[llvm-branch-commits] [llvm] AArch64: Use consistent atomicrmw expansion for FP operations (PR #103702)

[Eli Friedman via llvm-branch-commits]

efriedma-quic wrote:

Please also fix the comment before AArch64TargetLowering::shouldExpandAtomicRMWInIR .  (The comment is wrong; a floating-point "exception" isn't an exception in the sense that it would cancel an exclusive reservation.)

https://github.com/llvm/llvm-project/pull/103702