Fixes a performance problem with lowering of forall loops and creating

too many temporaries.

Fix clang-format errors.

Differential Revision: https://reviews.llvm.org/D125336

9 files changed, 373 insertions, 71 deletions

diff --git a/flang/lib/Lower/ConvertExpr.cpp b/flang/lib/Lower/ConvertExpr.cpp
index 0418740d890b..0e196693b393 100644
--- a/flang/lib/Lower/ConvertExpr.cpp
+++ b/flang/lib/Lower/ConvertExpr.cpp
@@ -2786,7 +2786,8 @@ public:
       mlir::Value zero = builder.createIntegerConstant(loc, len.getType(), 0);
       len = builder.create<mlir::arith::SelectOp>(loc, isPresent, len, zero);
       mlir::Value temp = builder.createTemporary(
-          loc, type, /*name=*/{}, /*shape=*/{}, mlir::ValueRange{len},
+          loc, type, /*name=*/{},
+          /*shape=*/{}, mlir::ValueRange{len},
           llvm::ArrayRef<mlir::NamedAttribute>{
               Fortran::lower::getAdaptToByRefAttr(builder)});
       return fir::CharBoxValue{temp, len};
@@ -4112,6 +4113,7 @@ private:
     return adjustedArrayElementType(pathTy);
   }
 
+  /// Lower rhs of an array expression.
   ExtValue lowerArrayExpression(const Fortran::lower::SomeExpr &exp) {
     mlir::Type resTy = converter.genType(exp);
     return std::visit(
@@ -4325,9 +4327,10 @@ private:
       // Adjust indices for any shift of the origin of the array.
       llvm::SmallVector<mlir::Value> indices = fir::factory::originateIndices(
           loc, *builder, tmp.getType(), shape, iters.iterVec());
-      auto addr = builder->create<fir::ArrayCoorOp>(
-          loc, eleRefTy, tmp, shape, /*slice=*/mlir::Value{}, indices,
-          /*typeParams=*/llvm::None);
+      auto addr =
+          builder->create<fir::ArrayCoorOp>(loc, eleRefTy, tmp, shape,
+                                            /*slice=*/mlir::Value{}, indices,
+                                            /*typeParams=*/llvm::None);
       auto load = builder->create<fir::LoadOp>(loc, addr);
       return builder->createConvert(loc, i1Ty, load);
     };
@@ -5080,9 +5083,15 @@ private:
   template <typename A>
   CC genarr(const Fortran::evaluate::Expr<A> &x) {
     LLVM_DEBUG(Fortran::lower::DumpEvaluateExpr::dump(llvm::dbgs(), x));
-    if (isArray(x) || explicitSpaceIsActive() ||
+    if (isArray(x) || (explicitSpaceIsActive() && isLeftHandSide()) ||
         isElementalProcWithArrayArgs(x))
       return std::visit([&](const auto &e) { return genarr(e); }, x.u);
+    if (explicitSpaceIsActive()) {
+      assert(!isArray(x) && !isLeftHandSide());
+      auto cc = std::visit([&](const auto &e) { return genarr(e); }, x.u);
+      auto result = cc(IterationSpace{});
+      return [=](IterSpace) { return result; };
+    }
     return genScalarAndForwardValue(x);
   }
 
@@ -5320,7 +5329,7 @@ private:
 
   template <typename A>
   CC genarr(const Fortran::evaluate::Constant<A> &x) {
-    if (/*explicitSpaceIsActive() &&*/ x.Rank() == 0)
+    if (x.Rank() == 0)
       return genScalarAndForwardValue(x);
     mlir::Location loc = getLoc();
     mlir::IndexType idxTy = builder.getIndexType();
@@ -6966,9 +6975,10 @@ private:
           if (components.hasExtendCoorRef())
             TODO(loc, "need to adjust typeparameter(s) to reflect the final "
                       "component");
-          mlir::Value embox = builder.create<fir::EmboxOp>(
-              loc, boxTy, ptrAddr, /*shape=*/mlir::Value{},
-              /*slice=*/mlir::Value{}, typeParams);
+          mlir::Value embox =
+              builder.create<fir::EmboxOp>(loc, boxTy, ptrAddr,
+                                           /*shape=*/mlir::Value{},
+                                           /*slice=*/mlir::Value{}, typeParams);
           return arrayLoadExtValue(builder, loc, load, iters.iterVec(), embox);
         }
       }
@@ -7183,6 +7193,12 @@ private:
     return semant == ConstituentSemantics::CustomCopyInCopyOut;
   }
 
+  /// Are we lowering in a left-hand side context?
+  inline bool isLeftHandSide() {
+    return isCopyInCopyOut() || isProjectedCopyInCopyOut() ||
+           isCustomCopyInCopyOut();
+  }
+
   /// Array appears in a context where it must be boxed.
   inline bool isBoxValue() { return semant == ConstituentSemantics::BoxValue; }
 
diff --git a/flang/test/Lower/forall/array-constructor.f90 b/flang/test/Lower/forall/array-constructor.f90
new file mode 100644
index 000000000000..ed28f498d2a3
--- /dev/null
+++ b/flang/test/Lower/forall/array-constructor.f90
@@ -0,0 +1,284 @@
+! RUN: bbc -emit-fir %s -o - | FileCheck %s
+
+subroutine ac1(arr,n)
+  integer :: arr(:), n
+  forall (i=1:n:2)
+     arr(i:i+2) = func((/i/))
+  end forall
+contains
+   pure integer function func(a)
+    integer, intent(in) :: a(:)
+    func = a(1)
+  end function func
+end subroutine ac1
+
+! CHECK-LABEL: func @_QPac1(
+! CHECK-SAME:               %[[VAL_0:.*]]: !fir.box<!fir.array<?xi32>> {fir.bindc_name = "arr"},
+! CHECK-SAME:               %[[VAL_1:.*]]: !fir.ref<i32> {fir.bindc_name = "n"}) {
+! CHECK:         %[[VAL_2:.*]] = fir.alloca index {bindc_name = ".buff.pos"}
+! CHECK:         %[[VAL_3:.*]] = fir.alloca index {bindc_name = ".buff.size"}
+! CHECK:         %[[VAL_4:.*]] = fir.alloca i32 {adapt.valuebyref, bindc_name = "i"}
+! CHECK:         %[[VAL_5:.*]] = arith.constant 1 : i32
+! CHECK:         %[[VAL_6:.*]] = fir.convert %[[VAL_5]] : (i32) -> index
+! CHECK:         %[[VAL_7:.*]] = fir.load %[[VAL_1]] : !fir.ref<i32>
+! CHECK:         %[[VAL_8:.*]] = fir.convert %[[VAL_7]] : (i32) -> index
+! CHECK:         %[[VAL_9:.*]] = arith.constant 2 : i32
+! CHECK:         %[[VAL_10:.*]] = fir.convert %[[VAL_9]] : (i32) -> index
+! CHECK:         %[[VAL_11:.*]] = fir.array_load %[[VAL_0]] : (!fir.box<!fir.array<?xi32>>) -> !fir.array<?xi32>
