[PATCH] D97547: [AMDGPU] Better codegen for i64 bitreverse

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added inline comments.


================
Comment at: llvm/lib/Target/AMDGPU/SIInstrInfo.cpp:5746
+    case AMDGPU::S_BREV_B64:
+      splitScalar64BitUnaryOp(Worklist, Inst, AMDGPU::S_BREV_B32, true);
+      Inst.eraseFromParent();
----------------
I don't see how you can just replace this with the two reversed halves. That would give you abcd -> badc instead of dcba


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D97547/new/

https://reviews.llvm.org/D97547