[PATCH] D127609: [AtomicExpand] Make floating point conversion happens before fence insertion

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added inline comments.


================
Comment at: llvm/lib/CodeGen/AtomicExpandPass.cpp:229
+      I = LI = convertAtomicLoadToIntegerType(LI);
+      assert(LI->getType()->isIntegerTy() && "invariant broken");
+      MadeChange = true;
----------------
Don't see much point in this assert (plus the others)


================
Comment at: llvm/lib/CodeGen/AtomicExpandPass.cpp:253-254
+      // extend convertCmpXchgToInteger for floating point too.
+      assert(!CASI->getCompareOperand()->getType()->isFloatingPointTy() &&
+             "unimplemented - floating point not legal at IR level");
+      if (CASI->getCompareOperand()->getType()->isPointerTy()) {
----------------
There's no reason to assert on things that fail the verifier


================
Comment at: llvm/test/CodeGen/PowerPC/cfence-double.ll:8
 define void @foo() {
+; CHECK-LABEL: foo:
+; CHECK:       # %bb.0: # %entry
----------------
Should have an IR test in test/Transforms/AtomicExpand too


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D127609/new/

https://reviews.llvm.org/D127609