[flang][hlfir] lower hlfir.all into runtime call

[jacob-crawley]

Summary

Depends on: D151111

[tblah]

Thanks for this. Just one nit.

flang/lib/Optimizer/HLFIR/Transforms/LowerHLFIRIntrinsics.cpp

205

You don't need this-> in c++

[jacob-crawley]

Thanks for the review!

Removed the use of "this->"

[jeanPerier]

LGTM, thanks

flang/lib/Optimizer/HLFIR/Transforms/LowerHLFIRIntrinsics.cpp

protected:
   }
 };
 
 template <class OP>
 class HlfirReductionIntrinsicConversion : public HlfirIntrinsicConversion<OP> {
   using HlfirIntrinsicConversion<OP>::HlfirIntrinsicConversion;
   using IntrinsicArgument =
       typename HlfirIntrinsicConversion<OP>::IntrinsicArgument;
+  using HlfirIntrinsicConversion<OP>::lowerArguments;
+  using HlfirIntrinsicConversion<OP>::processReturnValue;
+
+protected:
+  auto buildNumericalArgs(OP operation, mlir::Type i32, mlir::Type logicalType,
+                          mlir::PatternRewriter &rewriter,
+                          std::string opName) const {
+    llvm::SmallVector<IntrinsicArgument, 3> inArgs;
+    inArgs.push_back({operation.getArray(), operation.getArray().getType()});
+    inArgs.push_back({operation.getDim(), i32});
+    inArgs.push_back({operation.getMask(), logicalType});
+    auto *argLowering = fir::getIntrinsicArgumentLowering(opName);
+    return lowerArguments(operation, inArgs, rewriter, argLowering);
+  };
+
+  auto buildLogicalArgs(OP operation, mlir::Type i32, mlir::Type logicalType,
+                        mlir::PatternRewriter &rewriter,
+                        std::string opName) const {
+    llvm::SmallVector<IntrinsicArgument, 2> inArgs;
+    inArgs.push_back({operation.getMask(), logicalType});
+    inArgs.push_back({operation.getDim(), i32});
+    auto *argLowering = fir::getIntrinsicArgumentLowering(opName);
+    return lowerArguments(operation, inArgs, rewriter, argLowering);
+  };
 
 public:
   mlir::LogicalResult
   matchAndRewrite(OP operation,
                   mlir::PatternRewriter &rewriter) const override {
     std::string opName;
     if constexpr (std::is_same_v<OP, hlfir::SumOp>) {
       opName = "sum";
     } else if constexpr (std::is_same_v<OP, hlfir::ProductOp>) {
       opName = "product";
+    } else if constexpr (std::is_same_v<OP, hlfir::AnyOp>) {
+      opName = "any";
+    } else if constexpr (std::is_same_v<OP, hlfir::AllOp>) {
+      opName = "all";
     } else {
       return mlir::failure();
     }
+
     fir::KindMapping kindMapping{rewriter.getContext()};
     fir::FirOpBuilder builder{rewriter, kindMapping};
     const mlir::Location &loc = operation->getLoc();
 
     mlir::Type i32 = builder.getI32Type();
     mlir::Type logicalType = fir::LogicalType::get(
         builder.getContext(), builder.getKindMap().defaultLogicalKind());
-    llvm::SmallVector<IntrinsicArgument, 3> inArgs;
-    inArgs.push_back({operation.getArray(), operation.getArray().getType()});
-    inArgs.push_back({operation.getDim(), i32});
-    inArgs.push_back({operation.getMask(), logicalType});
 
-    auto *argLowering = fir::getIntrinsicArgumentLowering(opName);
-    llvm::SmallVector<fir::ExtendedValue, 3> args =
-        this->lowerArguments(operation, inArgs, rewriter, argLowering);
+    llvm::SmallVector<fir::ExtendedValue, 0> args;
+
+    if constexpr (std::is_same_v<OP, hlfir::SumOp> ||
+                  std::is_same_v<OP, hlfir::ProductOp>) {
+      args = buildNumericalArgs(operation, i32, logicalType, rewriter, opName);
+    } else {
+      args = buildLogicalArgs(operation, i32, logicalType, rewriter, opName);
+    }
 
     mlir::Type scalarResultType =
         hlfir::getFortranElementType(operation.getType());
 
     auto [resultExv, mustBeFreed] =
         fir::genIntrinsicCall(builder, loc, opName, scalarResultType, args);
 
