llvm-project/flang/lib/Evaluate/fold-integer.cpp

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

1118 lines
44 KiB
C++
Raw Normal View History

//===-- lib/Evaluate/fold-integer.cpp -------------------------------------===//
//
// 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
//
//===----------------------------------------------------------------------===//
#include "fold-implementation.h"
#include "fold-reduction.h"
#include "flang/Evaluate/check-expression.h"
namespace Fortran::evaluate {
// Given a collection of ConstantSubscripts values, package them as a Constant.
// Return scalar value if asScalar == true and shape-dim array otherwise.
template <typename T>
Expr<T> PackageConstantBounds(
const ConstantSubscripts &&bounds, bool asScalar = false) {
if (asScalar) {
return Expr<T>{Constant<T>{bounds.at(0)}};
} else {
// As rank-dim array
const int rank{GetRank(bounds)};
std::vector<Scalar<T>> packed(rank);
std::transform(bounds.begin(), bounds.end(), packed.begin(),
[](ConstantSubscript x) { return Scalar<T>(x); });
return Expr<T>{Constant<T>{std::move(packed), ConstantSubscripts{rank}}};
}
}
// Class to retrieve the constant bound of an expression which is an
// array that devolves to a type of Constant<T>
class GetConstantArrayBoundHelper {
public:
template <typename T>
static Expr<T> GetLbound(
const Expr<SomeType> &array, std::optional<int> dim) {
return PackageConstantBounds<T>(
GetConstantArrayBoundHelper(dim, /*getLbound=*/true).Get(array),
dim.has_value());
}
template <typename T>
static Expr<T> GetUbound(
const Expr<SomeType> &array, std::optional<int> dim) {
return PackageConstantBounds<T>(
GetConstantArrayBoundHelper(dim, /*getLbound=*/false).Get(array),
dim.has_value());
}
private:
GetConstantArrayBoundHelper(
std::optional<ConstantSubscript> dim, bool getLbound)
: dim_{dim}, getLbound_{getLbound} {}
template <typename T> ConstantSubscripts Get(const T &) {
// The method is needed for template expansion, but we should never get
// here in practice.
CHECK(false);
return {0};
}
template <typename T> ConstantSubscripts Get(const Constant<T> &x) {
if (getLbound_) {
// Return the lower bound
if (dim_) {
return {x.lbounds().at(*dim_)};
} else {
return x.lbounds();
}
} else {
// Return the upper bound
if (arrayFromParenthesesExpr) {
// Underlying array comes from (x) expression - return shapes
if (dim_) {
return {x.shape().at(*dim_)};
} else {
return x.shape();
}
} else {
return x.ComputeUbounds(dim_);
}
}
}
template <typename T> ConstantSubscripts Get(const Parentheses<T> &x) {
// Cause of temp variable inside parentheses - return [1, ... 1] for lower
// bounds and shape for upper bounds
if (getLbound_) {
return ConstantSubscripts(x.Rank(), ConstantSubscript{1});
} else {
// Indicate that underlying array comes from parentheses expression.
// Continue to unwrap expression until we hit a constant
arrayFromParenthesesExpr = true;
return Get(x.left());
}
}
template <typename T> ConstantSubscripts Get(const Expr<T> &x) {
// recurse through Expr<T>'a until we hit a constant
return common::visit([&](const auto &inner) { return Get(inner); },
// [&](const auto &) { return 0; },
x.u);
}
const std::optional<ConstantSubscript> dim_;
const bool getLbound_;
bool arrayFromParenthesesExpr{false};
};
template <int KIND>
Expr<Type<TypeCategory::Integer, KIND>> LBOUND(FoldingContext &context,
FunctionRef<Type<TypeCategory::Integer, KIND>> &&funcRef) {
using T = Type<TypeCategory::Integer, KIND>;
ActualArguments &args{funcRef.arguments()};
if (const auto *array{UnwrapExpr<Expr<SomeType>>(args[0])}) {
if (int rank{array->Rank()}; rank > 0) {
std::optional<int> dim;
if (funcRef.Rank() == 0) {
// Optional DIM= argument is present: result is scalar.
if (auto dim64{GetInt64Arg(args[1])}) {
if (*dim64 < 1 || *dim64 > rank) {
context.messages().Say("DIM=%jd dimension is out of range for "
"rank-%d array"_err_en_US,
*dim64, rank);
return MakeInvalidIntrinsic<T>(std::move(funcRef));
} else {
dim = *dim64 - 1; // 1-based to 0-based
}
} else {
// DIM= is present but not constant
return Expr<T>{std::move(funcRef)};
}
}
bool lowerBoundsAreOne{true};
if (auto named{ExtractNamedEntity(*array)}) {
const Symbol &symbol{named->GetLastSymbol()};
if (symbol.Rank() == rank) {
lowerBoundsAreOne = false;
if (dim) {
if (auto lb{GetLBOUND(context, *named, *dim)}) {
return Fold(context, ConvertToType<T>(std::move(*lb)));
}
} else if (auto extents{
AsExtentArrayExpr(GetLBOUNDs(context, *named))}) {
return Fold(context,
ConvertToType<T>(Expr<ExtentType>{std::move(*extents)}));
}
} else {
lowerBoundsAreOne = symbol.Rank() == 0; // LBOUND(array%component)
}
}
if (IsActuallyConstant(*array)) {
return GetConstantArrayBoundHelper::GetLbound<T>(*array, dim);
}
if (lowerBoundsAreOne) {
ConstantSubscripts ones(rank, ConstantSubscript{1});
return PackageConstantBounds<T>(std::move(ones), dim.has_value());
}
}
}
return Expr<T>{std::move(funcRef)};
}
template <int KIND>
Expr<Type<TypeCategory::Integer, KIND>> UBOUND(FoldingContext &context,
FunctionRef<Type<TypeCategory::Integer, KIND>> &&funcRef) {
using T = Type<TypeCategory::Integer, KIND>;
ActualArguments &args{funcRef.arguments()};
if (auto *array{UnwrapExpr<Expr<SomeType>>(args[0])}) {
if (int rank{array->Rank()}; rank > 0) {
std::optional<int> dim;
if (funcRef.Rank() == 0) {
// Optional DIM= argument is present: result is scalar.
