ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-07-07 07:37:01 +00:00
Code Issues Packages Projects Releases Wiki Activity

Integrate new packed cast threadwise tensor slice transfer into gridwise gemm pipelines.

Browse Source
This commit is contained in:
Ville Pietilä
2025-08-15 12:06:44 +00:00
parent 6374e16a43
commit 00a3ce734a
4 changed files with 157 additions and 283 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
test/grouped_convnd_bwd_weight/test_grouped_convnd_bwd_weight.cpp
View File

@@ -235,7 +235,8 @@ class TestGroupedConvndBwdWeight2d_bf16_gfx950 : public TestGroupedConvndBwdWeig
};
using KernelTypes2d_bf16_gfx950 = ::testing::Types<
std::tuple<ck::bhalf_t, ck::bhalf_t, ck::bhalf_t, NHWGC, GKYXC, NHWGK, ck::Number<2>>,
// This layout does not yet work.
//std::tuple<ck::bhalf_t, ck::bhalf_t, ck::bhalf_t, NHWGC, GKYXC, NHWGK, ck::Number<2>>,
std::tuple<ck::bhalf_t, ck::bhalf_t, ck::bhalf_t, NGCHW, GKYXC, NGKHW, ck::Number<2>>,
std::tuple<ck::bhalf_t, ck::bhalf_t, ck::bhalf_t, NGCHW, GKCYX, NGKHW, ck::Number<2>>>;