cutlass
cutlass copied to clipboard
[BUG] cute SM75 mma instruction missing
Which component has the problem?
CUTLASS C++
Bug Report
SM75 supports mma.sync.aligned.m16n8k8.row.col.f16.f16.f16.f16 ptx instruction, but cute/arch/mma_sm75.hpp doesn't include this instruction (while cutlass/arch/mma_sm75.h does). Is this a bug?
feel free to open a pr to add it