[fir] Add fircg.ext_array_coor conversion

[clementval]

Summary

This patch adds the conversion pattern for the fircg.ext_array_coor
operation. It applies the address arithmetic on a dynamically shaped, shifted
and/or sliced array.

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

Co-authored-by: Eric Schweitz <[e-mail]>
Co-authored-by: Jean Perier <[e-mail]>

[clementval]

Remove shape, shapeshift, sshift and slice conversion pattern that are not part of this patch

[kiranchandramohan]

Some minor comments.

flang/lib/Optimizer/CodeGen/CodeGen.cpp

138–139

Does 0 and 7 have a name defined elsewhere?

184–204

If this is common code with embox then it might be better to add this patch as a child patch of embox.

1784

Nit: Is something missing before etc?

1870

Can we do an early return here and remove the following else?

1877

Notify failure.

1911

Notify failure.

[clementval]

Address comment + rebase

[clementval]

flang/lib/Optimizer/CodeGen/CodeGen.cpp

138–139

7 has one. 0 is the base for GEP.

184–204

I'll rebase once fir.embox is in.

[kiranchandramohan]

flang/lib/Optimizer/CodeGen/CodeGen.cpp

1881

Please replace back with TODO.
See discussion in https://reviews.llvm.org/D114104#inline-1089507

1916

Please replace back with TODO.
See discussion in https://reviews.llvm.org/D114104#inline-1089507

[clementval]

Replace TODO + rebase

[kiranchandramohan]

I have some Nit comments.

flang/lib/Optimizer/CodeGen/CodeGen.cpp

1801–1804

Nit: Since these are all for the for loop, would it be better to initialize it in the loop init section?

1891

Nit: off or offset?

1900

1901

Nit: I think we need a test for this case.

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

1595

Nit: I think we need a test with a box.

1614

Nit: Not used?

1622

Nit: Can we have a non-zero shift and slice? Also can the slice be of a different dimension (3) than the array (1)? Or have I misunderstood?

1637

Nit: Not used?

[clementval]

Address review comments

flang/lib/Optimizer/CodeGen/CodeGen.cpp

1801–1804

This doesn't work AFAIK or am I missing something?

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

1622

The slice for a rank has 3 operands (lower, upper and step).

[clementval]

ping

[kiranchandramohan]

I have commented inline about places where we seem to be using the original operands. Please have a look and if required, switch to using the converted operands.

flang/lib/Optimizer/CodeGen/CodeGen.cpp

141

Nit: 2 -> kDimStridePos ?

1801–1804

OK. Can you move it closer (just above) to the for loop?

1817–1819

I think the indexOps and shiftOps are the original operands. We have to use the converted operands here I think.

1834

Here also use converted.

1852

Use converted in place of shapeOps.

1934

This should be the converted operands and not the original.

[clementval]

flang/lib/Optimizer/CodeGen/CodeGen.cpp

1817–1819

I don't think it would make a difference here in the end and probably harder to read to code with operands and indexing. WDYT?

[clementval]

Address some of the review comments

[kiranchandramohan]

flang/lib/Optimizer/CodeGen/CodeGen.cpp

1817–1819

If this is causing no issues now then it is OK with me. It might cause an issue in the future. Can you either add a TODO or create a ticket to clean it up in the future?

For the XReboxOp I had to use the converted operands to make it work.
https://reviews.llvm.org/D114709

[clementval]

flang/lib/Optimizer/CodeGen/CodeGen.cpp

1817–1819

I'm just gonna update it then.

[clementval]

Use converted operands

[clementval]

Use converted operands

[kiranchandramohan]

LGTM.

