[PATCH] D139422: [AMDGPU] Accelerate SIRegisterInfo::getPhysRegClass

[Jay Foad via Phabricator via llvm-commits]

foad added inline comments.


================
Comment at: llvm/lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp:1433
   auto RC =
-      TRI.getPhysRegClass(cast<RegisterSDNode>(Val.getOperand(1))->getReg());
+      TRI.getPhysRegBaseClass(cast<RegisterSDNode>(Val.getOperand(1))->getReg());
   return RC && TRI.isSGPRClass(RC);
----------------
This will sometimes call getPhysRegBaseClass on a virtual register, which I think should be disallowed.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D139422/new/

https://reviews.llvm.org/D139422