1>{})
<< std::endl;
std::cout << "WaveRepetitionPerBlock_M = " << Shape::WaveRepetitionPerBlock_M << " --> ("
<< Shape::Block_Tile_M << "/" << Shape::Waves_Per_Block_M << "*" << Shape::Wave_Tile_M
<< ")" << std::endl;