Fix Copy_Atom type mismatch in sgemm_sm80.cu (#2582)

This commit is contained in:
Lifu Huang
2025-09-04 16:56:17 -07:00
committed by GitHub
parent 2288c0c901
commit b6ccf34aef

View File

@ -496,8 +496,8 @@ gemm_nt(int m, int n, int k,
size(ceil_div(N, bN)));
gemm_device<<<dimGrid, dimBlock, smem_size, stream>>>
(prob_shape, cta_tiler,
A, dA, sA, copyA, AutoVectorizingCopy{},
B, dB, sB, copyB, AutoVectorizingCopy{},
A, dA, sA, copyA, Copy_Atom<AutoVectorizingCopy, TA>{},
B, dB, sB, copyB, Copy_Atom<AutoVectorizingCopy, TB>{},
C, dC, sC, mmaC,
alpha, beta);
}
@ -573,8 +573,8 @@ gemm_tn(int m, int n, int k,
size(ceil_div(N, bN)));
gemm_device<<<dimGrid, dimBlock, smem_size, stream>>>
(prob_shape, cta_tiler,
A, dA, sA, copyA, AutoVectorizingCopy{},
B, dB, sB, copyB, AutoVectorizingCopy{},
A, dA, sA, copyA, Copy_Atom<AutoVectorizingCopy, TA>{},
B, dB, sB, copyB, Copy_Atom<AutoVectorizingCopy, TB>{},
C, dC, sC, mmaC,
alpha, beta);
}