[fir] Add fircg.ext_embox conversion

[clementval]

Summary

Convert a fircg.ext_embox operation to LLVM IR dialect.
A fircg.ext_embox is converted to a sequence of operation that
create, allocate if needed, and populate a descriptor.

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]

Make test target agnostic

[kiranchandramohan]

Minor comments and request for two more tests.

flang/lib/Optimizer/CodeGen/CodeGen.cpp

1453–1469

Nit: add the name for the pos 7.

1555

Nit: else can be removed by moving the lowerBound pushing outside.

1579

Nit: remove commented code.

1660

Is this case covered in the test? (the char has a dynamic size)

1678

I think we need a test for derived type as well.

[clementval]

Address review comment

[kiranchandramohan]

flang/lib/Optimizer/CodeGen/CodeGen.cpp

1678

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

1682

Not sure about this one since it was not a TODO. But please see the following discussion.

https://reviews.llvm.org/D114104#inline-1089507

1787

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

1799

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

[clementval]

Keep TODOs

[clementval]

Put TODO back in place

flang/lib/Optimizer/CodeGen/CodeGen.cpp

1453–1469

updated. Added also some constexpr for the values in [dims].

1660

It is in the second test.

1678

Type conversion make it fails before we can reach the pattern

[kiranchandramohan]

flang/lib/Optimizer/CodeGen/CodeGen.cpp

1791

I think the last argument is using the original operand and not the converted operand.

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

1577

Nit: Consider this as a nit and I will leave it to you to decide whether we should have it in this patch or should do it later.

1. A more realistic test case will be helpful here. Consider the following, it has shape, slice and shift operands for the op. Since it has two dimensions, it will exercise the for loop inside the code.

// subroutine sb(n,sh1,sh2)
// integer::n,sh1,sh2
// double precision::arr(sh1:n,sh2:n)
// call xb(arr(2:n,4:n))
// end subroutine

! N is the upperbound, sh1 and sh2 are the shifts or lowerbounds
func @_QPsb(%N: index, %sh1: index, %sh2: index) {
  %c4 = arith.constant 4 : index
  %c1 = arith.constant 1 : index
  %c2 = arith.constant 2 : index
! Calculate nelems in dim1
  %n1_tmp = arith.subi %N, %sh1 : index
  %n1 = arith.addi %n1_tmp, %c1 : index
! Calculate nelems in dim2
  %n2_tmp = arith.subi %N, %sh2 : index
  %n2 = arith.addi %n2_tmp, %c1 : index
  %arr = fir.alloca !fir.array<?x?xf64>, %n1, %n2 {bindc_name = "arr", uniq_name = "_QFsbEarr"}
  %box = fircg.ext_embox %arr(%n1, %n2) origin %sh1, %sh2[%c2, %N, %c1, %c4, %N, %c1] : (!fir.ref<!fir.array<?x?xf64>>, index, index, index, index, index, index, index, index, index, index) -> !fir.box<!fir.array<?x?xf64>>
  fir.call @_QPxb(%box) : (!fir.box<!fir.array<?x?xf64>>) -> ()
  return
}      
func private @_QPxb(!fir.box<!fir.array<?x?xf64>>)

2. A test with a subcomponent if possible

1581

Does substr have a meaning here since it applies to chars?

[clementval]

Address review comments

[clementval]

flang/lib/Optimizer/CodeGen/CodeGen.cpp

1791

Not sure I follow. xbox.substr()[0] has no conversion before this use.

[kiranchandramohan]

flang/lib/Optimizer/CodeGen/CodeGen.cpp

1791

I meant xbox.substr()[0] is the original substr operand of the XEmbox operation and not the converted operand. In conversion patterns we should use the converted operands. So, shouldn't we access the converted operand using something like operands[xbox.substrOffset()]?

[clementval]

Use converted operand

[clementval]

flang/lib/Optimizer/CodeGen/CodeGen.cpp

1791

Ah ok! Makes perfect sens. I was looking for a fir.convert or smth similar and I didn't even think about operand conversion.

[clementval]

ping

[kiranchandramohan]

LGTM.

flang/lib/Optimizer/CodeGen/CodeGen.cpp

1658

Nit: mlir::Location

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

1581

Nit: Did you miss this point or is having substr here OK?

[clementval]

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

1581

Sorry I thought I did update that. Done.

[clementval]

Address last comments

flang/include/flang/Optimizer/Dialect/FIRType.h

 }
 
 inline mlir::Type unwrapRefType(mlir::Type t) {
   if (auto eleTy = dyn_cast_ptrEleTy(t))
     return eleTy;
   return t;
 }
 
+/// If `t` conforms with a pass-by-reference type (box, ref, ptr, etc.) then
+/// return the element type of `t`. Otherwise, return `t`.
+inline mlir::Type unwrapPassByRefType(mlir::Type t) {
+  if (auto eleTy = dyn_cast_ptrOrBoxEleTy(t))
+    return eleTy;
+  return t;
+}
+
 #ifndef NDEBUG
 // !fir.ptr<X> and !fir.heap<X> where X is !fir.ptr, !fir.heap, or !fir.ref
 // is undefined and disallowed.
 inline bool singleIndirectionLevel(mlir::Type ty) {
   return !fir::isa_ref_type(ty);
 }
 #endif
 
 /// Apply the components specified by `path` to `rootTy` to determine the type
 /// of the resulting component element. `rootTy` should be an aggregate type.
 /// Returns null on error.
 mlir::Type applyPathToType(mlir::Type rootTy, mlir::ValueRange path);
 
 } // namespace fir
 
 #endif // FORTRAN_OPTIMIZER_DIALECT_FIRTYPE_H

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"
mlir::Value insertField(mlir::ConversionPatternRewriter &rewriter,
     return rewriter.create<mlir::LLVM::InsertValueOp>(loc, boxTy, dest, value,
                                                       indexesAttr);
   }
 
   inline mlir::Value
   insertBaseAddress(mlir::ConversionPatternRewriter &rewriter,
                     mlir::Location loc, mlir::Value dest,
                     mlir::Value base) const {
-    return insertField(rewriter, loc, dest, {0}, base, /*bitCast=*/true);
+    return insertField(rewriter, loc, dest, {kAddrPosInBox}, base,
+                       /*bitCast=*/true);
+  }
+
+  inline mlir::Value insertLowerBound(mlir::ConversionPatternRewriter &rewriter,
+                                      mlir::Location loc, mlir::Value dest,
+                                      unsigned dim, mlir::Value lb) const {
+    return insertField(rewriter, loc, dest,
+                       {kDimsPosInBox, dim, kDimLowerBoundPos}, lb);
+  }
+
+  inline mlir::Value insertExtent(mlir::ConversionPatternRewriter &rewriter,
+                                  mlir::Location loc, mlir::Value dest,
+                                  unsigned dim, mlir::Value extent) const {
+    return insertField(rewriter, loc, dest, {kDimsPosInBox, dim, kDimExtentPos},
+                       extent);
+  }
+
+  inline mlir::Value insertStride(mlir::ConversionPatternRewriter &rewriter,
+                                  mlir::Location loc, mlir::Value dest,
+                                  unsigned dim, mlir::Value stride) const {
+    return insertField(rewriter, loc, dest, {kDimsPosInBox, dim, kDimStridePos},
+                       stride);
   }
 
   /// Get the address of the type descriptor global variable that was created by
   /// lowering for derived type \p recType.
   template <typename BOX>
   mlir::Value
   getTypeDescriptor(BOX box, mlir::ConversionPatternRewriter &rewriter,
                     mlir::Location loc, fir::RecordType recType) const {
     std::string name = recType.getLoweredName();
     auto module = box->template getParentOfType<mlir::ModuleOp>();
     if (auto global = module.template lookupSymbol<fir::GlobalOp>(name)) {
       auto ty = mlir::LLVM::LLVMPointerType::get(
           this->lowerTy().convertType(global.getType()));
       return rewriter.create<mlir::LLVM::AddressOfOp>(loc, ty,
                                                       global.sym_name());
     }
     if (auto global =
             module.template lookupSymbol<mlir::LLVM::GlobalOp>(name)) {
       // The global may have already been translated to LLVM.
       auto ty = mlir::LLVM::LLVMPointerType::get(global.getType());
       return rewriter.create<mlir::LLVM::AddressOfOp>(loc, ty,
                                                       global.sym_name());
     }
     // The global does not exist in the current translation unit, but may be
     // defined elsewhere (e.g., type defined in a module).
     // For now, create a extern_weak symbol (will become nullptr if unresolved)
     // to support generating code without the front-end generated symbols.
if (hasAddendum) {
       descriptor =
           insertField(rewriter, loc, descriptor, {typeDescFieldId}, typeDesc,
                       /*bitCast=*/true);
     }
 
     return {boxTy, descriptor, eleSize};
   }
 
+  /// Compute the base address of a substring given the base address of a scalar
+  /// string and the zero based string lower bound.
+  mlir::Value shiftSubstringBase(mlir::ConversionPatternRewriter &rewriter,
+                                 mlir::Location loc, mlir::Value base,
+                                 mlir::Value lowerBound) const {
+    llvm::SmallVector<mlir::Value> gepOperands;
+    auto baseType =
+        base.getType().cast<mlir::LLVM::LLVMPointerType>().getElementType();
+    if (baseType.isa<mlir::LLVM::LLVMArrayType>()) {
+      auto idxTy = this->lowerTy().indexType();
+      mlir::Value zero = genConstantIndex(loc, idxTy, rewriter, 0);
+      gepOperands.push_back(zero);
+    }
+    gepOperands.push_back(lowerBound);
+    return this->genGEP(loc, base.getType(), rewriter, base, gepOperands);
+  }
+
   /// If the embox is not in a globalOp body, allocate storage for the box;
   /// store the value inside and return the generated alloca. Return the input
   /// value otherwise.
   mlir::Value
   placeInMemoryIfNotGlobalInit(mlir::ConversionPatternRewriter &rewriter,
                                mlir::Location loc, mlir::Value boxValue) const {
     auto *thisBlock = rewriter.getInsertionBlock();
     if (thisBlock && mlir::isa<mlir::LLVM::GlobalOp>(thisBlock->getParentOp()))
       return boxValue;
     auto boxPtrTy = mlir::LLVM::LLVMPointerType::get(boxValue.getType());
     auto alloca = genAllocaWithType(loc, boxPtrTy, defaultAlign, rewriter);
     rewriter.create<mlir::LLVM::StoreOp>(loc, boxValue, alloca);
     return alloca;
   }
 };
 
+/// Compute the extent of a triplet slice (lb:ub:step).
+static mlir::Value
+computeTripletExtent(mlir::ConversionPatternRewriter &rewriter,
+                     mlir::Location loc, mlir::Value lb, mlir::Value ub,
+                     mlir::Value step, mlir::Value zero, mlir::Type type) {
+  mlir::Value extent = rewriter.create<mlir::LLVM::SubOp>(loc, type, ub, lb);
+  extent = rewriter.create<mlir::LLVM::AddOp>(loc, type, extent, step);
+  extent = rewriter.create<mlir::LLVM::SDivOp>(loc, type, extent, step);
+  // If the resulting extent is negative (`ub-lb` and `step` have different
+  // signs), zero must be returned instead.
+  auto cmp = rewriter.create<mlir::LLVM::ICmpOp>(
+      loc, mlir::LLVM::ICmpPredicate::sgt, extent, zero);
+  return rewriter.create<mlir::LLVM::SelectOp>(loc, cmp, extent, zero);
+}
+
+/// Helper function for generating the LLVM IR that computes the size
+/// in bytes for a derived type.
+static mlir::Value
+computeDerivedTypeSize(mlir::Location loc, mlir::Type ptrTy, mlir::Type idxTy,
+                       mlir::ConversionPatternRewriter &rewriter) {
+  auto nullPtr = rewriter.create<mlir::LLVM::NullOp>(loc, ptrTy);
+  mlir::Value one = genConstantIndex(loc, idxTy, rewriter, 1);
+  llvm::SmallVector<mlir::Value> args{nullPtr, one};
+  auto gep = rewriter.create<mlir::LLVM::GEPOp>(loc, ptrTy, args);
+  return rewriter.create<mlir::LLVM::PtrToIntOp>(loc, idxTy, gep);
+}
+
 /// Create a generic box on a memory reference. This conversions lowers the
 /// abstract box to the appropriate, initialized descriptor.
 struct EmboxOpConversion : public EmboxCommonConversion<fir::EmboxOp> {
   using EmboxCommonConversion::EmboxCommonConversion;
 
   mlir::LogicalResult
   matchAndRewrite(fir::EmboxOp embox, OpAdaptor adaptor,
                   mlir::ConversionPatternRewriter &rewriter) const override {
struct EmboxProcOpConversion : public FIROpConversion<fir::EmboxProcOp> {
   mlir::LogicalResult
   matchAndRewrite(fir::EmboxProcOp emboxproc, OpAdaptor adaptor,
                   mlir::ConversionPatternRewriter &rewriter) const override {
     return rewriter.notifyMatchFailure(
         emboxproc, "fir.emboxproc codegen is not implemented yet");
   }
 };
 
+/// Create a generic box on a memory reference.
+struct XEmboxOpConversion : public EmboxCommonConversion<fir::cg::XEmboxOp> {
+  using EmboxCommonConversion::EmboxCommonConversion;
+
+  mlir::LogicalResult
+  matchAndRewrite(fir::cg::XEmboxOp xbox, OpAdaptor adaptor,
+                  mlir::ConversionPatternRewriter &rewriter) const override {
+    auto [boxTy, dest, eleSize] = consDescriptorPrefix(
+        xbox, rewriter, xbox.getOutRank(),
+        adaptor.getOperands().drop_front(xbox.lenParamOffset()));
+    // Generate the triples in the dims field of the descriptor
+    mlir::ValueRange operands = adaptor.getOperands();
+    auto i64Ty = mlir::IntegerType::get(xbox.getContext(), 64);
+    mlir::Value base = operands[0];
+    assert(!xbox.shape().empty() && "must have a shape");
+    unsigned shapeOffset = xbox.shapeOffset();
+    bool hasShift = !xbox.shift().empty();
+    unsigned shiftOffset = xbox.shiftOffset();
+    bool hasSlice = !xbox.slice().empty();
+    unsigned sliceOffset = xbox.sliceOffset();
+    auto loc = xbox.getLoc();
+    mlir::Value zero = genConstantIndex(loc, i64Ty, rewriter, 0);
+    mlir::Value one = genConstantIndex(loc, i64Ty, rewriter, 1);
+    mlir::Value prevDim = integerCast(loc, rewriter, i64Ty, eleSize);
+    mlir::Value prevPtrOff = one;
+    mlir::Type eleTy = boxTy.getEleTy();
+    const unsigned rank = xbox.getRank();
+    llvm::SmallVector<mlir::Value> gepArgs;
+    unsigned constRows = 0;
+    mlir::Value ptrOffset = zero;
+    if (auto memEleTy = fir::dyn_cast_ptrEleTy(xbox.memref().getType()))
+      if (auto seqTy = memEleTy.dyn_cast<fir::SequenceType>()) {
+        mlir::Type seqEleTy = seqTy.getEleTy();
+        // Adjust the element scaling factor if the element is a dependent type.
+        if (fir::hasDynamicSize(seqEleTy)) {
+          if (fir::isa_char(seqEleTy)) {
+            assert(xbox.lenParams().size() == 1);
+            prevPtrOff = integerCast(loc, rewriter, i64Ty,
+                                     operands[xbox.lenParamOffset()]);
+          } else if (seqEleTy.isa<fir::RecordType>()) {
+            TODO(loc, "generate call to calculate size of PDT");
+          } else {
+            return rewriter.notifyMatchFailure(xbox, "unexpected dynamic type");
+          }
+        } else {
+          constRows = seqTy.getConstantRows();
+        }
+      }
+
+    bool hasSubcomp = !xbox.subcomponent().empty();
+    mlir::Value stepExpr;
+    if (hasSubcomp) {
+      // We have a subcomponent. The step value needs to be the number of
+      // bytes per element (which is a derived type).
+      mlir::Type ty0 = base.getType();
+      [[maybe_unused]] auto ptrTy = ty0.dyn_cast<mlir::LLVM::LLVMPointerType>();
+      assert(ptrTy && "expected pointer type");
+      mlir::Type memEleTy = fir::dyn_cast_ptrEleTy(xbox.memref().getType());
+      assert(memEleTy && "expected fir pointer type");
+      auto seqTy = memEleTy.dyn_cast<fir::SequenceType>();
+      assert(seqTy && "expected sequence type");
+      mlir::Type seqEleTy = seqTy.getEleTy();
+      auto eleTy = mlir::LLVM::LLVMPointerType::get(convertType(seqEleTy));
+      stepExpr = computeDerivedTypeSize(loc, eleTy, i64Ty, rewriter);
+    }
+
+    // Process the array subspace arguments (shape, shift, etc.), if any,
+    // translating everything to values in the descriptor wherever the entity
+    // has a dynamic array dimension.
+    for (unsigned di = 0, descIdx = 0; di < rank; ++di) {
+      mlir::Value extent = operands[shapeOffset];
+      mlir::Value outerExtent = extent;
+      bool skipNext = false;
+      if (hasSlice) {
+        mlir::Value off = operands[sliceOffset];
+        mlir::Value adj = one;
+        if (hasShift)
+          adj = operands[shiftOffset];
+        auto ao = rewriter.create<mlir::LLVM::SubOp>(loc, i64Ty, off, adj);
+        if (constRows > 0) {
+          gepArgs.push_back(ao);
+          --constRows;
+        } else {
+          auto dimOff =
+              rewriter.create<mlir::LLVM::MulOp>(loc, i64Ty, ao, prevPtrOff);
+          ptrOffset =
+              rewriter.create<mlir::LLVM::AddOp>(loc, i64Ty, dimOff, ptrOffset);
+        }
+        if (mlir::isa_and_nonnull<fir::UndefOp>(
+                xbox.slice()[3 * di + 1].getDefiningOp())) {
+          // This dimension contains a scalar expression in the array slice op.
+          // The dimension is loop invariant, will be dropped, and will not
+          // appear in the descriptor.
+          skipNext = true;
+        }
+      }
+      if (!skipNext) {
+        // store lower bound (normally 0)
+        mlir::Value lb = zero;
+        if (eleTy.isa<fir::PointerType>() || eleTy.isa<fir::HeapType>()) {
+          lb = one;
+          if (hasShift)
+            lb = operands[shiftOffset];
+        }
+        dest = insertLowerBound(rewriter, loc, dest, descIdx, lb);
+
+        // store extent
+        if (hasSlice)
+          extent = computeTripletExtent(rewriter, loc, operands[sliceOffset],
+                                        operands[sliceOffset + 1],
+                                        operands[sliceOffset + 2], zero, i64Ty);
+        dest = insertExtent(rewriter, loc, dest, descIdx, extent);
+
+        // store step (scaled by shaped extent)
+
+        mlir::Value step = hasSubcomp ? stepExpr : prevDim;
+        if (hasSlice)
+          step = rewriter.create<mlir::LLVM::MulOp>(loc, i64Ty, step,
+                                                    operands[sliceOffset + 2]);
+        dest = insertStride(rewriter, loc, dest, descIdx, step);
+        ++descIdx;
+      }
+
+      // compute the stride and offset for the next natural dimension
+      prevDim =
+          rewriter.create<mlir::LLVM::MulOp>(loc, i64Ty, prevDim, outerExtent);
+      if (constRows == 0)
+        prevPtrOff = rewriter.create<mlir::LLVM::MulOp>(loc, i64Ty, prevPtrOff,
+                                                        outerExtent);
+
+      // increment iterators
+      ++shapeOffset;
+      if (hasShift)
+        ++shiftOffset;
+      if (hasSlice)
+        sliceOffset += 3;
+    }
+    if (hasSlice || hasSubcomp || !xbox.substr().empty()) {
+      llvm::SmallVector<mlir::Value> args = {base, ptrOffset};
+      args.append(gepArgs.rbegin(), gepArgs.rend());
+      if (hasSubcomp) {
+        // For each field in the path add the offset to base via the args list.
+        // In the most general case, some offsets must be computed since
+        // they are not be known until runtime.
+        if (fir::hasDynamicSize(fir::unwrapSequenceType(
+                fir::unwrapPassByRefType(xbox.memref().getType()))))
+          TODO(loc, "fir.embox codegen dynamic size component in derived type");
+        args.append(operands.begin() + xbox.subcomponentOffset(),
+                    operands.begin() + xbox.subcomponentOffset() +
+                        xbox.subcomponent().size());
+      }
+      base = rewriter.create<mlir::LLVM::GEPOp>(loc, base.getType(), args);
+      if (!xbox.substr().empty())
+        base = shiftSubstringBase(rewriter, loc, base, xbox.substr()[0]);
+    }
+    dest = insertBaseAddress(rewriter, loc, dest, base);
+    if (isDerivedTypeWithLenParams(boxTy))
+      TODO(loc, "fir.embox codegen of derived with length parameters");
+
+    mlir::Value result = placeInMemoryIfNotGlobalInit(rewriter, loc, dest);
+    rewriter.replaceOp(xbox, result);
+    return success();
+  }
+};
 