+! CHECK:         %[[VAL_12:.*]] = fir.do_loop %[[VAL_13:.*]] = %[[VAL_6]] to %[[VAL_8]] step %[[VAL_10]] unordered iter_args(%[[VAL_14:.*]] = %[[VAL_11]]) -> (!fir.array<?xi32>) {
+! CHECK:           %[[VAL_15:.*]] = fir.convert %[[VAL_13]] : (index) -> i32
+! CHECK:           fir.store %[[VAL_15]] to %[[VAL_4]] : !fir.ref<i32>
+! CHECK:           %[[VAL_16:.*]] = arith.constant 1 : index
+! CHECK:           %[[VAL_17:.*]] = fir.load %[[VAL_4]] : !fir.ref<i32>
+! CHECK:           %[[VAL_18:.*]] = fir.convert %[[VAL_17]] : (i32) -> i64
+! CHECK:           %[[VAL_19:.*]] = fir.convert %[[VAL_18]] : (i64) -> index
+! CHECK:           %[[VAL_20:.*]] = arith.constant 1 : i64
+! CHECK:           %[[VAL_21:.*]] = fir.convert %[[VAL_20]] : (i64) -> index
+! CHECK:           %[[VAL_22:.*]] = fir.load %[[VAL_4]] : !fir.ref<i32>
+! CHECK:           %[[VAL_23:.*]] = arith.constant 2 : i32
+! CHECK:           %[[VAL_24:.*]] = arith.addi %[[VAL_22]], %[[VAL_23]] : i32
+! CHECK:           %[[VAL_25:.*]] = fir.convert %[[VAL_24]] : (i32) -> i64
+! CHECK:           %[[VAL_26:.*]] = fir.convert %[[VAL_25]] : (i64) -> index
+! CHECK:           %[[VAL_27:.*]] = arith.constant 0 : index
+! CHECK:           %[[VAL_28:.*]] = arith.subi %[[VAL_26]], %[[VAL_19]] : index
+! CHECK:           %[[VAL_29:.*]] = arith.addi %[[VAL_28]], %[[VAL_21]] : index
+! CHECK:           %[[VAL_30:.*]] = arith.divsi %[[VAL_29]], %[[VAL_21]] : index
+! CHECK:           %[[VAL_31:.*]] = arith.cmpi sgt, %[[VAL_30]], %[[VAL_27]] : index
+! CHECK:           %[[VAL_32:.*]] = arith.select %[[VAL_31]], %[[VAL_30]], %[[VAL_27]] : index
+! CHECK:           %[[VAL_33:.*]] = arith.constant 1 : index
+! CHECK:           %[[VAL_34:.*]] = arith.constant 0 : index
+! CHECK:           fir.store %[[VAL_34]] to %[[VAL_2]] : !fir.ref<index>
+! CHECK:           %[[VAL_35:.*]] = fir.allocmem !fir.array<1xi32>
+! CHECK:           %[[VAL_36:.*]] = arith.constant 1 : index
+! CHECK:           fir.store %[[VAL_36]] to %[[VAL_3]] : !fir.ref<index>
+! CHECK:           %[[VAL_37:.*]] = fir.load %[[VAL_4]] : !fir.ref<i32>
+! CHECK:           %[[VAL_38:.*]] = arith.constant 1 : index
+! CHECK:           %[[VAL_39:.*]] = fir.zero_bits !fir.ref<!fir.array<1xi32>>
+! CHECK:           %[[VAL_40:.*]] = fir.coordinate_of %[[VAL_39]], %[[VAL_38]] : (!fir.ref<!fir.array<1xi32>>, index) -> !fir.ref<i32>
+! CHECK:           %[[VAL_41:.*]] = fir.convert %[[VAL_40]] : (!fir.ref<i32>) -> index
+! CHECK:           %[[VAL_42:.*]] = fir.load %[[VAL_2]] : !fir.ref<index>
+! CHECK:           %[[VAL_43:.*]] = fir.load %[[VAL_3]] : !fir.ref<index>
+! CHECK:           %[[VAL_44:.*]] = arith.constant 1 : index
+! CHECK:           %[[VAL_45:.*]] = arith.addi %[[VAL_42]], %[[VAL_44]] : index
+! CHECK:           %[[VAL_46:.*]] = arith.cmpi sle, %[[VAL_43]], %[[VAL_45]] : index
+! CHECK:           %[[VAL_47:.*]] = fir.if %[[VAL_46]] -> (!fir.heap<!fir.array<1xi32>>) {
+! CHECK:             %[[VAL_48:.*]] = arith.constant 2 : index
+! CHECK:             %[[VAL_49:.*]] = arith.muli %[[VAL_45]], %[[VAL_48]] : index
+! CHECK:             fir.store %[[VAL_49]] to %[[VAL_3]] : !fir.ref<index>
+! CHECK:             %[[VAL_50:.*]] = arith.muli %[[VAL_49]], %[[VAL_41]] : index
+! CHECK:             %[[VAL_51:.*]] = fir.convert %[[VAL_35]] : (!fir.heap<!fir.array<1xi32>>) -> !fir.ref<i8>
+! CHECK:             %[[VAL_52:.*]] = fir.convert %[[VAL_50]] : (index) -> i64
+! CHECK:             %[[VAL_53:.*]] = fir.call @realloc(%[[VAL_51]], %[[VAL_52]]) : (!fir.ref<i8>, i64) -> !fir.ref<i8>
+! CHECK:             %[[VAL_54:.*]] = fir.convert %[[VAL_53]] : (!fir.ref<i8>) -> !fir.heap<!fir.array<1xi32>>
+! CHECK:             fir.result %[[VAL_54]] : !fir.heap<!fir.array<1xi32>>
+! CHECK:           } else {
+! CHECK:             fir.result %[[VAL_35]] : !fir.heap<!fir.array<1xi32>>
+! CHECK:           }
+! CHECK:           %[[VAL_55:.*]] = fir.coordinate_of %[[VAL_56:.*]], %[[VAL_42]] : (!fir.heap<!fir.array<1xi32>>, index) -> !fir.ref<i32>
+! CHECK:           fir.store %[[VAL_37]] to %[[VAL_55]] : !fir.ref<i32>
+! CHECK:           fir.store %[[VAL_45]] to %[[VAL_2]] : !fir.ref<index>
+! CHECK:           %[[VAL_57:.*]] = fir.load %[[VAL_2]] : !fir.ref<index>
+! CHECK:           %[[VAL_58:.*]] = fir.shape %[[VAL_57]] : (index) -> !fir.shape<1>
+! CHECK:           %[[VAL_59:.*]] = fir.array_load %[[VAL_56]](%[[VAL_58]]) : (!fir.heap<!fir.array<1xi32>>, !fir.shape<1>) -> !fir.array<1xi32>
+! CHECK:           %[[VAL_60:.*]] = fir.allocmem !fir.array<1xi32>
+! CHECK:           %[[VAL_61:.*]] = fir.shape %[[VAL_33]] : (index) -> !fir.shape<1>
+! CHECK:           %[[VAL_62:.*]] = fir.array_load %[[VAL_60]](%[[VAL_61]]) : (!fir.heap<!fir.array<1xi32>>, !fir.shape<1>) -> !fir.array<1xi32>
+! CHECK:           %[[VAL_63:.*]] = arith.constant 1 : index
+! CHECK:           %[[VAL_64:.*]] = arith.constant 0 : index
+! CHECK:           %[[VAL_65:.*]] = arith.subi %[[VAL_33]], %[[VAL_63]] : index
+! CHECK:           %[[VAL_66:.*]] = fir.do_loop %[[VAL_67:.*]] = %[[VAL_64]] to %[[VAL_65]] step %[[VAL_63]] unordered iter_args(%[[VAL_68:.*]] = %[[VAL_62]]) -> (!fir.array<1xi32>) {
+! CHECK:             %[[VAL_69:.*]] = fir.array_fetch %[[VAL_59]], %[[VAL_67]] : (!fir.array<1xi32>, index) -> i32
+! CHECK:             %[[VAL_70:.*]] = fir.array_update %[[VAL_68]], %[[VAL_69]], %[[VAL_67]] : (!fir.array<1xi32>, i32, index) -> !fir.array<1xi32>
+! CHECK:             fir.result %[[VAL_70]] : !fir.array<1xi32>
+! CHECK:           }
+! CHECK:           fir.array_merge_store %[[VAL_62]], %[[VAL_71:.*]] to %[[VAL_60]] : !fir.array<1xi32>, !fir.array<1xi32>, !fir.heap<!fir.array<1xi32>>
+! CHECK:           %[[VAL_72:.*]] = fir.shape %[[VAL_33]] : (index) -> !fir.shape<1>
