abc4f74df7
This is a first simple patch to introduce a new FIR attribute to carry the CUDA variable attribute information to hlfir.declare and fir.declare operations. It currently lowers this information for local variables. The texture attribute is omitted since it is rejected by semantic and will not make its way to MLIR. This new attribute is added as optional attribute to the hlfir.declare and fir.declare operations.
795 lines
36 KiB
C++
795 lines
36 KiB
C++
//===- ConvertToFIR.cpp - Convert HLFIR to FIR ----------------------------===//
|
|
//
|
|
// 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
|
|
//
|
|
//===----------------------------------------------------------------------===//
|
|
// This file defines a pass to lower HLFIR to FIR
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
#include "flang/Optimizer/Builder/Character.h"
|
|
#include "flang/Optimizer/Builder/FIRBuilder.h"
|
|
#include "flang/Optimizer/Builder/HLFIRTools.h"
|
|
#include "flang/Optimizer/Builder/MutableBox.h"
|
|
#include "flang/Optimizer/Builder/Runtime/Assign.h"
|
|
#include "flang/Optimizer/Builder/Runtime/Derived.h"
|
|
#include "flang/Optimizer/Builder/Runtime/Inquiry.h"
|
|
#include "flang/Optimizer/Builder/Todo.h"
|
|
#include "flang/Optimizer/Dialect/FIRDialect.h"
|
|
#include "flang/Optimizer/Dialect/FIROps.h"
|
|
#include "flang/Optimizer/Dialect/FIRType.h"
|
|
#include "flang/Optimizer/Dialect/Support/FIRContext.h"
|
|
#include "flang/Optimizer/HLFIR/HLFIROps.h"
|
|
#include "flang/Optimizer/HLFIR/Passes.h"
|
|
#include "mlir/Transforms/DialectConversion.h"
|
|
|
|
namespace hlfir {
|
|
#define GEN_PASS_DEF_CONVERTHLFIRTOFIR
|
|
#include "flang/Optimizer/HLFIR/Passes.h.inc"
|
|
} // namespace hlfir
|
|
|
|
using namespace mlir;
|
|
|
|
static mlir::Value genAllocatableTempFromSourceBox(mlir::Location loc,
|
|
fir::FirOpBuilder &builder,
|
|
mlir::Value sourceBox) {
|
|
assert(sourceBox.getType().isa<fir::BaseBoxType>() &&
|
|
"must be a base box type");
|
|
// Use the runtime to make a quick and dirty temp with the rhs value.
|
|
// Overkill for scalar rhs that could be done in much more clever ways.
|
|
// Note that temp descriptor must have the allocatable flag set so that
|
|
// the runtime will allocate it with the shape and type parameters of
|
|
// the RHS.
|
|
// This has the huge benefit of dealing with all cases, including
|
|
// polymorphic entities.
|
|
mlir::Type fromHeapType = fir::HeapType::get(fir::unwrapRefType(
|
|
sourceBox.getType().cast<fir::BaseBoxType>().getEleTy()));
|
|
mlir::Type fromBoxHeapType = fir::BoxType::get(fromHeapType);
|
|
mlir::Value fromMutableBox =
|
|
fir::factory::genNullBoxStorage(builder, loc, fromBoxHeapType);
|
|
fir::runtime::genAssignTemporary(builder, loc, fromMutableBox, sourceBox);
|
|
mlir::Value copy = builder.create<fir::LoadOp>(loc, fromMutableBox);
|
|
return copy;
|
|
}
|
|
|
|
namespace {
|
|
/// May \p lhs alias with \p rhs?
|
|
/// TODO: implement HLFIR alias analysis.
|
|
class AssignOpConversion : public mlir::OpRewritePattern<hlfir::AssignOp> {
|
|
public:
|
|
explicit AssignOpConversion(mlir::MLIRContext *ctx) : OpRewritePattern{ctx} {}
|
|
|
|
mlir::LogicalResult
|
|
matchAndRewrite(hlfir::AssignOp assignOp,
|
|
mlir::PatternRewriter &rewriter) const override {
|
|
mlir::Location loc = assignOp->getLoc();
|
|
hlfir::Entity lhs(assignOp.getLhs());
|
|
hlfir::Entity rhs(assignOp.getRhs());
|
|
auto module = assignOp->getParentOfType<mlir::ModuleOp>();
|
|
fir::FirOpBuilder builder(rewriter, module);
|
|
|
|
if (rhs.getType().isa<hlfir::ExprType>()) {
|
|
mlir::emitError(loc, "hlfir must be bufferized with --bufferize-hlfir "
|
|
"pass before being converted to FIR");
|
|
return mlir::failure();
|
|
}
|
|
auto [rhsExv, rhsCleanUp] =
|
|
hlfir::translateToExtendedValue(loc, builder, rhs);
|
|
auto [lhsExv, lhsCleanUp] =
|
|
hlfir::translateToExtendedValue(loc, builder, lhs);
|
|
assert(!lhsCleanUp && !rhsCleanUp &&
|
|
"variable to fir::ExtendedValue must not require cleanup");
|
|
|
|
auto emboxRHS = [&](fir::ExtendedValue &rhsExv) -> mlir::Value {
|
|
// There may be overlap between lhs and rhs. The runtime is able to detect
|
|
// and to make a copy of the rhs before modifying the lhs if needed.
|
|
// The code below relies on this and does not do any compile time alias
|
|
// analysis.
