[PATCH] D82709: [MachineLICM] [PowerPC] hoisting rematerializable cheap instructions based on register pressure.

[ChenZheng via Phabricator via llvm-commits]

shchenz added a comment.

OK, I will look into RA and see what's wrong here on PowerPC target. It seems RA can not sink any remat instructions at all on PowerPC as the case here should be very common.

Also thanks very much for your info about how ARM handles big IMM. To be honest, that's my most favorite solution either when I come to this issue. On PowerPC we expand the big IMM in ISEL in different ways for different IMM. This will expand a remat instruction to 2 or more non-remat instructions. But I gave up because of the big effort to refactor these.

Anyway, thanks very much for your good comments.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D82709/new/

https://reviews.llvm.org/D82709