-    this->processReturnValue(operation, resultExv, mustBeFreed, builder,
-                             rewriter);
+    processReturnValue(operation, resultExv, mustBeFreed, builder, rewriter);
     return mlir::success();
   }
 };
 
 using SumOpConversion = HlfirReductionIntrinsicConversion<hlfir::SumOp>;
 
 using ProductOpConversion = HlfirReductionIntrinsicConversion<hlfir::ProductOp>;
 
-struct AnyOpConversion : public HlfirIntrinsicConversion<hlfir::AnyOp> {
-  using HlfirIntrinsicConversion<hlfir::AnyOp>::HlfirIntrinsicConversion;
-
-  mlir::LogicalResult
-  matchAndRewrite(hlfir::AnyOp any,
-                  mlir::PatternRewriter &rewriter) const override {
-    fir::KindMapping kindMapping{rewriter.getContext()};
-    fir::FirOpBuilder builder{rewriter, kindMapping};
-    const mlir::Location &loc = any->getLoc();
-
-    mlir::Type i32 = builder.getI32Type();
-    mlir::Type logicalType = fir::LogicalType::get(
-        builder.getContext(), builder.getKindMap().defaultLogicalKind());
-    llvm::SmallVector<IntrinsicArgument, 2> inArgs;
-    inArgs.push_back({any.getMask(), logicalType});
-    inArgs.push_back({any.getDim(), i32});
-
-    auto *argLowering = fir::getIntrinsicArgumentLowering("any");
-    llvm::SmallVector<fir::ExtendedValue, 2> args =
-        this->lowerArguments(any, inArgs, rewriter, argLowering);
-
-    mlir::Type resultType = hlfir::getFortranElementType(any.getType());
+using AnyOpConversion = HlfirReductionIntrinsicConversion<hlfir::AnyOp>;
 
-    auto [resultExv, mustBeFreed] =
-        fir::genIntrinsicCall(builder, loc, "any", resultType, args);
-
-    this->processReturnValue(any, resultExv, mustBeFreed, builder, rewriter);
-
-    return mlir::success();
-  }
-};
+using AllOpConversion = HlfirReductionIntrinsicConversion<hlfir::AllOp>;
 
 struct MatmulOpConversion : public HlfirIntrinsicConversion<hlfir::MatmulOp> {
   using HlfirIntrinsicConversion<hlfir::MatmulOp>::HlfirIntrinsicConversion;
 
   mlir::LogicalResult
   matchAndRewrite(hlfir::MatmulOp matmul,
                   mlir::PatternRewriter &rewriter) const override {
     fir::KindMapping kindMapping{rewriter.getContext()};
void runOnOperation() override {
     // scheduled on FuncOp. A solution could be to provide an optional mutex
     // when building a FirOpBuilder and locking around FuncOp and GlobalOp
     // creation, but this needs a bit more thinking, so at this point the pass
     // is scheduled on the moduleOp.
     mlir::ModuleOp module = this->getOperation();
     mlir::MLIRContext *context = &getContext();
     mlir::RewritePatternSet patterns(context);
     patterns.insert<MatmulOpConversion, MatmulTransposeOpConversion,
-                    AnyOpConversion, SumOpConversion, ProductOpConversion,
-                    TransposeOpConversion>(context);
+                    AllOpConversion, AnyOpConversion, SumOpConversion,
+                    ProductOpConversion, TransposeOpConversion>(context);
     mlir::ConversionTarget target(*context);
     target.addLegalDialect<mlir::BuiltinDialect, mlir::arith::ArithDialect,
                            mlir::func::FuncDialect, fir::FIROpsDialect,
                            hlfir::hlfirDialect>();
     target.addIllegalOp<hlfir::MatmulOp, hlfir::MatmulTransposeOp, hlfir::SumOp,
-                        hlfir::ProductOp, hlfir::TransposeOp, hlfir::AnyOp>();
+                        hlfir::ProductOp, hlfir::TransposeOp, hlfir::AnyOp,
+                        hlfir::AllOp>();
     target.markUnknownOpDynamicallyLegal(
         [](mlir::Operation *) { return true; });
     if (mlir::failed(
             mlir::applyFullConversion(module, target, std::move(patterns)))) {
       mlir::emitError(mlir::UnknownLoc::get(context),
                       "failure in HLFIR intrinsic lowering");
       signalPassFailure();
     }
   }
 };
 } // namespace
 
