[PATCH] D153007: [AMDGPU] Switch to the new cl option amdgpu-atomic-optimizer-strategy.

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added inline comments.


================
Comment at: llvm/lib/Target/AMDGPU/AMDGPUAtomicOptimizer.cpp:164
+  // Scan option None disables the Pass
+  assert(ScanImpl != ScanOptions::None);
+
----------------
Just early return, running the pass standalone should work


================
Comment at: llvm/lib/Target/AMDGPU/AMDGPUAtomicOptimizer.cpp:867
             B.CreateIntrinsic(Intrinsic::amdgcn_strict_wwm, Ty, ExclScan);
-      } else {
+      } else if (ScanImpl == ScanOptions::Iterative) {
         LaneOffset = ExclScan;
----------------
So the pass behavior with none is to just crash?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D153007/new/

https://reviews.llvm.org/D153007