llvm-project/llvm/lib/Frontend/OpenMP/OMPContext.cpp

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

544 lines
20 KiB
C++
Raw Normal View History

//===- OMPContext.cpp ------ Collection of helpers for OpenMP contexts ----===//
//
// 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
//
//===----------------------------------------------------------------------===//
/// \file
///
/// This file implements helper functions and classes to deal with OpenMP
/// contexts as used by `[begin/end] declare variant` and `metadirective`.
///
//===----------------------------------------------------------------------===//
#include "llvm/Frontend/OpenMP/OMPContext.h"
#include "llvm/ADT/SetOperations.h"
#include "llvm/ADT/StringSwitch.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/raw_ostream.h"
#define DEBUG_TYPE "openmp-ir-builder"
using namespace llvm;
using namespace omp;
OMPContext::OMPContext(bool IsDeviceCompilation, Triple TargetTriple) {
// Add the appropriate device kind trait based on the triple and the
// IsDeviceCompilation flag.
ActiveTraits.set(unsigned(IsDeviceCompilation
? TraitProperty::device_kind_nohost
: TraitProperty::device_kind_host));
switch (TargetTriple.getArch()) {
case Triple::arm:
case Triple::armeb:
case Triple::aarch64:
case Triple::aarch64_be:
case Triple::aarch64_32:
case Triple::mips:
case Triple::mipsel:
case Triple::mips64:
case Triple::mips64el:
case Triple::ppc:
case Triple::ppcle:
case Triple::ppc64:
case Triple::ppc64le:
case Triple::x86:
case Triple::x86_64:
ActiveTraits.set(unsigned(TraitProperty::device_kind_cpu));
break;
case Triple::amdgcn:
case Triple::nvptx:
case Triple::nvptx64:
ActiveTraits.set(unsigned(TraitProperty::device_kind_gpu));
break;
default:
break;
}
// Add the appropriate device architecture trait based on the triple.
#define OMP_TRAIT_PROPERTY(Enum, TraitSetEnum, TraitSelectorEnum, Str) \
if (TraitSelector::TraitSelectorEnum == TraitSelector::device_arch) \
if (TargetTriple.getArch() == TargetTriple.getArchTypeForLLVMName(Str)) \
ActiveTraits.set(unsigned(TraitProperty::Enum));
#include "llvm/Frontend/OpenMP/OMPKinds.def"
// TODO: What exactly do we want to see as device ISA trait?
// The discussion on the list did not seem to have come to an agreed
// upon solution.
// LLVM is the "OpenMP vendor" but we could also interpret vendor as the
// target vendor.
ActiveTraits.set(unsigned(TraitProperty::implementation_vendor_llvm));
// The user condition true is accepted but not false.
ActiveTraits.set(unsigned(TraitProperty::user_condition_true));
// This is for sure some device.
ActiveTraits.set(unsigned(TraitProperty::device_kind_any));
LLVM_DEBUG({
dbgs() << "[" << DEBUG_TYPE
<< "] New OpenMP context with the following properties:\n";
for (unsigned Bit : ActiveTraits.set_bits()) {
TraitProperty Property = TraitProperty(Bit);
dbgs() << "\t " << getOpenMPContextTraitPropertyFullName(Property)
<< "\n";
}
});
}
/// Return true if \p C0 is a subset of \p C1. Note that both arrays are
/// expected to be sorted.
template <typename T> static bool isSubset(ArrayRef<T> C0, ArrayRef<T> C1) {
#ifdef EXPENSIVE_CHECKS
assert(llvm::is_sorted(C0) && llvm::is_sorted(C1) &&
"Expected sorted arrays!");
#endif
if (C0.size() > C1.size())
return false;
auto It0 = C0.begin(), End0 = C0.end();
auto It1 = C1.begin(), End1 = C1.end();
while (It0 != End0) {
if (It1 == End1)
return false;
if (*It0 == *It1) {
++It0;
++It1;
continue;
}
++It0;
}
return true;
}
/// Return true if \p C0 is a strict subset of \p C1. Note that both arrays are
/// expected to be sorted.
template <typename T>
static bool isStrictSubset(ArrayRef<T> C0, ArrayRef<T> C1) {
if (C0.size() >= C1.size())
return false;
return isSubset<T>(C0, C1);
}
static bool isStrictSubset(const VariantMatchInfo &VMI0,
const VariantMatchInfo &VMI1) {
// If all required traits are a strict subset and the ordered vectors storing
// the construct traits, we say it is a strict subset. Note that the latter
// relation is not required to be strict.