+! CHECK:           %[[VAL_73:.*]] = fir.embox %[[VAL_60]](%[[VAL_72]]) : (!fir.heap<!fir.array<1xi32>>, !fir.shape<1>) -> !fir.box<!fir.array<1xi32>>
+! CHECK:           %[[VAL_74:.*]] = fir.convert %[[VAL_73]] : (!fir.box<!fir.array<1xi32>>) -> !fir.box<!fir.array<?xi32>>
+! CHECK:           %[[VAL_75:.*]] = fir.call @_QFac1Pfunc(%[[VAL_74]]) : (!fir.box<!fir.array<?xi32>>) -> i32
+! CHECK:           %[[VAL_76:.*]] = arith.constant 1 : index
+! CHECK:           %[[VAL_77:.*]] = arith.constant 0 : index
+! CHECK:           %[[VAL_78:.*]] = arith.subi %[[VAL_32]], %[[VAL_76]] : index
+! CHECK:           %[[VAL_79:.*]] = fir.do_loop %[[VAL_80:.*]] = %[[VAL_77]] to %[[VAL_78]] step %[[VAL_76]] unordered iter_args(%[[VAL_81:.*]] = %[[VAL_14]]) -> (!fir.array<?xi32>) {
+! CHECK:             %[[VAL_82:.*]] = arith.subi %[[VAL_19]], %[[VAL_16]] : index
+! CHECK:             %[[VAL_83:.*]] = arith.muli %[[VAL_80]], %[[VAL_21]] : index
+! CHECK:             %[[VAL_84:.*]] = arith.addi %[[VAL_82]], %[[VAL_83]] : index
+! CHECK:             %[[VAL_85:.*]] = fir.array_update %[[VAL_81]], %[[VAL_75]], %[[VAL_84]] : (!fir.array<?xi32>, i32, index) -> !fir.array<?xi32>
+! CHECK:             fir.result %[[VAL_85]] : !fir.array<?xi32>
+! CHECK:           }
+! CHECK:           fir.freemem %[[VAL_60]] : !fir.heap<!fir.array<1xi32>>
+! CHECK:           fir.freemem %[[VAL_56]] : !fir.heap<!fir.array<1xi32>>
+! CHECK:           fir.result %[[VAL_86:.*]] : !fir.array<?xi32>
+! CHECK:         }
+! CHECK:         fir.array_merge_store %[[VAL_11]], %[[VAL_87:.*]] to %[[VAL_0]] : !fir.array<?xi32>, !fir.array<?xi32>, !fir.box<!fir.array<?xi32>>
+! CHECK:         return
+! CHECK:       }
+
+! CHECK-LABEL: func @_QFac1Pfunc(
+! CHECK-SAME:                    %[[VAL_0:.*]]: !fir.box<!fir.array<?xi32>> {fir.bindc_name = "a"}) -> i32 {
+! CHECK:         %[[VAL_1:.*]] = fir.alloca i32 {bindc_name = "func", uniq_name = "_QFfuncEfunc"}
+! CHECK:         %[[VAL_2:.*]] = arith.constant 1 : i64
+! CHECK:         %[[VAL_3:.*]] = arith.constant 1 : i64
+! CHECK:         %[[VAL_4:.*]] = arith.subi %[[VAL_2]], %[[VAL_3]] : i64
+! CHECK:         %[[VAL_5:.*]] = fir.coordinate_of %[[VAL_0]], %[[VAL_4]] : (!fir.box<!fir.array<?xi32>>, i64) -> !fir.ref<i32>
+! CHECK:         %[[VAL_6:.*]] = fir.load %[[VAL_5]] : !fir.ref<i32>
+! CHECK:         fir.store %[[VAL_6]] to %[[VAL_1]] : !fir.ref<i32>
+! CHECK:         %[[VAL_7:.*]] = fir.load %[[VAL_1]] : !fir.ref<i32>
+! CHECK:         return %[[VAL_7]] : i32
+! CHECK:       }
+
+subroutine ac2(arr,n)
+  integer :: arr(:), n
+  forall (i=1:n:2)
+     arr(i:i+2) = func((/i/))
+  end forall
+contains
+  pure function func(a)
+    integer :: func(3)
+    integer, intent(in) :: a(:)
+    func = a(1:3)
+  end function func
+end subroutine ac2
+
+! CHECK-LABEL: func @_QPac2(
+! CHECK-SAME:               %[[VAL_0:.*]]: !fir.box<!fir.array<?xi32>> {fir.bindc_name = "arr"},
+! CHECK-SAME:               %[[VAL_1:.*]]: !fir.ref<i32> {fir.bindc_name = "n"}) {
+! CHECK:         %[[VAL_2:.*]] = fir.alloca !fir.array<3xi32> {bindc_name = ".result"}
+! CHECK:         %[[VAL_3:.*]] = fir.alloca index {bindc_name = ".buff.pos"}
+! CHECK:         %[[VAL_4:.*]] = fir.alloca index {bindc_name = ".buff.size"}
+! CHECK:         %[[VAL_5:.*]] = fir.alloca i32 {adapt.valuebyref, bindc_name = "i"}
+! CHECK:         %[[VAL_6:.*]] = arith.constant 1 : i32
+! CHECK:         %[[VAL_7:.*]] = fir.convert %[[VAL_6]] : (i32) -> index
+! CHECK:         %[[VAL_8:.*]] = fir.load %[[VAL_1]] : !fir.ref<i32>
+! CHECK:         %[[VAL_9:.*]] = fir.convert %[[VAL_8]] : (i32) -> index
+! CHECK:         %[[VAL_10:.*]] = arith.constant 2 : i32
+! CHECK:         %[[VAL_11:.*]] = fir.convert %[[VAL_10]] : (i32) -> index
+! CHECK:         %[[VAL_12:.*]] = fir.array_load %[[VAL_0]] : (!fir.box<!fir.array<?xi32>>) -> !fir.array<?xi32>
+! CHECK:         %[[VAL_13:.*]] = fir.do_loop %[[VAL_14:.*]] = %[[VAL_7]] to %[[VAL_9]] step %[[VAL_11]] unordered iter_args(%[[VAL_15:.*]] = %[[VAL_12]]) -> (!fir.array<?xi32>) {
+! CHECK:           %[[VAL_16:.*]] = fir.convert %[[VAL_14]] : (index) -> i32
+! CHECK:           fir.store %[[VAL_16]] to %[[VAL_5]] : !fir.ref<i32>
+! CHECK:           %[[VAL_17:.*]] = arith.constant 1 : index
+! CHECK:           %[[VAL_18:.*]] = fir.load %[[VAL_5]] : !fir.ref<i32>
+! CHECK:           %[[VAL_19:.*]] = fir.convert %[[VAL_18]] : (i32) -> i64
+! CHECK:           %[[VAL_20:.*]] = fir.convert %[[VAL_19]] : (i64) -> index
+! CHECK:           %[[VAL_21:.*]] = arith.constant 1 : i64
+! CHECK:           %[[VAL_22:.*]] = fir.convert %[[VAL_21]] : (i64) -> index
+! CHECK:           %[[VAL_23:.*]] = fir.load %[[VAL_5]] : !fir.ref<i32>
+! CHECK:           %[[VAL_24:.*]] = arith.constant 2 : i32
+! CHECK:           %[[VAL_25:.*]] = arith.addi %[[VAL_23]], %[[VAL_24]] : i32
+! CHECK:           %[[VAL_26:.*]] = fir.convert %[[VAL_25]] : (i32) -> i64
+! CHECK:           %[[VAL_27:.*]] = fir.convert %[[VAL_26]] : (i64) -> index
+! CHECK:           %[[VAL_28:.*]] = arith.constant 0 : index
+! CHECK:           %[[VAL_29:.*]] = arith.subi %[[VAL_27]], %[[VAL_20]] : index
+! CHECK:           %[[VAL_30:.*]] = arith.addi %[[VAL_29]], %[[VAL_22]] : index
+! CHECK:           %[[VAL_31:.*]] = arith.divsi %[[VAL_30]], %[[VAL_22]] : index
+! CHECK:           %[[VAL_32:.*]] = arith.cmpi sgt, %[[VAL_31]], %[[VAL_28]] : index
+! CHECK:           %[[VAL_33:.*]] = arith.select %[[VAL_32]], %[[VAL_31]], %[[VAL_28]] : index
+! CHECK:           %[[VAL_34:.*]] = arith.constant 1 : index
+! CHECK:           %[[VAL_35:.*]] = arith.constant 0 : index
+! CHECK:           fir.store %[[VAL_35]] to %[[VAL_3]] : !fir.ref<index>
+! CHECK:           %[[VAL_36:.*]] = fir.allocmem !fir.array<1xi32>
+! CHECK:           %[[VAL_37:.*]] = arith.constant 1 : index
