[PATCH] D153151: [EarlyCSE] Do not CSE convergent calls with memory effects

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added inline comments.


================
Comment at: llvm/lib/Transforms/Scalar/EarlyCSE.cpp:336
   // implicitly depend on the set of threads that is currently executing.
-  if (CallInst *CI = dyn_cast<CallInst>(Inst); CI && CI->isConvergent()) {
-    return hash_combine(
-        Inst->getOpcode(), Inst->getParent(),
-        hash_combine_range(Inst->value_op_begin(), Inst->value_op_end()));
-  }
+  if (CallInst *CI = dyn_cast<CallInst>(Inst))
+    return hashCallInst(CI);
----------------
CallBase? Are invokes ever CSEable?


================
Comment at: llvm/test/Transforms/EarlyCSE/AMDGPU/convergent-call.ll:27
 if:
   ; %y = ballot operation over lanes satisfying %cond.
   %y1 = call i32 @llvm.read_register.i32(metadata !{!"exec_lo"}) convergent
----------------
Should still add another test that doesn't involve read_register and uses a proper convergent call


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D153151/new/

https://reviews.llvm.org/D153151