 // Code shared between insert_value and extract_value Ops.
 struct ValueOpCommon {
   // Translate the arguments pertaining to any multidimensional array to
   // row-major order for LLVM-IR.
   static void toRowMajor(SmallVectorImpl<mlir::Attribute> &attrs,
                          mlir::Type ty) {
     assert(ty && "type is null");
pattern.insert<
         FirEndOpConversion, HasValueOpConversion, GenTypeDescOpConversion,
         GlobalLenOpConversion, GlobalOpConversion, InsertOnRangeOpConversion,
         InsertValueOpConversion, IsPresentOpConversion, LoadOpConversion,
         NegcOpConversion, MulcOpConversion, SelectCaseOpConversion,
         SelectOpConversion, SelectRankOpConversion, SelectTypeOpConversion,
         ShapeOpConversion, ShapeShiftOpConversion, ShiftOpConversion,
         SliceOpConversion, StoreOpConversion, StringLitOpConversion,
         SubcOpConversion, UnboxCharOpConversion, UnboxProcOpConversion,
-        UndefOpConversion, UnreachableOpConversion, ZeroOpConversion>(
-        typeConverter);
+        UndefOpConversion, UnreachableOpConversion, XEmboxOpConversion,
+        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/lib/Optimizer/CodeGen/TypeConverter.h

 static constexpr unsigned kVersionPosInBox = 2;
 static constexpr unsigned kRankPosInBox = 3;
 static constexpr unsigned kTypePosInBox = 4;
 static constexpr unsigned kAttributePosInBox = 5;
 static constexpr unsigned kF18AddendumPosInBox = 6;
 static constexpr unsigned kDimsPosInBox = 7;
 static constexpr unsigned kOptTypePtrPosInBox = 8;
 static constexpr unsigned kOptRowTypePosInBox = 9;
+// Position of the different values in [dims]
+static constexpr unsigned kDimLowerBoundPos = 0;
+static constexpr unsigned kDimExtentPos = 1;
+static constexpr unsigned kDimStridePos = 2;
 
 namespace fir {
 
 /// FIR type converter
 /// This converts FIR types to LLVM types (for now)
 class LLVMTypeConverter : public mlir::LLVMTypeConverter {
 public:
   LLVMTypeConverter(mlir::ModuleOp module)

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

func @field_index_dynamic_size() -> () {
   %2 = fir.field_index field_2, !fir.type<custom_type{field_1:i32, field_2:!fir.array<?xf32>}>
   return
 }
 