+! CHECK:           fir.store %[[VAL_37]] to %[[VAL_4]] : !fir.ref<index>
+! CHECK:           %[[VAL_38:.*]] = fir.load %[[VAL_5]] : !fir.ref<i32>
+! CHECK:           %[[VAL_39:.*]] = arith.constant 1 : index
+! CHECK:           %[[VAL_40:.*]] = fir.zero_bits !fir.ref<!fir.array<1xi32>>
+! CHECK:           %[[VAL_41:.*]] = fir.coordinate_of %[[VAL_40]], %[[VAL_39]] : (!fir.ref<!fir.array<1xi32>>, index) -> !fir.ref<i32>
+! CHECK:           %[[VAL_42:.*]] = fir.convert %[[VAL_41]] : (!fir.ref<i32>) -> index
+! CHECK:           %[[VAL_43:.*]] = fir.load %[[VAL_3]] : !fir.ref<index>
+! CHECK:           %[[VAL_44:.*]] = fir.load %[[VAL_4]] : !fir.ref<index>
+! CHECK:           %[[VAL_45:.*]] = arith.constant 1 : index
+! CHECK:           %[[VAL_46:.*]] = arith.addi %[[VAL_43]], %[[VAL_45]] : index
+! CHECK:           %[[VAL_47:.*]] = arith.cmpi sle, %[[VAL_44]], %[[VAL_46]] : index
+! CHECK:           %[[VAL_48:.*]] = fir.if %[[VAL_47]] -> (!fir.heap<!fir.array<1xi32>>) {
+! CHECK:             %[[VAL_49:.*]] = arith.constant 2 : index
+! CHECK:             %[[VAL_50:.*]] = arith.muli %[[VAL_46]], %[[VAL_49]] : index
+! CHECK:             fir.store %[[VAL_50]] to %[[VAL_4]] : !fir.ref<index>
+! CHECK:             %[[VAL_51:.*]] = arith.muli %[[VAL_50]], %[[VAL_42]] : index
+! CHECK:             %[[VAL_52:.*]] = fir.convert %[[VAL_36]] : (!fir.heap<!fir.array<1xi32>>) -> !fir.ref<i8>
+! CHECK:             %[[VAL_53:.*]] = fir.convert %[[VAL_51]] : (index) -> i64
+! CHECK:             %[[VAL_54:.*]] = fir.call @realloc(%[[VAL_52]], %[[VAL_53]]) : (!fir.ref<i8>, i64) -> !fir.ref<i8>
+! CHECK:             %[[VAL_55:.*]] = fir.convert %[[VAL_54]] : (!fir.ref<i8>) -> !fir.heap<!fir.array<1xi32>>
+! CHECK:             fir.result %[[VAL_55]] : !fir.heap<!fir.array<1xi32>>
+! CHECK:           } else {
+! CHECK:             fir.result %[[VAL_36]] : !fir.heap<!fir.array<1xi32>>
+! CHECK:           }
+! CHECK:           %[[VAL_56:.*]] = fir.coordinate_of %[[VAL_57:.*]], %[[VAL_43]] : (!fir.heap<!fir.array<1xi32>>, index) -> !fir.ref<i32>
+! CHECK:           fir.store %[[VAL_38]] to %[[VAL_56]] : !fir.ref<i32>
+! CHECK:           fir.store %[[VAL_46]] to %[[VAL_3]] : !fir.ref<index>
+! CHECK:           %[[VAL_58:.*]] = fir.load %[[VAL_3]] : !fir.ref<index>
+! CHECK:           %[[VAL_59:.*]] = fir.shape %[[VAL_58]] : (index) -> !fir.shape<1>
+! CHECK:           %[[VAL_60:.*]] = fir.array_load %[[VAL_57]](%[[VAL_59]]) : (!fir.heap<!fir.array<1xi32>>, !fir.shape<1>) -> !fir.array<1xi32>
+! CHECK:           %[[VAL_61:.*]] = fir.allocmem !fir.array<1xi32>
+! CHECK:           %[[VAL_62:.*]] = fir.shape %[[VAL_34]] : (index) -> !fir.shape<1>
+! CHECK:           %[[VAL_63:.*]] = fir.array_load %[[VAL_61]](%[[VAL_62]]) : (!fir.heap<!fir.array<1xi32>>, !fir.shape<1>) -> !fir.array<1xi32>
+! CHECK:           %[[VAL_64:.*]] = arith.constant 1 : index
+! CHECK:           %[[VAL_65:.*]] = arith.constant 0 : index
+! CHECK:           %[[VAL_66:.*]] = arith.subi %[[VAL_34]], %[[VAL_64]] : index
+! CHECK:           %[[VAL_67:.*]] = fir.do_loop %[[VAL_68:.*]] = %[[VAL_65]] to %[[VAL_66]] step %[[VAL_64]] unordered iter_args(%[[VAL_69:.*]] = %[[VAL_63]]) -> (!fir.array<1xi32>) {
+! CHECK:             %[[VAL_70:.*]] = fir.array_fetch %[[VAL_60]], %[[VAL_68]] : (!fir.array<1xi32>, index) -> i32
+! CHECK:             %[[VAL_71:.*]] = fir.array_update %[[VAL_69]], %[[VAL_70]], %[[VAL_68]] : (!fir.array<1xi32>, i32, index) -> !fir.array<1xi32>
+! CHECK:             fir.result %[[VAL_71]] : !fir.array<1xi32>
+! CHECK:           }
+! CHECK:           fir.array_merge_store %[[VAL_63]], %[[VAL_72:.*]] to %[[VAL_61]] : !fir.array<1xi32>, !fir.array<1xi32>, !fir.heap<!fir.array<1xi32>>
+! CHECK:           %[[VAL_73:.*]] = fir.shape %[[VAL_34]] : (index) -> !fir.shape<1>
+! CHECK:           %[[VAL_74:.*]] = fir.embox %[[VAL_61]](%[[VAL_73]]) : (!fir.heap<!fir.array<1xi32>>, !fir.shape<1>) -> !fir.box<!fir.array<1xi32>>
+! CHECK:           %[[VAL_75:.*]] = arith.constant 3 : i64
+! CHECK:           %[[VAL_76:.*]] = arith.constant 1 : i64
+! CHECK:           %[[VAL_77:.*]] = arith.subi %[[VAL_75]], %[[VAL_76]] : i64
+! CHECK:           %[[VAL_78:.*]] = arith.constant 1 : i64
+! CHECK:           %[[VAL_79:.*]] = arith.addi %[[VAL_77]], %[[VAL_78]] : i64
+! CHECK:           %[[VAL_80:.*]] = fir.convert %[[VAL_79]] : (i64) -> index
+! CHECK:           %[[VAL_81:.*]] = fir.call @llvm.stacksave() : () -> !fir.ref<i8>
+! CHECK:           %[[VAL_82:.*]] = fir.shape %[[VAL_80]] : (index) -> !fir.shape<1>
+! CHECK:           %[[VAL_83:.*]] = fir.convert %[[VAL_74]] : (!fir.box<!fir.array<1xi32>>) -> !fir.box<!fir.array<?xi32>>
+! CHECK:           %[[VAL_84:.*]] = fir.call @_QFac2Pfunc(%[[VAL_83]]) : (!fir.box<!fir.array<?xi32>>) -> !fir.array<3xi32>
+! CHECK:           fir.save_result %[[VAL_84]] to %[[VAL_2]](%[[VAL_82]]) : !fir.array<3xi32>, !fir.ref<!fir.array<3xi32>>, !fir.shape<1>
+! CHECK:           %[[VAL_85:.*]] = fir.shape %[[VAL_80]] : (index) -> !fir.shape<1>
+! CHECK:           %[[VAL_86:.*]] = fir.array_load %[[VAL_2]](%[[VAL_85]]) : (!fir.ref<!fir.array<3xi32>>, !fir.shape<1>) -> !fir.array<3xi32>
+! CHECK:           %[[VAL_87:.*]] = arith.constant 1 : index
+! CHECK:           %[[VAL_88:.*]] = arith.constant 0 : index
+! CHECK:           %[[VAL_89:.*]] = arith.subi %[[VAL_33]], %[[VAL_87]] : index
+! CHECK:           %[[VAL_90:.*]] = fir.do_loop %[[VAL_91:.*]] = %[[VAL_88]] to %[[VAL_89]] step %[[VAL_87]] unordered iter_args(%[[VAL_92:.*]] = %[[VAL_15]]) -> (!fir.array<?xi32>) {
+! CHECK:             %[[VAL_93:.*]] = fir.array_fetch %[[VAL_86]], %[[VAL_91]] : (!fir.array<3xi32>, index) -> i32
+! CHECK:             %[[VAL_94:.*]] = arith.subi %[[VAL_20]], %[[VAL_17]] : index
+! CHECK:             %[[VAL_95:.*]] = arith.muli %[[VAL_91]], %[[VAL_22]] : index
