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

[ChenZheng via Phabricator via llvm-commits]

shchenz added a comment.

Hi @qcolombet , I saw you added comments for other patch about Greedy Register Pressure like "The greedy allocator is already very complicated". I am not sure my proposal change in above comment in greedyRA will increase the complexity or if it is the right way to go. Could you please help to confirm this? Thanks.

> We should add logic inside selectOrSplitImpl like: when we try to spill a non-remat instruction, we should check if there is any remat instruction assigned with a physical register. If there is one, we should first remat the remat instruction and assign the physical register to the spilled one?

If we can not depend on register allocator to rematerialize all required rematerializable instructions, the way proposed in this PR should be reasonable. Instead of letting the register allocator rematerialize the instructions after we hoist all rematerializable instructions without considering register pressure in machineLICM, now we hoist the rematerializable instructions also based on register pressure.

Sorry for pinging this patch for a long time. We need this patch on PowerPC target. I think it should also benefit other targets.


Repository:
  rG LLVM Github Monorepo

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

https://reviews.llvm.org/D82709