 std::unique_ptr<mlir::Pass> hlfir::createLowerHLFIRIntrinsicsPass() {
   return std::make_unique<LowerHLFIRIntrinsics>();
 }

flang/test/HLFIR/all-lowering.fir

+// Test hlfir.all operation lowering to fir runtime call
+// RUN: fir-opt %s -lower-hlfir-intrinsics | FileCheck %s
+
+func.func @_QPall1(%arg0: !fir.box<!fir.array<?x!fir.logical<4>>> {fir.bindc_name = "a"}, %arg1: !fir.ref<!fir.logical<4>> {fir.bindc_name = "s"}) {
+  %0:2 = hlfir.declare %arg0 {uniq_name = "_QFall1Ea"} : (!fir.box<!fir.array<?x!fir.logical<4>>>) -> (!fir.box<!fir.array<?x!fir.logical<4>>>, !fir.box<!fir.array<?x!fir.logical<4>>>)
+  %1:2 = hlfir.declare %arg1 {uniq_name = "_QFall1Es"} : (!fir.ref<!fir.logical<4>>) -> (!fir.ref<!fir.logical<4>>, !fir.ref<!fir.logical<4>>)
+  %2 = hlfir.all %0#0 : (!fir.box<!fir.array<?x!fir.logical<4>>>) -> !fir.logical<4>
+  hlfir.assign %2 to %1#0 : !fir.logical<4>, !fir.ref<!fir.logical<4>>
+  return
+}
+// CHECK-LABEL: func.func @_QPall1(
+// CHECK:           %[[ARG0:.*]]: !fir.box<!fir.array<?x!fir.logical<4>>> {fir.bindc_name = "a"}
+// CHECK:           %[[ARG1:.*]]: !fir.ref<!fir.logical<4>>
+// CHECK-DAG:     %[[MASK:.*]]:2 = hlfir.declare %[[ARG0]]
+// CHECK-DAG:     %[[RES:.*]]:2 = hlfir.declare %[[ARG1]]
+// CHECK-DAG:     %[[MASK_ARG:.*]] = fir.convert %[[MASK]]#1 : (!fir.box<!fir.array<?x!fir.logical<4>>>) -> !fir.box<none>
+// CHECK:         %[[RET_ARG:.*]] = fir.call @_FortranAAll(%[[MASK_ARG]], %[[LOC_STR:.*]], %[[LOC_N:.*]], %[[C1:.*]]) : (!fir.box<none>, !fir.ref<i8>, i32, i32) -> i1
+// CHECK-NEXT:    %[[RET:.*]] = fir.convert %[[RET_ARG]] : (i1) -> !fir.logical<4>
+// CHECK-NEXT:    hlfir.assign %[[RET]] to %[[RES]]#0 : !fir.logical<4>, !fir.ref<!fir.logical<4>>
+// CHECK-NEXT:    return
+// CHECK-NEXT:  }
+
+func.func @_QPall2(%arg0: !fir.box<!fir.array<?x?x!fir.logical<4>>> {fir.bindc_name = "a"}, %arg1: !fir.box<!fir.array<?x!fir.logical<4>>> {fir.bindc_name = "s"}, %arg2: !fir.ref<i32> {fir.bindc_name = "d"}) {
+  %0:2 = hlfir.declare %arg0 {uniq_name = "_QFall2Ea"} : (!fir.box<!fir.array<?x?x!fir.logical<4>>>) -> (!fir.box<!fir.array<?x?x!fir.logical<4>>>, !fir.box<!fir.array<?x?x!fir.logical<4>>>)