 // CHECK-LABEL: @field_index_dynamic_size
 // CHECK-NEXT: %{{.*}} = llvm.call @custom_typeP.field_1.offset() {field = 0 : i64} : () -> i32
 // CHECK-NEXT: %{{.*}} = llvm.call @custom_typeP.field_2.offset() {field = 1 : i64} : () -> i32
 // CHECK-NEXT:  llvm.return
+
+// -----
+
+// Test `fircg.ext_embox` conversion.
+
+// Check complete `fircg.ext_embox`.
+
+func @xembox0(%arg0: !fir.ref<!fir.array<?xi32>>) {
+  %c0 = arith.constant 0 : i64
+  %0 = fircg.ext_embox %arg0(%c0) origin %c0[%c0, %c0, %c0] substr %c0, %c0 : (!fir.ref<!fir.array<?xi32>>, i64, i64, i64, i64, i64, i64, i64) -> !fir.box<!fir.array<?xi32>>
+  return
+}
+
+// CHECK-LABEL: llvm.func @xembox0(
+// CHECK-SAME:                     %[[ARG0:.*]]: !llvm.ptr<i32>
+// CHECK:         %[[ALLOCA_SIZE:.*]] = llvm.mlir.constant(1 : i32) : i32
+// CHECK:         %[[ALLOCA:.*]] = llvm.alloca %[[ALLOCA_SIZE]] x !llvm.struct<(ptr<i32>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<1 x array<3 x i64>>)> {alignment = 8 : i64} : (i32) -> !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:         %[[BOX0:.*]] = llvm.mlir.undef : !llvm.struct<(ptr<i32>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<1 x array<3 x i64>>)>
+// CHECK:         %[[ELEM_LEN:.*]] = llvm.mlir.constant(4 : i32) : i32
+// CHECK:         %[[TYPE:.*]] = llvm.mlir.constant(9 : i32) : i32
+// CHECK:         %[[ELEM_LEN_I64:.*]] = llvm.sext %[[ELEM_LEN]] : i32 to i64
+// CHECK:         %[[BOX1:.*]] = llvm.insertvalue %[[ELEM_LEN_I64]], %[[BOX0]][1 : i32] : !llvm.struct<(ptr<i32>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<1 x array<3 x i64>>)>
+// CHECK:         %[[VERSION:.*]] = llvm.mlir.constant(20180515 : i32) : i32
+// CHECK:         %[[BOX2:.*]] = llvm.insertvalue %[[VERSION]], %[[BOX1]][2 : i32] : !llvm.struct<(ptr<i32>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<1 x array<3 x i64>>)>
+// CHECK:         %[[RANK:.*]] = llvm.mlir.constant(1 : i32) : i32
+// CHECK:         %[[RANK_I8:.*]] = llvm.trunc %[[RANK]] : i32 to i8
+// CHECK:         %[[BOX3:.*]] = llvm.insertvalue %[[RANK_I8]], %[[BOX2]][3 : i32] : !llvm.struct<(ptr<i32>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<1 x array<3 x i64>>)>
+// CHECK:         %[[TYPE_I8:.*]] = llvm.trunc %[[TYPE]] : i32 to i8
+// CHECK:         %[[BOX4:.*]] = llvm.insertvalue %[[TYPE_I8]], %[[BOX3]][4 : i32] : !llvm.struct<(ptr<i32>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<1 x array<3 x i64>>)>
+// CHECK:         %[[ATTR:.*]] = llvm.mlir.constant(0 : i32) : i32
+// CHECK:         %[[ATTR_I8:.*]] = llvm.trunc %[[ATTR]] : i32 to i8
+// CHECK:         %[[BOX5:.*]] = llvm.insertvalue %[[ATTR_I8]], %[[BOX4]][5 : i32] : !llvm.struct<(ptr<i32>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<1 x array<3 x i64>>)>
+// CHECK:         %[[F18ADDENDUM:.*]] = llvm.mlir.constant(0 : i32) : i32
+// CHECK:         %[[F18ADDENDUM_I8:.*]] = llvm.trunc %[[F18ADDENDUM]] : i32 to i8
+// CHECK:         %[[BOX6:.*]] = llvm.insertvalue %[[F18ADDENDUM_I8]], %[[BOX5]][6 : i32] : !llvm.struct<(ptr<i32>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<1 x array<3 x i64>>)>
+// CHECK:         %[[ZERO:.*]] = llvm.mlir.constant(0 : i64) : i64
+// CHECK:         %[[ONE:.*]] = llvm.mlir.constant(1 : i64) : i64
+// CHECK:         %[[ELEM_LEN_I64:.*]] = llvm.sext %[[ELEM_LEN]] : i32 to i64
+// CHECK:         %[[ADJUSTED_OFFSET:.*]] = llvm.sub %[[C0]], %[[C0]]  : i64
+// CHECK:         %[[DIM_OFFSET:.*]] = llvm.mul %[[ADJUSTED_OFFSET]], %[[ONE]]  : i64
+// CHECK:         %[[PTR_OFFSET:.*]] = llvm.add %[[DIM_OFFSET]], %[[ZERO]]  : i64
+// CHECK:         %[[BOX7:.*]] = llvm.insertvalue %[[ZERO]], %[[BOX6]][7 : i32, 0 : i32, 0 : i32] : !llvm.struct<(ptr<i32>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<1 x array<3 x i64>>)>
+// CHECK:         %[[EXTENT0:.*]] = llvm.sub %[[C0]], %[[C0]]  : i64
+// CHECK:         %[[EXTENT1:.*]] = llvm.add %[[EXTENT0]], %[[C0]]  : i64
+// CHECK:         %[[EXTENT2:.*]] = llvm.sdiv %[[EXTENT1]], %[[C0]]  : i64
+// CHECK:         %[[EXTENT_CMP:.*]] = llvm.icmp "sgt" %[[EXTENT2]], %[[ZERO]] : i64
+// CHECK:         %[[EXTENT:.*]] = llvm.select %[[EXTENT_CMP]], %[[EXTENT2]], %[[ZERO]] : i1, i64
+// CHECK:         %[[BOX8:.*]] = llvm.insertvalue %[[EXTENT]], %[[BOX7]][7 : i32, 0 : i32, 1 : i32] : !llvm.struct<(ptr<i32>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<1 x array<3 x i64>>)>
+// CHECK:         %[[STRIDE:.*]] = llvm.mul %[[ELEM_LEN_I64]], %[[C0]]  : i64
+// CHECK:         %[[BOX9:.*]] = llvm.insertvalue %[[STRIDE]], %[[BOX8]][7 : i32, 0 : i32, 2 : i32] : !llvm.struct<(ptr<i32>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<1 x array<3 x i64>>)>
+// CHECK:         %[[PREV_DIM:.*]] = llvm.mul %[[ELEM_LEN_I64]], %[[C0]]  : i64
+// CHECK:         %[[PREV_PTROFF:.*]] = llvm.mul %[[ONE]], %[[C0]]  : i64
+// CHECK:         %[[BASE_PTR:.*]] = llvm.getelementptr %[[ARG0]][%[[PTR_OFFSET]]] : (!llvm.ptr<i32>, i64) -> !llvm.ptr<i32>
+// CHECK:         %[[BASE:.*]] = llvm.getelementptr %[[BASE_PTR]][%[[C0]]] : (!llvm.ptr<i32>, i64) -> !llvm.ptr<i32>
+// CHECK:         %[[ADDR_BITCAST:.*]] = llvm.bitcast %[[BASE]] : !llvm.ptr<i32> to !llvm.ptr<i32>
+// CHECK:         %[[BOX10:.*]] = llvm.insertvalue %[[ADDR_BITCAST]], %[[BOX9]][0 : i32] : !llvm.struct<(ptr<i32>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<1 x array<3 x i64>>)>
+// CHECK:         llvm.store %[[BOX10]], %[[ALLOCA]] : !llvm.ptr<struct<(ptr<i32>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<1 x array<3 x i64>>)>>
+
+// Check adjustment of element scaling factor.
+
+func @xembox1(%arg0: !fir.ref<!fir.array<?x!fir.char<1, 10>>>) {
+  %c0 = arith.constant 0 : i64
+  %0 = fircg.ext_embox %arg0(%c0) origin %c0[%c0, %c0, %c0] : (!fir.ref<!fir.array<?x!fir.char<1, 10>>>, i64, i64, i64, i64, i64) -> !fir.box<!fir.array<?x!fir.char<1, 10>>>
+  return
+}
+
+// CHECK-LABEL: llvm.func @xembox1(%{{.*}}: !llvm.ptr<array<10 x i8>>) {
+// CHECK:         %[[C0:.*]] = llvm.mlir.constant(0 : i64) : i64
+// CHECK:         %[[ELEM_LEN:.*]] = llvm.mlir.constant(10 : i32) : i32
+// CHECK:         %[[ELEM_LEN_I64:.*]] = llvm.sext %[[ELEM_LEN]] : i32 to i64
+// CHECK:         %{{.*}} = llvm.insertvalue %[[ELEM_LEN_I64]], %{{.*}}[1 : i32] : !llvm.struct<(ptr<array<10 x i8>>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<1 x array<3 x i64>>)>
+// CHECK:         %[[PTR_OFFSET:.*]] = llvm.sext %[[ELEM_LEN]] : i32 to i64
+// CHECK:         %[[PREV_PTROFF:.*]] = llvm.mul %[[PTR_OFFSET]], %[[C0]]  : i64
+
+// Fortran realistic use case extracted from the following snippet:
+//
+// ```
+// subroutine sb(n,sh1,sh2)
+//   integer::n,sh1,sh2
+//   double precision::arr(sh1:n,sh2:n)
+//   call xb(arr(2:n,4:n))
+// end subroutine 
+// ```
+
+// N is the upperbound, sh1 and sh2 are the shifts or lowerbounds
+func @_QPsb(%N: index, %sh1: index, %sh2: index) {
+  %c4 = arith.constant 4 : index
+  %c1 = arith.constant 1 : index
+  %c2 = arith.constant 2 : index
+  // Calculate nelems in dim1
+  %n1_tmp = arith.subi %N, %sh1 : index
+  %n1 = arith.addi %n1_tmp, %c1 : index
+  // Calculate nelems in dim2
+  %n2_tmp = arith.subi %N, %sh2 : index
+  %n2 = arith.addi %n2_tmp, %c1 : index
+  %arr = fir.alloca !fir.array<?x?xf64>, %n1, %n2 {bindc_name = "arr", uniq_name = "_QFsbEarr"}
+  %box = fircg.ext_embox %arr(%n1, %n2) origin %sh1, %sh2[%c2, %N, %c1, %c4, %N, %c1] : (!fir.ref<!fir.array<?x?xf64>>, index, index, index, index, index, index, index, index, index, index) -> !fir.box<!fir.array<?x?xf64>>
+  fir.call @_QPxb(%box) : (!fir.box<!fir.array<?x?xf64>>) -> ()
+  return
+}      
+func private @_QPxb(!fir.box<!fir.array<?x?xf64>>)
+
+// CHECK-LABEL: llvm.func @_QPsb(
+// CHECK-SAME:                   %[[N:.*]]: i64, %[[SH1:.*]]: i64, %[[SH2:.*]]: i64) {
+// CHECK:         %[[ALLOCA_SIZE:.*]] = llvm.mlir.constant(1 : i32) : i32
+// CHECK:         %[[ALLOCA:.*]] = llvm.alloca %[[ALLOCA_SIZE]] x !llvm.struct<(ptr<f64>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<2 x array<3 x i64>>)> {alignment = 8 : i64} : (i32) -> !llvm.ptr<struct<(ptr<f64>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<2 x array<3 x i64>>)>>
+// CHECK:         %[[C4:.*]] = llvm.mlir.constant(4 : index) : i64
+// CHECK:         %[[C1:.*]] = llvm.mlir.constant(1 : index) : i64
+// CHECK:         %[[C2:.*]] = llvm.mlir.constant(2 : index) : i64
+// CHECK:         %[[N1_TMP:.*]] = llvm.sub %[[N]], %[[SH1]]  : i64
+// CHECK:         %[[N1:.*]] = llvm.add %[[N1_TMP]], %[[C1]]  : i64
+// CHECK:         %[[N2_TMP:.*]] = llvm.sub %[[N]], %[[SH2]]  : i64
+// CHECK:         %[[N2:.*]] = llvm.add %[[N2_TMP]], %[[C1]]  : i64
+// CHECK:         %[[C1_0:.*]] = llvm.mlir.constant(1 : i64) : i64
+// CHECK:         %[[C1_1:.*]] = llvm.mlir.constant(1 : i64) : i64
+// CHECK:         %[[ARR_SIZE_TMP0:.*]] = llvm.mul %[[C1_0]], %[[C1_1]]  : i64
+// CHECK:         %[[ARR_SIZE_TMP1:.*]] = llvm.mul %[[ARR_SIZE_TMP0]], %[[N1]]  : i64
+// CHECK:         %[[ARR_SIZE:.*]] = llvm.mul %[[ARR_SIZE_TMP1]], %[[N2]]  : i64
+// CHECK:         %[[ARR:.*]] = llvm.alloca %[[ARR_SIZE]] x f64 {bindc_name = "arr", in_type = !fir.array<?x?xf64>, operand_segment_sizes = dense<[0, 2]> : vector<2xi32>, uniq_name = "_QFsbEarr"} : (i64) -> !llvm.ptr<f64>
+// CHECK:         %[[BOX0:.*]] = llvm.mlir.undef : !llvm.struct<(ptr<f64>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<2 x array<3 x i64>>)>
+// CHECK:         %[[ELEM_LEN:.*]] = llvm.mlir.constant(8 : i32) : i32
+// CHECK:         %[[TYPE_CODE:.*]] = llvm.mlir.constant(26 : i32) : i32
+// CHECK:         %[[ELEM_LEN_I64:.*]] = llvm.sext %[[ELEM_LEN]] : i32 to i64
+// CHECK:         %[[BOX1:.*]] = llvm.insertvalue %[[ELEM_LEN_I64]], %[[BOX0]][1 : i32] : !llvm.struct<(ptr<f64>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<2 x array<3 x i64>>)>
+// CHECK:         %[[VERSION:.*]] = llvm.mlir.constant(20180515 : i32) : i32
+// CHECK:         %[[BOX2:.*]] = llvm.insertvalue %[[VERSION]], %[[BOX1]][2 : i32] : !llvm.struct<(ptr<f64>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<2 x array<3 x i64>>)>
+// CHECK:         %[[RANK:.*]] = llvm.mlir.constant(2 : i32) : i32
+// CHECK:         %[[RANK_I8:.*]] = llvm.trunc %[[RANK]] : i32 to i8
+// CHECK:         %[[BOX3:.*]] = llvm.insertvalue %[[RANK_I8]], %[[BOX2]][3 : i32] : !llvm.struct<(ptr<f64>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<2 x array<3 x i64>>)>
+// CHECK:         %[[TYPE_I8:.*]] = llvm.trunc %[[TYPE_CODE]] : i32 to i8
+// CHECK:         %[[BOX4:.*]] = llvm.insertvalue %[[TYPE_I8]], %[[BOX3]][4 : i32] : !llvm.struct<(ptr<f64>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<2 x array<3 x i64>>)>
+// CHECK:         %[[ATTR:.*]] = llvm.mlir.constant(0 : i32) : i32
+// CHECK:         %[[ATTR_I8:.*]] = llvm.trunc %[[ATTR]] : i32 to i8
+// CHECK:         %[[BOX5:.*]] = llvm.insertvalue %[[ATTR_I8]], %[[BOX4]][5 : i32] : !llvm.struct<(ptr<f64>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<2 x array<3 x i64>>)>
+// CHECK:         %[[F18ADDENDUM:.*]] = llvm.mlir.constant(0 : i32) : i32
+// CHECK:         %[[F18ADDENDUM_I8:.*]] = llvm.trunc %[[F18ADDENDUM]] : i32 to i8
