ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-03 13:11:25 +00:00
Code Issues Packages Projects Releases Wiki Activity

Work around develop validation failure (#513)

Browse Source 
* workaround bf16 atten fwd issue on gfx908

* typo
This commit is contained in:
Anthony Chang
2022-11-18 00:38:13 +08:00
committed by GitHub
parent 4c4c7328a6
commit 892a8d769d
3 changed files with 16 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
include/ck/tensor_operation/gpu/grid/gridwise_batched_gemm_softmax_gemm_xdl_cshuffle_v1.hpp
View File

@@ -874,6 +874,14 @@ struct GridwiseBatchedGemmSoftmaxGemm_Xdl_CShuffle
}
} // end gemm1
// workaround compiler issue; see ck/ck.hpp
if constexpr(CK_WORKAROUND_SWDEV_XXXXXX_BF16_ATTEN_FWD_GFX908_ISSUE == 1 &&
is_same_v<FloatAB, bhalf_t> && MPerBlock == 256 && NPerBlock == 128 &&
Gemm1NPerBlock == 128)
{
__builtin_amdgcn_sched_barrier(0);
}
constexpr auto c_thread_desc_m0_n0_m1_n1_m2_n2_n3_n4 =
gemm1_blockwise_gemm.GetCThreadDescriptor_M0_N0_M1_N1_M2_N2_N3_N4();
constexpr auto cm0 = c_thread_desc_m0_n0_m1_n1_m2_n2_n3_n4.GetLength(I0);