|
|
const bool rhsIsValue = fir::isa_trivial(fir::getBase(rhsExv).getType());
|
|
if (rhsIsValue) {
|
|
// createBox can only be called for fir::ExtendedValue that are
|
|
// already in memory. Place the integer/real/complex/logical scalar
|
|
// in memory.
|
|
// The RHS might be i1, which is not supported for emboxing.
|
|
// If LHS is not polymorphic, we may cast the RHS to the LHS type
|
|
// before emboxing. If LHS is polymorphic we have to figure out
|
|
// the data type for RHS emboxing anyway.
|
|
// It is probably a good idea to make sure that the data type
|
|
// of the RHS is always a valid Fortran storage data type.
|
|
// For the time being, just handle i1 explicitly here.
|
|
mlir::Type rhsType = rhs.getFortranElementType();
|
|
mlir::Value rhsVal = fir::getBase(rhsExv);
|
|
if (rhsType == builder.getI1Type()) {
|
|
rhsType = fir::LogicalType::get(builder.getContext(), 4);
|
|
rhsVal = builder.createConvert(loc, rhsType, rhsVal);
|
|
}
|
|
mlir::Value temp = builder.create<fir::AllocaOp>(loc, rhsType);
|
|
builder.create<fir::StoreOp>(loc, rhsVal, temp);
|
|
rhsExv = temp;
|
|
}
|
|
return fir::getBase(builder.createBox(loc, rhsExv));
|
|
};
|
|
|
|
if (assignOp.isAllocatableAssignment()) {
|
|
// Whole allocatable assignment: use the runtime to deal with the
|
|
// reallocation.
|
|
mlir::Value from = emboxRHS(rhsExv);
|
|
mlir::Value to = fir::getBase(lhsExv);
|
|
if (assignOp.mustKeepLhsLengthInAllocatableAssignment()) {
|
|
// Indicate the runtime that it should not reallocate in case of length
|
|
// mismatch, and that it should use the LHS explicit/assumed length if
|
|
// allocating/reallocation the LHS.
|
|
// Note that AssignExplicitLengthCharacter() must be used
|
|
// when isTemporaryLHS() is true here: the LHS is known to be
|
|
// character allocatable in this case, so finalization will not
|
|
// happen (as implied by temporary_lhs attribute), and LHS
|
|
// must keep its length (as implied by keep_lhs_length_if_realloc).
|
|
fir::runtime::genAssignExplicitLengthCharacter(builder, loc, to, from);
|
|
} else if (assignOp.isTemporaryLHS()) {
|
|
// Use AssignTemporary, when the LHS is a compiler generated temporary.
|
|
// Note that it also works properly for polymorphic LHS (i.e. the LHS
|
|
// will have the RHS dynamic type after the assignment).
|
|
fir::runtime::genAssignTemporary(builder, loc, to, from);
|
|
} else if (lhs.isPolymorphic()) {
|
|
// Indicate the runtime that the LHS must have the RHS dynamic type
|
|
// after the assignment.
|
|
fir::runtime::genAssignPolymorphic(builder, loc, to, from);
|
|
} else {
|
|
fir::runtime::genAssign(builder, loc, to, from);
|
|
}
|
|
} else if (lhs.isArray() ||
|
|
// Special case for element-by-element (or scalar) assignments
|
|
// generated for creating polymorphic expressions.
|
|
// The LHS of these assignments is a box describing just
|
|
// a single element, not the whole allocatable temp.
|
|
// They do not have 'realloc' attribute, because reallocation
|
|
// must not happen. The only expected effect of such an
|
|
// assignment is the copy of the contents, because the dynamic
|
|
// types of the LHS and the RHS must match already. We use the
|
|
// runtime in this case so that the polymorphic (including
|
|
// unlimited) content is copied properly.
|
|
(lhs.isPolymorphic() && assignOp.isTemporaryLHS())) {
|
|
// Use the runtime for simplicity. An optimization pass will be added to
|
|
// inline array assignment when profitable.
|
|
mlir::Value from = emboxRHS(rhsExv);
|
|
mlir::Value to = fir::getBase(builder.createBox(loc, lhsExv));
|
|
// This is not a whole allocatable assignment: the runtime will not
|
|
// reallocate and modify "toMutableBox" even if it is taking it by
|
|
// reference.
|
|
auto toMutableBox = builder.createTemporary(loc, to.getType());
|
|
builder.create<fir::StoreOp>(loc, to, toMutableBox);
|
|
if (assignOp.isTemporaryLHS())
|
|
fir::runtime::genAssignTemporary(builder, loc, toMutableBox, from);
|
|
else
|
|
fir::runtime::genAssign(builder, loc, toMutableBox, from);
|
|
} else {
|
|
// TODO: use the type specification to see if IsFinalizable is set,
|
|
// or propagate IsFinalizable attribute from lowering.
|
|
bool needFinalization =
|
|
!assignOp.isTemporaryLHS() &&
|
|
mlir::isa<fir::RecordType>(fir::getElementTypeOf(lhsExv));
|
|
|
|
// genScalarAssignment() must take care of potential overlap
|
|
// between LHS and RHS. Note that the overlap is possible
|
|
// also for components of LHS/RHS, and the Assign() runtime
|
|
// must take care of it.