+  %1:2 = hlfir.declare %arg2 {uniq_name = "_QFall2Ed"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+  %2:2 = hlfir.declare %arg1 {uniq_name = "_QFall2Es"} : (!fir.box<!fir.array<?x!fir.logical<4>>>) -> (!fir.box<!fir.array<?x!fir.logical<4>>>, !fir.box<!fir.array<?x!fir.logical<4>>>)
+  %3 = fir.load %1#0 : !fir.ref<i32>
+  %4 = hlfir.all %0#0 dim %3 : (!fir.box<!fir.array<?x?x!fir.logical<4>>>, i32) -> !hlfir.expr<?x!fir.logical<4>>
+  hlfir.assign %4 to %2#0 : !hlfir.expr<?x!fir.logical<4>>, !fir.box<!fir.array<?x!fir.logical<4>>>
+  hlfir.destroy %4 : !hlfir.expr<?x!fir.logical<4>>
+  return
+}
+// CHECK-LABEL: func.func @_QPall2(
+// CHECK:           %[[ARG0:.*]]: !fir.box<!fir.array<?x?x!fir.logical<4>>>
+// CHECK:           %[[ARG1:.*]]: !fir.box<!fir.array<?x!fir.logical<4>>>
+// CHECK:           %[[ARG2:.*]]: !fir.ref<i32>
+// CHECK-DAG:     %[[MASK:.*]]:2 = hlfir.declare %[[ARG0]]
+// CHECK-DAG:     %[[DIM_VAR:.*]]:2 = hlfir.declare %[[ARG2]]
+// CHECK-DAG:     %[[RES:.*]]:2 = hlfir.declare %[[ARG1]]
+
+// CHECK-DAG:     %[[RET_BOX:.*]] = fir.alloca !fir.box<!fir.heap<!fir.array<?x!fir.logical<4>>>>
+// CHECK-DAG:     %[[RET_ADDR:.*]] = fir.zero_bits !fir.heap<!fir.array<?x!fir.logical<4>>>
+// CHECK-DAG:     %[[C0:.*]] = arith.constant 0 : index
+// CHECK-DAG:     %[[RET_SHAPE:.*]] = fir.shape %[[C0]] : (index) -> !fir.shape<1>
+// CHECK-DAG:     %[[RET_EMBOX:.*]] = fir.embox %[[RET_ADDR]](%[[RET_SHAPE]])
+// CHECK-DAG:     fir.store %[[RET_EMBOX]] to %[[RET_BOX]]
+
+// CHECK-DAG:     %[[DIM:.*]] = fir.load %[[DIM_VAR]]#0 : !fir.ref<i32>
+// CHECK-DAG:     %[[RET_ARG:.*]] = fir.convert %[[RET_BOX]]
+// CHECK-DAG:     %[[MASK_ARG:.*]] = fir.convert %[[MASK]]#1
+
+// CHECK:         %[[NONE:.*]] = fir.call @_FortranAAllDim(%[[RET_ARG]], %[[MASK_ARG]], %[[DIM]], %[[LOC_STR:.*]], %[[LOC_N:.*]]) : (!fir.ref<!fir.box<none>>, !fir.box<none>, i32, !fir.ref<i8>, i32) -> none
+// CHECK:         %[[RET:.*]] = fir.load %[[RET_BOX]]
+// CHECK:         %[[BOX_DIMS:.*]]:3 = fir.box_dims %[[RET]]
+// CHECK-NEXT:    %[[ADDR:.*]] = fir.box_addr %[[RET]]
+// CHECK-NEXT:    %[[SHIFT:.*]] = fir.shape_shift %[[BOX_DIMS]]#0, %[[BOX_DIMS]]#1
+// CHECK-NEXT:    %[[TMP:.*]]:2 = hlfir.declare %[[ADDR]](%[[SHIFT]]) {uniq_name = ".tmp.intrinsic_result"}
+// CHECK:         %[[TRUE:.*]] = arith.constant true
+// CHECK:         %[[EXPR:.*]] = hlfir.as_expr %[[TMP]]#0 move %[[TRUE]] : (!fir.box<!fir.array<?x!fir.logical<4>>>, i1) -> !hlfir.expr<?x!fir.logical<4>>