if (auto dim64{GetInt64Arg(args[1])}) {
if (*dim64 < 1 || *dim64 > rank) {
context.messages().Say("DIM=%jd dimension is out of range for "
"rank-%d array"_err_en_US,
*dim64, rank);
return MakeInvalidIntrinsic<T>(std::move(funcRef));
} else {
dim = *dim64 - 1; // 1-based to 0-based
}
} else {
// DIM= is present but not constant
return Expr<T>{std::move(funcRef)};
}
}
bool takeBoundsFromShape{true};
if (auto named{ExtractNamedEntity(*array)}) {
const Symbol &symbol{named->GetLastSymbol()};
if (symbol.Rank() == rank) {
takeBoundsFromShape = false;
if (dim) {
if (semantics::IsAssumedSizeArray(symbol) && *dim == rank - 1) {
context.messages().Say("DIM=%jd dimension is out of range for "
"rank-%d assumed-size array"_err_en_US,
rank, rank);
return MakeInvalidIntrinsic<T>(std::move(funcRef));
} else if (auto ub{GetUBOUND(context, *named, *dim)}) {
return Fold(context, ConvertToType<T>(std::move(*ub)));
}
} else {
Shape ubounds{GetUBOUNDs(context, *named)};
if (semantics::IsAssumedSizeArray(symbol)) {
CHECK(!ubounds.back());
ubounds.back() = ExtentExpr{-1};
}
if (auto extents{AsExtentArrayExpr(ubounds)}) {
return Fold(context,
ConvertToType<T>(Expr<ExtentType>{std::move(*extents)}));
}
}
} else {
takeBoundsFromShape = symbol.Rank() == 0; // UBOUND(array%component)
}
}
if (IsActuallyConstant(*array)) {
return GetConstantArrayBoundHelper::GetUbound<T>(*array, dim);
}
if (takeBoundsFromShape) {
[flang] Do not lose call in shape inquiry on function reference Currently, something like `print *, size(foo(n,m))` was rewritten to `print *, size(foo_result_symbol)` when foo result is a non constant shape array. This cannot be processed by lowering or reprocessed by a Fortran compiler since the syntax is wrong (`foo_result_symbol` is unknown on the caller side) and the arguments are lost when they might be required to compute the result shape. It is not possible (and probably not desired) to make GetShape fail in general in such case since returning nullopt seems only expected for scalars or assumed rank (see GetRank usage in lib/Semantics/check-call.cpp), and returning a vector with nullopt extent may trigger some checks to believe they are facing an assumed size (like here in intrinsic argument checks: https://github.com/llvm/llvm-project/blob/196204c72c68a577c72af95d70f18e3550939a5e/flang/lib/Evaluate/intrinsics.cpp#L1530). Hence, I went for a solution that limits the rewrite change to folding (where the original expression is returned if the shape depends on a non constant shape from a call). I added a non default option to GetShapeHelper that prevents the rewrite of shape inquiry on calls to descriptor inquiries. At first I wanted to avoid touching GetShapeHelper, but it would require to re-implement all its logic to determine if the shape comes from a function call or not (the expression could be `size(1+foo(n,m))`). So added an alternate entry point to GetShapeHelper seemed the cleanest solution to me. Differential Revision: https://reviews.llvm.org/D116933
2022-01-11 02:09:45 +08:00
if (auto shape{GetContextFreeShape(context, *array)}) {
if (dim) {
if (auto &dimSize{shape->at(*dim)}) {
return Fold(context,
ConvertToType<T>(Expr<ExtentType>{std::move(*dimSize)}));
}
} else if (auto shapeExpr{AsExtentArrayExpr(*shape)}) {
return Fold(context, ConvertToType<T>(std::move(*shapeExpr)));
}
}
}
}
}
return Expr<T>{std::move(funcRef)};
}
// COUNT()
template <typename T>
static Expr<T> FoldCount(FoldingContext &context, FunctionRef<T> &&ref) {
static_assert(T::category == TypeCategory::Integer);
ActualArguments &arg{ref.arguments()};
if (const Constant<LogicalResult> *mask{arg.empty()
? nullptr
: Folder<LogicalResult>{context}.Folding(arg[0])}) {
std::optional<int> dim;
if (CheckReductionDIM(dim, context, arg, 1, mask->Rank())) {
auto accumulator{[&](Scalar<T> &element, const ConstantSubscripts &at) {
if (mask->At(at).IsTrue()) {
element = element.AddSigned(Scalar<T>{1}).value;
}
}};
return Expr<T>{DoReduction<T>(*mask, dim, Scalar<T>{}, accumulator)};
}
}
return Expr<T>{std::move(ref)};
}
// FINDLOC(), MAXLOC(), & MINLOC()
enum class WhichLocation { Findloc, Maxloc, Minloc };
template <WhichLocation WHICH> class LocationHelper {
public:
LocationHelper(
DynamicType &&type, ActualArguments &arg, FoldingContext &context)
: type_{type}, arg_{arg}, context_{context} {}
using Result = std::optional<Constant<SubscriptInteger>>;
using Types = std::conditional_t<WHICH == WhichLocation::Findloc,
AllIntrinsicTypes, RelationalTypes>;
template <typename T> Result Test() const {
if (T::category != type_.category() || T::kind != type_.kind()) {
return std::nullopt;
}
CHECK(arg_.size() == (WHICH == WhichLocation::Findloc ? 6 : 5));
Folder<T> folder{context_};
Constant<T> *array{folder.Folding(arg_[0])};
if (!array) {
return std::nullopt;
}
std::optional<Constant<T>> value;
if constexpr (WHICH == WhichLocation::Findloc) {
if (const Constant<T> *p{folder.Folding(arg_[1])}) {
value.emplace(*p);
} else {
return std::nullopt;
}
}
std::optional<int> dim;
Constant<LogicalResult> *mask{
GetReductionMASK(arg_[maskArg], array->shape(), context_)};
if ((!mask && arg_[maskArg]) ||
!CheckReductionDIM(dim, context_, arg_, dimArg, array->Rank())) {
return std::nullopt;
}
bool back{false};
if (arg_[backArg]) {
const auto *backConst{
Folder<LogicalResult>{context_}.Folding(arg_[backArg])};
if (backConst) {
back = backConst->GetScalarValue().value().IsTrue();
} else {
return std::nullopt;
}
}
const RelationalOperator relation{WHICH == WhichLocation::Findloc
? RelationalOperator::EQ
: WHICH == WhichLocation::Maxloc
? (back ? RelationalOperator::GE : RelationalOperator::GT)
: back ? RelationalOperator::LE
: RelationalOperator::LT};
// Use lower bounds of 1 exclusively.
array->SetLowerBoundsToOne();
ConstantSubscripts at{array->lbounds()}, maskAt, resultIndices, resultShape;
if (mask) {
mask->SetLowerBoundsToOne();
maskAt = mask->lbounds();
}
if (dim) { // DIM=
if (*dim < 1 || *dim > array->Rank()) {
context_.messages().Say("DIM=%d is out of range"_err_en_US, *dim);
return std::nullopt;
}
int zbDim{*dim - 1};
resultShape = array->shape();
resultShape.erase(
resultShape.begin() + zbDim); // scalar if array is vector
ConstantSubscript dimLength{array->shape()[zbDim]};
ConstantSubscript n{GetSize(resultShape)};
for (ConstantSubscript j{0}; j < n; ++j) {
ConstantSubscript hit{0};
if constexpr (WHICH == WhichLocation::Maxloc ||
WHICH == WhichLocation::Minloc) {
value.reset();
}
for (ConstantSubscript k{0}; k < dimLength;
++k, ++at[zbDim], mask && ++maskAt[zbDim]) {
if ((!mask || mask->At(maskAt).IsTrue()) &&
IsHit(array->At(at), value, relation)) {
hit = at[zbDim];
if constexpr (WHICH == WhichLocation::Findloc) {
if (!back) {
break;
}
}
}
}
resultIndices.emplace_back(hit);
at[zbDim] = std::max<ConstantSubscript>(dimLength, 1);
array->IncrementSubscripts(at);
at[zbDim] = 1;
if (mask) {
maskAt[zbDim] = mask->lbounds()[zbDim] +
std::max<ConstantSubscript>(dimLength, 1) - 1;
mask->IncrementSubscripts(maskAt);
maskAt[zbDim] = mask->lbounds()[zbDim];
}
}
} else { // no DIM=
resultShape = ConstantSubscripts{array->Rank()}; // always a vector
ConstantSubscript n{GetSize(array->shape())};
resultIndices = ConstantSubscripts(array->Rank(), 0);
for (ConstantSubscript j{0}; j < n; ++j, array->IncrementSubscripts(at),
mask && mask->IncrementSubscripts(maskAt)) {
if ((!mask || mask->At(maskAt).IsTrue()) &&
IsHit(array->At(at), value, relation)) {
resultIndices = at;
if constexpr (WHICH == WhichLocation::Findloc) {
if (!back) {
break;
}
}
}
}
}
std::vector<Scalar<SubscriptInteger>> resultElements;
for (ConstantSubscript j : resultIndices) {
resultElements.emplace_back(j);
}
return Constant<SubscriptInteger>{
std::move(resultElements), std::move(resultShape)};
}
private:
template <typename T>
bool IsHit(typename Constant<T>::Element element,
std::optional<Constant<T>> &value,
[[maybe_unused]] RelationalOperator relation) const {
std::optional<Expr<LogicalResult>> cmp;
bool result{true};
if (value) {
if constexpr (T::category == TypeCategory::Logical) {
// array(at) .EQV. value?