flang/include/flang/Optimizer/CodeGen/CGOps.td

def fircg_XArrayCoorOp : fircg_Op<"ext_array_coor", [AttrSizedOperandSegments]> {
   let assemblyFormat = [{
     $memref (`(`$shape^`)`)? (`origin` $shift^)? (`[`$slice^`]`)?
       (`path` $subcomponent^)? `<`$indices`>` (`typeparams` $lenParams^)?
       attr-dict `:` functional-type(operands, results)
   }];
 
   let extraClassDeclaration = [{
     unsigned getRank();
+
+    // Shape is optional, but if it exists, it will be at offset 1.
+    unsigned shapeOffset() { return 1; }
+    unsigned shiftOffset() { return shapeOffset() + shape().size(); }
+    unsigned sliceOffset() { return shiftOffset() + shift().size(); }
+    unsigned subcomponentOffset() { return sliceOffset() + slice().size(); }
+    unsigned indicesOffset() {
+      return subcomponentOffset() + subcomponent().size();
+    }
+    unsigned lenParamsOffset() { return indicesOffset() + indices().size(); }
   }];
 }
 
 #endif

flang/lib/Optimizer/CodeGen/CodeGen.cpp

 //===-- CodeGen.cpp -- bridge to lower to LLVM ----------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
 // Coding style: https://mlir.llvm.org/getting_started/DeveloperGuide/
 //
 //===----------------------------------------------------------------------===//
 
 #include "flang/Optimizer/CodeGen/CodeGen.h"
+#include "CGOps.h"
 #include "PassDetail.h"
 #include "flang/ISO_Fortran_binding.h"
 #include "flang/Optimizer/Dialect/FIRAttr.h"
 #include "flang/Optimizer/Dialect/FIROps.h"
 #include "flang/Optimizer/Support/TypeCode.h"
 #include "mlir/Conversion/ArithmeticToLLVM/ArithmeticToLLVM.h"
 #include "mlir/Conversion/LLVMCommon/Pattern.h"
 #include "mlir/Conversion/StandardToLLVM/ConvertStandardToLLVM.h"
public:
   explicit FIROpConversion(fir::LLVMTypeConverter &lowering)
       : mlir::ConvertOpToLLVMPattern<FromOp>(lowering) {}
 
 protected:
   mlir::Type convertType(mlir::Type ty) const {
     return lowerTy().convertType(ty);
   }
 
+  mlir::Type getVoidPtrType() const {
+    return mlir::LLVM::LLVMPointerType::get(
+        mlir::IntegerType::get(&lowerTy().getContext(), 8));
+  }
+
   mlir::LLVM::ConstantOp
   genI32Constant(mlir::Location loc, mlir::ConversionPatternRewriter &rewriter,
                  int value) const {
     mlir::Type i32Ty = rewriter.getI32Type();
     mlir::IntegerAttr attr = rewriter.getI32IntegerAttr(value);
     return rewriter.create<mlir::LLVM::ConstantOp>(loc, i32Ty, attr);
   }
 
loadFromOffset(mlir::Location loc, mlir::Value a, mlir::LLVM::ConstantOp c0,
                  mlir::Type ty,
                  mlir::ConversionPatternRewriter &rewriter) const {
     auto pty = mlir::LLVM::LLVMPointerType::get(ty);
     mlir::LLVM::ConstantOp c = genConstantOffset(loc, rewriter, off);
     mlir::LLVM::GEPOp p = genGEP(loc, pty, rewriter, a, c0, cDims, dim, c);
     return rewriter.create<mlir::LLVM::LoadOp>(loc, ty, p);
   }
 
+  mlir::Value
+  loadStrideFromBox(mlir::Location loc, mlir::Value box, unsigned dim,
+                    mlir::ConversionPatternRewriter &rewriter) const {
+    auto idxTy = lowerTy().indexType();
+    auto c0 = genConstantOffset(loc, rewriter, 0);
+    auto cDims = genConstantOffset(loc, rewriter, kDimsPosInBox);
+    auto dimValue = genConstantIndex(loc, idxTy, rewriter, dim);
+    return loadFromOffset(loc, box, c0, cDims, dimValue, 2, idxTy, rewriter);
+  }
+
   /// Read base address from a fir.box. Returned address has type ty.
   mlir::Value
   loadBaseAddrFromBox(mlir::Location loc, mlir::Type ty, mlir::Value box,
                       mlir::ConversionPatternRewriter &rewriter) const {
     mlir::LLVM::ConstantOp c0 = genConstantOffset(loc, rewriter, 0);
     mlir::LLVM::ConstantOp cAddr =
         genConstantOffset(loc, rewriter, kAddrPosInBox);
     auto pty = mlir::LLVM::LLVMPointerType::get(ty);
mlir::LLVM::ConstantOp attrMask =
         genConstantOffset(loc, rewriter, maskValue);
     auto maskRes =
         rewriter.create<mlir::LLVM::AndOp>(loc, attrTy, attribute, attrMask);
     mlir::LLVM::ConstantOp c0 = genConstantOffset(loc, rewriter, 0);
     return rewriter.create<mlir::LLVM::ICmpOp>(
         loc, mlir::LLVM::ICmpPredicate::ne, maskRes, c0);
   }
 