+// CHECK:         hlfir.assign %[[EXPR]] to %[[RES]]#0
+// CHECK:         hlfir.destroy %[[EXPR]]
+// CHECK-NEXT:    return
+// CHECK-NEXT:  }
+
+func.func @_QPall3(%arg0: !fir.ref<!fir.array<2x!fir.logical<4>>> {fir.bindc_name = "s"}) {
+  %0 = fir.address_of(@_QFall3Ea) : !fir.ref<!fir.array<2x2x!fir.logical<4>>>
+  %c2 = arith.constant 2 : index
+  %c2_0 = arith.constant 2 : index
+  %1 = fir.shape %c2, %c2_0 : (index, index) -> !fir.shape<2>
+  %2:2 = hlfir.declare %0(%1) {uniq_name = "_QFall3Ea"} : (!fir.ref<!fir.array<2x2x!fir.logical<4>>>, !fir.shape<2>) -> (!fir.ref<!fir.array<2x2x!fir.logical<4>>>, !fir.ref<!fir.array<2x2x!fir.logical<4>>>)
+  %c2_1 = arith.constant 2 : index
+  %3 = fir.shape %c2_1 : (index) -> !fir.shape<1>
+  %4:2 = hlfir.declare %arg0(%3) {uniq_name = "_QFall3Es"} : (!fir.ref<!fir.array<2x!fir.logical<4>>>, !fir.shape<1>) -> (!fir.ref<!fir.array<2x!fir.logical<4>>>, !fir.ref<!fir.array<2x!fir.logical<4>>>)
+  %c1_i32 = arith.constant 1 : i32
+  %5 = hlfir.all %2#0 dim %c1_i32 : (!fir.ref<!fir.array<2x2x!fir.logical<4>>>, i32) -> !hlfir.expr<2x!fir.logical<4>>
+  hlfir.assign %5 to %4#0 : !hlfir.expr<2x!fir.logical<4>>, !fir.ref<!fir.array<2x!fir.logical<4>>>
+  hlfir.destroy %5 : !hlfir.expr<2x!fir.logical<4>>
+  return
+}
+// CHECK-LABEL:  func.func @_QPall3(
+// CHECK:           %[[ARG0:.*]]: !fir.ref<!fir.array<2x!fir.logical<4>>>
+// CHECK-DAG:     %[[RET_BOX:.*]] = fir.alloca !fir.box<!fir.heap<!fir.array<?x!fir.logical<4>>>>
+// CHECK-DAG:     %[[RET_ADDR:.*]] = fir.zero_bits !fir.heap<!fir.array<?x!fir.logical<4>>>
+// CHECK-DAG:     %[[C0:.*]] = arith.constant 0 : index
+// CHECK-DAG:     %[[RET_SHAPE:.*]] = fir.shape %[[C0]] : (index) -> !fir.shape<1>
+// CHECK-DAG:     %[[RET_EMBOX:.*]] = fir.embox %[[RET_ADDR]](%[[RET_SHAPE]])
+// CHECK-DAG:     fir.store %[[RET_EMBOX]] to %[[RET_BOX]]
+// CHECK-DAG:     %[[RES:.*]]:2 = hlfir.declare %[[ARG0]](%[[RES_SHAPE:.*]])
+
+// CHECK-DAG:     %[[MASK_ADDR:.*]] = fir.address_of
+// CHECK-DAG:     %[[MASK_VAR:.*]]:2 = hlfir.declare %[[MASK_ADDR]](%[[MASK_SHAPE:.*]])
+// CHECK-DAG:     %[[MASK_BOX:.*]] = fir.embox %[[MASK_VAR]]#1(%[[MASK_SHAPE:.*]])
+
+// CHECK-DAG:     %[[DIM:.*]] = arith.constant 1 : i32
+
