[Mlir-commits] [flang] [llvm] [mlir] [flang][MLIR][OpenMP] make reduction by-ref toggled per variable (PR #92244)

[Kiran Chandramohan]

================
@@ -919,13 +923,21 @@ inlineOmpRegionCleanup(llvm::SmallVectorImpl<Region *> &cleanupRegions,
   return success();
 }
 
+static ArrayRef<bool> getIsByRef(std::optional<ArrayRef<bool>> attr) {
+  if (!attr)
+    return {};
+  return *attr;
+}
+
 /// Converts an OpenMP workshare loop into LLVM IR using OpenMPIRBuilder.
 static LogicalResult
 convertOmpWsloop(Operation &opInst, llvm::IRBuilderBase &builder,
                  LLVM::ModuleTranslation &moduleTranslation) {
   auto wsloopOp = cast<omp::WsloopOp>(opInst);
   auto loopOp = cast<omp::LoopNestOp>(wsloopOp.getWrappedLoop());
-  const bool isByRef = wsloopOp.getByref();
+
+  llvm::ArrayRef<bool> isByRef = getIsByRef(wsloopOp.getReductionVarsByref());
+  assert(isByRef.size() == wsloopOp.getNumReductionVars());
----------------
kiranchandramohan wrote:

Isn't this guaranteed by the verifier?

https://github.com/llvm/llvm-project/pull/92244