   // Get the element type given an LLVM type that is of the form
   // [llvm.ptr](array|struct|vector)+ and the provided indexes.
   static mlir::Type getBoxEleTy(mlir::Type type,
                                 llvm::ArrayRef<unsigned> indexes) {
     if (auto t = type.dyn_cast<mlir::LLVM::LLVMPointerType>())
       type = t.getElementType();
     for (auto i : indexes) {
       if (auto t = type.dyn_cast<mlir::LLVM::LLVMStructType>()) {
         assert(!t.isOpaque() && i < t.getBody().size());
         type = t.getBody()[i];
       } else if (auto t = type.dyn_cast<mlir::LLVM::LLVMArrayType>()) {
         type = t.getElementType();
       } else if (auto t = type.dyn_cast<mlir::VectorType>()) {
         type = t.getElementType();
       } else {
         fir::emitFatalError(mlir::UnknownLoc::get(type.getContext()),
                             "request for invalid box element type");
       }
     }
     return type;
   }
 
+  // Return LLVM type of the base address given the LLVM type
+  // of the related descriptor (lowered fir.box type).
+  static mlir::Type getBaseAddrTypeFromBox(mlir::Type type) {
+    return getBoxEleTy(type, {kAddrPosInBox});
+  }
+
   template <typename... ARGS>
   mlir::LLVM::GEPOp genGEP(mlir::Location loc, mlir::Type ty,
                            mlir::ConversionPatternRewriter &rewriter,
                            mlir::Value base, ARGS... args) const {
     SmallVector<mlir::Value> cv{args...};
     return rewriter.create<mlir::LLVM::GEPOp>(loc, ty, base, cv);
   }
 
doRewrite(fir::InsertOnRangeOp range, mlir::Type ty, OpAdaptor adaptor,
     rewriter.replaceOpWithNewOp<mlir::LLVM::InsertValueOp>(
         range, ty, lastOp, insertVal,
         ArrayAttr::get(range.getContext(), arrayRef));
 
     return success();
   }
 };
 
+/// XArrayCoor is the address arithmetic on a dynamically shaped, sliced,
+/// shifted etc. array.
+/// (See the static restriction on coordinate_of.) array_coor determines the
+/// coordinate (location) of a specific element.
+struct XArrayCoorOpConversion
+    : public FIROpAndTypeConversion<fir::cg::XArrayCoorOp> {
+  using FIROpAndTypeConversion::FIROpAndTypeConversion;
+
+  mlir::LogicalResult
+  doRewrite(fir::cg::XArrayCoorOp coor, mlir::Type ty, OpAdaptor adaptor,
+            mlir::ConversionPatternRewriter &rewriter) const override {
+    auto loc = coor.getLoc();
+    unsigned rank = coor.getRank();
+    assert(coor.indices().size() == rank);
+    assert(coor.shape().empty() || coor.shape().size() == rank);
+    assert(coor.shift().empty() || coor.shift().size() == rank);
+    assert(coor.slice().empty() || coor.slice().size() == 3 * rank);
+    auto indexOps = coor.indices().begin();
+    auto shapeOps = coor.shape().begin();
+    auto shiftOps = coor.shift().begin();
+    auto sliceOps = coor.slice().begin();
+    mlir::Type idxTy = lowerTy().indexType();
+    mlir::Value one = genConstantIndex(loc, idxTy, rewriter, 1);
+    mlir::Value prevExt = one;
+    mlir::Value zero = genConstantIndex(loc, idxTy, rewriter, 0);
+    mlir::Value offset = zero;
+    const bool isShifted = !coor.shift().empty();
+    const bool isSliced = !coor.slice().empty();
+    const bool baseIsBoxed = coor.memref().getType().isa<fir::BoxType>();
+
+    // For each dimension of the array, generate the offset calculation.
+    for (unsigned i = 0; i < rank;
+         ++i, ++indexOps, ++shapeOps, ++shiftOps, sliceOps += 3) {
+      mlir::Value index = integerCast(loc, rewriter, idxTy, *indexOps);
+      mlir::Value lb =
+          isShifted ? integerCast(loc, rewriter, idxTy, *shiftOps) : one;
+      mlir::Value step = one;
+      bool normalSlice = isSliced;
+      // Compute zero based index in dimension i of the element, applying
+      // potential triplets and lower bounds.
+      if (isSliced) {
+        mlir::Value ub = *(sliceOps + 1);
+        normalSlice = !mlir::isa_and_nonnull<fir::UndefOp>(ub.getDefiningOp());
+        if (normalSlice)
+          step = integerCast(loc, rewriter, idxTy, *(sliceOps + 2));
+      }
+      auto idx = rewriter.create<mlir::LLVM::SubOp>(loc, idxTy, index, lb);
+      mlir::Value diff =
+          rewriter.create<mlir::LLVM::MulOp>(loc, idxTy, idx, step);
+      if (normalSlice) {
+        mlir::Value sliceLb = integerCast(loc, rewriter, idxTy, *sliceOps);
+        auto adj = rewriter.create<mlir::LLVM::SubOp>(loc, idxTy, sliceLb, lb);
+        diff = rewriter.create<mlir::LLVM::AddOp>(loc, idxTy, diff, adj);
+      }
+      // Update the offset given the stride and the zero based index `diff`
+      // that was just computed.
+      if (baseIsBoxed) {
+        // Use stride in bytes from the descriptor.
+        mlir::Value stride =
+            loadStrideFromBox(loc, adaptor.getOperands()[0], i, rewriter);
+        auto sc = rewriter.create<mlir::LLVM::MulOp>(loc, idxTy, diff, stride);
+        offset = rewriter.create<mlir::LLVM::AddOp>(loc, idxTy, sc, offset);
+      } else {
+        // Use stride computed at last iteration.
+        auto sc = rewriter.create<mlir::LLVM::MulOp>(loc, idxTy, diff, prevExt);
+        offset = rewriter.create<mlir::LLVM::AddOp>(loc, idxTy, sc, offset);
+        // Compute next stride assuming contiguity of the base array
+        // (in element number).
+        auto nextExt = integerCast(loc, rewriter, idxTy, *shapeOps);
+        prevExt =
+            rewriter.create<mlir::LLVM::MulOp>(loc, idxTy, prevExt, nextExt);
+      }
+    }
+
+    // Add computed offset to the base address.
+    if (baseIsBoxed) {
+      // Working with byte offsets. The base address is read from the fir.box.
+      // and need to be casted to i8* to do the pointer arithmetic.
+      mlir::Type baseTy =
+          getBaseAddrTypeFromBox(adaptor.getOperands()[0].getType());
+      mlir::Value base =
+          loadBaseAddrFromBox(loc, baseTy, adaptor.getOperands()[0], rewriter);
+      mlir::Type voidPtrTy = getVoidPtrType();
+      base = rewriter.create<mlir::LLVM::BitcastOp>(loc, voidPtrTy, base);
+      llvm::SmallVector<mlir::Value> args{base, offset};
+      auto addr = rewriter.create<mlir::LLVM::GEPOp>(loc, voidPtrTy, args);
+      if (coor.subcomponent().empty()) {
+        rewriter.replaceOpWithNewOp<mlir::LLVM::BitcastOp>(coor, baseTy, addr);
+        return success();
+      }
+      auto casted = rewriter.create<mlir::LLVM::BitcastOp>(loc, baseTy, addr);
+      args.clear();
+      args.push_back(casted);
+      args.push_back(zero);
+      if (!coor.lenParams().empty()) {
+        // If type parameters are present, then we don't want to use a GEPOp
+        // as below, as the LLVM struct type cannot be statically defined.
+        TODO(loc, "derived type with type parameters");
+      }
+      // TODO: array offset subcomponents must be converted to LLVM's
+      // row-major layout here.
+      for (auto i = coor.subcomponentOffset(); i != coor.indicesOffset(); ++i)
+        args.push_back(adaptor.getOperands()[i]);
+      rewriter.replaceOpWithNewOp<mlir::LLVM::GEPOp>(coor, baseTy, args);
+      return success();
+    }
+
+    // The array was not boxed, so it must be contiguous. offset is therefore an
+    // element offset and the base type is kept in the GEP unless the element
+    // type size is itself dynamic.
+    mlir::Value base;
+    if (coor.subcomponent().empty()) {
+      // No subcomponent.
+      if (!coor.lenParams().empty()) {
+        // Type parameters. Adjust element size explicitly.
+        auto eleTy = fir::dyn_cast_ptrEleTy(coor.getType());
+        assert(eleTy && "result must be a reference-like type");
+        if (fir::characterWithDynamicLen(eleTy)) {
+          assert(coor.lenParams().size() == 1);
+          auto bitsInChar = lowerTy().getKindMap().getCharacterBitsize(
+              eleTy.cast<fir::CharacterType>().getFKind());
+          auto scaling = genConstantIndex(loc, idxTy, rewriter, bitsInChar / 8);
+          auto scaledBySize =
+              rewriter.create<mlir::LLVM::MulOp>(loc, idxTy, offset, scaling);
+          auto length =
+              integerCast(loc, rewriter, idxTy,
+                          adaptor.getOperands()[coor.lenParamsOffset()]);
+          offset = rewriter.create<mlir::LLVM::MulOp>(loc, idxTy, scaledBySize,
+                                                      length);
+        } else {
+          TODO(loc, "compute size of derived type with type parameters");
+        }
+      }
+      // Cast the base address to a pointer to T.
+      base = rewriter.create<mlir::LLVM::BitcastOp>(loc, ty,
+                                                    adaptor.getOperands()[0]);
+    } else {
+      // Operand #0 must have a pointer type. For subcomponent slicing, we
+      // want to cast away the array type and have a plain struct type.
+      mlir::Type ty0 = adaptor.getOperands()[0].getType();
+      auto ptrTy = ty0.dyn_cast<mlir::LLVM::LLVMPointerType>();
+      assert(ptrTy && "expected pointer type");
+      mlir::Type eleTy = ptrTy.getElementType();
+      while (auto arrTy = eleTy.dyn_cast<mlir::LLVM::LLVMArrayType>())
+        eleTy = arrTy.getElementType();
+      auto newTy = mlir::LLVM::LLVMPointerType::get(eleTy);
+      base = rewriter.create<mlir::LLVM::BitcastOp>(loc, newTy,
+                                                    adaptor.getOperands()[0]);
+    }
+    SmallVector<mlir::Value> args = {base, offset};
+    args.append(coor.subcomponent().begin(), coor.subcomponent().end());
+    rewriter.replaceOpWithNewOp<mlir::LLVM::GEPOp>(coor, ty, args);
+    return success();
+  }
+};
+
 //
 // Primitive operations on Complex types
 //
 
