Remove wrong comment about fp exceptions

Author: arsenm

llvm/lib/Target/AArch64/AArch64ISelLowering.cpp

@@ -27100,9 +27100,6 @@ AArch64TargetLowering::shouldExpandAtomicLoadInIR(LoadInst *LI) const {
 // However, with the LSE instructions (or outline-atomics mode, which provides
 // library routines in place of the LSE-instructions), we can directly emit many
 // operations instead.
-//
-// Floating-point operations are always emitted to a cmpxchg loop, because they
-// may trigger a trap which aborts an LLSC sequence.
 TargetLowering::AtomicExpansionKind
 AArch64TargetLowering::shouldExpandAtomicRMWInIR(AtomicRMWInst *AI) const {
   unsigned Size = AI->getType()->getPrimitiveSizeInBits();