|
|
fir::factory::genScalarAssignment(builder, loc, lhsExv, rhsExv,
|
|
needFinalization,
|
|
assignOp.isTemporaryLHS());
|
|
}
|
|
rewriter.eraseOp(assignOp);
|
|
return mlir::success();
|
|
}
|
|
};
|
|
|
|
class CopyInOpConversion : public mlir::OpRewritePattern<hlfir::CopyInOp> {
|
|
public:
|
|
explicit CopyInOpConversion(mlir::MLIRContext *ctx) : OpRewritePattern{ctx} {}
|
|
|
|
struct CopyInResult {
|
|
mlir::Value addr;
|
|
mlir::Value wasCopied;
|
|
};
|
|
|
|
static CopyInResult genNonOptionalCopyIn(mlir::Location loc,
|
|
fir::FirOpBuilder &builder,
|
|
hlfir::CopyInOp copyInOp) {
|
|
mlir::Value inputVariable = copyInOp.getVar();
|
|
mlir::Type resultAddrType = copyInOp.getCopiedIn().getType();
|
|
mlir::Value isContiguous =
|
|
fir::runtime::genIsContiguous(builder, loc, inputVariable);
|
|
mlir::Value addr =
|
|
builder
|
|
.genIfOp(loc, {resultAddrType}, isContiguous,
|
|
/*withElseRegion=*/true)
|
|
.genThen(
|
|
[&]() { builder.create<fir::ResultOp>(loc, inputVariable); })
|
|
.genElse([&] {
|
|
// Create temporary on the heap. Note that the runtime is used and
|
|
// that is desired: since the data copy happens under a runtime
|
|
// check (for IsContiguous) the copy loops can hardly provide any
|
|
// value to optimizations, instead, the optimizer just wastes
|
|
// compilation time on these loops.
|
|
mlir::Value temp =
|
|
genAllocatableTempFromSourceBox(loc, builder, inputVariable);
|
|
// Get rid of allocatable flag in the fir.box.
|
|
temp = builder.create<fir::ReboxOp>(loc, resultAddrType, temp,
|
|
/*shape=*/mlir::Value{},
|
|
/*slice=*/mlir::Value{});
|
|
builder.create<fir::ResultOp>(loc, temp);
|
|
})
|
|
.getResults()[0];
|
|
return {addr, builder.genNot(loc, isContiguous)};
|
|
}
|
|
|
|
static CopyInResult genOptionalCopyIn(mlir::Location loc,
|
|
fir::FirOpBuilder &builder,
|
|
hlfir::CopyInOp copyInOp) {
|
|
mlir::Type resultAddrType = copyInOp.getCopiedIn().getType();
|
|
mlir::Value isPresent = copyInOp.getVarIsPresent();
|
|
auto res =
|
|
builder
|
|
.genIfOp(loc, {resultAddrType, builder.getI1Type()}, isPresent,
|
|
/*withElseRegion=*/true)
|
|
.genThen([&]() {
|
|
CopyInResult res = genNonOptionalCopyIn(loc, builder, copyInOp);
|
|
builder.create<fir::ResultOp>(
|
|
loc, mlir::ValueRange{res.addr, res.wasCopied});
|
|
})
|
|
.genElse([&] {
|
|
mlir::Value absent =
|
|
builder.create<fir::AbsentOp>(loc, resultAddrType);
|
|
builder.create<fir::ResultOp>(
|
|
loc, mlir::ValueRange{absent, isPresent});
|
|
})
|
|
.getResults();
|
|
return {res[0], res[1]};
|
|
}
|
|
|
|
mlir::LogicalResult
|
|
matchAndRewrite(hlfir::CopyInOp copyInOp,
|
|
mlir::PatternRewriter &rewriter) const override {
|
|
mlir::Location loc = copyInOp.getLoc();
|
|
fir::FirOpBuilder builder(rewriter, copyInOp.getOperation());
|
|
CopyInResult result = copyInOp.getVarIsPresent()
|
|
? genOptionalCopyIn(loc, builder, copyInOp)
|
|
: genNonOptionalCopyIn(loc, builder, copyInOp);
|
|
rewriter.replaceOp(copyInOp, {result.addr, result.wasCopied});
|
|
return mlir::success();
|
|
}
|
|
};
|
|
|
|
class CopyOutOpConversion : public mlir::OpRewritePattern<hlfir::CopyOutOp> {
|
|
public:
|
|
explicit CopyOutOpConversion(mlir::MLIRContext *ctx)
|
|
: OpRewritePattern{ctx} {}
|
|
|
|
mlir::LogicalResult
|
|
matchAndRewrite(hlfir::CopyOutOp copyOutOp,
|
|
mlir::PatternRewriter &rewriter) const override {
|
|
mlir::Location loc = copyOutOp.getLoc();
|
|
fir::FirOpBuilder builder(rewriter, copyOutOp.getOperation());
|
|
|
|
builder.genIfThen(loc, copyOutOp.getWasCopied())
|
|
.genThen([&]() {
|
|
mlir::Value temp = copyOutOp.getTemp();
|
|
if (mlir::Value var = copyOutOp.getVar()) {
|
|
auto mutableBoxTo = builder.createTemporary(loc, var.getType());
|
|
builder.create<fir::StoreOp>(loc, var, mutableBoxTo);
|
|
// Generate CopyOutAssign() call to copy data from the temporary
|
|
// to the actualArg. Note that in case the actual argument
|
|
// is ALLOCATABLE/POINTER the CopyOutAssign() implementation
|
|
// should not engage its reallocation, because the temporary
|
|
// is rank, shape and type compatible with it.
