Skip to content

GitLab

Commit 8f520532 authored by Jing Zhang's avatar Jing Zhang
Browse files

fixed

parent fb7b4609
Hide whitespace changes
Inline Side-by-side
Showing with 9 additions and 9 deletions
composable_kernel/include/tensor_operation/gridwise_gemm_dlops_v2.hpp
View file @ 8f520532
...@@ -118,8 +118,8 @@ template <index_t BlockSize, ...@@ -118,8 +118,8 @@ template <index_t BlockSize,
index_t HoPerThread, index_t HoPerThread,
index_t WoPerThread, index_t WoPerThread,
index_t EPerThread, index_t EPerThread,
typename ABlockTransferThreadSliceLengths_E_K, typename ABlockTransferThreadSliceLengths_E0_E1_K,
typename ABlockTransferThreadClusterLengths_E_K, typename ABlockTransferThreadClusterLengths_E0_E1_K,
typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder, typename ABlockTransferSrcAccessOrder,
index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcVectorDim,
...@@ -187,11 +187,11 @@ struct GridwiseGemmDlops_km_kn_mn_v3 ...@@ -187,11 +187,11 @@ struct GridwiseGemmDlops_km_kn_mn_v3
// const auto K = a_e0_e1_k_global_desc.GetLength(I1); // const auto K = a_e0_e1_k_global_desc.GetLength(I1);
// const auto N = b_e0_e1_n_ho_wo_global_desc.GetLength(I1); // const auto N = b_e0_e1_n_ho_wo_global_desc.GetLength(I1);
const auto Ho = b_e0_e1_n_ho_wo_global_desc.GetLength(I2); const auto Ho = b_e0_e1_n_ho_wo_global_desc.GetLength(I3);
const auto Wo = b_e0_e1_n_ho_wo_global_desc.GetLength(I3); const auto Wo = b_e0_e1_n_ho_wo_global_desc.GetLength(I4);
// divide block work by [M, N] // divide block work by [M, N]
#if 0 #if 1
const auto ho_block_work_num = Ho / Number<HoPerBlock>{}; const auto ho_block_work_num = Ho / Number<HoPerBlock>{};
const auto wo_block_work_num = Wo / Number<WoPerBlock>{}; const auto wo_block_work_num = Wo / Number<WoPerBlock>{};
const auto hwo_block_work_num = ho_block_work_num * wo_block_work_num; const auto hwo_block_work_num = ho_block_work_num * wo_block_work_num;
...@@ -269,17 +269,17 @@ struct GridwiseGemmDlops_km_kn_mn_v3 ...@@ -269,17 +269,17 @@ struct GridwiseGemmDlops_km_kn_mn_v3
BlockwiseTensorSliceTransfer_v4<BlockSize, BlockwiseTensorSliceTransfer_v4<BlockSize,
InMemoryDataOperationEnum_t::Set, InMemoryDataOperationEnum_t::Set,
Sequence<I1, E1, KPerBlock>, Sequence<I1, E1, KPerBlock>,
ABlockTransferThreadSliceLengths_E_K, ABlockTransferThreadSliceLengths_E0_E1_K,
ABlockTransferThreadClusterLengths_E_K, ABlockTransferThreadClusterLengths_E0_E1_K,
ABlockTransferThreadClusterArrangeOrder, ABlockTransferThreadClusterArrangeOrder,
FloatAB, FloatAB,
FloatAB, FloatAB,
decltype(a_e0_e1_k_global_desc), decltype(a_e0_e1_k_global_desc),
decltype(a_e0_e1_k_block_desc), decltype(a_e0_e1_k_block_desc),
ABlockTransferSrcAccessOrder, ABlockTransferSrcAccessOrder,
Sequence<0, 1, 2>, Sequence<0, 1, 2>, // ABlockTransferDstAccessOrder
ABlockTransferSrcVectorDim, ABlockTransferSrcVectorDim,
2, 2, // ABlockTransferDstVectorDim
ABlockTransferSrcScalarPerVector, ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_K, ABlockTransferDstScalarPerVector_K,
1, 1,
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment