[Mlir-commits] [flang] [llvm] [mlir] [openmp][mlir] Expand use of atomicrmw for atomic update (PR #215654)

[Sairudra More]

================
@@ -11444,15 +11451,26 @@ Expected<std::pair<Value *, Value *>> OpenMPIRBuilder::emitAtomicUpdate(
   case AtomicRMWInst::Or:
   case AtomicRMWInst::Xor:
   case AtomicRMWInst::Xchg:
+  case AtomicRMWInst::Max:
+  case AtomicRMWInst::Min:
+  case AtomicRMWInst::UMax:
+  case AtomicRMWInst::UMin:
+  case AtomicRMWInst::FAdd:
+  case AtomicRMWInst::FMax:
+  case AtomicRMWInst::FMin:
+  case AtomicRMWInst::FMaximum:
+  case AtomicRMWInst::FMinimum:
     emitRMWOp = XElemTy;
     break;
   case AtomicRMWInst::Sub:
+  case AtomicRMWInst::FSub:
     emitRMWOp = (IsXBinopExpr && XElemTy);
     break;
   default:
     emitRMWOp = false;
   }
-  emitRMWOp &= XElemTy->isIntegerTy();
+  emitRMWOp &=
+      (XElemTy->isIntegerTy() || XElemTy->isFloatTy() || XElemTy->isDoubleTy());
----------------
Saieiei wrote:

All the new and updated float tests use f32, so nothing covers the `isDoubleTy()` path. Could you add an f64 update to openmp-llvm.mlir?

https://github.com/llvm/llvm-project/pull/215654