|
|
// Moreover, CopyOutAssign() guarantees that there will be no
|
|
// finalization for the LHS even if it is of a derived type
|
|
// with finalization.
|
|
fir::runtime::genCopyOutAssign(builder, loc, mutableBoxTo, temp,
|
|
/*skipToInit=*/true);
|
|
}
|
|
// Destroy components of the temporary (if any).
|
|
fir::runtime::genDerivedTypeDestroyWithoutFinalization(builder, loc,
|
|
temp);
|
|
mlir::Type heapType =
|
|
fir::HeapType::get(fir::dyn_cast_ptrOrBoxEleTy(temp.getType()));
|
|
mlir::Value tempAddr =
|
|
builder.create<fir::BoxAddrOp>(loc, heapType, temp);
|
|
|
|
// Deallocate the top-level entity of the temporary.
|
|
//
|
|
// Note that this FreeMemOp is coupled with the runtime
|
|
// allocation engaged by the code generated by
|
|
// genAllocatableTempFromSourceBox().
|
|
builder.create<fir::FreeMemOp>(loc, tempAddr);
|
|
})
|
|
.end();
|
|
rewriter.eraseOp(copyOutOp);
|
|
return mlir::success();
|
|
}
|
|
};
|
|
|
|
class DeclareOpConversion : public mlir::OpRewritePattern<hlfir::DeclareOp> {
|
|
public:
|
|
explicit DeclareOpConversion(mlir::MLIRContext *ctx)
|
|
: OpRewritePattern{ctx} {}
|
|
|
|
mlir::LogicalResult
|
|
matchAndRewrite(hlfir::DeclareOp declareOp,
|
|
mlir::PatternRewriter &rewriter) const override {
|
|
mlir::Location loc = declareOp->getLoc();
|
|
mlir::Value memref = declareOp.getMemref();
|
|
fir::FortranVariableFlagsAttr fortranAttrs;
|
|
fir::CUDAAttributeAttr cudaAttr;
|
|
if (auto attrs = declareOp.getFortranAttrs())
|
|
fortranAttrs =
|
|
fir::FortranVariableFlagsAttr::get(rewriter.getContext(), *attrs);
|
|
if (auto attr = declareOp.getCudaAttr())
|
|
cudaAttr = fir::CUDAAttributeAttr::get(rewriter.getContext(), *attr);
|
|
auto firDeclareOp = rewriter.create<fir::DeclareOp>(
|
|
loc, memref.getType(), memref, declareOp.getShape(),
|
|
declareOp.getTypeparams(), declareOp.getUniqName(), fortranAttrs,
|
|
cudaAttr);
|
|
|
|
// Propagate other attributes from hlfir.declare to fir.declare.
|
|
// OpenACC's acc.declare is one example. Right now, the propagation
|
|
// is verbatim.
|
|
mlir::NamedAttrList elidedAttrs =
|
|
mlir::NamedAttrList{firDeclareOp->getAttrs()};
|
|
for (const mlir::NamedAttribute &attr : declareOp->getAttrs())
|
|
if (!elidedAttrs.get(attr.getName()))
|
|
firDeclareOp->setAttr(attr.getName(), attr.getValue());
|
|
|
|
auto firBase = firDeclareOp.getResult();
|
|
mlir::Value hlfirBase;
|
|
mlir::Type hlfirBaseType = declareOp.getBase().getType();
|
|
if (hlfirBaseType.isa<fir::BaseBoxType>()) {
|
|
fir::FirOpBuilder builder(rewriter, declareOp.getOperation());
|
|
// Helper to generate the hlfir fir.box with the local lower bounds and
|
|
// type parameters.
|
|
auto genHlfirBox = [&]() -> mlir::Value {
|
|
if (!firBase.getType().isa<fir::BaseBoxType>()) {
|
|
llvm::SmallVector<mlir::Value> typeParams;
|
|
auto maybeCharType =
|
|
fir::unwrapSequenceType(fir::unwrapPassByRefType(hlfirBaseType))
|
|
.dyn_cast<fir::CharacterType>();
|
|
if (!maybeCharType || maybeCharType.hasDynamicLen())
|
|
typeParams.append(declareOp.getTypeparams().begin(),
|
|
declareOp.getTypeparams().end());
|
|
return builder.create<fir::EmboxOp>(
|
|
loc, hlfirBaseType, firBase, declareOp.getShape(),
|
|
/*slice=*/mlir::Value{}, typeParams);
|
|
} else {
|
|
// Rebox so that lower bounds are correct.
|
|
return builder.create<fir::ReboxOp>(loc, hlfirBaseType, firBase,
|
|
declareOp.getShape(),
|
|
/*slice=*/mlir::Value{});
|
|
}
|
|
};
|
|
if (!mlir::cast<fir::FortranVariableOpInterface>(declareOp.getOperation())
|
|
.isOptional()) {
|
|
hlfirBase = genHlfirBox();
|
|
// If the original base is a box too, we could as well
|
|
// use the HLFIR box as the FIR base: otherwise, the two
|
|
// boxes are "alive" at the same time, and the FIR box
|
|
// is used for accessing the base_addr and the HLFIR box
|
|
// is used for accessing the bounds etc. Using the HLFIR box,
|
|
// that holds the same base_addr at this point, makes
|
|
// the representation a little bit more clear.