+// CHECK:         %[[BOX6:.*]] = llvm.insertvalue %[[F18ADDENDUM_I8]], %[[BOX5]][6 : i32] : !llvm.struct<(ptr<f64>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<2 x array<3 x i64>>)>
+// CHECK:         %[[ZERO:.*]] = llvm.mlir.constant(0 : i64) : i64
+// CHECK:         %[[ONE:.*]] = llvm.mlir.constant(1 : i64) : i64
+// CHECK:         %[[ELEM_LEN_I64:.*]] = llvm.sext %[[ELEM_LEN]] : i32 to i64
+// CHECK:         %[[ADJUSTED_OFFSET:.*]] = llvm.sub %[[C2]], %[[SH1]] : i64
+// CHECK:         %[[DIM_OFFSET:.*]] = llvm.mul %[[ADJUSTED_OFFSET]], %[[ONE]]  : i64
+// CHECK:         %[[PTR_OFFSET:.*]] = llvm.add %[[DIM_OFFSET]], %[[ZERO]]  : i64
+// CHECK:         %[[BOX7:.*]] = llvm.insertvalue %[[ZERO]], %[[BOX6]][7 : i32, 0 : i32, 0 : i32] : !llvm.struct<(ptr<f64>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<2 x array<3 x i64>>)>
+// CHECK:         %[[EXTENT0:.*]] = llvm.sub %[[ARG0]], %[[C2]]  : i64
+// CHECK:         %[[EXTENT1:.*]] = llvm.add %[[EXTENT0]], %[[C1]]  : i64
+// CHECK:         %[[EXTENT2:.*]] = llvm.sdiv %[[EXTENT1]], %[[C1]]  : i64
+// CHECK:         %[[EXTENT_CMP:.*]] = llvm.icmp "sgt" %[[EXTENT2]], %[[ZERO]] : i64
+// CHECK:         %[[EXTENT:.*]] = llvm.select %[[EXTENT_CMP]], %[[EXTENT2]], %[[ZERO]] : i1, i64
+// CHECK:         %[[BOX8:.*]] = llvm.insertvalue %[[EXTENT]], %[[BOX7]][7 : i32, 0 : i32, 1 : i32] : !llvm.struct<(ptr<f64>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<2 x array<3 x i64>>)>
+// CHECK:         %[[STRIDE:.*]] = llvm.mul %[[ELEM_LEN_I64]], %[[C1]]  : i64
+// CHECK:         %[[BOX9:.*]] = llvm.insertvalue %[[STRIDE]], %[[BOX8]][7 : i32, 0 : i32, 2 : i32] : !llvm.struct<(ptr<f64>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<2 x array<3 x i64>>)>
+// CHECK:         %[[PREV_DIM:.*]] = llvm.mul %[[ELEM_LEN_I64]], %[[N1]]  : i64
+// CHECK:         %[[PREV_PTROFF:.*]] = llvm.mul %[[ONE]], %[[N1]]  : i64
+// CHECK:         %[[ADJUSTED_OFFSET:.*]] = llvm.sub %[[C4]], %[[SH2]]  : i64
+// CHECK:         %[[DIM_OFFSET:.*]] = llvm.mul %[[ADJUSTED_OFFSET]], %[[PREV_PTROFF]]  : i64
+// CHECK:         %[[PTR_OFFSET0:.*]] = llvm.add %[[DIM_OFFSET]], %[[PTR_OFFSET]]  : i64
+// CHECK:         %[[BOX10:.*]] = llvm.insertvalue %[[ZERO]], %[[BOX9]][7 : i32, 1 : i32, 0 : i32] : !llvm.struct<(ptr<f64>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<2 x array<3 x i64>>)>
+// CHECK:         %[[EXT_SUB:.*]] = llvm.sub %[[N]], %[[C4]]  : i64
+// CHECK:         %[[EXT_ADD:.*]] = llvm.add %[[EXT_SUB]], %[[C1]]  : i64
+// CHECK:         %[[EXT_SDIV:.*]] = llvm.sdiv %[[EXT_ADD]], %[[C1]]  : i64
+// CHECK:         %[[EXT_ICMP:.*]] = llvm.icmp "sgt" %[[EXT_SDIV]], %[[ZERO]] : i64
+// CHECK:         %[[EXT_SELECT:.*]] = llvm.select %[[EXT_ICMP]], %[[EXT_SDIV]], %[[ZERO]] : i1, i64
+// CHECK:         %[[BOX11:.*]] = llvm.insertvalue %[[EXT_SELECT]], %[[BOX10]][7 : i32, 1 : i32, 1 : i32] : !llvm.struct<(ptr<f64>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<2 x array<3 x i64>>)>
+// CHECK:         %[[STRIDE_MUL:.*]] = llvm.mul %[[PREV_DIM]], %[[C1]]  : i64
+// CHECK:         %[[BOX12:.*]] = llvm.insertvalue %[[STRIDE_MUL]], %[[BOX11]][7 : i32, 1 : i32, 2 : i32] : !llvm.struct<(ptr<f64>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<2 x array<3 x i64>>)>
+// CHECK:         %[[BASE_PTR:.*]] = llvm.getelementptr %[[ARR]][%[[PTR_OFFSET0]]] : (!llvm.ptr<f64>, i64) -> !llvm.ptr<f64>
+// CHECK:         %[[ADDR_BITCAST:.*]] = llvm.bitcast %[[BASE_PTR]] : !llvm.ptr<f64> to !llvm.ptr<f64>
+// CHECK:         %[[BOX13:.*]] = llvm.insertvalue %[[ADDR_BITCAST]], %[[BOX12]][0 : i32] : !llvm.struct<(ptr<f64>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<2 x array<3 x i64>>)>
+// CHECK:         llvm.store %[[BOX13]], %[[ALLOCA]] : !llvm.ptr<struct<(ptr<f64>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<2 x array<3 x i64>>)>>
+
+// Conversion with a subcomponent.
+
+func @_QPtest_dt_slice() {
+  %c20 = arith.constant 20 : index
+  %c1_i64 = arith.constant 1 : i64
+  %c10_i64 = arith.constant 10 : i64
+  %c2_i64 = arith.constant 2 : i64
+  %0 = fir.alloca i32 {bindc_name = "v", uniq_name = "_QFtest_dt_sliceEv"}
+  %1 = fir.alloca !fir.array<20x!fir.type<_QFtest_dt_sliceTt{i:i32,j:i32}>> {bindc_name = "x", uniq_name = "_QFtest_dt_sliceEx"}
+  %2 = fir.field_index i, !fir.type<_QFtest_dt_sliceTt{i:i32,j:i32}>
+  %5 = fircg.ext_embox %1(%c20)[%c1_i64, %c10_i64, %c2_i64] path %2 : (!fir.ref<!fir.array<20x!fir.type<_QFtest_dt_sliceTt{i:i32,j:i32}>>>, index, i64, i64, i64, !fir.field) -> !fir.box<!fir.array<?xi32>>
+  fir.call @_QPtest_dt_callee(%5) : (!fir.box<!fir.array<?xi32>>) -> ()
+  return
+}
+func private @_QPtest_dt_callee(%arg0: !fir.box<!fir.array<?xi32>>)
+
+// CHECK-LABEL: llvm.func @_QPtest_dt_slice
+// CHECK:         %[[ALLOCA_SIZE:.*]] = llvm.mlir.constant(1 : i32) : i32
+// CHECK:         %[[ALLOCA:.*]] = llvm.alloca %[[ALLOCA_SIZE]] x !llvm.struct<(ptr<i32>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<1 x array<3 x i64>>)> {alignment = 8 : i64} : (i32) -> !llvm.ptr<struct<(ptr<i32>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<1 x array<3 x i64>>)>>
+// CHECK:         %[[C20:.*]] = llvm.mlir.constant(20 : index) : i64
+// CHECK:         %[[C1:.*]] = llvm.mlir.constant(1 : i64) : i64
+// CHECK:         %[[C10:.*]] = llvm.mlir.constant(10 : i64) : i64
+// CHECK:         %[[C2:.*]] = llvm.mlir.constant(2 : i64) : i64
+// CHECK:         %[[ALLOCA_SIZE_V:.*]] = llvm.mlir.constant(1 : i64) : i64
+// CHECK:         %[[V:.*]] = llvm.alloca %[[ALLOCA_SIZE_V]] x i32 {bindc_name = "v", in_type = i32, operand_segment_sizes = dense<0> : vector<2xi32>, uniq_name = "_QFtest_dt_sliceEv"} : (i64) -> !llvm.ptr<i32>
