Skip to content

GitLab

Commit 7d50e8a5 authored by letaoqin's avatar letaoqin
Browse files

d0 use D0BlockTransferSrcScalarPerVector load data

parent 339b86e9
Show whitespace changes
Inline Side-by-side
Showing with 4 additions and 4 deletions
include/ck/tensor_operation/gpu/grid/gridwise_batched_mha_bwd_xdl_cshuffle_qloop_b2t_light_v1.hpp
View file @ 7d50e8a5
...@@ -1294,7 +1294,7 @@ struct GridwiseBatchedMultiheadAttentionBackward_Qloop_Xdl_CShuffle_Light_V1 ...@@ -1294,7 +1294,7 @@ struct GridwiseBatchedMultiheadAttentionBackward_Qloop_Xdl_CShuffle_Light_V1
Sequence<0, 1, 2, 4, 3, 5>, // DstDimAccessOrder Sequence<0, 1, 2, 4, 3, 5>, // DstDimAccessOrder
4, // SrcVectorDim 4, // SrcVectorDim
5, // DstVectorDim 5, // DstVectorDim
4, // SrcScalarPerVector D0BlockTransferSrcScalarPerVector, // SrcScalarPerVector
4, // DstScalarPerVector 4, // DstScalarPerVector
1, 1,
1, 1,
......
include/ck/tensor_operation/gpu/grid/gridwise_batched_mha_bwd_xdl_cshuffle_qloop_b2t_light_v2.hpp
View file @ 7d50e8a5
...@@ -1374,7 +1374,7 @@ struct GridwiseBatchedMultiheadAttentionBackward_Qloop_Xdl_CShuffle_Light_V2 ...@@ -1374,7 +1374,7 @@ struct GridwiseBatchedMultiheadAttentionBackward_Qloop_Xdl_CShuffle_Light_V2
Sequence<0, 1, 2, 4, 3, 5>, // DstDimAccessOrder Sequence<0, 1, 2, 4, 3, 5>, // DstDimAccessOrder
4, // SrcVectorDim 4, // SrcVectorDim
5, // DstVectorDim 5, // DstVectorDim
4, // SrcScalarPerVector D0BlockTransferSrcScalarPerVector, // SrcScalarPerVector
4, // DstScalarPerVector 4, // DstScalarPerVector
1, 1,
1, 1,
......
include/ck/tensor_operation/gpu/grid/gridwise_batched_mha_bwd_xdl_cshuffle_qloop_b2t_v1.hpp
View file @ 7d50e8a5
...@@ -1362,7 +1362,7 @@ struct GridwiseBatchedMultiheadAttentionBackward_Qloop_Xdl_CShuffle_V1 ...@@ -1362,7 +1362,7 @@ struct GridwiseBatchedMultiheadAttentionBackward_Qloop_Xdl_CShuffle_V1
Sequence<0, 1, 2, 4, 3, 5>, // DstDimAccessOrder Sequence<0, 1, 2, 4, 3, 5>, // DstDimAccessOrder
4, // SrcVectorDim 4, // SrcVectorDim
5, // DstVectorDim 5, // DstVectorDim
4, // SrcScalarPerVector D0BlockTransferSrcScalarPerVector, // SrcScalarPerVector
4, // DstScalarPerVector 4, // DstScalarPerVector
1, 1,
1, 1,
......
include/ck/tensor_operation/gpu/grid/gridwise_batched_mha_bwd_xdl_cshuffle_qloop_b2t_v2.hpp
View file @ 7d50e8a5
...@@ -1428,7 +1428,7 @@ struct GridwiseBatchedMultiheadAttentionBackward_Qloop_Xdl_CShuffle_V2 ...@@ -1428,7 +1428,7 @@ struct GridwiseBatchedMultiheadAttentionBackward_Qloop_Xdl_CShuffle_V2
Sequence<0, 1, 2, 4, 3, 5>, // DstDimAccessOrder Sequence<0, 1, 2, 4, 3, 5>, // DstDimAccessOrder
4, // SrcVectorDim 4, // SrcVectorDim
5, // DstVectorDim 5, // DstVectorDim
4, // SrcScalarPerVector D0BlockTransferSrcScalarPerVector, // SrcScalarPerVector
4, // DstScalarPerVector 4, // DstScalarPerVector
1, 1,
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