[PATCH] D120199: [X86] Use bit test instructions to optimize some logic atomic operations

[Craig Topper via Phabricator via llvm-commits]

craig.topper added inline comments.


================
Comment at: llvm/lib/Target/X86/X86ISelLowering.cpp:5451
+      Info.align = Align(Size);
+      Info.flags |= MachineMemOperand::MOLoad | MachineMemOperand::MOStore;
+      return true;
----------------
pengfei wrote:
> craig.topper wrote:
> > Does this need MOVolatile?
> Maybe not. `MOVolatile` is usually set when `isVolatile()` return true in load and store instructions. We don't have such info in a target intrinsic.
> But seems we never worried about it, and there're few place it is set in X86 code. So I guess X86 is fine with or without `MOVolatile`?
This information is used to create a MachineMemOperand. I think that MachineMemOperand needs to know this access can't undergo certain optimizations because it represents an atomic access. I don't think this interface can create an atomic MachineMemOperand, but it can create a volatile one.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D120199/new/

https://reviews.llvm.org/D120199