if (VMI0.RequiredTraits.count() >= VMI1.RequiredTraits.count())
return false;
for (unsigned Bit : VMI0.RequiredTraits.set_bits())
if (!VMI1.RequiredTraits.test(Bit))
return false;
if (!isSubset<TraitProperty>(VMI0.ConstructTraits, VMI1.ConstructTraits))
return false;
return true;
}
static int isVariantApplicableInContextHelper(
const VariantMatchInfo &VMI, const OMPContext &Ctx,
SmallVectorImpl<unsigned> *ConstructMatches, bool DeviceSetOnly) {
// The match kind determines if we need to match all traits, any of the
// traits, or none of the traits for it to be an applicable context.
enum MatchKind { MK_ALL, MK_ANY, MK_NONE };
MatchKind MK = MK_ALL;
// Determine the match kind the user wants, "all" is the default and provided
// to the user only for completeness.
if (VMI.RequiredTraits.test(
unsigned(TraitProperty::implementation_extension_match_any)))
MK = MK_ANY;
if (VMI.RequiredTraits.test(
unsigned(TraitProperty::implementation_extension_match_none)))
MK = MK_NONE;
// Helper to deal with a single property that was (not) found in the OpenMP
// context based on the match kind selected by the user via
// `implementation={extensions(match_[all,any,none])}'
auto HandleTrait = [MK](TraitProperty Property,
bool WasFound) -> Optional<bool> /* Result */ {
// For kind "any" a single match is enough but we ignore non-matched
// properties.
if (MK == MK_ANY) {
if (WasFound)
return true;
return None;
}
// In "all" or "none" mode we accept a matching or non-matching property
// respectively and move on. We are not done yet!
if ((WasFound && MK == MK_ALL) || (!WasFound && MK == MK_NONE))
return None;
// We missed a property, provide some debug output and indicate failure.
LLVM_DEBUG({
if (MK == MK_ALL)
dbgs() << "[" << DEBUG_TYPE << "] Property "
<< getOpenMPContextTraitPropertyName(Property, "")
<< " was not in the OpenMP context but match kind is all.\n";
if (MK == MK_NONE)
dbgs() << "[" << DEBUG_TYPE << "] Property "
<< getOpenMPContextTraitPropertyName(Property, "")
<< " was in the OpenMP context but match kind is none.\n";
});
return false;
};
for (unsigned Bit : VMI.RequiredTraits.set_bits()) {
TraitProperty Property = TraitProperty(Bit);
if (DeviceSetOnly &&
getOpenMPContextTraitSetForProperty(Property) != TraitSet::device)
continue;
// So far all extensions are handled elsewhere, we skip them here as they
// are not part of the OpenMP context.
if (getOpenMPContextTraitSelectorForProperty(Property) ==
TraitSelector::implementation_extension)
continue;
bool IsActiveTrait = Ctx.ActiveTraits.test(unsigned(Property));
// We overwrite the isa trait as it is actually up to the OMPContext hook to
// check the raw string(s).
if (Property == TraitProperty::device_isa___ANY)
IsActiveTrait = llvm::all_of(VMI.ISATraits, [&](StringRef RawString) {
return Ctx.matchesISATrait(RawString);
});
Optional<bool> Result = HandleTrait(Property, IsActiveTrait);
if (Result.hasValue())
return Result.getValue();
}
if (!DeviceSetOnly) {
// We could use isSubset here but we also want to record the match
// locations.
unsigned ConstructIdx = 0, NoConstructTraits = Ctx.ConstructTraits.size();
for (TraitProperty Property : VMI.ConstructTraits) {
assert(getOpenMPContextTraitSetForProperty(Property) ==
TraitSet::construct &&
"Variant context is ill-formed!");
// Verify the nesting.
bool FoundInOrder = false;
while (!FoundInOrder && ConstructIdx != NoConstructTraits)
FoundInOrder = (Ctx.ConstructTraits[ConstructIdx++] == Property);
if (ConstructMatches)
ConstructMatches->push_back(ConstructIdx - 1);
Optional<bool> Result = HandleTrait(Property, FoundInOrder);
if (Result.hasValue())
return Result.getValue();
if (!FoundInOrder) {
LLVM_DEBUG(dbgs() << "[" << DEBUG_TYPE << "] Construct property "
<< getOpenMPContextTraitPropertyName(Property, "")
<< " was not nested properly.\n");
return false;
}
// TODO: Verify SIMD
}
assert(isSubset<TraitProperty>(VMI.ConstructTraits, Ctx.ConstructTraits) &&
"Broken invariant!");
}
if (MK == MK_ANY) {
LLVM_DEBUG(dbgs() << "[" << DEBUG_TYPE
<< "] None of the properties was in the OpenMP context "
"but match kind is any.\n");
return false;
}
return true;
}
bool llvm::omp::isVariantApplicableInContext(const VariantMatchInfo &VMI,
const OMPContext &Ctx,
bool DeviceSetOnly) {
return isVariantApplicableInContextHelper(
VMI, Ctx, /* ConstructMatches */ nullptr, DeviceSetOnly);
}
static APInt getVariantMatchScore(const VariantMatchInfo &VMI,
const OMPContext &Ctx,
SmallVectorImpl<unsigned> &ConstructMatches) {
APInt Score(64, 1);
unsigned NoConstructTraits = VMI.ConstructTraits.size();
for (unsigned Bit : VMI.RequiredTraits.set_bits()) {
TraitProperty Property = TraitProperty(Bit);
// If there is a user score attached, use it.