static_assert(WHICH == WhichLocation::Findloc);
cmp.emplace(ConvertToType<LogicalResult>(
Expr<T>{LogicalOperation<T::kind>{LogicalOperator::Eqv,
Expr<T>{Constant<T>{element}}, Expr<T>{Constant<T>{*value}}}}));
} else { // compare array(at) to value
cmp.emplace(PackageRelation(relation, Expr<T>{Constant<T>{element}},
Expr<T>{Constant<T>{*value}}));
}
Expr<LogicalResult> folded{Fold(context_, std::move(*cmp))};
result = GetScalarConstantValue<LogicalResult>(folded).value().IsTrue();
} else {
// first unmasked element for MAXLOC/MINLOC - always take it
}
if constexpr (WHICH == WhichLocation::Maxloc ||
WHICH == WhichLocation::Minloc) {
if (result) {
value.emplace(std::move(element));
}
}
return result;
}
static constexpr int dimArg{WHICH == WhichLocation::Findloc ? 2 : 1};
static constexpr int maskArg{dimArg + 1};
static constexpr int backArg{maskArg + 2};
DynamicType type_;
ActualArguments &arg_;
FoldingContext &context_;
};
template <WhichLocation which>
static std::optional<Constant<SubscriptInteger>> FoldLocationCall(
ActualArguments &arg, FoldingContext &context) {
if (arg[0]) {
if (auto type{arg[0]->GetType()}) {
return common::SearchTypes(
LocationHelper<which>{std::move(*type), arg, context});
}
}
return std::nullopt;
}
template <WhichLocation which, typename T>
static Expr<T> FoldLocation(FoldingContext &context, FunctionRef<T> &&ref) {
static_assert(T::category == TypeCategory::Integer);
if (std::optional<Constant<SubscriptInteger>> found{
FoldLocationCall<which>(ref.arguments(), context)}) {
return Expr<T>{Fold(
context, ConvertToType<T>(Expr<SubscriptInteger>{std::move(*found)}))};
} else {
return Expr<T>{std::move(ref)};
}
}
// for IALL, IANY, & IPARITY
template <typename T>
static Expr<T> FoldBitReduction(FoldingContext &context, FunctionRef<T> &&ref,
Scalar<T> (Scalar<T>::*operation)(const Scalar<T> &) const,
Scalar<T> identity) {
static_assert(T::category == TypeCategory::Integer);
std::optional<int> dim;
if (std::optional<Constant<T>> array{
ProcessReductionArgs<T>(context, ref.arguments(), dim, identity,
/*ARRAY=*/0, /*DIM=*/1, /*MASK=*/2)}) {
auto accumulator{[&](Scalar<T> &element, const ConstantSubscripts &at) {
element = (element.*operation)(array->At(at));
}};
return Expr<T>{DoReduction<T>(*array, dim, identity, accumulator)};
}
return Expr<T>{std::move(ref)};
}
template <int KIND>
Expr<Type<TypeCategory::Integer, KIND>> FoldIntrinsicFunction(
FoldingContext &context,
FunctionRef<Type<TypeCategory::Integer, KIND>> &&funcRef) {
using T = Type<TypeCategory::Integer, KIND>;
using Int4 = Type<TypeCategory::Integer, 4>;
ActualArguments &args{funcRef.arguments()};
auto *intrinsic{std::get_if<SpecificIntrinsic>(&funcRef.proc().u)};
CHECK(intrinsic);
std::string name{intrinsic->name};
if (name == "abs") { // incl. babs, iiabs, jiaabs, & kiabs
return FoldElementalIntrinsic<T, T>(context, std::move(funcRef),
ScalarFunc<T, T>([&context](const Scalar<T> &i) -> Scalar<T> {
typename Scalar<T>::ValueWithOverflow j{i.ABS()};
if (j.overflow) {
context.messages().Say(
"abs(integer(kind=%d)) folding overflowed"_warn_en_US, KIND);
}
return j.value;
}));
} else if (name == "bit_size") {
return Expr<T>{Scalar<T>::bits};
} else if (name == "ceiling" || name == "floor" || name == "nint") {
if (const auto *cx{UnwrapExpr<Expr<SomeReal>>(args[0])}) {
// NINT rounds ties away from zero, not to even
common::RoundingMode mode{name == "ceiling" ? common::RoundingMode::Up
: name == "floor" ? common::RoundingMode::Down
: common::RoundingMode::TiesAwayFromZero};
return common::visit(
[&](const auto &kx) {
using TR = ResultType<decltype(kx)>;
return FoldElementalIntrinsic<T, TR>(context, std::move(funcRef),
ScalarFunc<T, TR>([&](const Scalar<TR> &x) {
auto y{x.template ToInteger<Scalar<T>>(mode)};
if (y.flags.test(RealFlag::Overflow)) {
context.messages().Say(
"%s intrinsic folding overflow"_warn_en_US, name);
}
return y.value;
}));
},
cx->u);
}
} else if (name == "count") {
return FoldCount<T>(context, std::move(funcRef));
} else if (name == "digits") {
if (const auto *cx{UnwrapExpr<Expr<SomeInteger>>(args[0])}) {
return Expr<T>{common::visit(
[](const auto &kx) {
return Scalar<ResultType<decltype(kx)>>::DIGITS;
},
cx->u)};
} else if (const auto *cx{UnwrapExpr<Expr<SomeReal>>(args[0])}) {
return Expr<T>{common::visit(
[](const auto &kx) {
return Scalar<ResultType<decltype(kx)>>::DIGITS;
},
cx->u)};
} else if (const auto *cx{UnwrapExpr<Expr<SomeComplex>>(args[0])}) {
return Expr<T>{common::visit(
[](const auto &kx) {
return Scalar<typename ResultType<decltype(kx)>::Part>::DIGITS;
},
cx->u)};
}
} else if (name == "dim") {
return FoldElementalIntrinsic<T, T, T>(
context, std::move(funcRef), &Scalar<T>::DIM);
} else if (name == "dshiftl" || name == "dshiftr") {
const auto fptr{
name == "dshiftl" ? &Scalar<T>::DSHIFTL : &Scalar<T>::DSHIFTR};
// Third argument can be of any kind. However, it must be smaller or equal
// than BIT_SIZE. It can be converted to Int4 to simplify.
