ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-07-15 11:34:54 +00:00
Code Issues Packages Projects Releases Wiki Activity

tempsave

Browse Source
This commit is contained in:
aska-0096
2024-09-03 07:53:04 +00:00
parent 5d9c964ece
commit 7c8e92face
2 changed files with 12 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
example/01_gemm/gemm_xdl_bf16_v3.cpp
View File

@@ -28,13 +28,13 @@ using DeviceGemmV2Instance =
ADataType, BDataType, CDataType, AccDataType, CShuffleDataType,
PassThrough, PassThrough, PassThrough, GemmDefault,
256,
128, 128,
224, 256,
64, 8, 1,
16, 16,
4, 4,
7, 8,
S<8, 32, 1>, S<1, 0, 2>, S<1, 0, 2>,
2, 8, 8, 0,
S<16, 16, 1>, S<0, 2, 1>, S<0, 2, 1>,
S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>,
1, 8, 8, 1,
1, 2, S<1, 32, 1, 8>, 8,
ck::BlockGemmPipelineScheduler::Intrawave,ck::BlockGemmPipelineVersion::v3>;

12
include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v3.hpp
View File

@@ -752,12 +752,18 @@ struct GridwiseGemm_xdl_cshuffle_v3
__device__ static constexpr auto GetBBlockDescriptor_BK0PerBlock_NPerBlock_BK1()
{
// B matrix in LDS memory, dst of blockwise copy
if constexpr(BBlockLdsExtraN)
// if constexpr(BBlockLdsExtraN)
// {
// return make_naive_tensor_descriptor(
// make_tuple(BK0Number, Number<NPerBlock>{}, BK1Number),
// make_tuple(BK1Number, Number<KPerBlock + BBlockLdsExtraN>{}, I1));
// }
// else
if constexpr(BBlockLdsExtraN && is_same<tensor_layout::gemm::RowMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(
make_tuple(BK0Number, Number<NPerBlock>{}, BK1Number),
// make_tuple(BK1Number, Number<KPerBlock + BBlockLdsExtraN>{}, I1));
make_tuple(BK1Number, Number<KPerBlock>{}, I1));
make_tuple(BK1Number * Number<NPerBlock>{}, I1, Number<NPerBlock>{}));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, BLayout>::value)
{