Merge commit 'f240ae32487219b4dd9d3152b816f87166e20feb' into develop

This commit is contained in:
assistant-librarian[bot]
2025-07-07 18:07:20 +00:00
parent f8ee69963d
commit 5c05211959
12 changed files with 225 additions and 143 deletions

View File

@@ -53,16 +53,17 @@ bool run(const ck_tile::ArgParser& arg_parser)
x_buf.ToDevice(x_host.data());
using BlockWaves = ck_tile::sequence<2, 1>;
using BlockTile = ck_tile::sequence<64, 8>;
using WaveTile = ck_tile::sequence<64, 8>;
using Vector = ck_tile::sequence<1, 4>;
using BlockWaves = ck_tile::sequence<2, 1>;
using BlockTile = ck_tile::sequence<64, 8>;
using WaveTile = ck_tile::sequence<64, 8>;
using Vector = ck_tile::sequence<1, 2>;
constexpr bool AsyncCopy = true;
ck_tile::index_t kGridSize = (m / BlockTile::at(ck_tile::number<0>{}));
std::cout << "grid size " << kGridSize << std::endl;
using Shape = ck_tile::TileCopyShape<BlockWaves, BlockTile, WaveTile, Vector>;
using Problem = ck_tile::TileCopyProblem<XDataType, Shape>;
using Problem = ck_tile::TileCopyProblem<XDataType, Shape, AsyncCopy>;
using Kernel = ck_tile::TileCopy<Problem>;
constexpr ck_tile::index_t kBlockSize = 128;