return FoldElementalIntrinsic<T, T, T, Int4>(context, std::move(funcRef),
ScalarFunc<T, T, T, Int4>(
[&fptr](const Scalar<T> &i, const Scalar<T> &j,
const Scalar<Int4> &shift) -> Scalar<T> {
return std::invoke(fptr, i, j, static_cast<int>(shift.ToInt64()));
}));
} else if (name == "exponent") {
if (auto *sx{UnwrapExpr<Expr<SomeReal>>(args[0])}) {
return common::visit(
[&funcRef, &context](const auto &x) -> Expr<T> {
using TR = typename std::decay_t<decltype(x)>::Result;
return FoldElementalIntrinsic<T, TR>(context, std::move(funcRef),
&Scalar<TR>::template EXPONENT<Scalar<T>>);
},
sx->u);
} else {
DIE("exponent argument must be real");
}
} else if (name == "findloc") {
return FoldLocation<WhichLocation::Findloc, T>(context, std::move(funcRef));
} else if (name == "huge") {
return Expr<T>{Scalar<T>::HUGE()};
} else if (name == "iachar" || name == "ichar") {
auto *someChar{UnwrapExpr<Expr<SomeCharacter>>(args[0])};
CHECK(someChar);
if (auto len{ToInt64(someChar->LEN())}) {
if (len.value() != 1) {
// Do not die, this was not checked before
context.messages().Say(
"Character in intrinsic function %s must have length one"_warn_en_US,
name);
} else {
return common::visit(
[&funcRef, &context](const auto &str) -> Expr<T> {
using Char = typename std::decay_t<decltype(str)>::Result;
return FoldElementalIntrinsic<T, Char>(context,
std::move(funcRef),
ScalarFunc<T, Char>([](const Scalar<Char> &c) {
return Scalar<T>{CharacterUtils<Char::kind>::ICHAR(c)};
}));
},
someChar->u);
}
}
} else if (name == "iand" || name == "ior" || name == "ieor") {
auto fptr{&Scalar<T>::IAND};
if (name == "iand") { // done in fptr declaration
} else if (name == "ior") {
fptr = &Scalar<T>::IOR;
} else if (name == "ieor") {
fptr = &Scalar<T>::IEOR;
} else {
common::die("missing case to fold intrinsic function %s", name.c_str());
}
return FoldElementalIntrinsic<T, T, T>(
context, std::move(funcRef), ScalarFunc<T, T, T>(fptr));
} else if (name == "iall") {
return FoldBitReduction(
context, std::move(funcRef), &Scalar<T>::IAND, Scalar<T>{}.NOT());
} else if (name == "iany") {
return FoldBitReduction(
context, std::move(funcRef), &Scalar<T>::IOR, Scalar<T>{});
} else if (name == "ibclr" || name == "ibset") {
// Second argument can be of any kind. However, it must be smaller
// than BIT_SIZE. It can be converted to Int4 to simplify.
auto fptr{&Scalar<T>::IBCLR};
if (name == "ibclr") { // done in fptr definition
} else if (name == "ibset") {
fptr = &Scalar<T>::IBSET;
} else {
common::die("missing case to fold intrinsic function %s", name.c_str());
}
return FoldElementalIntrinsic<T, T, Int4>(context, std::move(funcRef),
ScalarFunc<T, T, Int4>([&](const Scalar<T> &i,
const Scalar<Int4> &pos) -> Scalar<T> {
auto posVal{static_cast<int>(pos.ToInt64())};
if (posVal < 0) {
context.messages().Say(
"bit position for %s (%d) is negative"_err_en_US, name, posVal);
} else if (posVal >= i.bits) {
context.messages().Say(
"bit position for %s (%d) is not less than %d"_err_en_US, name,
posVal, i.bits);
}
return std::invoke(fptr, i, posVal);
}));
} else if (name == "ibits") {
return FoldElementalIntrinsic<T, T, Int4, Int4>(context, std::move(funcRef),
ScalarFunc<T, T, Int4, Int4>([&](const Scalar<T> &i,
const Scalar<Int4> &pos,
const Scalar<Int4> &len) -> Scalar<T> {
auto posVal{static_cast<int>(pos.ToInt64())};
auto lenVal{static_cast<int>(len.ToInt64())};
if (posVal < 0) {
context.messages().Say(
"bit position for IBITS(POS=%d,LEN=%d) is negative"_err_en_US,
posVal, lenVal);
} else if (lenVal < 0) {
context.messages().Say(
"bit length for IBITS(POS=%d,LEN=%d) is negative"_err_en_US,
posVal, lenVal);
} else if (posVal + lenVal > i.bits) {
context.messages().Say(
"IBITS(POS=%d,LEN=%d) must have POS+LEN no greater than %d"_err_en_US,
posVal + lenVal, i.bits);
}
return i.IBITS(posVal, lenVal);
}));
} else if (name == "index" || name == "scan" || name == "verify") {
if (auto *charExpr{UnwrapExpr<Expr<SomeCharacter>>(args[0])}) {
return common::visit(
[&](const auto &kch) -> Expr<T> {
using TC = typename std::decay_t<decltype(kch)>::Result;
if (UnwrapExpr<Expr<SomeLogical>>(args[2])) { // BACK=
return FoldElementalIntrinsic<T, TC, TC, LogicalResult>(context,
std::move(funcRef),
ScalarFunc<T, TC, TC, LogicalResult>{
[&name](const Scalar<TC> &str, const Scalar<TC> &other,
const Scalar<LogicalResult> &back) -> Scalar<T> {
return name == "index"
? CharacterUtils<TC::kind>::INDEX(
str, other, back.IsTrue())
: name == "scan" ? CharacterUtils<TC::kind>::SCAN(
str, other, back.IsTrue())
: CharacterUtils<TC::kind>::VERIFY(
str, other, back.IsTrue());
}});
} else {
return FoldElementalIntrinsic<T, TC, TC>(context,
std::move(funcRef),
ScalarFunc<T, TC, TC>{
[&name](const Scalar<TC> &str,
const Scalar<TC> &other) -> Scalar<T> {
return name == "index"
? CharacterUtils<TC::kind>::INDEX(str, other)
: name == "scan"
? CharacterUtils<TC::kind>::SCAN(str, other)
: CharacterUtils<TC::kind>::VERIFY(str, other);
}});
}
},
charExpr->u);
} else {
DIE("first argument must be CHARACTER");
}
} else if (name == "int") {
if (auto *expr{UnwrapExpr<Expr<SomeType>>(args[0])}) {
return common::visit(
[&](auto &&x) -> Expr<T> {
using From = std::decay_t<decltype(x)>;
if constexpr (std::is_same_v<From, BOZLiteralConstant> ||
IsNumericCategoryExpr<From>()) {
return Fold(context, ConvertToType<T>(std::move(x)));
}
DIE("int() argument type not valid");
},
std::move(expr->u));
}
} else if (name == "int_ptr_kind") {
return Expr<T>{8};
} else if (name == "kind") {
if constexpr (common::HasMember<T, IntegerTypes>) {
return Expr<T>{args[0].value().GetType()->kind()};
} else {
DIE("kind() result not integral");
}
} else if (name == "iparity") {
return FoldBitReduction(
context, std::move(funcRef), &Scalar<T>::IEOR, Scalar<T>{});
} else if (name == "ishft") {
return FoldElementalIntrinsic<T, T, Int4>(context, std::move(funcRef),
ScalarFunc<T, T, Int4>([&](const Scalar<T> &i,
const Scalar<Int4> &pos) -> Scalar<T> {
auto posVal{static_cast<int>(pos.ToInt64())};
if (posVal < -i.bits) {
context.messages().Say(
"SHIFT=%d count for ishft is less than %d"_err_en_US, posVal,
-i.bits);
} else if (posVal > i.bits) {
context.messages().Say(
"SHIFT=%d count for ishft is greater than %d"_err_en_US, posVal,
i.bits);
}
return i.ISHFT(posVal);
}));
} else if (name == "lbound") {
