[PATCH] D129757: [RISCV] Optimize SELECT_CC when the true value of select is Constant

[Alex Bradbury via Phabricator via llvm-commits]

asb added a comment.

Did you investigate any cases where this results in slightly worse codegen? one example would be 20100416-1.c from the GCC torture suite (-march=rv64imafdc -mabi=lp64d -O3). Overall the change seems to be positive, but there do seem to be a few cases like this where enabling this transformation results in an increase in instructions.

It needn't necessarily block this patch, as it does seem like a sensible transformation to make - but it would be useful to understand why these regressions occur, and capture some representative test cases.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D129757/new/

https://reviews.llvm.org/D129757