[PATCH] D78132: [PowerPC] Refactor PPCInstrVSX.td

[Jinsong Ji via Phabricator via llvm-commits]

jsji added a comment.

> I think we need to be a little more selective here. If we set that to all instructions that modify the FPSCR, that would mean it has to be set on all arithmetic, conversion/rounding and comparison operations. This would exclude all of them from being considered by LICM. Setting it on comparisons seems reasonable. But I don't think we should set it for anything that modifies the FPSCR - we can model whatever aspects need to be modeled rather than just throwing our hands up and saying "It has unmodeled side effects."

Yes, agree. But before we model what needs to be modeled, maybe don't unset the flag in this refactor patch?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D78132/new/

https://reviews.llvm.org/D78132