return LBOUND(context, std::move(funcRef));
} else if (name == "leadz" || name == "trailz" || name == "poppar" ||
name == "popcnt") {
if (auto *sn{UnwrapExpr<Expr<SomeInteger>>(args[0])}) {
return common::visit(
[&funcRef, &context, &name](const auto &n) -> Expr<T> {
using TI = typename std::decay_t<decltype(n)>::Result;
if (name == "poppar") {
return FoldElementalIntrinsic<T, TI>(context, std::move(funcRef),
ScalarFunc<T, TI>([](const Scalar<TI> &i) -> Scalar<T> {
return Scalar<T>{i.POPPAR() ? 1 : 0};
}));
}
auto fptr{&Scalar<TI>::LEADZ};
if (name == "leadz") { // done in fptr definition
} else if (name == "trailz") {
fptr = &Scalar<TI>::TRAILZ;
} else if (name == "popcnt") {
fptr = &Scalar<TI>::POPCNT;
} else {
common::die(
"missing case to fold intrinsic function %s", name.c_str());
}
return FoldElementalIntrinsic<T, TI>(context, std::move(funcRef),
ScalarFunc<T, TI>([&fptr](const Scalar<TI> &i) -> Scalar<T> {
return Scalar<T>{std::invoke(fptr, i)};
}));
},
sn->u);
} else {
DIE("leadz argument must be integer");
}
} else if (name == "len") {
if (auto *charExpr{UnwrapExpr<Expr<SomeCharacter>>(args[0])}) {
return common::visit(
[&](auto &kx) {
if (auto len{kx.LEN()}) {
if (IsScopeInvariantExpr(*len)) {
return Fold(context, ConvertToType<T>(*std::move(len)));
} else {
return Expr<T>{std::move(funcRef)};
}
} else {
return Expr<T>{std::move(funcRef)};
}
},
charExpr->u);
} else {
DIE("len() argument must be of character type");
}
} else if (name == "len_trim") {
if (auto *charExpr{UnwrapExpr<Expr<SomeCharacter>>(args[0])}) {
return common::visit(
[&](const auto &kch) -> Expr<T> {
using TC = typename std::decay_t<decltype(kch)>::Result;
return FoldElementalIntrinsic<T, TC>(context, std::move(funcRef),
ScalarFunc<T, TC>{[](const Scalar<TC> &str) -> Scalar<T> {
return CharacterUtils<TC::kind>::LEN_TRIM(str);
}});
},
charExpr->u);
} else {
DIE("len_trim() argument must be of character type");
}
} else if (name == "maskl" || name == "maskr") {
// Argument can be of any kind but value has to be smaller than BIT_SIZE.
// It can be safely converted to Int4 to simplify.
const auto fptr{name == "maskl" ? &Scalar<T>::MASKL : &Scalar<T>::MASKR};
return FoldElementalIntrinsic<T, Int4>(context, std::move(funcRef),
ScalarFunc<T, Int4>([&fptr](const Scalar<Int4> &places) -> Scalar<T> {
return fptr(static_cast<int>(places.ToInt64()));
}));
} else if (name == "max") {
return FoldMINorMAX(context, std::move(funcRef), Ordering::Greater);
} else if (name == "max0" || name == "max1") {
return RewriteSpecificMINorMAX(context, std::move(funcRef));
} else if (name == "maxexponent") {
if (auto *sx{UnwrapExpr<Expr<SomeReal>>(args[0])}) {
return common::visit(
[](const auto &x) {
using TR = typename std::decay_t<decltype(x)>::Result;
return Expr<T>{Scalar<TR>::MAXEXPONENT};
},
sx->u);
}
} else if (name == "maxloc") {
return FoldLocation<WhichLocation::Maxloc, T>(context, std::move(funcRef));
} else if (name == "maxval") {
return FoldMaxvalMinval<T>(context, std::move(funcRef),
RelationalOperator::GT, T::Scalar::Least());
} else if (name == "merge") {
return FoldMerge<T>(context, std::move(funcRef));
} else if (name == "merge_bits") {
return FoldElementalIntrinsic<T, T, T, T>(
context, std::move(funcRef), &Scalar<T>::MERGE_BITS);
} else if (name == "min") {
return FoldMINorMAX(context, std::move(funcRef), Ordering::Less);
} else if (name == "min0" || name == "min1") {
return RewriteSpecificMINorMAX(context, std::move(funcRef));
} else if (name == "minexponent") {
if (auto *sx{UnwrapExpr<Expr<SomeReal>>(args[0])}) {
return common::visit(
[](const auto &x) {
using TR = typename std::decay_t<decltype(x)>::Result;
return Expr<T>{Scalar<TR>::MINEXPONENT};
},
sx->u);
}
} else if (name == "minloc") {
return FoldLocation<WhichLocation::Minloc, T>(context, std::move(funcRef));
} else if (name == "minval") {
return FoldMaxvalMinval<T>(
context, std::move(funcRef), RelationalOperator::LT, T::Scalar::HUGE());
} else if (name == "mod") {
return FoldElementalIntrinsic<T, T, T>(context, std::move(funcRef),
ScalarFuncWithContext<T, T, T>(
[](FoldingContext &context, const Scalar<T> &x,
const Scalar<T> &y) -> Scalar<T> {
auto quotRem{x.DivideSigned(y)};
if (quotRem.divisionByZero) {
context.messages().Say("mod() by zero"_warn_en_US);
} else if (quotRem.overflow) {
context.messages().Say("mod() folding overflowed"_warn_en_US);
}
return quotRem.remainder;
}));
} else if (name == "modulo") {
return FoldElementalIntrinsic<T, T, T>(context, std::move(funcRef),
ScalarFuncWithContext<T, T, T>(
[](FoldingContext &context, const Scalar<T> &x,
const Scalar<T> &y) -> Scalar<T> {
auto result{x.MODULO(y)};
if (result.overflow) {
context.messages().Say(
"modulo() folding overflowed"_warn_en_US);
}
return result.value;
}));
} else if (name == "not") {
return FoldElementalIntrinsic<T, T>(
context, std::move(funcRef), &Scalar<T>::NOT);
} else if (name == "precision") {
if (const auto *cx{UnwrapExpr<Expr<SomeReal>>(args[0])}) {
return Expr<T>{common::visit(
[](const auto &kx) {
return Scalar<ResultType<decltype(kx)>>::PRECISION;
},
cx->u)};
} else if (const auto *cx{UnwrapExpr<Expr<SomeComplex>>(args[0])}) {
return Expr<T>{common::visit(
[](const auto &kx) {
return Scalar<typename ResultType<decltype(kx)>::Part>::PRECISION;
},
cx->u)};
}
} else if (name == "product") {
return FoldProduct<T>(context, std::move(funcRef), Scalar<T>{1});
} else if (name == "radix") {
return Expr<T>{2};
} else if (name == "range") {
if (const auto *cx{UnwrapExpr<Expr<SomeInteger>>(args[0])}) {
return Expr<T>{common::visit(
[](const auto &kx) {
return Scalar<ResultType<decltype(kx)>>::RANGE;
},
cx->u)};
} else if (const auto *cx{UnwrapExpr<Expr<SomeReal>>(args[0])}) {
return Expr<T>{common::visit(
[](const auto &kx) {
return Scalar<ResultType<decltype(kx)>>::RANGE;
},
cx->u)};
} else if (const auto *cx{UnwrapExpr<Expr<SomeComplex>>(args[0])}) {
return Expr<T>{common::visit(
[](const auto &kx) {
return Scalar<typename ResultType<decltype(kx)>::Part>::RANGE;
},
cx->u)};
}
} else if (name == "rank") {
if (const auto *array{UnwrapExpr<Expr<SomeType>>(args[0])}) {
if (auto named{ExtractNamedEntity(*array)}) {
const Symbol &symbol{named->GetLastSymbol()};
if (IsAssumedRank(symbol)) {
// DescriptorInquiry can only be placed in expression of kind
// DescriptorInquiry::Result::kind.