 /// Generate inline code for complex addition/subtraction
 template <typename LLVMOP, typename OPTY>
 mlir::LLVM::InsertValueOp complexSum(OPTY sumop, mlir::ValueRange opnds,
                                      mlir::ConversionPatternRewriter &rewriter,
pattern.insert<
         GlobalLenOpConversion, GlobalOpConversion, InsertOnRangeOpConversion,
         InsertValueOpConversion, IsPresentOpConversion, LoadOpConversion,
         NegcOpConversion, NoReassocOpConversion, MulcOpConversion,
         SelectCaseOpConversion, SelectOpConversion, SelectRankOpConversion,
         SelectTypeOpConversion, ShapeOpConversion, ShapeShiftOpConversion,
         ShiftOpConversion, SliceOpConversion, StoreOpConversion,
         StringLitOpConversion, SubcOpConversion, UnboxCharOpConversion,
         UnboxProcOpConversion, UndefOpConversion, UnreachableOpConversion,
-        ZeroOpConversion>(typeConverter);
+        XArrayCoorOpConversion, ZeroOpConversion>(typeConverter);
     mlir::populateStdToLLVMConversionPatterns(typeConverter, pattern);
     mlir::arith::populateArithmeticToLLVMConversionPatterns(typeConverter,
                                                             pattern);
     mlir::ConversionTarget target{*context};
     target.addLegalDialect<mlir::LLVM::LLVMDialect>();
 
