[fir] Add fir.no_reassoc conversion

Currently `fir.no_reassoc` is just removed in the conversion.

This patch is part of the upstreaming effort from fir-dev branch.

Reviewed By: kiranchandramohan

Differential Revision: https://reviews.llvm.org/D114154

Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>

flang/lib/Optimizer/CodeGen/CodeGen.cpp

@@ -1145,6 +1145,20 @@ struct LoadOpConversion : public FIROpConversion<fir::LoadOp> {
   }
 };
 
+/// Lower `fir.no_reassoc` to LLVM IR dialect.
+/// TODO: how do we want to enforce this in LLVM-IR? Can we manipulate the fast
+/// math flags?
+struct NoReassocOpConversion : public FIROpConversion<fir::NoReassocOp> {
+  using FIROpConversion::FIROpConversion;
+
+  mlir::LogicalResult
+  matchAndRewrite(fir::NoReassocOp noreassoc, OpAdaptor adaptor,
+                  mlir::ConversionPatternRewriter &rewriter) const override {
+    rewriter.replaceOp(noreassoc, adaptor.getOperands()[0]);
+    return success();
+  }
+};
+
 /// Lower `fir.select_type` to LLVM IR dialect.
 struct SelectTypeOpConversion : public FIROpConversion<fir::SelectTypeOp> {
   using FIROpConversion::FIROpConversion;
@@ -2166,13 +2180,13 @@ class FIRToLLVMLowering : public fir::FIRToLLVMLoweringBase<FIRToLLVMLowering> {
         FirEndOpConversion, HasValueOpConversion, GenTypeDescOpConversion,
         GlobalLenOpConversion, GlobalOpConversion, InsertOnRangeOpConversion,
         InsertValueOpConversion, IsPresentOpConversion, LoadOpConversion,
-        NegcOpConversion, MulcOpConversion, SelectCaseOpConversion,
-        SelectOpConversion, SelectRankOpConversion, SelectTypeOpConversion,
-        ShapeOpConversion, ShapeShiftOpConversion, ShiftOpConversion,
-        SliceOpConversion, StoreOpConversion, StringLitOpConversion,
-        SubcOpConversion, UnboxCharOpConversion, UnboxProcOpConversion,
-        UndefOpConversion, UnreachableOpConversion, ZeroOpConversion>(
-        typeConverter);
+        NegcOpConversion, NoReassocOpConversion, MulcOpConversion,
+        SelectCaseOpConversion, SelectOpConversion, SelectRankOpConversion,
+        SelectTypeOpConversion, ShapeOpConversion, ShapeShiftOpConversion,
+        ShiftOpConversion, SliceOpConversion, StoreOpConversion,
+        StringLitOpConversion, SubcOpConversion, UnboxCharOpConversion,
+        UnboxProcOpConversion, UndefOpConversion, UnreachableOpConversion,
+        ZeroOpConversion>(typeConverter);
     mlir::populateStdToLLVMConversionPatterns(typeConverter, pattern);
     mlir::arith::populateArithmeticToLLVMConversionPatterns(typeConverter,
                                                             pattern);

flang/test/Fir/convert-to-llvm.fir

@@ -1569,3 +1569,23 @@ func @field_index_dynamic_size() -> () {
 // CHECK-NEXT: %{{.*}} = llvm.call @custom_typeP.field_1.offset() {field = 0 : i64} : () -> i32
 // CHECK-NEXT: %{{.*}} = llvm.call @custom_typeP.field_2.offset() {field = 1 : i64} : () -> i32
 // CHECK-NEXT:  llvm.return
+
+// -----
+
+// Check `fir.no_reassoc` conversion to LLVM IR dialect
+
+func @no_reassoc(%arg0: !fir.ref<i32>) {
+  %0 = fir.alloca i32
+  %1 = fir.load %arg0 : !fir.ref<i32>
+  %2 = fir.no_reassoc %1 : i32
+  fir.store %2 to %0 : !fir.ref<i32>
+  return
+}
+
+// CHECK-LABEL: llvm.func @no_reassoc(
+// CHECK-SAME:                        %[[ARG0:.*]]: !llvm.ptr<i32>) {
+// CHECK:         %[[C1:.*]] = llvm.mlir.constant(1 : i64) : i64
+// CHECK:         %[[ALLOC:.*]] = llvm.alloca %[[C1]] x i32 {in_type = i32, operand_segment_sizes = dense<0> : vector<2xi32>} : (i64) -> !llvm.ptr<i32>
+// CHECK:         %[[LOAD:.*]] = llvm.load %[[ARG0]] : !llvm.ptr<i32>
+// CHECK:         llvm.store %[[LOAD]], %[[ALLOC]] : !llvm.ptr<i32>
+// CHECK:         llvm.return