return ConvertToType<T>(Expr<
Type<TypeCategory::Integer, DescriptorInquiry::Result::kind>>{
DescriptorInquiry{*named, DescriptorInquiry::Field::Rank}});
}
}
return Expr<T>{args[0].value().Rank()};
}
return Expr<T>{args[0].value().Rank()};
} else if (name == "selected_char_kind") {
if (const auto *chCon{UnwrapExpr<Constant<TypeOf<std::string>>>(args[0])}) {
if (std::optional<std::string> value{chCon->GetScalarValue()}) {
int defaultKind{
context.defaults().GetDefaultKind(TypeCategory::Character)};
return Expr<T>{SelectedCharKind(*value, defaultKind)};
}
}
} else if (name == "selected_int_kind") {
if (auto p{GetInt64Arg(args[0])}) {
return Expr<T>{SelectedIntKind(*p)};
}
} else if (name == "selected_real_kind" ||
name == "__builtin_ieee_selected_real_kind") {
if (auto p{GetInt64ArgOr(args[0], 0)}) {
if (auto r{GetInt64ArgOr(args[1], 0)}) {
if (auto radix{GetInt64ArgOr(args[2], 2)}) {
return Expr<T>{SelectedRealKind(*p, *r, *radix)};
}
}
}
} else if (name == "shape") {
[flang] Do not lose call in shape inquiry on function reference Currently, something like `print *, size(foo(n,m))` was rewritten to `print *, size(foo_result_symbol)` when foo result is a non constant shape array. This cannot be processed by lowering or reprocessed by a Fortran compiler since the syntax is wrong (`foo_result_symbol` is unknown on the caller side) and the arguments are lost when they might be required to compute the result shape. It is not possible (and probably not desired) to make GetShape fail in general in such case since returning nullopt seems only expected for scalars or assumed rank (see GetRank usage in lib/Semantics/check-call.cpp), and returning a vector with nullopt extent may trigger some checks to believe they are facing an assumed size (like here in intrinsic argument checks: https://github.com/llvm/llvm-project/blob/196204c72c68a577c72af95d70f18e3550939a5e/flang/lib/Evaluate/intrinsics.cpp#L1530). Hence, I went for a solution that limits the rewrite change to folding (where the original expression is returned if the shape depends on a non constant shape from a call). I added a non default option to GetShapeHelper that prevents the rewrite of shape inquiry on calls to descriptor inquiries. At first I wanted to avoid touching GetShapeHelper, but it would require to re-implement all its logic to determine if the shape comes from a function call or not (the expression could be `size(1+foo(n,m))`). So added an alternate entry point to GetShapeHelper seemed the cleanest solution to me. Differential Revision: https://reviews.llvm.org/D116933
2022-01-11 02:09:45 +08:00
if (auto shape{GetContextFreeShape(context, args[0])}) {
if (auto shapeExpr{AsExtentArrayExpr(*shape)}) {
return Fold(context, ConvertToType<T>(std::move(*shapeExpr)));
}
}
} else if (name == "shifta" || name == "shiftr" || name == "shiftl") {
// Second argument can be of any kind. However, it must be smaller or
// equal than BIT_SIZE. It can be converted to Int4 to simplify.
auto fptr{&Scalar<T>::SHIFTA};
if (name == "shifta") { // done in fptr definition
} else if (name == "shiftr") {
fptr = &Scalar<T>::SHIFTR;
} else if (name == "shiftl") {
fptr = &Scalar<T>::SHIFTL;
} else {
common::die("missing case to fold intrinsic function %s", name.c_str());
}
return FoldElementalIntrinsic<T, T, Int4>(context, std::move(funcRef),
ScalarFunc<T, T, Int4>([&](const Scalar<T> &i,
const Scalar<Int4> &pos) -> Scalar<T> {
auto posVal{static_cast<int>(pos.ToInt64())};
if (posVal < 0) {
context.messages().Say(
"SHIFT=%d count for %s is negative"_err_en_US, posVal, name);
} else if (posVal > i.bits) {
context.messages().Say(
"SHIFT=%d count for %s is greater than %d"_err_en_US, posVal,
name, i.bits);
}
return std::invoke(fptr, i, posVal);
}));
} else if (name == "sign") {
return FoldElementalIntrinsic<T, T, T>(context, std::move(funcRef),
ScalarFunc<T, T, T>(
[&context](const Scalar<T> &j, const Scalar<T> &k) -> Scalar<T> {
typename Scalar<T>::ValueWithOverflow result{j.SIGN(k)};
if (result.overflow) {
context.messages().Say(
"sign(integer(kind=%d)) folding overflowed"_warn_en_US,
KIND);
}
return result.value;
}));
} else if (name == "size") {
[flang] Do not lose call in shape inquiry on function reference Currently, something like `print *, size(foo(n,m))` was rewritten to `print *, size(foo_result_symbol)` when foo result is a non constant shape array. This cannot be processed by lowering or reprocessed by a Fortran compiler since the syntax is wrong (`foo_result_symbol` is unknown on the caller side) and the arguments are lost when they might be required to compute the result shape. It is not possible (and probably not desired) to make GetShape fail in general in such case since returning nullopt seems only expected for scalars or assumed rank (see GetRank usage in lib/Semantics/check-call.cpp), and returning a vector with nullopt extent may trigger some checks to believe they are facing an assumed size (like here in intrinsic argument checks: https://github.com/llvm/llvm-project/blob/196204c72c68a577c72af95d70f18e3550939a5e/flang/lib/Evaluate/intrinsics.cpp#L1530). Hence, I went for a solution that limits the rewrite change to folding (where the original expression is returned if the shape depends on a non constant shape from a call). I added a non default option to GetShapeHelper that prevents the rewrite of shape inquiry on calls to descriptor inquiries. At first I wanted to avoid touching GetShapeHelper, but it would require to re-implement all its logic to determine if the shape comes from a function call or not (the expression could be `size(1+foo(n,m))`). So added an alternate entry point to GetShapeHelper seemed the cleanest solution to me. Differential Revision: https://reviews.llvm.org/D116933
2022-01-11 02:09:45 +08:00
if (auto shape{GetContextFreeShape(context, args[0])}) {
if (auto &dimArg{args[1]}) { // DIM= is present, get one extent
if (auto dim{GetInt64Arg(args[1])}) {
int rank{GetRank(*shape)};
if (*dim >= 1 && *dim <= rank) {
const Symbol *symbol{UnwrapWholeSymbolDataRef(args[0])};
