This is just for consistency with FLAT_Global_Atomic_Pseudo.
Details
Details
- Reviewers
piotr rampitec - Group Reviewers
Restricted Project - Commits
- rGd6c3d7924fee: [AMDGPU] Refactor multiclass FLAT_Atomic_Pseudo. NFC.
Diff Detail
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
llvm/lib/Target/AMDGPU/FLATInstructions.td | ||
---|---|---|
469 | This can be just a class now? |
llvm/lib/Target/AMDGPU/FLATInstructions.td | ||
---|---|---|
469 | Yes, and I did try that, but I preferred the consistency of having everything here be a multiclass. Note that both FLAT_Global_Atomic_Pseudo_NO_RTN and FLAT_Global_Atomic_Pseudo_RTN have to be multiclasses because they have to define additional _SADDR variants. Do you feel strongly that it should be a class? |
This can be just a class now?