Slightly improving 4 bit mat mul performance through better engage ALU pipes by splitting uint to float conversion operation.#15447

Merged

meta-codesync[bot] merged 1 commit intopytorch:mainfrom

trivedivivek:export-D85779855

Nov 1, 2025

Commits on Nov 1, 2025

Slightly improving 4 bit mat mul performance through better engage ALU pipes by splitting uint to float conversion operation. (#15447)

trivedivivek
authored and
facebook-github-bot
committed