+// CHECK:         %[[ALLOCA_SIZE_X:.*]] = llvm.mlir.constant(1 : i64) : i64
+// CHECK:         %[[X:.*]] = llvm.alloca %[[ALLOCA_SIZE_X]] x !llvm.array<20 x struct<"_QFtest_dt_sliceTt", (i32, i32)>> {bindc_name = "x", in_type = !fir.array<20x!fir.type<_QFtest_dt_sliceTt{i:i32,j:i32}>>, operand_segment_sizes = dense<0> : vector<2xi32>, uniq_name = "_QFtest_dt_sliceEx"} : (i64) -> !llvm.ptr<array<20 x struct<"_QFtest_dt_sliceTt", (i32, i32)>>>
+// CHECK:         %[[C0:.*]] = llvm.mlir.constant(0 : i32) : i32
+// CHECK:         %[[BOX0:.*]] = llvm.mlir.undef : !llvm.struct<(ptr<i32>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<1 x array<3 x i64>>)>
+// CHECK:         %[[ELEM_LEN:.*]] = llvm.mlir.constant(4 : i32) : i32
+// CHECK:         %[[TYPE_CODE:.*]] = llvm.mlir.constant(9 : i32) : i32
+// CHECK:         %[[ELEM_LEN_I64:.*]] = llvm.sext %[[ELEM_LEN]] : i32 to i64
+// CHECK:         %[[BOX1:.*]] = llvm.insertvalue %[[ELEM_LEN_I64]], %[[BOX0]][1 : i32] : !llvm.struct<(ptr<i32>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<1 x array<3 x i64>>)>
+// CHECK:         %[[VERSION:.*]] = llvm.mlir.constant(20180515 : i32) : i32
+// CHECK:         %[[BOX2:.*]] = llvm.insertvalue %[[VERSION]], %[[BOX1]][2 : i32] : !llvm.struct<(ptr<i32>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<1 x array<3 x i64>>)>
+// CHECK:         %[[RANK:.*]] = llvm.mlir.constant(1 : i32) : i32
+// CHECK:         %[[RANK_I8:.*]] = llvm.trunc %[[RANK]] : i32 to i8
+// CHECK:         %[[BOX3:.*]] = llvm.insertvalue %[[RANK_I8]], %[[BOX2]][3 : i32] : !llvm.struct<(ptr<i32>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<1 x array<3 x i64>>)>
+// CHECK:         %[[TYPE_CODE_I8:.*]] = llvm.trunc %[[TYPE_CODE]] : i32 to i8
+// CHECK:         %[[BOX4:.*]] = llvm.insertvalue %[[TYPE_CODE_I8]], %[[BOX3]][4 : i32] : !llvm.struct<(ptr<i32>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<1 x array<3 x i64>>)>
+// CHECK:         %[[ATTR:.*]] = llvm.mlir.constant(0 : i32) : i32
+// CHECK:         %[[ATTR_I8:.*]] = llvm.trunc %[[ATTR]] : i32 to i8
+// CHECK:         %[[BOX5:.*]] = llvm.insertvalue %[[ATTR_I8]], %[[BOX4]][5 : i32] : !llvm.struct<(ptr<i32>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<1 x array<3 x i64>>)>
+// CHECK:         %[[F18ADDENDUM:.*]] = llvm.mlir.constant(0 : i32) : i32
+// CHECK:         %[[F18ADDENDUM_I8:.*]] = llvm.trunc %[[F18ADDENDUM]] : i32 to i8
+// CHECK:         %[[BOX6:.*]] = llvm.insertvalue %[[F18ADDENDUM_I8]], %[[BOX5]][6 : i32] : !llvm.struct<(ptr<i32>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<1 x array<3 x i64>>)>
+// CHECK:         %[[ZERO:.*]] = llvm.mlir.constant(0 : i64) : i64
+// CHECK:         %[[ONE:.*]] = llvm.mlir.constant(1 : i64) : i64
+// CHECK:         %[[ELEM_LEN_I64:.*]] = llvm.sext %[[ELEM_LEN]] : i32 to i64    
+// CHECK:         %[[ELE_TYPE:.*]] = llvm.mlir.null : !llvm.ptr<struct<"_QFtest_dt_sliceTt", (i32, i32)>>
+// CHECK:         %[[C1_0:.*]] = llvm.mlir.constant(1 : i64) : i64
+// CHECK:         %[[GEP_DTYPE_SIZE:.*]] = llvm.getelementptr %[[ELE_TYPE]][%[[C1_0]]] : (!llvm.ptr<struct<"_QFtest_dt_sliceTt", (i32, i32)>>, i64) -> !llvm.ptr<struct<"_QFtest_dt_sliceTt", (i32, i32)>>
+// CHECK:         %[[PTRTOINT_DTYPE_SIZE:.*]] = llvm.ptrtoint %[[GEP_DTYPE_SIZE]] : !llvm.ptr<struct<"_QFtest_dt_sliceTt", (i32, i32)>> to i64
+// CHECK:         %[[ADJUSTED_OFFSET:.*]] = llvm.sub %3, %30  : i64
+// CHECK:         %[[BOX7:.*]] = llvm.insertvalue %[[ZERO]], %[[BOX6]][7 : i32, 0 : i32, 0 : i32] : !llvm.struct<(ptr<i32>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<1 x array<3 x i64>>)>
+// CHECK:         %[[EXT_SUB:.*]] = llvm.sub %[[C10]], %[[C1]]  : i64
+// CHECK:         %[[EXT_ADD:.*]] = llvm.add %[[EXT_SUB]], %[[C2]]  : i64
+// CHECK:         %[[EXT_SDIV:.*]] = llvm.sdiv %[[EXT_ADD]], %[[C2]]  : i64
+// CHECK:         %[[EXT_ICMP:.*]] = llvm.icmp "sgt" %[[EXT_SDIV]], %[[ZERO]] : i64
+// CHECK:         %[[EXT_SELECT:.*]] = llvm.select %[[EXT_ICMP]], %[[EXT_SDIV]], %[[ZERO]] : i1, i64
+// CHECK:         %[[BOX8:.*]] = llvm.insertvalue %[[EXT_SELECT]], %[[BOX7]][7 : i32, 0 : i32, 1 : i32] : !llvm.struct<(ptr<i32>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<1 x array<3 x i64>>)>
+// CHECK:         %[[STRIDE_MUL:.*]] = llvm.mul %[[PTRTOINT_DTYPE_SIZE]], %[[C2]]  : i64
+// CHECK:         %[[BOX9:.*]] = llvm.insertvalue %[[STRIDE_MUL]], %[[BOX8]][7 : i32, 0 : i32, 2 : i32] : !llvm.struct<(ptr<i32>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<1 x array<3 x i64>>)>
+// CHECK:         %[[BASE_PTR:.*]] = llvm.getelementptr %[[X]][%[[ZERO]], %[[ADJUSTED_OFFSET]], %[[C0]]] : (!llvm.ptr<array<20 x struct<"_QFtest_dt_sliceTt", (i32, i32)>>>, i64, i64, i32) -> !llvm.ptr<array<20 x struct<"_QFtest_dt_sliceTt", (i32, i32)>>>
+// CHECK:         %[[ADDR_BITCAST:.*]] = llvm.bitcast %[[BASE_PTR]] : !llvm.ptr<array<20 x struct<"_QFtest_dt_sliceTt", (i32, i32)>>> to !llvm.ptr<i32>
+// CHECK:         %[[BOX10:.*]] = llvm.insertvalue %[[ADDR_BITCAST]], %[[BOX9]][0 : i32] : !llvm.struct<(ptr<i32>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<1 x array<3 x i64>>)>
+// CHECK:         llvm.store %[[BOX10]], %[[ALLOCA]] : !llvm.ptr<struct<(ptr<i32>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<1 x array<3 x i64>>)>>
+// CHECK:         llvm.call @_QPtest_dt_callee(%1) : (!llvm.ptr<struct<(ptr<i32>, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, i{{.*}}, array<1 x array<3 x i64>>)>>) -> ()