ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-07-11 01:27:34 +00:00
Code Issues Packages Projects Releases Wiki Activity

fixed building errors

Browse Source
This commit is contained in:
Jakub Piasecki
2026-01-30 19:22:34 +00:00
parent ae2d2d9f2c
commit 2086516deb
2 changed files with 5 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_bwd_data_multiple_d_xdl_cshuffle_v3.hpp
View File

@@ -71,7 +71,7 @@ __launch_bounds__(CK_MAX_THREAD_PER_BLOCK, MinimumOccupancy)
const long_index_t e_batch_offset =
amd_wave_read_first_lane(compute_ptr_offset_of_batch.GetEPtrOffset(g_idx));
__shared__ char p_shared[GridwiseGemm::GetSharedMemoryNumberOfByte(get_device_arch())];
__shared__ char p_shared[GridwiseGemm::GetSharedMemoryNumberOfByte()];
index_t left = 0;
index_t right = gemms_count;

8
include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_conv_v3.hpp
View File

@@ -705,12 +705,12 @@ struct GridwiseGemm_xdl_cshuffle_conv_v3
BDataType,
ComputeTypeA,
AccDataType,
decltype(GetABlockDescriptor_AK0PerBlock_MPerBlock_AK1(get_device_arch())),
decltype(GetBBlockDescriptor_BK0PerBlock_NPerBlock_BK1(get_device_arch())),
decltype(GetABlockDescriptor_AK0PerBlock_MPerBlock_AK1()),
decltype(GetBBlockDescriptor_BK0PerBlock_NPerBlock_BK1()),
decltype(MakeAMmaTileDescriptor_M0_M1_M2_K(
GetABlockDescriptor_AK0PerBlock_MPerBlock_AK1(get_device_arch()))),
GetABlockDescriptor_AK0PerBlock_MPerBlock_AK1())),
decltype(MakeBMmaTileDescriptor_N0_N1_N2_K(
GetBBlockDescriptor_BK0PerBlock_NPerBlock_BK1(get_device_arch()))),
GetBBlockDescriptor_BK0PerBlock_NPerBlock_BK1())),
ABlockTransferSrcScalarPerVector,
BBlockTransferSrcScalarPerVector,
MPerBlock,