[TESTING] [DO NOT MERGE] Triton perf improvements for release/2.8#2676

iupaikov-amd wants to merge 2 commits intorelease/2.8ROCm/pytorch:release/2.8from

iupaikov_charai_perf_release2.8ROCm/pytorch:iupaikov_charai_perf_release2.8Copy head branch name to clipboard