+// CHECK-DAG:     %[[RET_ARG:.*]] = fir.convert %[[RET_BOX]]
+// CHECK-DAG:     %[[MASK_ARG:.*]] = fir.convert %[[MASK_BOX]] : (!fir.box<!fir.array<2x2x!fir.logical<4>>>) -> !fir.box<none>
+// CHECK:         %[[NONE:.*]] = fir.call @_FortranAAllDim(%[[RET_ARG]], %[[MASK_ARG]], %[[DIM]], %[[LOC_STR:.*]], %[[LOC_N:.*]])
+// CHECK:         %[[RET:.*]] = fir.load %[[RET_BOX]]
+// CHECK:         %[[BOX_DIMS:.*]]:3 = fir.box_dims %[[RET]]
+// CHECK-NEXT:    %[[ADDR:.*]] = fir.box_addr %[[RET]]
+// CHECK-NEXT:    %[[SHIFT:.*]] = fir.shape_shift %[[BOX_DIMS]]#0, %[[BOX_DIMS]]#1
+// CHECK-NEXT:    %[[TMP:.*]]:2 = hlfir.declare %[[ADDR]](%[[SHIFT]]) {uniq_name = ".tmp.intrinsic_result"}
+// CHECK:         %[[TRUE:.*]] = arith.constant true
+// CHECK:         %[[EXPR:.*]] = hlfir.as_expr %[[TMP]]#0 move %[[TRUE]] : (!fir.box<!fir.array<?x!fir.logical<4>>>, i1) -> !hlfir.expr<?x!fir.logical<4>>
+// CHECK:         hlfir.assign %[[EXPR]] to %[[RES]]
+// CHECK:         hlfir.destroy %[[EXPR]]
+// CHECK-NEXT:    return
+// CHECK-NEXT:  }
+
+func.func @_QPall4(%arg0: !fir.box<!fir.array<?x?x!fir.logical<4>>> {fir.bindc_name = "a"}, %arg1: !fir.box<!fir.array<?x!fir.logical<4>>> {fir.bindc_name = "s"}, %arg2: !fir.ref<!fir.box<!fir.ptr<i32>>> {fir.bindc_name = "d"}) {
+  %0:2 = hlfir.declare %arg0 {uniq_name = "_QFall4Ea"} : (!fir.box<!fir.array<?x?x!fir.logical<4>>>) -> (!fir.box<!fir.array<?x?x!fir.logical<4>>>, !fir.box<!fir.array<?x?x!fir.logical<4>>>)
+  %1:2 = hlfir.declare %arg2 {fortran_attrs = #fir.var_attrs<pointer>, uniq_name = "_QFall4Ed"} : (!fir.ref<!fir.box<!fir.ptr<i32>>>) -> (!fir.ref<!fir.box<!fir.ptr<i32>>>, !fir.ref<!fir.box<!fir.ptr<i32>>>)
+  %2:2 = hlfir.declare %arg1 {uniq_name = "_QFall4Es"} : (!fir.box<!fir.array<?x!fir.logical<4>>>) -> (!fir.box<!fir.array<?x!fir.logical<4>>>, !fir.box<!fir.array<?x!fir.logical<4>>>)
+  %3 = fir.load %1#0 : !fir.ref<!fir.box<!fir.ptr<i32>>>
+  %4 = fir.box_addr %3 : (!fir.box<!fir.ptr<i32>>) -> !fir.ptr<i32>
+  %5 = fir.load %4 : !fir.ptr<i32>
+  %6 = hlfir.no_reassoc %5 : i32
+  %7 = hlfir.all %0#0 dim %6 : (!fir.box<!fir.array<?x?x!fir.logical<4>>>, i32) -> !hlfir.expr<?x!fir.logical<4>>
+  hlfir.assign %7 to %2#0 : !hlfir.expr<?x!fir.logical<4>>, !fir.box<!fir.array<?x!fir.logical<4>>>
+  hlfir.destroy %7 : !hlfir.expr<?x!fir.logical<4>>