if (symbol && IsAssumedSizeArray(*symbol) && *dim == rank) {
context.messages().Say(
"size(array,dim=%jd) of last dimension is not available for rank-%d assumed-size array dummy argument"_err_en_US,
*dim, rank);
return MakeInvalidIntrinsic<T>(std::move(funcRef));
} else if (auto &extent{shape->at(*dim - 1)}) {
return Fold(context, ConvertToType<T>(std::move(*extent)));
}
} else {
context.messages().Say(
"size(array,dim=%jd) dimension is out of range for rank-%d array"_warn_en_US,
*dim, rank);
}
}
} else if (auto extents{common::AllElementsPresent(std::move(*shape))}) {
// DIM= is absent; compute PRODUCT(SHAPE())
ExtentExpr product{1};
for (auto &&extent : std::move(*extents)) {
product = std::move(product) * std::move(extent);
}
return Expr<T>{ConvertToType<T>(Fold(context, std::move(product)))};
}
}
} else if (name == "sizeof") { // in bytes; extension
if (auto info{
characteristics::TypeAndShape::Characterize(args[0], context)}) {
if (auto bytes{info->MeasureSizeInBytes(context)}) {
return Expr<T>{Fold(context, ConvertToType<T>(std::move(*bytes)))};
}
}
} else if (name == "storage_size") { // in bits
if (auto info{
characteristics::TypeAndShape::Characterize(args[0], context)}) {
if (auto bytes{info->MeasureElementSizeInBytes(context, true)}) {
return Expr<T>{
Fold(context, Expr<T>{8} * ConvertToType<T>(std::move(*bytes)))};
}
}
} else if (name == "sum") {
return FoldSum<T>(context, std::move(funcRef));
} else if (name == "ubound") {
return UBOUND(context, std::move(funcRef));
}
// TODO: dot_product, ishftc, matmul, sign, transfer
return Expr<T>{std::move(funcRef)};
}
[flang] Fix spurious errors from runtime derived type table construction Andrezj W. @ Arm discovered that the runtime derived type table building code in semantics was detecting fatal errors in the tests that the f18 driver wasn't printing. This patch fixes f18 so that these messages are printed; however, the messages were not valid user errors, and the rest of this patch fixes them up. There were two sources of the bogus errors. One was that the runtime derived type information table builder was calculating the shapes of allocatable and pointer array components in derived types, and then complaining that they weren't constant or LEN parameter values, which of course they couldn't be since they have to have deferred shapes and those bounds were expressions like LBOUND(component,dim=1). The second was that f18 was forwarding the actual LEN type parameter expressions of a type instantiation too far into the uses of those parameters in various expressions in the declarations of components; when an actual LEN type parameter is not a constant value, it needs to remain a "bare" type parameter inquiry so that it will be lowered to a descriptor inquiry and acquire a captured expression value. Fixing this up properly involved: moving some code into new utility function templates in Evaluate/tools.h, tweaking the rewriting of conversions in expression folding to elide needless integer kind conversions of type parameter inquiries, making type parameter inquiry folding *not* replace bare LEN type parameters with non-constant actual parameter values, and cleaning up some altered test results. Differential Revision: https://reviews.llvm.org/D101001
2021-04-22 06:12:07 +08:00
// Substitutes a bare type parameter reference with its value if it has one now
// in an instantiation. Bare LEN type parameters are substituted only when
// the known value is constant.
Expr<TypeParamInquiry::Result> FoldOperation(
FoldingContext &context, TypeParamInquiry &&inquiry) {
std::optional<NamedEntity> base{inquiry.base()};
parser::CharBlock parameterName{inquiry.parameter().name()};
if (base) {
// Handling "designator%typeParam". Get the value of the type parameter
// from the instantiation of the base
if (const semantics::DeclTypeSpec *
declType{base->GetLastSymbol().GetType()}) {
if (const semantics::ParamValue *
paramValue{
declType->derivedTypeSpec().FindParameter(parameterName)}) {
const semantics::MaybeIntExpr &paramExpr{paramValue->GetExplicit()};
if (paramExpr && IsConstantExpr(*paramExpr)) {
Expr<SomeInteger> intExpr{*paramExpr};
return Fold(context,
ConvertToType<TypeParamInquiry::Result>(std::move(intExpr)));
}
}
}
} else {
[flang] Fix spurious errors from runtime derived type table construction Andrezj W. @ Arm discovered that the runtime derived type table building code in semantics was detecting fatal errors in the tests that the f18 driver wasn't printing. This patch fixes f18 so that these messages are printed; however, the messages were not valid user errors, and the rest of this patch fixes them up. There were two sources of the bogus errors. One was that the runtime derived type information table builder was calculating the shapes of allocatable and pointer array components in derived types, and then complaining that they weren't constant or LEN parameter values, which of course they couldn't be since they have to have deferred shapes and those bounds were expressions like LBOUND(component,dim=1). The second was that f18 was forwarding the actual LEN type parameter expressions of a type instantiation too far into the uses of those parameters in various expressions in the declarations of components; when an actual LEN type parameter is not a constant value, it needs to remain a "bare" type parameter inquiry so that it will be lowered to a descriptor inquiry and acquire a captured expression value. Fixing this up properly involved: moving some code into new utility function templates in Evaluate/tools.h, tweaking the rewriting of conversions in expression folding to elide needless integer kind conversions of type parameter inquiries, making type parameter inquiry folding *not* replace bare LEN type parameters with non-constant actual parameter values, and cleaning up some altered test results. Differential Revision: https://reviews.llvm.org/D101001
2021-04-22 06:12:07 +08:00
// A "bare" type parameter: replace with its value, if that's now known
// in a current derived type instantiation, for KIND type parameters.