+! CHECK:             %[[VAL_96:.*]] = arith.addi %[[VAL_94]], %[[VAL_95]] : index
+! CHECK:             %[[VAL_97:.*]] = fir.array_update %[[VAL_92]], %[[VAL_93]], %[[VAL_96]] : (!fir.array<?xi32>, i32, index) -> !fir.array<?xi32>
+! CHECK:             fir.result %[[VAL_97]] : !fir.array<?xi32>
+! CHECK:           }
+! CHECK:           fir.call @llvm.stackrestore(%[[VAL_81]]) : (!fir.ref<i8>) -> ()
+! CHECK:           fir.freemem %[[VAL_61]] : !fir.heap<!fir.array<1xi32>>
+! CHECK:           fir.freemem %[[VAL_57]] : !fir.heap<!fir.array<1xi32>>
+! CHECK:           fir.result %[[VAL_98:.*]] : !fir.array<?xi32>
+! CHECK:         }
+! CHECK:         fir.array_merge_store %[[VAL_12]], %[[VAL_99:.*]] to %[[VAL_0]] : !fir.array<?xi32>, !fir.array<?xi32>, !fir.box<!fir.array<?xi32>>
+! CHECK:         return
+! CHECK:       }
+
+! CHECK-LABEL: func @_QFac2Pfunc(
+! CHECK-SAME:                    %[[VAL_0:.*]]: !fir.box<!fir.array<?xi32>> {fir.bindc_name = "a"}) -> !fir.array<3xi32> {
+! CHECK:         %[[VAL_1:.*]] = arith.constant 3 : index
+! CHECK:         %[[VAL_2:.*]] = fir.alloca !fir.array<3xi32> {bindc_name = "func", uniq_name = "_QFfuncEfunc"}
+! CHECK:         %[[VAL_3:.*]] = fir.shape %[[VAL_1]] : (index) -> !fir.shape<1>
+! CHECK:         %[[VAL_4:.*]] = fir.array_load %[[VAL_2]](%[[VAL_3]]) : (!fir.ref<!fir.array<3xi32>>, !fir.shape<1>) -> !fir.array<3xi32>
+! CHECK:         %[[VAL_5:.*]] = arith.constant 1 : i64
+! CHECK:         %[[VAL_6:.*]] = fir.convert %[[VAL_5]] : (i64) -> index
+! CHECK:         %[[VAL_7:.*]] = arith.constant 1 : i64
+! CHECK:         %[[VAL_8:.*]] = fir.convert %[[VAL_7]] : (i64) -> index
+! CHECK:         %[[VAL_9:.*]] = arith.constant 3 : i64
+! CHECK:         %[[VAL_10:.*]] = fir.convert %[[VAL_9]] : (i64) -> index
+! CHECK:         %[[VAL_11:.*]] = fir.slice %[[VAL_6]], %[[VAL_10]], %[[VAL_8]] : (index, index, index) -> !fir.slice<1>
+! CHECK:         %[[VAL_12:.*]] = fir.array_load %[[VAL_0]] {{\[}}%[[VAL_11]]] : (!fir.box<!fir.array<?xi32>>, !fir.slice<1>) -> !fir.array<?xi32>
+! CHECK:         %[[VAL_13:.*]] = arith.constant 1 : index
+! CHECK:         %[[VAL_14:.*]] = arith.constant 0 : index
+! CHECK:         %[[VAL_15:.*]] = arith.subi %[[VAL_1]], %[[VAL_13]] : index
+! CHECK:         %[[VAL_16:.*]] = fir.do_loop %[[VAL_17:.*]] = %[[VAL_14]] to %[[VAL_15]] step %[[VAL_13]] unordered iter_args(%[[VAL_18:.*]] = %[[VAL_4]]) -> (!fir.array<3xi32>) {
+! CHECK:           %[[VAL_19:.*]] = fir.array_fetch %[[VAL_12]], %[[VAL_17]] : (!fir.array<?xi32>, index) -> i32
+! CHECK:           %[[VAL_20:.*]] = fir.array_update %[[VAL_18]], %[[VAL_19]], %[[VAL_17]] : (!fir.array<3xi32>, i32, index) -> !fir.array<3xi32>
+! CHECK:           fir.result %[[VAL_20]] : !fir.array<3xi32>
+! CHECK:         }
+! CHECK:         fir.array_merge_store %[[VAL_4]], %[[VAL_21:.*]] to %[[VAL_2]] : !fir.array<3xi32>, !fir.array<3xi32>, !fir.ref<!fir.array<3xi32>>
+! CHECK:         %[[VAL_22:.*]] = fir.load %[[VAL_2]] : !fir.ref<!fir.array<3xi32>>
+! CHECK:         return %[[VAL_22]] : !fir.array<3xi32>
+! CHECK:       }
diff --git a/flang/test/Lower/forall/forall-2.f90 b/flang/test/Lower/forall/forall-2.f90
index 9f4ff6ed6c2e..1a07b082f598 100644
--- a/flang/test/Lower/forall/forall-2.f90
+++ b/flang/test/Lower/forall/forall-2.f90
@@ -131,13 +131,13 @@ end subroutine forall_pointer_assign
 ! CHECK:           %[[VAL_30:.*]] = fir.convert %[[VAL_29]] : (i32) -> i64
 ! CHECK:           %[[VAL_31:.*]] = fir.convert %[[VAL_30]] : (i64) -> index
 ! CHECK:           %[[VAL_32:.*]] = arith.subi %[[VAL_31]], %[[VAL_15]] : index
+! CHECK:           %[[VAL_39:.*]] = fir.load %[[VAL_0]] : !fir.ref<i32>
+! CHECK:           %[[VAL_40:.*]] = arith.constant 0 : i32
+! CHECK:           %[[VAL_41:.*]] = arith.subi %[[VAL_40]], %[[VAL_39]] : i32
 ! CHECK:           %[[VAL_33:.*]] = arith.constant 1 : index
 ! CHECK:           %[[VAL_34:.*]] = arith.constant 0 : index
 ! CHECK:           %[[VAL_35:.*]] = arith.subi %[[VAL_28]], %[[VAL_33]] : index
 ! CHECK:           %[[VAL_36:.*]] = fir.do_loop %[[VAL_37:.*]] = %[[VAL_34]] to %[[VAL_35]] step %[[VAL_33]] unordered iter_args(%[[VAL_38:.*]] = %[[VAL_13]]) -> (!fir.array<10x10xi32>) {
-! CHECK:             %[[VAL_39:.*]] = fir.load %[[VAL_0]] : !fir.ref<i32>
-! CHECK:             %[[VAL_40:.*]] = arith.constant 0 : i32
-! CHECK:             %[[VAL_41:.*]] = arith.subi %[[VAL_40]], %[[VAL_39]] : i32
 ! CHECK:             %[[VAL_42:.*]] = arith.subi %[[VAL_17]], %[[VAL_15]] : index
 ! CHECK:             %[[VAL_43:.*]] = arith.muli %[[VAL_37]], %[[VAL_19]] : index
 ! CHECK:             %[[VAL_44:.*]] = arith.addi %[[VAL_42]], %[[VAL_43]] : index
@@ -174,9 +174,8 @@ end subroutine slice_with_explicit_iters
 ! CHECK:         %[[VAL_13:.*]] = fir.do_loop %[[VAL_14:.*]] = %[[VAL_7]] to %[[VAL_9]] step %[[VAL_10]] unordered iter_args(%[[VAL_15:.*]] = %[[VAL_12]]) -> (!fir.array<1xi32>) {
 ! CHECK:           %[[VAL_16:.*]] = fir.convert %[[VAL_14]] : (index) -> i32
 ! CHECK:           fir.store %[[VAL_16]] to %[[VAL_2]] : !fir.ref<i32>
-! CHECK:           %[[VAL_17:.*]] = arith.constant 1 : i32
-! CHECK:           %[[VAL_18:.*]] = arith.constant 1 : index
-! CHECK:           %[[VAL_19:.*]] = arith.constant 1 : i64
+! CHECK-DAG:       %[[VAL_18:.*]] = arith.constant 1 : index
+! CHECK-DAG:       %[[VAL_19:.*]] = arith.constant 1 : i64
 ! CHECK:           %[[VAL_20:.*]] = fir.convert %[[VAL_19]] : (i64) -> index
 ! CHECK:           %[[VAL_21:.*]] = arith.addi %[[VAL_18]], %[[VAL_4]] : index
 ! CHECK:           %[[VAL_22:.*]] = arith.subi %[[VAL_21]], %[[VAL_18]] : index
@@ -187,6 +186,7 @@ end subroutine slice_with_explicit_iters
