Details
- Reviewers
arsenm - Commits
- rG7d16a22eb0c1: [AMDGPU] Peephole adjacent equivalent S_SET_GPR_IDX_ON
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
llvm/lib/Target/AMDGPU/SIPreEmitPeephole.cpp | ||
---|---|---|
225–228 | I'm confused by this usage of threshold. You never break from the loop on it and it's not part of the loop condition |
llvm/lib/Target/AMDGPU/SIPreEmitPeephole.cpp | ||
---|---|---|
225–228 | Right, the block will be scanned once. However, it will not send two S_SET_GPR_IDX_ON instructions into the optimizeSetGPR() if they are too far apart. That is the loop from the first to next S_SET_GPR_IDX_ON in the optimizeSetGPR() which performs expensive checks. |
llvm/lib/Target/AMDGPU/SIPreEmitPeephole.cpp | ||
---|---|---|
225–228 | This could uses a comment |
Register() instead of NoRegister. When does the NoRegister case happen?