if (const auto *pdt{context.pdtInstance()}) {
[flang] Fix spurious errors from runtime derived type table construction Andrezj W. @ Arm discovered that the runtime derived type table building code in semantics was detecting fatal errors in the tests that the f18 driver wasn't printing. This patch fixes f18 so that these messages are printed; however, the messages were not valid user errors, and the rest of this patch fixes them up. There were two sources of the bogus errors. One was that the runtime derived type information table builder was calculating the shapes of allocatable and pointer array components in derived types, and then complaining that they weren't constant or LEN parameter values, which of course they couldn't be since they have to have deferred shapes and those bounds were expressions like LBOUND(component,dim=1). The second was that f18 was forwarding the actual LEN type parameter expressions of a type instantiation too far into the uses of those parameters in various expressions in the declarations of components; when an actual LEN type parameter is not a constant value, it needs to remain a "bare" type parameter inquiry so that it will be lowered to a descriptor inquiry and acquire a captured expression value. Fixing this up properly involved: moving some code into new utility function templates in Evaluate/tools.h, tweaking the rewriting of conversions in expression folding to elide needless integer kind conversions of type parameter inquiries, making type parameter inquiry folding *not* replace bare LEN type parameters with non-constant actual parameter values, and cleaning up some altered test results. Differential Revision: https://reviews.llvm.org/D101001
2021-04-22 06:12:07 +08:00
bool isLen{false};
if (const semantics::Scope * scope{context.pdtInstance()->scope()}) {
auto iter{scope->find(parameterName)};
if (iter != scope->end()) {
const Symbol &symbol{*iter->second};
const auto *details{symbol.detailsIf<semantics::TypeParamDetails>()};
if (details) {
[flang] Fix spurious errors from runtime derived type table construction Andrezj W. @ Arm discovered that the runtime derived type table building code in semantics was detecting fatal errors in the tests that the f18 driver wasn't printing. This patch fixes f18 so that these messages are printed; however, the messages were not valid user errors, and the rest of this patch fixes them up. There were two sources of the bogus errors. One was that the runtime derived type information table builder was calculating the shapes of allocatable and pointer array components in derived types, and then complaining that they weren't constant or LEN parameter values, which of course they couldn't be since they have to have deferred shapes and those bounds were expressions like LBOUND(component,dim=1). The second was that f18 was forwarding the actual LEN type parameter expressions of a type instantiation too far into the uses of those parameters in various expressions in the declarations of components; when an actual LEN type parameter is not a constant value, it needs to remain a "bare" type parameter inquiry so that it will be lowered to a descriptor inquiry and acquire a captured expression value. Fixing this up properly involved: moving some code into new utility function templates in Evaluate/tools.h, tweaking the rewriting of conversions in expression folding to elide needless integer kind conversions of type parameter inquiries, making type parameter inquiry folding *not* replace bare LEN type parameters with non-constant actual parameter values, and cleaning up some altered test results. Differential Revision: https://reviews.llvm.org/D101001
2021-04-22 06:12:07 +08:00
isLen = details->attr() == common::TypeParamAttr::Len;
const semantics::MaybeIntExpr &initExpr{details->init()};
[flang] Fix spurious errors from runtime derived type table construction Andrezj W. @ Arm discovered that the runtime derived type table building code in semantics was detecting fatal errors in the tests that the f18 driver wasn't printing. This patch fixes f18 so that these messages are printed; however, the messages were not valid user errors, and the rest of this patch fixes them up. There were two sources of the bogus errors. One was that the runtime derived type information table builder was calculating the shapes of allocatable and pointer array components in derived types, and then complaining that they weren't constant or LEN parameter values, which of course they couldn't be since they have to have deferred shapes and those bounds were expressions like LBOUND(component,dim=1). The second was that f18 was forwarding the actual LEN type parameter expressions of a type instantiation too far into the uses of those parameters in various expressions in the declarations of components; when an actual LEN type parameter is not a constant value, it needs to remain a "bare" type parameter inquiry so that it will be lowered to a descriptor inquiry and acquire a captured expression value. Fixing this up properly involved: moving some code into new utility function templates in Evaluate/tools.h, tweaking the rewriting of conversions in expression folding to elide needless integer kind conversions of type parameter inquiries, making type parameter inquiry folding *not* replace bare LEN type parameters with non-constant actual parameter values, and cleaning up some altered test results. Differential Revision: https://reviews.llvm.org/D101001
2021-04-22 06:12:07 +08:00
if (initExpr && IsConstantExpr(*initExpr) &&
(!isLen || ToInt64(*initExpr))) {
Expr<SomeInteger> expr{*initExpr};
return Fold(context,
ConvertToType<TypeParamInquiry::Result>(std::move(expr)));
}
}
}
}
if (const auto *value{pdt->FindParameter(parameterName)}) {
if (value->isExplicit()) {
[flang] Fix spurious errors from runtime derived type table construction Andrezj W. @ Arm discovered that the runtime derived type table building code in semantics was detecting fatal errors in the tests that the f18 driver wasn't printing. This patch fixes f18 so that these messages are printed; however, the messages were not valid user errors, and the rest of this patch fixes them up. There were two sources of the bogus errors. One was that the runtime derived type information table builder was calculating the shapes of allocatable and pointer array components in derived types, and then complaining that they weren't constant or LEN parameter values, which of course they couldn't be since they have to have deferred shapes and those bounds were expressions like LBOUND(component,dim=1). The second was that f18 was forwarding the actual LEN type parameter expressions of a type instantiation too far into the uses of those parameters in various expressions in the declarations of components; when an actual LEN type parameter is not a constant value, it needs to remain a "bare" type parameter inquiry so that it will be lowered to a descriptor inquiry and acquire a captured expression value. Fixing this up properly involved: moving some code into new utility function templates in Evaluate/tools.h, tweaking the rewriting of conversions in expression folding to elide needless integer kind conversions of type parameter inquiries, making type parameter inquiry folding *not* replace bare LEN type parameters with non-constant actual parameter values, and cleaning up some altered test results. Differential Revision: https://reviews.llvm.org/D101001
2021-04-22 06:12:07 +08:00
auto folded{Fold(context,
AsExpr(ConvertToType<TypeParamInquiry::Result>(
[flang] Fix spurious errors from runtime derived type table construction Andrezj W. @ Arm discovered that the runtime derived type table building code in semantics was detecting fatal errors in the tests that the f18 driver wasn't printing. This patch fixes f18 so that these messages are printed; however, the messages were not valid user errors, and the rest of this patch fixes them up. There were two sources of the bogus errors. One was that the runtime derived type information table builder was calculating the shapes of allocatable and pointer array components in derived types, and then complaining that they weren't constant or LEN parameter values, which of course they couldn't be since they have to have deferred shapes and those bounds were expressions like LBOUND(component,dim=1). The second was that f18 was forwarding the actual LEN type parameter expressions of a type instantiation too far into the uses of those parameters in various expressions in the declarations of components; when an actual LEN type parameter is not a constant value, it needs to remain a "bare" type parameter inquiry so that it will be lowered to a descriptor inquiry and acquire a captured expression value. Fixing this up properly involved: moving some code into new utility function templates in Evaluate/tools.h, tweaking the rewriting of conversions in expression folding to elide needless integer kind conversions of type parameter inquiries, making type parameter inquiry folding *not* replace bare LEN type parameters with non-constant actual parameter values, and cleaning up some altered test results. Differential Revision: https://reviews.llvm.org/D101001
2021-04-22 06:12:07 +08:00
Expr<SomeInteger>{value->GetExplicit().value()})))};
if (!isLen || ToInt64(folded)) {
return folded;
}
}
}
}
}
return AsExpr(std::move(inquiry));
}
std::optional<std::int64_t> ToInt64(const Expr<SomeInteger> &expr) {
return common::visit(
[](const auto &kindExpr) { return ToInt64(kindExpr); }, expr.u);
}
std::optional<std::int64_t> ToInt64(const Expr<SomeType> &expr) {
if (const auto *intExpr{UnwrapExpr<Expr<SomeInteger>>(expr)}) {
return ToInt64(*intExpr);
} else {
return std::nullopt;
}
}
#ifdef _MSC_VER // disable bogus warning about missing definitions
#pragma warning(disable : 4661)
#endif
FOR_EACH_INTEGER_KIND(template class ExpressionBase, )
template class ExpressionBase<SomeInteger>;
} // namespace Fortran::evaluate