ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-12 09:16:52 +00:00
Code Issues Packages Projects Releases Wiki Activity

Improve parallelism in the benchmark

Browse Source
This commit is contained in:
Damien Lejeune
2026-02-09 13:16:21 +00:00
parent ec1e8ec58e
commit 8b59a1e192
2 changed files with 11 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
example/ck_tile/42_mhc/mhc_v3_bf16_benchmark.cpp
View File

@@ -94,9 +94,10 @@ bool run_mhc_benchmark(const ck_tile::ArgParser& arg_parser)
d_phi_mem.ToDevice(h_phi.data());
d_output_mem.ToDevice(h_output.data());
// Define block shape - 128 threads (2 warps) to match BlockGemmShape configuration
using BlockShape = ck_tile::Generic2dBlockShape<ck_tile::sequence<1, 128>,
ck_tile::sequence<1, 128>,
// Define block shape - 64 threads (1 warp) to match BlockGemmShape configuration
// This matches a 16x16 block tile with 1 warp (1x1 warp layout)
using BlockShape = ck_tile::Generic2dBlockShape<ck_tile::sequence<1, 64>,
ck_tile::sequence<1, 64>,
ck_tile::sequence<1, 1>>;
using Problem = ck_tile::MHCProblem<XDataType, ComputeDataType, YDataType, BlockShape>;