     // required NOPs for applying a full conversion
     target.addLegalOp<mlir::ModuleOp>();

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

 
 // 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
+
+// -----
+
+// Test `fircg.ext_array_coor` conversion.
+
+// Conversion with only shape and indice.
+
+func @ext_array_coor0(%arg0: !fir.ref<!fir.array<?xi32>>) {
+  %c0 = arith.constant 0 : i64
+  %1 = fircg.ext_array_coor %arg0(%c0) <%c0> : (!fir.ref<!fir.array<?xi32>>, i64, i64) -> !fir.ref<i32>
+  return
+}
+
+// CHECK-LABEL: llvm.func @ext_array_coor0(
+// CHECK-SAME:                             %[[ARG0:.*]]: !llvm.ptr<i32>)
+// CHECK:         %[[C0:.*]] = llvm.mlir.constant(0 : i64) : i64
+// CHECK:         %[[C1:.*]] = llvm.mlir.constant(1 : i64) : i64
+// CHECK:         %[[C0_1:.*]] = llvm.mlir.constant(0 : i64) : i64
+// CHECK:         %[[IDX:.*]] = llvm.sub %[[C0]], %[[C1]] : i64
+// CHECK:         %[[DIFF0:.*]] = llvm.mul %[[IDX]], %[[C1]] : i64
+// CHECK:         %[[SC:.*]] = llvm.mul %[[DIFF0]], %[[C1]]  : i64
+// CHECK:         %[[OFFSET:.*]] = llvm.add %[[SC]], %[[C0_1]]  : i64
+// CHECK:         %[[BITCAST:.*]] = llvm.bitcast %[[ARG0]] : !llvm.ptr<i32> to !llvm.ptr<i32>
+// CHECK:         %{{.*}} = llvm.getelementptr %[[BITCAST]][%[[OFFSET]]] : (!llvm.ptr<i32>, i64) -> !llvm.ptr<i32>
+
+// Conversion with shift and slice.
+
+func @ext_array_coor1(%arg0: !fir.ref<!fir.array<?xi32>>) {
+  %c0 = arith.constant 0 : i64
+  %1 = fircg.ext_array_coor %arg0(%c0) origin %c0[%c0, %c0, %c0]<%c0> : (!fir.ref<!fir.array<?xi32>>, i64, i64, i64, i64, i64, i64) -> !fir.ref<i32>
+  return
+}
+
+// CHECK-LABEL: llvm.func @ext_array_coor1(
+// CHECK-SAME:                             %[[ARG0:.*]]: !llvm.ptr<i32>)
+// CHECK:         %[[C0:.*]] = llvm.mlir.constant(0 : i64) : i64
+// CHECK:         %[[C1:.*]] = llvm.mlir.constant(1 : i64) : i64
+// CHECK:         %[[C0_1:.*]] = llvm.mlir.constant(0 : i64) : i64
+// CHECK:         %[[IDX:.*]] = llvm.sub %[[C0]], %[[C0]] : i64
+// CHECK:         %[[DIFF0:.*]] = llvm.mul %[[IDX]], %[[C0]] : i64
+// CHECK:         %[[ADJ:.*]] = llvm.sub %[[C0]], %[[C0]]  : i64
+// CHECK:         %[[DIFF1:.*]] = llvm.add %[[DIFF0]], %[[ADJ]]  : i64
+// CHECK:         %[[STRIDE:.*]] = llvm.mul %[[DIFF1]], %[[C1]]  : i64
+// CHECK:         %[[OFFSET:.*]] = llvm.add %[[STRIDE]], %[[C0_1]]  : i64
+// CHECK:         %[[BITCAST:.*]] = llvm.bitcast %[[ARG0]] : !llvm.ptr<i32> to !llvm.ptr<i32>
+// CHECK:         %{{.*}} = llvm.getelementptr %[[BITCAST]][%[[OFFSET]]] : (!llvm.ptr<i32>, i64) -> !llvm.ptr<i32>
+
+// Conversion for a dynamic length char.
+
+func @ext_array_coor2(%arg0: !fir.ref<!fir.array<?x!fir.char<1,?>>>) {
+  %c0 = arith.constant 0 : i64
+  %1 = fircg.ext_array_coor %arg0(%c0) <%c0> : (!fir.ref<!fir.array<?x!fir.char<1,?>>>, i64, i64) -> !fir.ref<i32>
+  return
+}
+
+// CHECK-LABEL: llvm.func @ext_array_coor2(
+// CHECK-SAME:                             %[[ARG0:.*]]: !llvm.ptr<i8>)
+// CHECK:         %[[C0:.*]] = llvm.mlir.constant(0 : i64) : i64
+// CHECK:         %[[C1:.*]] = llvm.mlir.constant(1 : i64) : i64
+// CHECK:         %[[C0_1:.*]] = llvm.mlir.constant(0 : i64) : i64
+// CHECK:         %[[IDX:.*]] = llvm.sub %[[C0]], %[[C1]] : i64
+// CHECK:         %[[DIFF0:.*]] = llvm.mul %[[IDX]], %[[C1]] : i64
+// CHECK:         %[[SC:.*]] = llvm.mul %[[DIFF0]], %[[C1]]  : i64
+// CHECK:         %[[OFFSET:.*]] = llvm.add %[[SC]], %[[C0_1]]  : i64
+// CHECK:         %[[BITCAST:.*]] = llvm.bitcast %[[ARG0]] : !llvm.ptr<i8> to !llvm.ptr<i32>
+// CHECK:         %{{.*}} = llvm.getelementptr %[[BITCAST]][%[[OFFSET]]] : (!llvm.ptr<i32>, i64) -> !llvm.ptr<i32>
+
+// Conversion for a `fir.box`.
+
+func @ext_array_coor3(%arg0: !fir.box<!fir.array<?xi32>>) {
+  %c0 = arith.constant 0 : i64
