[AMD] Use warp shuffle for MFMA to Dot operand layout conversion (FP8)#5139

antiagainst merged 11 commits into

triton-lang:maintriton-lang/triton:mainfrom

ilia-cher:amd_mfma_dot_fp8_cvtCopy head branch name to clipboard