 ! CHECK:           %[[VAL_27:.*]] = fir.slice %[[VAL_18]], %[[VAL_22]], %[[VAL_20]], %[[VAL_24]], %[[VAL_25]], %[[VAL_25]] : (index, index, index, i64, index, index) -> !fir.slice<2>
 ! CHECK:           %[[VAL_28:.*]] = fir.embox %[[VAL_1]](%[[VAL_26]]) {{\[}}%[[VAL_27]]] : (!fir.ref<!fir.array<2x2xi32>>, !fir.shape<2>, !fir.slice<2>) -> !fir.box<!fir.array<?xi32>>
 ! CHECK:           %[[VAL_29:.*]] = fir.call @_QPe(%[[VAL_28]]) : (!fir.box<!fir.array<?xi32>>) -> i32
+! CHECK:           %[[VAL_17:.*]] = arith.constant 1 : i32
 ! CHECK:           %[[VAL_30:.*]] = arith.addi %[[VAL_29]], %[[VAL_17]] : i32
 ! CHECK:           %[[VAL_31:.*]] = arith.constant 1 : index
 ! CHECK:           %[[VAL_32:.*]] = fir.load %[[VAL_2]] : !fir.ref<i32>
diff --git a/flang/test/Lower/forall/forall-construct-2.f90 b/flang/test/Lower/forall/forall-construct-2.f90
index dff089837555..b82935cf8b3b 100644
--- a/flang/test/Lower/forall/forall-construct-2.f90
+++ b/flang/test/Lower/forall/forall-construct-2.f90
@@ -52,6 +52,7 @@ end subroutine test2_forall_construct
 ! CHECK:             %[[VAL_40:.*]] = fir.convert %[[VAL_39]] : (i32) -> i64
 ! CHECK:             %[[VAL_41:.*]] = fir.convert %[[VAL_40]] : (i64) -> index
 ! CHECK:             %[[VAL_42:.*]] = arith.subi %[[VAL_41]], %[[VAL_34]] : index
+! CHECK:             %[[VAL_54:.*]] = fir.array_fetch %[[VAL_23]], %[[VAL_38]], %[[VAL_42]] : (!fir.array<200x200xf32>, index, index) -> f32
 ! CHECK:             %[[VAL_43:.*]] = arith.constant 1 : index
 ! CHECK-DAG:         %[[VAL_44:.*]] = fir.load %[[VAL_5]] : !fir.ref<i32>
 ! CHECK-DAG:         %[[VAL_45:.*]] = arith.constant 1 : i32
@@ -63,7 +64,6 @@ end subroutine test2_forall_construct
 ! CHECK:             %[[VAL_51:.*]] = fir.convert %[[VAL_50]] : (i32) -> i64
 ! CHECK:             %[[VAL_52:.*]] = fir.convert %[[VAL_51]] : (i64) -> index
 ! CHECK:             %[[VAL_53:.*]] = arith.subi %[[VAL_52]], %[[VAL_43]] : index
-! CHECK:             %[[VAL_54:.*]] = fir.array_fetch %[[VAL_23]], %[[VAL_38]], %[[VAL_42]] : (!fir.array<200x200xf32>, index, index) -> f32
 ! CHECK:             %[[VAL_55:.*]] = fir.array_fetch %[[VAL_25]], %[[VAL_49]], %[[VAL_53]] : (!fir.array<200x200xf32>, index, index) -> f32
 ! CHECK:             %[[VAL_56:.*]] = arith.addf %[[VAL_54]], %[[VAL_55]] : f32
 ! CHECK:             %[[VAL_57:.*]] = arith.constant 1 : index
diff --git a/flang/test/Lower/forall/forall-construct-3.f90 b/flang/test/Lower/forall/forall-construct-3.f90
index 823b18872b77..9d3c0405d8d6 100644
--- a/flang/test/Lower/forall/forall-construct-3.f90
+++ b/flang/test/Lower/forall/forall-construct-3.f90
@@ -65,6 +65,7 @@ end subroutine test3_forall_construct
 ! CHECK:               %[[VAL_53:.*]] = fir.convert %[[VAL_52]] : (i32) -> i64
 ! CHECK:               %[[VAL_54:.*]] = fir.convert %[[VAL_53]] : (i64) -> index
 ! CHECK:               %[[VAL_55:.*]] = arith.subi %[[VAL_54]], %[[VAL_47]] : index
+! CHECK:               %[[VAL_67:.*]] = fir.array_fetch %[[VAL_24]], %[[VAL_51]], %[[VAL_55]] : (!fir.array<200x200xf32>, index, index) -> f32
 ! CHECK:               %[[VAL_56:.*]] = arith.constant 1 : index
 ! CHECK-DAG:           %[[VAL_57:.*]] = fir.load %[[VAL_6]] : !fir.ref<i32>
 ! CHECK-DAG:           %[[VAL_58:.*]] = arith.constant 1 : i32
@@ -76,7 +77,6 @@ end subroutine test3_forall_construct
 ! CHECK:               %[[VAL_64:.*]] = fir.convert %[[VAL_63]] : (i32) -> i64
 ! CHECK:               %[[VAL_65:.*]] = fir.convert %[[VAL_64]] : (i64) -> index
 ! CHECK:               %[[VAL_66:.*]] = arith.subi %[[VAL_65]], %[[VAL_56]] : index
-! CHECK:               %[[VAL_67:.*]] = fir.array_fetch %[[VAL_24]], %[[VAL_51]], %[[VAL_55]] : (!fir.array<200x200xf32>, index, index) -> f32
 ! CHECK:               %[[VAL_68:.*]] = fir.array_fetch %[[VAL_26]], %[[VAL_62]], %[[VAL_66]] : (!fir.array<200x200xf32>, index, index) -> f32
 ! CHECK:               %[[VAL_69:.*]] = arith.addf %[[VAL_67]], %[[VAL_68]] : f32
 ! CHECK:               %[[VAL_70:.*]] = arith.constant 1 : index
diff --git a/flang/test/Lower/forall/forall-construct.f90 b/flang/test/Lower/forall/forall-construct.f90
index 08415bfea1eb..3b3d4e2d33cf 100644
--- a/flang/test/Lower/forall/forall-construct.f90
+++ b/flang/test/Lower/forall/forall-construct.f90
@@ -71,8 +71,8 @@ end subroutine test_forall_construct
 ! CHECK:               %[[VAL_59:.*]] = fir.convert %[[VAL_58]] : (i32) -> i64
 ! CHECK:               %[[VAL_60:.*]] = fir.convert %[[VAL_59]] : (i64) -> index
 ! CHECK:               %[[VAL_61:.*]] = arith.subi %[[VAL_60]], %[[VAL_53]] : index
-! CHECK:               %[[VAL_62:.*]] = arith.constant 3.140000e+00 : f32
-! CHECK:               %[[VAL_63:.*]] = fir.array_fetch %[[VAL_31]], %[[VAL_57]], %[[VAL_61]] : (!fir.array<?x?xf32>, index, index) -> f32
+! CHECK-DAG:           %[[VAL_62:.*]] = arith.constant 3.140000e+00 : f32
+! CHECK-DAG:           %[[VAL_63:.*]] = fir.array_fetch %[[VAL_31]], %[[VAL_57]], %[[VAL_61]] : (!fir.array<?x?xf32>, index, index) -> f32
 ! CHECK:               %[[VAL_64:.*]] = arith.divf %[[VAL_63]], %[[VAL_62]] : f32
 ! CHECK:               %[[VAL_65:.*]] = arith.constant 1 : index
 ! CHECK:               %[[VAL_66:.*]] = fir.load %[[VAL_3]] : !fir.ref<i32>
diff --git a/flang/test/Lower/forall/forall-ranked.f90 b/flang/test/Lower/forall/forall-ranked.f90
index 4af3fa3fda24..dd492c45bafa 100644
--- a/flang/test/Lower/forall/forall-ranked.f90
+++ b/flang/test/Lower/forall/forall-ranked.f90
@@ -40,11 +40,11 @@
 ! CHECK:           %[[VAL_36:.*]] = fir.convert %[[VAL_35]] : (i32) -> i64
 ! CHECK:           %[[VAL_37:.*]] = fir.convert %[[VAL_36]] : (i64) -> index
 ! CHECK:           %[[VAL_38:.*]] = arith.subi %[[VAL_37]], %[[VAL_16]] : index
+! CHECK:           %[[VAL_45:.*]] = fir.call @_QPf(%[[VAL_0]]) : (!fir.ref<i32>) -> i32
 ! CHECK:           %[[VAL_39:.*]] = arith.constant 1 : index