|
|
if (hlfirBase.getType() == firBase.getType())
|
|
firBase = hlfirBase;
|
|
} else {
|
|
// Need to conditionally rebox/embox the optional: the input fir.box
|
|
// may be null and the rebox would be illegal. It is also important to
|
|
// preserve the optional aspect: the hlfir fir.box should be null if
|
|
// the entity is absent so that later fir.is_present on the hlfir base
|
|
// are valid.
|
|
mlir::Value isPresent =
|
|
builder.create<fir::IsPresentOp>(loc, builder.getI1Type(), firBase);
|
|
hlfirBase = builder
|
|
.genIfOp(loc, {hlfirBaseType}, isPresent,
|
|
/*withElseRegion=*/true)
|
|
.genThen([&] {
|
|
builder.create<fir::ResultOp>(loc, genHlfirBox());
|
|
})
|
|
.genElse([&]() {
|
|
mlir::Value absent =
|
|
builder.create<fir::AbsentOp>(loc, hlfirBaseType);
|
|
builder.create<fir::ResultOp>(loc, absent);
|
|
})
|
|
.getResults()[0];
|
|
}
|
|
} else if (hlfirBaseType.isa<fir::BoxCharType>()) {
|
|
assert(declareOp.getTypeparams().size() == 1 &&
|
|
"must contain character length");
|
|
hlfirBase = rewriter.create<fir::EmboxCharOp>(
|
|
loc, hlfirBaseType, firBase, declareOp.getTypeparams()[0]);
|
|
} else {
|
|
if (hlfirBaseType != firBase.getType()) {
|
|
declareOp.emitOpError()
|
|
<< "unhandled HLFIR variable type '" << hlfirBaseType << "'\n";
|
|
return mlir::failure();
|
|
}
|
|
hlfirBase = firBase;
|
|
}
|
|
rewriter.replaceOp(declareOp, {hlfirBase, firBase});
|
|
return mlir::success();
|
|
}
|
|
};
|
|
|
|
class DesignateOpConversion
|
|
: public mlir::OpRewritePattern<hlfir::DesignateOp> {
|
|
// Helper method to generate the coordinate of the first element
|
|
// of an array section. It is also called for cases of non-section
|
|
// array element addressing.
|
|
static mlir::Value genSubscriptBeginAddr(
|
|
fir::FirOpBuilder &builder, mlir::Location loc,
|
|
hlfir::DesignateOp designate, mlir::Type baseEleTy, mlir::Value base,
|
|
mlir::Value shape,
|
|
const llvm::SmallVector<mlir::Value> &firBaseTypeParameters) {
|
|
assert(!designate.getIndices().empty());
|
|
llvm::SmallVector<mlir::Value> firstElementIndices;
|
|
auto indices = designate.getIndices();
|
|
int i = 0;
|
|
for (auto isTriplet : designate.getIsTripletAttr().asArrayRef()) {
|
|
// Coordinate of the first element are the index and triplets lower
|
|
// bounds
|
|
firstElementIndices.push_back(indices[i]);
|
|
i = i + (isTriplet ? 3 : 1);
|
|
}
|
|
mlir::Type arrayCoorType = fir::ReferenceType::get(baseEleTy);
|
|
base = builder.create<fir::ArrayCoorOp>(
|
|
loc, arrayCoorType, base, shape,
|
|
/*slice=*/mlir::Value{}, firstElementIndices, firBaseTypeParameters);
|
|
return base;
|
|
}
|
|
|
|
public:
|
|
explicit DesignateOpConversion(mlir::MLIRContext *ctx)
|
|
: OpRewritePattern{ctx} {}
|
|
|
|
mlir::LogicalResult
|
|
matchAndRewrite(hlfir::DesignateOp designate,
|
|
mlir::PatternRewriter &rewriter) const override {
|
|
mlir::Location loc = designate.getLoc();
|
|
fir::FirOpBuilder builder(rewriter, designate.getOperation());
|
|
|
|
hlfir::Entity baseEntity(designate.getMemref());
|
|
|
|
if (baseEntity.isMutableBox())
|
|
TODO(loc, "hlfir::designate load of pointer or allocatable");
|
|
|
|
mlir::Type designateResultType = designate.getResult().getType();
|
|
llvm::SmallVector<mlir::Value> firBaseTypeParameters;
|
|
auto [base, shape] = hlfir::genVariableFirBaseShapeAndParams(
|
|
loc, builder, baseEntity, firBaseTypeParameters);
|
|
mlir::Type baseEleTy = hlfir::getFortranElementType(base.getType());
|
|
mlir::Type resultEleTy = hlfir::getFortranElementType(designateResultType);
|
|
|
|
mlir::Value fieldIndex;
|
|
if (designate.getComponent()) {
|
|
mlir::Type baseRecordType = baseEntity.getFortranElementType();
|
|
if (fir::isRecordWithTypeParameters(baseRecordType))
|
|
TODO(loc, "hlfir.designate with a parametrized derived type base");
|
|
fieldIndex = builder.create<fir::FieldIndexOp>(
|
|
loc, fir::FieldType::get(builder.getContext()),
|
|
designate.getComponent().value(), baseRecordType,
|
|
/*typeParams=*/mlir::ValueRange{});
|
|
if (baseEntity.isScalar()) {
|
|
// Component refs of scalar base right away:
|
|
// - scalar%scalar_component [substring|complex_part] or
|
|
// - scalar%static_size_array_comp
|
|
// - scalar%array(indices) [substring| complex part]
|
|
mlir::Type componentType = baseEleTy.cast<fir::RecordType>().getType(
|
|
designate.getComponent().value());
|
|
mlir::Type coorTy = fir::ReferenceType::get(componentType);
|
|
base = builder.create<fir::CoordinateOp>(loc, coorTy, base, fieldIndex);
|
|
if (componentType.isa<fir::BaseBoxType>()) {
|
|
auto variableInterface = mlir::cast<fir::FortranVariableOpInterface>(
|
|
designate.getOperation());
|
|
if (variableInterface.isAllocatable() ||
|
|
variableInterface.isPointer()) {
|
|
rewriter.replaceOp(designate, base);
|
|
return mlir::success();
|
|
}
|
|
TODO(loc,
|
|
"addressing parametrized derived type automatic components");
|
|
}
|
|
baseEleTy = hlfir::getFortranElementType(componentType);
|
|
shape = designate.getComponentShape();
|
|
} else {
|
|
// array%component[(indices) substring|complex part] cases.