+  %1 = fircg.ext_array_coor %arg0(%c0) <%c0> : (!fir.box<!fir.array<?xi32>>, i64, i64) -> !fir.ref<i32>
+  return
+}
+
+// CHECK-LABEL: llvm.func @ext_array_coor3(
+// CHECK-SAME:                             %[[ARG0:.*]]: !llvm.ptr<struct<(ptr<i32>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<1 x array<3 x i64>>)>>) {
+// CHECK:         %[[C0:.*]] = llvm.mlir.constant(0 : i64) : i64
+// CHECK:         %[[C1:.*]] = llvm.mlir.constant(1 : i64) : i64
+// CHECK:         %[[C0_1:.*]] = llvm.mlir.constant(0 : i64) : i64
+// CHECK:         %[[IDX:.*]] = llvm.sub %[[C0]], %[[C1]] : i64
+// CHECK:         %[[DIFF0:.*]] = llvm.mul %[[IDX]], %[[C1]] : i64
+// CHECK:         %[[C0_2:.*]] = llvm.mlir.constant(0 : i32) : i32
+// CHECK:         %[[DIMPOSINBOX:.*]] = llvm.mlir.constant(7 : i32) : i32
+// CHECK:         %[[DIMOFFSET:.*]] = llvm.mlir.constant(0 : i64) : i64
+// CHECK:         %[[STRIDPOS:.*]] = llvm.mlir.constant(2 : i32) : i32
+// CHECK:         %[[GEPSTRIDE:.*]] = llvm.getelementptr %[[ARG0]][%[[C0_2]], %[[DIMPOSINBOX]], %[[DIMOFFSET]], %[[STRIDPOS]]] : (!llvm.ptr<struct<(ptr<i32>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<1 x array<3 x i64>>)>>, i32, i32, i64, i32) -> !llvm.ptr<i64>
+// CHECK:         %[[LOADEDSTRIDE:.*]] = llvm.load %[[GEPSTRIDE]] : !llvm.ptr<i64>
+// CHECK:         %[[SC:.*]] = llvm.mul %[[DIFF0]], %[[LOADEDSTRIDE]]  : i64
+// CHECK:         %[[OFFSET:.*]] = llvm.add %[[SC]], %[[C0_1]]  : i64
+// CHECK:         %[[C0_3:.*]] = llvm.mlir.constant(0 : i32) : i32
+// CHECK:         %[[ADDRPOSINBOX:.*]] = llvm.mlir.constant(0 : i32) : i32
+// CHECK:         %[[GEPADDR:.*]] = llvm.getelementptr %[[ARG0]][%[[C0_3]], %[[ADDRPOSINBOX]]] : (!llvm.ptr<struct<(ptr<i32>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<1 x array<3 x i64>>)>>, i32, i32) -> !llvm.ptr<ptr<i32>>
+// CHECK:         %[[LOADEDADDR:.*]] = llvm.load %[[GEPADDR]] : !llvm.ptr<ptr<i32>>
+// CHECK:         %[[LOADEDADDRBITCAST:.*]] = llvm.bitcast %[[LOADEDADDR]] : !llvm.ptr<i32> to !llvm.ptr<i8>
+// CHECK:         %[[GEPADDROFFSET:.*]] = llvm.getelementptr %[[LOADEDADDRBITCAST]][%[[OFFSET]]] : (!llvm.ptr<i8>, i64) -> !llvm.ptr<i8>
+// CHECK:         %{{.*}} = llvm.bitcast %[[GEPADDROFFSET]] : !llvm.ptr<i8> to !llvm.ptr<i32>
+
+// Conversion with non zero shift and slice.
+
+func @ext_array_coor4(%arg0: !fir.ref<!fir.array<100xi32>>) {
+  %c0 = arith.constant 0 : i64
+  %c10 = arith.constant 10 : i64
+  %c20 = arith.constant 20 : i64
+  %c1 = arith.constant 1 : i64
+  %1 = fircg.ext_array_coor %arg0(%c0) origin %c0[%c10, %c20, %c1]<%c1> : (!fir.ref<!fir.array<100xi32>>, i64, i64, i64, i64, i64, i64) -> !fir.ref<i32>
+  return
+}
+
+// CHECK-LABEL: llvm.func @ext_array_coor4(
+// CHECK:                                  %[[ARG0:.*]]: !llvm.ptr<array<100 x i32>>) {
+// CHECK:         %[[C0:.*]] = llvm.mlir.constant(0 : i64) : i64
+// CHECK:         %[[C10:.*]] = llvm.mlir.constant(10 : i64) : i64
+// CHECK:         %[[C20:.*]] = llvm.mlir.constant(20 : i64) : i64
+// CHECK:         %[[C1:.*]] = llvm.mlir.constant(1 : i64) : i64
+// CHECK:         %[[C1_1:.*]] = llvm.mlir.constant(1 : i64) : i64
+// CHECK:         %[[C0_1:.*]] = llvm.mlir.constant(0 : i64) : i64
+// CHECK:         %[[IDX:.*]] = llvm.sub %[[C1]], %[[C0]]  : i64
+// CHECK:         %[[DIFF0:.*]] = llvm.mul %[[IDX]], %[[C1]]  : i64
+// CHECK:         %[[ADJ:.*]] = llvm.sub %[[C10]], %[[C0]]  : i64
+// CHECK:         %[[DIFF1:.*]] = llvm.add %[[DIFF0]], %[[ADJ]]  : i64
+// CHECK:         %[[STRIDE:.*]] = llvm.mul %[[DIFF1]], %[[C1_1]] : i64
+// CHECK:         %[[OFFSET:.*]] = llvm.add %[[STRIDE]], %[[C0_1]]  : i64
+// CHECK:         %[[BITCAST:.*]] = llvm.bitcast %[[ARG0]] : !llvm.ptr<array<100 x i32>> to !llvm.ptr<i32>
+// CHECK:         %{{.*}} = llvm.getelementptr %[[BITCAST]][%[[OFFSET]]] : (!llvm.ptr<i32>, i64) -> !llvm.ptr<i32>