[flang-commits] [flang] [flang][fir] Lower `do concurrent` loop nests to `fir.do_concurrent` (PR #132904)

Tom Eccles via flang-commits flang-commits at lists.llvm.org
Wed Mar 26 10:13:20 PDT 2025


================
@@ -2277,96 +2283,77 @@ class FirConverter : public Fortran::lower::AbstractConverter {
       llvm::SmallVectorImpl<const Fortran::parser::CompilerDirective *> &dirs) {
     assert(!incrementLoopNestInfo.empty() && "empty loop nest");
     mlir::Location loc = toLocation();
-    mlir::Operation *boundsAndStepIP = nullptr;
     mlir::arith::IntegerOverflowFlags iofBackup{};
 
-    for (IncrementLoopInfo &info : incrementLoopNestInfo) {
-      mlir::Value lowerValue;
-      mlir::Value upperValue;
-      mlir::Value stepValue;
-
-      {
-        mlir::OpBuilder::InsertionGuard guard(*builder);
+    llvm::SmallVector<mlir::Value> nestLBs;
+    llvm::SmallVector<mlir::Value> nestUBs;
+    llvm::SmallVector<mlir::Value> nestSts;
+    llvm::SmallVector<mlir::Value> nestReduceOperands;
+    llvm::SmallVector<mlir::Attribute> nestReduceAttrs;
+    bool genDoConcurrent = false;
 
-        // Set the IP before the first loop in the nest so that all nest bounds
-        // and step values are created outside the nest.
-        if (boundsAndStepIP)
-          builder->setInsertionPointAfter(boundsAndStepIP);
+    for (IncrementLoopInfo &info : incrementLoopNestInfo) {
+      genDoConcurrent = info.isStructured() && info.isUnordered;
----------------
tblah wrote:

The `isUnordered` field has a comment saying that it is also used for `FORALL` (although that may not still be true in the HLFIR lowering path). I think it would be better to have a field that specifically says this is supposed to be a `do concurrent`.

https://github.com/llvm/llvm-project/pull/132904


More information about the flang-commits mailing list