+  return
+}
+// CHECK-LABEL:  func.func @_QPall4(
+// CHECK:           %[[ARG0:.*]]: !fir.box<!fir.array<?x?x!fir.logical<4>>>
+// CHECK:           %[[ARG1:.*]]: !fir.box<!fir.array<?x!fir.logical<4>>>
+// CHECK:           %[[ARG2:.*]]: !fir.ref<!fir.box<!fir.ptr<i32>>>
+// CHECK-DAG:     %[[MASK:.*]]:2 = hlfir.declare %[[ARG0]]
+// CHECK-DAG:     %[[DIM_ARG:.*]]:2 = hlfir.declare %[[ARG2]]
+// CHECK-DAG:     %[[RES:.*]]:2 = hlfir.declare %[[ARG1]]
+
+// CHECK-DAG:     %[[RET_BOX:.*]] = fir.alloca !fir.box<!fir.heap<!fir.array<?x!fir.logical<4>>>>
+// CHECK-DAG:     %[[DIM_PTR:.*]] = fir.load %[[DIM_ARG]]#0 : !fir.ref<!fir.box<!fir.ptr<i32>>>
+// CHECK-DAG:     %[[DIM_ADDR:.*]] = fir.box_addr %[[DIM_PTR]]
+// CHECK-DAG:     %[[DIM_VAR:.*]] = fir.load %[[DIM_ADDR]]
+// CHECK-DAG:     %[[DIM:.*]] = hlfir.no_reassoc %[[DIM_VAR]]
+
+// CHECK-DAG:     %[[C0:.*]] = arith.constant 0 : index
+// CHECK-DAG:     %[[RET_ADDR:.*]] = fir.zero_bits !fir.heap<!fir.array<?x!fir.logical<4>>>
+// CHECK-DAG:     %[[RET_SHAPE:.*]] = fir.shape %[[C0]] : (index) -> !fir.shape<1>
+// CHECK-DAG:     %[[RET_EMBOX:.*]] = fir.embox %[[RET_ADDR]](%[[RET_SHAPE]])
+// CHECK-DAG:     fir.store %[[RET_EMBOX]] to %[[RET_BOX]]
+// CHECK-DAG:     %[[RET_ARG:.*]] = fir.convert %[[RET_BOX]]
+// CHECK-DAG:     %[[MASK_ARG:.*]] = fir.convert %[[MASK]]#1
+
+// CHECK:         %[[NONE:.*]] = fir.call @_FortranAAllDim(%[[RET_ARG]], %[[MASK_ARG]], %[[DIM]], %[[LOC_STR:.*]], %[[LOC_N:.*]]) : (!fir.ref<!fir.box<none>>, !fir.box<none>, i32, !fir.ref<i8>, i32) -> none
+// CHECK:         %[[RET:.*]] = fir.load %[[RET_BOX]]
+// CHECK:         %[[BOX_DIMS:.*]]:3 = fir.box_dims %[[RET]]
+// CHECK-NEXT:    %[[ADDR:.*]] = fir.box_addr %[[RET]]
+// CHECK-NEXT:    %[[SHIFT:.*]] = fir.shape_shift %[[BOX_DIMS]]#0, %[[BOX_DIMS]]#1
+// CHECK-NEXT:    %[[TMP:.*]]:2 = hlfir.declare %[[ADDR]](%[[SHIFT]]) {uniq_name = ".tmp.intrinsic_result"}
+// CHECK:         %[[TRUE:.*]] = arith.constant true
+// CHECK:         %[[EXPR:.*]] = hlfir.as_expr %[[TMP]]#0 move %[[TRUE]] : (!fir.box<!fir.array<?x!fir.logical<4>>>, i1) -> !hlfir.expr<?x!fir.logical<4>>
+// CHECK:         hlfir.assign %[[EXPR]] to %[[RES]]
+// CHECK:         hlfir.destroy %[[EXPR]]
+// CHECK-NEXT:    return
+// CHECK-NEXT:  }
 No newline at end of file