|
|
// Component ref of array bases are dealt with below in embox/rebox.
|
|
assert(designateResultType.isa<fir::BaseBoxType>());
|
|
}
|
|
}
|
|
|
|
if (designateResultType.isa<fir::BaseBoxType>()) {
|
|
// Generate embox or rebox.
|
|
mlir::Type eleTy = fir::unwrapPassByRefType(designateResultType);
|
|
bool isScalarDesignator = !eleTy.isa<fir::SequenceType>();
|
|
mlir::Value sourceBox;
|
|
if (isScalarDesignator) {
|
|
// The base box will be used for emboxing the scalar element.
|
|
sourceBox = base;
|
|
// Generate the coordinate of the element.
|
|
base = genSubscriptBeginAddr(builder, loc, designate, baseEleTy, base,
|
|
shape, firBaseTypeParameters);
|
|
shape = nullptr;
|
|
// Type information will be taken from the source box,
|
|
// so the type parameters are not needed.
|
|
firBaseTypeParameters.clear();
|
|
}
|
|
llvm::SmallVector<mlir::Value> triples;
|
|
llvm::SmallVector<mlir::Value> sliceFields;
|
|
mlir::Type idxTy = builder.getIndexType();
|
|
auto subscripts = designate.getIndices();
|
|
if (fieldIndex && baseEntity.isArray()) {
|
|
// array%scalar_comp or array%array_comp(indices)
|
|
// Generate triples for array(:, :, ...).
|
|
triples = genFullSliceTriples(builder, loc, baseEntity);
|
|
sliceFields.push_back(fieldIndex);
|
|
// Add indices in the field path for "array%array_comp(indices)"
|
|
// case. The indices of components provided to the sliceOp must
|
|
// be zero based (fir.slice has no knowledge of the component
|
|
// lower bounds). The component lower bounds are applied here.
|
|
if (!subscripts.empty()) {
|
|
llvm::SmallVector<mlir::Value> lbounds = hlfir::genLowerbounds(
|
|
loc, builder, designate.getComponentShape(), subscripts.size());
|
|
for (auto [i, lb] : llvm::zip(subscripts, lbounds)) {
|
|
mlir::Value iIdx = builder.createConvert(loc, idxTy, i);
|
|
mlir::Value lbIdx = builder.createConvert(loc, idxTy, lb);
|
|
sliceFields.emplace_back(
|
|
builder.create<mlir::arith::SubIOp>(loc, iIdx, lbIdx));
|
|
}
|
|
}
|
|
} else if (!isScalarDesignator) {
|
|
// Otherwise, this is an array section with triplets.
|
|
auto undef = builder.create<fir::UndefOp>(loc, idxTy);
|
|
unsigned i = 0;
|
|
for (auto isTriplet : designate.getIsTriplet()) {
|
|
triples.push_back(subscripts[i++]);
|
|
if (isTriplet) {
|
|
triples.push_back(subscripts[i++]);
|
|
triples.push_back(subscripts[i++]);
|
|
} else {
|
|
triples.push_back(undef);
|
|
triples.push_back(undef);
|
|
}
|
|
}
|
|
}
|
|
llvm::SmallVector<mlir::Value, 2> substring;
|
|
if (!designate.getSubstring().empty()) {
|
|
substring.push_back(designate.getSubstring()[0]);
|
|
mlir::Type idxTy = builder.getIndexType();
|
|
// fir.slice op substring expects the zero based lower bound.
|
|
mlir::Value one = builder.createIntegerConstant(loc, idxTy, 1);
|
|
substring[0] = builder.createConvert(loc, idxTy, substring[0]);
|
|
substring[0] =
|
|
builder.create<mlir::arith::SubIOp>(loc, substring[0], one);
|
|
substring.push_back(designate.getTypeparams()[0]);
|
|
}
|
|
if (designate.getComplexPart()) {
|
|
if (triples.empty())
|
|
triples = genFullSliceTriples(builder, loc, baseEntity);
|
|
sliceFields.push_back(builder.createIntegerConstant(
|
|
loc, idxTy, *designate.getComplexPart()));
|
|
}
|
|
mlir::Value slice;
|
|
if (!triples.empty())
|
|
slice =
|
|
builder.create<fir::SliceOp>(loc, triples, sliceFields, substring);
|
|
else
|
|
assert(sliceFields.empty() && substring.empty());
|
|
llvm::SmallVector<mlir::Type> resultType{designateResultType};
|
|
mlir::Value resultBox;
|
|
if (base.getType().isa<fir::BaseBoxType>())
|
|
resultBox =
|
|
builder.create<fir::ReboxOp>(loc, resultType, base, shape, slice);
|
|
else
|
|
resultBox =
|
|
builder.create<fir::EmboxOp>(loc, resultType, base, shape, slice,
|
|
firBaseTypeParameters, sourceBox);
|
|
rewriter.replaceOp(designate, resultBox);
|
|
return mlir::success();
|
|
}
|
|
|
|
// Otherwise, the result is the address of a scalar, or the address of the
|
|
// first element of a contiguous array section with compile time constant
|
|
// shape. The base may be an array, or a scalar.