 ! CHECK:           %[[VAL_40:.*]] = arith.constant 0 : index
 ! CHECK:           %[[VAL_41:.*]] = arith.subi %[[VAL_31]], %[[VAL_39]] : index
 ! CHECK:           %[[VAL_42:.*]] = fir.do_loop %[[VAL_43:.*]] = %[[VAL_40]] to %[[VAL_41]] step %[[VAL_39]] unordered iter_args(%[[VAL_44:.*]] = %[[VAL_13]]) -> (!fir.array<10x10x!fir.type<_QFtest_forall_with_ranked_dimensionTt{arr:!fir.array<11xi32>}>>) {
-! CHECK:             %[[VAL_45:.*]] = fir.call @_QPf(%[[VAL_0]]) : (!fir.ref<i32>) -> i32
 ! CHECK:             %[[VAL_46:.*]] = arith.subi %[[VAL_17]], %[[VAL_17]] : index
 ! CHECK:             %[[VAL_47:.*]] = arith.muli %[[VAL_43]], %[[VAL_23]] : index
 ! CHECK:             %[[VAL_48:.*]] = arith.addi %[[VAL_46]], %[[VAL_47]] : index
diff --git a/flang/test/Lower/forall/forall-slice.f90 b/flang/test/Lower/forall/forall-slice.f90
index 77f977ca2b8f..e34303873757 100644
--- a/flang/test/Lower/forall/forall-slice.f90
+++ b/flang/test/Lower/forall/forall-slice.f90
@@ -53,11 +53,11 @@
 ! CHECK:             %[[VAL_50:.*]] = arith.divsi %[[VAL_49]], %[[VAL_43]] : index
 ! CHECK:             %[[VAL_51:.*]] = arith.cmpi sgt, %[[VAL_50]], %[[VAL_47]] : index
 ! CHECK:             %[[VAL_52:.*]] = arith.select %[[VAL_51]], %[[VAL_50]], %[[VAL_47]] : index
+! CHECK:             %[[VAL_59:.*]] = fir.call @_QPf(%[[VAL_3]]) : (!fir.ref<i32>) -> i32
 ! CHECK:             %[[VAL_53:.*]] = arith.constant 1 : index
 ! CHECK:             %[[VAL_54:.*]] = arith.constant 0 : index
 ! CHECK:             %[[VAL_55:.*]] = arith.subi %[[VAL_52]], %[[VAL_53]] : index
 ! CHECK:             %[[VAL_56:.*]] = fir.do_loop %[[VAL_57:.*]] = %[[VAL_54]] to %[[VAL_55]] step %[[VAL_53]] unordered iter_args(%[[VAL_58:.*]] = %[[VAL_25]]) -> (!fir.array<10x10x!fir.type<_QFtest_forall_with_sliceTt{arr:!fir.array<11xi32>}>>) {
-! CHECK:               %[[VAL_59:.*]] = fir.call @_QPf(%[[VAL_3]]) : (!fir.ref<i32>) -> i32
 ! CHECK:               %[[VAL_60:.*]] = arith.subi %[[VAL_40]], %[[VAL_37]] : index
 ! CHECK:               %[[VAL_61:.*]] = arith.muli %[[VAL_57]], %[[VAL_43]] : index
 ! CHECK:               %[[VAL_62:.*]] = arith.addi %[[VAL_60]], %[[VAL_61]] : index
diff --git a/flang/test/Lower/forall/test9.f90 b/flang/test/Lower/forall/test9.f90
index 6ac862b5cb39..c8a8e135ef99 100644
--- a/flang/test/Lower/forall/test9.f90
+++ b/flang/test/Lower/forall/test9.f90
@@ -16,59 +16,61 @@ subroutine test9(a,b,n)
 end subroutine test9
 
 ! CHECK-LABEL: func @_QPtest9(
-! CHECK-SAME:    %[[VAL_0:.*]]: !fir.ref<!fir.array<?xf32>>{{.*}}, %[[VAL_1:.*]]: !fir.ref<!fir.array<?xf32>>{{.*}}, %[[VAL_2:.*]]: !fir.ref<i32>{{.*}}) {
+! CHECK-SAME:                 %[[VAL_0:.*]]: !fir.ref<!fir.array<?xf32>> {fir.bindc_name = "a"},
+! CHECK-SAME:                 %[[VAL_1:.*]]: !fir.ref<!fir.array<?xf32>> {fir.bindc_name = "b"},
+! CHECK-SAME:                 %[[VAL_2:.*]]: !fir.ref<i32> {fir.bindc_name = "n"}) {
 ! CHECK:         %[[VAL_3:.*]] = fir.alloca i32 {adapt.valuebyref, bindc_name = "i"}
 ! CHECK:         %[[VAL_4:.*]] = fir.load %[[VAL_2]] : !fir.ref<i32>
 ! CHECK:         %[[VAL_5:.*]] = fir.convert %[[VAL_4]] : (i32) -> i64
-! CHECK:         %[[VAL_6A:.*]] = fir.convert %[[VAL_5]] : (i64) -> index
-! CHECK:         %[[C0:.*]] = arith.constant 0 : index
-! CHECK:         %[[CMP:.*]] = arith.cmpi sgt, %[[VAL_6A]], %[[C0]] : index
-! CHECK:         %[[VAL_6:.*]] = arith.select %[[CMP]], %[[VAL_6A]], %[[C0]] : index
-! CHECK:         %[[VAL_7:.*]] = fir.load %[[VAL_2]] : !fir.ref<i32>
-! CHECK:         %[[VAL_8:.*]] = fir.convert %[[VAL_7]] : (i32) -> i64
-! CHECK:         %[[VAL_9A:.*]] = fir.convert %[[VAL_8]] : (i64) -> index
-! CHECK:         %[[C0_2:.*]] = arith.constant 0 : index
-! CHECK:         %[[CMP_2:.*]] = arith.cmpi sgt, %[[VAL_9A]], %[[C0_2]] : index
-! CHECK:         %[[VAL_9:.*]] = arith.select %[[CMP_2]], %[[VAL_9A]], %[[C0_2]] : index
-! CHECK:         %[[VAL_10:.*]] = arith.constant 1 : i32
-! CHECK:         %[[VAL_11:.*]] = fir.convert %[[VAL_10]] : (i32) -> index
-! CHECK:         %[[VAL_12:.*]] = fir.load %[[VAL_2]] : !fir.ref<i32>
-! CHECK:         %[[VAL_13:.*]] = arith.constant 1 : i32
-! CHECK:         %[[VAL_14:.*]] = arith.subi %[[VAL_12]], %[[VAL_13]] : i32
-! CHECK:         %[[VAL_15:.*]] = fir.convert %[[VAL_14]] : (i32) -> index
-! CHECK:         %[[VAL_16:.*]] = arith.constant 1 : index
-! CHECK:         %[[VAL_17:.*]] = fir.shape %[[VAL_6]] : (index) -> !fir.shape<1>
-! CHECK:         %[[VAL_18:.*]] = fir.array_load %[[VAL_0]](%[[VAL_17]]) : (!fir.ref<!fir.array<?xf32>>, !fir.shape<1>) -> !fir.array<?xf32>
-! CHECK:         %[[VAL_19:.*]] = fir.shape %[[VAL_6]] : (index) -> !fir.shape<1>
-! CHECK:         %[[VAL_20:.*]] = fir.array_load %[[VAL_0]](%[[VAL_19]]) : (!fir.ref<!fir.array<?xf32>>, !fir.shape<1>) -> !fir.array<?xf32>
-! CHECK:         %[[VAL_21:.*]] = fir.shape %[[VAL_9]] : (index) -> !fir.shape<1>
-! CHECK:         %[[VAL_22:.*]] = fir.array_load %[[VAL_1]](%[[VAL_21]]) : (!fir.ref<!fir.array<?xf32>>, !fir.shape<1>) -> !fir.array<?xf32>
-! CHECK:         %[[VAL_23:.*]] = fir.do_loop %[[VAL_24:.*]] = %[[VAL_11]] to %[[VAL_15]] step %[[VAL_16]] unordered iter_args(%[[VAL_25:.*]] = %[[VAL_18]]) -> (!fir.array<?xf32>) {
-! CHECK:           %[[VAL_26:.*]] = fir.convert %[[VAL_24]] : (index) -> i32
-! CHECK:           fir.store %[[VAL_26]] to %[[VAL_3]] : !fir.ref<i32>
-! CHECK:           %[[VAL_27:.*]] = arith.constant 1 : index
-! CHECK:           %[[VAL_28:.*]] = fir.load %[[VAL_3]] : !fir.ref<i32>
-! CHECK:           %[[VAL_29:.*]] = fir.convert %[[VAL_28]] : (i32) -> i64
-! CHECK:           %[[VAL_30:.*]] = fir.convert %[[VAL_29]] : (i64) -> index
-! CHECK:           %[[VAL_31:.*]] = arith.subi %[[VAL_30]], %[[VAL_27]] : index
