[llvm] [AMDGPU] Support double type in atomic optimizer. (PR #84307)

[Matt Arsenault via llvm-commits]

================
@@ -209,8 +209,9 @@ void AMDGPUAtomicOptimizerImpl::visitAtomicRMWInst(AtomicRMWInst &I) {
     break;
   }
 
-  // Only 32-bit floating point atomic ops are supported.
-  if (AtomicRMWInst::isFPOperation(Op) && !I.getType()->isFloatTy()) {
+  // Only 32 and 64 bit floating point atomic ops are supported.
----------------
arsenm wrote:

Should really handle f16 too, in a separate patch 

https://github.com/llvm/llvm-project/pull/84307