|
|
mlir::Type resultAddressType = designateResultType;
|
|
if (auto boxCharType = designateResultType.dyn_cast<fir::BoxCharType>())
|
|
resultAddressType = fir::ReferenceType::get(boxCharType.getEleTy());
|
|
|
|
// Array element indexing.
|
|
if (!designate.getIndices().empty()) {
|
|
// - array(indices) [substring|complex_part] or
|
|
// - scalar%array_comp(indices) [substring|complex_part]
|
|
// This may be a ranked contiguous array section in which case
|
|
// The first element address is being computed.
|
|
base = genSubscriptBeginAddr(builder, loc, designate, baseEleTy, base,
|
|
shape, firBaseTypeParameters);
|
|
}
|
|
|
|
// Scalar substring (potentially on the previously built array element or
|
|
// component reference).
|
|
if (!designate.getSubstring().empty())
|
|
base = fir::factory::CharacterExprHelper{builder, loc}.genSubstringBase(
|
|
base, designate.getSubstring()[0], resultAddressType);
|
|
|
|
// Scalar complex part ref
|
|
if (designate.getComplexPart()) {
|
|
// Sequence types should have already been handled by this point
|
|
assert(!designateResultType.isa<fir::SequenceType>());
|
|
auto index = builder.createIntegerConstant(loc, builder.getIndexType(),
|
|
*designate.getComplexPart());
|
|
auto coorTy = fir::ReferenceType::get(resultEleTy);
|
|
base = builder.create<fir::CoordinateOp>(loc, coorTy, base, index);
|
|
}
|
|
|
|
// Cast/embox the computed scalar address if needed.
|
|
if (designateResultType.isa<fir::BoxCharType>()) {
|
|
assert(designate.getTypeparams().size() == 1 &&
|
|
"must have character length");
|
|
auto emboxChar = builder.create<fir::EmboxCharOp>(
|
|
loc, designateResultType, base, designate.getTypeparams()[0]);
|
|
rewriter.replaceOp(designate, emboxChar.getResult());
|
|
} else {
|
|
base = builder.createConvert(loc, designateResultType, base);
|
|
rewriter.replaceOp(designate, base);
|
|
}
|
|
return mlir::success();
|
|
}
|
|
|
|
private:
|
|
// Generates triple for full slice
|
|
// Used for component and complex part slices when a triple is
|
|
// not specified
|
|
static llvm::SmallVector<mlir::Value>
|
|
genFullSliceTriples(fir::FirOpBuilder &builder, mlir::Location loc,
|
|
hlfir::Entity baseEntity) {
|
|
llvm::SmallVector<mlir::Value> triples;
|
|
mlir::Type idxTy = builder.getIndexType();
|
|
auto one = builder.createIntegerConstant(loc, idxTy, 1);
|
|
for (auto [lb, ub] : hlfir::genBounds(loc, builder, baseEntity)) {
|
|
triples.push_back(builder.createConvert(loc, idxTy, lb));
|
|
triples.push_back(builder.createConvert(loc, idxTy, ub));
|
|
triples.push_back(one);
|
|
}
|
|
return triples;
|
|
}
|
|
};
|
|
|
|
class ParentComponentOpConversion
|
|
: public mlir::OpRewritePattern<hlfir::ParentComponentOp> {
|
|
public:
|
|
explicit ParentComponentOpConversion(mlir::MLIRContext *ctx)
|
|
: OpRewritePattern{ctx} {}
|
|
|
|
mlir::LogicalResult
|
|
matchAndRewrite(hlfir::ParentComponentOp parentComponent,
|
|
mlir::PatternRewriter &rewriter) const override {
|
|
mlir::Location loc = parentComponent.getLoc();
|
|
mlir::Type resultType = parentComponent.getType();
|
|
if (!parentComponent.getType().isa<fir::BoxType>()) {
|
|
mlir::Value baseAddr = parentComponent.getMemref();
|
|
// Scalar parent component ref without any length type parameters. The
|
|
// input may be a fir.class if it is polymorphic, since this is a scalar
|
|
// and the output will be monomorphic, the base address can be extracted
|
|
// from the fir.class.
|
|
if (baseAddr.getType().isa<fir::BaseBoxType>())
|
|
baseAddr = rewriter.create<fir::BoxAddrOp>(loc, baseAddr);
|
|
rewriter.replaceOpWithNewOp<fir::ConvertOp>(parentComponent, resultType,
|
|
baseAddr);
|
|
return mlir::success();
|
|
}
|
|
// Array parent component ref or PDTs.