-! CHECK:           %[[VAL_32:.*]] = arith.constant 1 : index
-! CHECK:           %[[VAL_33:.*]] = fir.load %[[VAL_3]] : !fir.ref<i32>
-! CHECK:           %[[VAL_34:.*]] = fir.convert %[[VAL_33]] : (i32) -> i64
-! CHECK:           %[[VAL_35:.*]] = fir.convert %[[VAL_34]] : (i64) -> index
-! CHECK:           %[[VAL_36:.*]] = arith.subi %[[VAL_35]], %[[VAL_32]] : index
-! CHECK:           %[[VAL_37:.*]] = fir.array_fetch %[[VAL_20]], %[[VAL_31]] : (!fir.array<?xf32>, index) -> f32
-! CHECK:           %[[VAL_38:.*]] = fir.array_fetch %[[VAL_22]], %[[VAL_36]] : (!fir.array<?xf32>, index) -> f32
-! CHECK:           %[[VAL_39:.*]] = arith.addf %[[VAL_37]], %[[VAL_38]] : f32
-! CHECK:           %[[VAL_40:.*]] = arith.constant 1 : index
-! CHECK-DAG:       %[[VAL_41:.*]] = fir.load %[[VAL_3]] : !fir.ref<i32>
-! CHECK-DAG:       %[[VAL_42:.*]] = arith.constant 1 : i32
-! CHECK:           %[[VAL_43:.*]] = arith.addi %[[VAL_41]], %[[VAL_42]] : i32
-! CHECK:           %[[VAL_44:.*]] = fir.convert %[[VAL_43]] : (i32) -> i64
-! CHECK:           %[[VAL_45:.*]] = fir.convert %[[VAL_44]] : (i64) -> index
-! CHECK:           %[[VAL_46:.*]] = arith.subi %[[VAL_45]], %[[VAL_40]] : index
-! CHECK:           %[[VAL_47:.*]] = fir.array_update %[[VAL_25]], %[[VAL_39]], %[[VAL_46]] : (!fir.array<?xf32>, f32, index) -> !fir.array<?xf32>
-! CHECK:           fir.result %[[VAL_47]] : !fir.array<?xf32>
+! CHECK:         %[[VAL_6:.*]] = fir.convert %[[VAL_5]] : (i64) -> index
+! CHECK:         %[[VAL_7:.*]] = arith.constant 0 : index
+! CHECK:         %[[VAL_8:.*]] = arith.cmpi sgt, %[[VAL_6]], %[[VAL_7]] : index
+! CHECK:         %[[VAL_9:.*]] = arith.select %[[VAL_8]], %[[VAL_6]], %[[VAL_7]] : index
+! CHECK:         %[[VAL_10:.*]] = fir.load %[[VAL_2]] : !fir.ref<i32>
+! CHECK:         %[[VAL_11:.*]] = fir.convert %[[VAL_10]] : (i32) -> i64
+! CHECK:         %[[VAL_12:.*]] = fir.convert %[[VAL_11]] : (i64) -> index
+! CHECK:         %[[VAL_13:.*]] = arith.constant 0 : index
+! CHECK:         %[[VAL_14:.*]] = arith.cmpi sgt, %[[VAL_12]], %[[VAL_13]] : index
+! CHECK:         %[[VAL_15:.*]] = arith.select %[[VAL_14]], %[[VAL_12]], %[[VAL_13]] : index
+! CHECK:         %[[VAL_16:.*]] = arith.constant 1 : i32
+! CHECK:         %[[VAL_17:.*]] = fir.convert %[[VAL_16]] : (i32) -> index
+! CHECK:         %[[VAL_18:.*]] = fir.load %[[VAL_2]] : !fir.ref<i32>
+! CHECK:         %[[VAL_19:.*]] = arith.constant 1 : i32
+! CHECK:         %[[VAL_20:.*]] = arith.subi %[[VAL_18]], %[[VAL_19]] : i32
+! CHECK:         %[[VAL_21:.*]] = fir.convert %[[VAL_20]] : (i32) -> index
+! CHECK:         %[[VAL_22:.*]] = arith.constant 1 : index
+! CHECK:         %[[VAL_23:.*]] = fir.shape %[[VAL_9]] : (index) -> !fir.shape<1>
+! CHECK:         %[[VAL_24:.*]] = fir.array_load %[[VAL_0]](%[[VAL_23]]) : (!fir.ref<!fir.array<?xf32>>, !fir.shape<1>) -> !fir.array<?xf32>
+! CHECK:         %[[VAL_25:.*]] = fir.shape %[[VAL_9]] : (index) -> !fir.shape<1>
+! CHECK:         %[[VAL_26:.*]] = fir.array_load %[[VAL_0]](%[[VAL_25]]) : (!fir.ref<!fir.array<?xf32>>, !fir.shape<1>) -> !fir.array<?xf32>
+! CHECK:         %[[VAL_27:.*]] = fir.shape %[[VAL_15]] : (index) -> !fir.shape<1>
+! CHECK:         %[[VAL_28:.*]] = fir.array_load %[[VAL_1]](%[[VAL_27]]) : (!fir.ref<!fir.array<?xf32>>, !fir.shape<1>) -> !fir.array<?xf32>
+! CHECK:         %[[VAL_29:.*]] = fir.do_loop %[[VAL_30:.*]] = %[[VAL_17]] to %[[VAL_21]] step %[[VAL_22]] unordered iter_args(%[[VAL_31:.*]] = %[[VAL_24]]) -> (!fir.array<?xf32>) {
+! CHECK:           %[[VAL_32:.*]] = fir.convert %[[VAL_30]] : (index) -> i32
+! CHECK:           fir.store %[[VAL_32]] to %[[VAL_3]] : !fir.ref<i32>
+! CHECK:           %[[VAL_33:.*]] = arith.constant 1 : index
+! CHECK:           %[[VAL_34:.*]] = fir.load %[[VAL_3]] : !fir.ref<i32>
+! CHECK:           %[[VAL_35:.*]] = fir.convert %[[VAL_34]] : (i32) -> i64
+! CHECK:           %[[VAL_36:.*]] = fir.convert %[[VAL_35]] : (i64) -> index
+! CHECK:           %[[VAL_37:.*]] = arith.subi %[[VAL_36]], %[[VAL_33]] : index
+! CHECK:           %[[VAL_38:.*]] = fir.array_fetch %[[VAL_26]], %[[VAL_37]] : (!fir.array<?xf32>, index) -> f32
+! CHECK:           %[[VAL_39:.*]] = arith.constant 1 : index
+! CHECK:           %[[VAL_40:.*]] = fir.load %[[VAL_3]] : !fir.ref<i32>
+! CHECK:           %[[VAL_41:.*]] = fir.convert %[[VAL_40]] : (i32) -> i64
+! CHECK:           %[[VAL_42:.*]] = fir.convert %[[VAL_41]] : (i64) -> index
+! CHECK:           %[[VAL_43:.*]] = arith.subi %[[VAL_42]], %[[VAL_39]] : index
+! CHECK:           %[[VAL_44:.*]] = fir.array_fetch %[[VAL_28]], %[[VAL_43]] : (!fir.array<?xf32>, index) -> f32
+! CHECK:           %[[VAL_45:.*]] = arith.addf %[[VAL_38]], %[[VAL_44]] : f32
+! CHECK:           %[[VAL_46:.*]] = arith.constant 1 : index
+! CHECK:           %[[VAL_47:.*]] = fir.load %[[VAL_3]] : !fir.ref<i32>
+! CHECK:           %[[VAL_48:.*]] = arith.constant 1 : i32
+! CHECK:           %[[VAL_49:.*]] = arith.addi %[[VAL_47]], %[[VAL_48]] : i32
+! CHECK:           %[[VAL_50:.*]] = fir.convert %[[VAL_49]] : (i32) -> i64
+! CHECK:           %[[VAL_51:.*]] = fir.convert %[[VAL_50]] : (i64) -> index
+! CHECK:           %[[VAL_52:.*]] = arith.subi %[[VAL_51]], %[[VAL_46]] : index
+! CHECK:           %[[VAL_53:.*]] = fir.array_update %[[VAL_31]], %[[VAL_45]], %[[VAL_52]] : (!fir.array<?xf32>, f32, index) -> !fir.array<?xf32>
+! CHECK:           fir.result %[[VAL_53]] : !fir.array<?xf32>
 ! CHECK:         }
-! CHECK:         fir.array_merge_store %[[VAL_18]], %[[VAL_48:.*]] to %[[VAL_0]] : !fir.array<?xf32>, !fir.array<?xf32>, !fir.ref<!fir.array<?xf32>>
+! CHECK:         fir.array_merge_store %[[VAL_24]], %[[VAL_54:.*]] to %[[VAL_0]] : !fir.array<?xf32>, !fir.array<?xf32>, !fir.ref<!fir.array<?xf32>>
 ! CHECK:         return
 ! CHECK:       }