[Mlir-commits] [mlir] [MLIR] Support non-atomic RMW option for emulated vector stores (PR #124887)

[Alan Li]

================
@@ -118,6 +119,11 @@ struct TestEmulateNarrowTypePass
       *this, "skip-memref-type-conversion",
       llvm::cl::desc("disable memref type conversion (to test failures)"),
       llvm::cl::init(false)};
+
+  Option<bool> atomicStore{
+      *this, "atomic-store",
+      llvm::cl::desc("use atomic store instead of load-modify-write"),
+      llvm::cl::init(true)};
----------------
lialan wrote:

That sounds like a good idea. I updated the name, also replaced `useAtomicRMW` with `disableAtomicRMW`

https://github.com/llvm/llvm-project/pull/124887