ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-04-19 22:39:03 +00:00
Code Issues Packages Projects Releases Wiki Activity

Batched gemm softmax gemm descriptor fix (#3564)

Browse Source 
* Add rocm to prefix path for codegen

* Fix issue with c0_matrix_mask construction
This commit is contained in:
music-dino
2026-01-20 16:25:30 +01:00
committed by GitHub
parent b09121f860
commit 6300ad3c62
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

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

@@ -1059,7 +1059,7 @@ struct DeviceBatchedGemmSoftmaxGemm_Xdl_CShuffle
c_grid_desc_m_n)},
has_main_k_block_loop{GridwiseGemm64::CalculateHasMainKBlockLoop(
a_grid_desc_ak0_m_ak1.GetLength(I0) * a_grid_desc_ak0_m_ak1.GetLength(I2))},
c0_matrix_mask{c.GetLength(I1)},
c0_matrix_mask{b.GetLength(I0)},
a_element_op{a_element_op_},
b_element_op{b_element_op_},
b1_element_op{b1_element_op_},