|
|
hlfir::Entity base{parentComponent.getMemref()};
|
|
mlir::Value baseAddr = base.getBase();
|
|
if (!baseAddr.getType().isa<fir::BaseBoxType>()) {
|
|
// Embox cannot directly be used to address parent components: it expects
|
|
// the output type to match the input type when there are no slices. When
|
|
// the types have at least one component, a slice to the first element can
|
|
// be built, and the result set to the parent component type. Just create
|
|
// a fir.box with the base for now since this covers all cases.
|
|
mlir::Type baseBoxType =
|
|
fir::BoxType::get(base.getElementOrSequenceType());
|
|
assert(!base.hasLengthParameters() &&
|
|
"base must be a box if it has any type parameters");
|
|
baseAddr = rewriter.create<fir::EmboxOp>(
|
|
loc, baseBoxType, baseAddr, parentComponent.getShape(),
|
|
/*slice=*/mlir::Value{}, /*typeParams=*/mlir::ValueRange{});
|
|
}
|
|
rewriter.replaceOpWithNewOp<fir::ReboxOp>(parentComponent, resultType,
|
|
baseAddr,
|
|
/*shape=*/mlir::Value{},
|
|
/*slice=*/mlir::Value{});
|
|
return mlir::success();
|
|
}
|
|
};
|
|
|
|
class NoReassocOpConversion
|
|
: public mlir::OpRewritePattern<hlfir::NoReassocOp> {
|
|
public:
|
|
explicit NoReassocOpConversion(mlir::MLIRContext *ctx)
|
|
: OpRewritePattern{ctx} {}
|
|
|
|
mlir::LogicalResult
|
|
matchAndRewrite(hlfir::NoReassocOp noreassoc,
|
|
mlir::PatternRewriter &rewriter) const override {
|
|
rewriter.replaceOpWithNewOp<fir::NoReassocOp>(noreassoc,
|
|
noreassoc.getVal());
|
|
return mlir::success();
|
|
}
|
|
};
|
|
|
|
class NullOpConversion : public mlir::OpRewritePattern<hlfir::NullOp> {
|
|
public:
|
|
explicit NullOpConversion(mlir::MLIRContext *ctx) : OpRewritePattern{ctx} {}
|
|
|
|
mlir::LogicalResult
|
|
matchAndRewrite(hlfir::NullOp nullop,
|
|
mlir::PatternRewriter &rewriter) const override {
|
|
rewriter.replaceOpWithNewOp<fir::ZeroOp>(nullop, nullop.getType());
|
|
return mlir::success();
|
|
}
|
|
};
|
|
|
|
class GetExtentOpConversion
|
|
: public mlir::OpRewritePattern<hlfir::GetExtentOp> {
|
|
public:
|
|
using mlir::OpRewritePattern<hlfir::GetExtentOp>::OpRewritePattern;
|
|
|
|
mlir::LogicalResult
|
|
matchAndRewrite(hlfir::GetExtentOp getExtentOp,
|
|
mlir::PatternRewriter &rewriter) const override {
|
|
mlir::Value shape = getExtentOp.getShape();
|
|
mlir::Operation *shapeOp = shape.getDefiningOp();
|
|
// the hlfir.shape_of operation which led to the creation of this get_extent
|
|
// operation should now have been lowered to a fir.shape operation
|
|
if (auto s = mlir::dyn_cast_or_null<fir::ShapeOp>(shapeOp)) {
|
|
fir::ShapeType shapeTy = shape.getType().cast<fir::ShapeType>();
|
|
llvm::APInt dim = getExtentOp.getDim();
|
|
uint64_t dimVal = dim.getLimitedValue(shapeTy.getRank());
|
|
mlir::Value extent = s.getExtents()[dimVal];
|
|
rewriter.replaceOp(getExtentOp, extent);
|
|
return mlir::success();
|
|
}
|
|
return mlir::failure();
|
|
}
|
|
};
|
|
|
|
class ConvertHLFIRtoFIR
|
|
: public hlfir::impl::ConvertHLFIRtoFIRBase<ConvertHLFIRtoFIR> {
|
|
public:
|
|
void runOnOperation() override {
|
|
// TODO: like "bufferize-hlfir" pass, runtime signature may be added
|
|
// by this pass. This requires the pass to run on the ModuleOp. It would
|
|
// probably be more optimal to have it run on FuncOp and find a way to
|
|
// generate the signatures in a thread safe way.
|
|
auto module = this->getOperation();
|
|
auto *context = &getContext();
|
|
mlir::RewritePatternSet patterns(context);
|
|
patterns.insert<AssignOpConversion, CopyInOpConversion, CopyOutOpConversion,
|
|
DeclareOpConversion, DesignateOpConversion,
|
|
GetExtentOpConversion, NoReassocOpConversion,
|
|
NullOpConversion, ParentComponentOpConversion>(context);
|
|
mlir::ConversionTarget target(*context);
|
|
target.addIllegalDialect<hlfir::hlfirDialect>();
|
|
target.markUnknownOpDynamicallyLegal(
|
|
[](mlir::Operation *) { return true; });
|
|
if (mlir::failed(mlir::applyPartialConversion(module, target,
|
|
std::move(patterns)))) {
|
|
mlir::emitError(mlir::UnknownLoc::get(context),
|
|
"failure in HLFIR to FIR conversion pass");
|
|
signalPassFailure();
|
|
}
|
|
}
|
|
};
|
|
|
|
} // namespace
|
|
|
|
std::unique_ptr<mlir::Pass> hlfir::createConvertHLFIRtoFIRPass() {
|
|
return std::make_unique<ConvertHLFIRtoFIR>();
|
|
}
|