if (VMI.ScoreMap.count(Property)) {
const APInt &UserScore = VMI.ScoreMap.lookup(Property);
assert(UserScore.uge(0) && "Expect non-negative user scores!");
Score += UserScore.getZExtValue();
continue;
}
switch (getOpenMPContextTraitSetForProperty(Property)) {
case TraitSet::construct:
// We handle the construct traits later via the VMI.ConstructTraits
// container.
continue;
case TraitSet::implementation:
// No effect on the score (implementation defined).
continue;
case TraitSet::user:
// No effect on the score.
continue;
case TraitSet::device:
// Handled separately below.
break;
case TraitSet::invalid:
llvm_unreachable("Unknown trait set is not to be used!");
}
// device={kind(any)} is "as if" no kind selector was specified.
if (Property == TraitProperty::device_kind_any)
continue;
switch (getOpenMPContextTraitSelectorForProperty(Property)) {
case TraitSelector::device_kind:
Score += (1ULL << (NoConstructTraits + 0));
continue;
case TraitSelector::device_arch:
Score += (1ULL << (NoConstructTraits + 1));
continue;
case TraitSelector::device_isa:
Score += (1ULL << (NoConstructTraits + 2));
continue;
default:
continue;
}
}
unsigned ConstructIdx = 0;
assert(NoConstructTraits == ConstructMatches.size() &&
"Mismatch in the construct traits!");
for (TraitProperty Property : VMI.ConstructTraits) {
assert(getOpenMPContextTraitSetForProperty(Property) ==
TraitSet::construct &&
"Ill-formed variant match info!");
(void)Property;
// ConstructMatches is the position p - 1 and we need 2^(p-1).
Score += (1ULL << ConstructMatches[ConstructIdx++]);
}
LLVM_DEBUG(dbgs() << "[" << DEBUG_TYPE << "] Variant has a score of " << Score
<< "\n");
return Score;
}
int llvm::omp::getBestVariantMatchForContext(
const SmallVectorImpl<VariantMatchInfo> &VMIs, const OMPContext &Ctx) {
APInt BestScore(64, 0);
int BestVMIIdx = -1;
const VariantMatchInfo *BestVMI = nullptr;
for (unsigned u = 0, e = VMIs.size(); u < e; ++u) {
const VariantMatchInfo &VMI = VMIs[u];
SmallVector<unsigned, 8> ConstructMatches;
// If the variant is not applicable its not the best.
if (!isVariantApplicableInContextHelper(VMI, Ctx, &ConstructMatches,
/* DeviceSetOnly */ false))
continue;
// Check if its clearly not the best.
APInt Score = getVariantMatchScore(VMI, Ctx, ConstructMatches);
if (Score.ult(BestScore))
continue;
// Equal score need subset checks.
if (Score.eq(BestScore)) {
// Strict subset are never best.
if (isStrictSubset(VMI, *BestVMI))
continue;
// Same score and the current best is no strict subset so we keep it.
if (!isStrictSubset(*BestVMI, VMI))
continue;
}
// New best found.
BestVMI = &VMI;
BestVMIIdx = u;
BestScore = Score;
}
return BestVMIIdx;
}
TraitSet llvm::omp::getOpenMPContextTraitSetKind(StringRef S) {
return StringSwitch<TraitSet>(S)
#define OMP_TRAIT_SET(Enum, Str) .Case(Str, TraitSet::Enum)
#include "llvm/Frontend/OpenMP/OMPKinds.def"
.Default(TraitSet::invalid);
}
[OpenMP][Part 2] Use reusable OpenMP context/traits handling This patch implements an almost complete handling of OpenMP contexts/traits such that we can reuse most of the logic in Flang through the OMPContext.{h,cpp} in llvm/Frontend/OpenMP. All but construct SIMD specifiers, e.g., inbranch, and the device ISA selector are define in `llvm/lib/Frontend/OpenMP/OMPKinds.def`. From these definitions we generate the enum classes `TraitSet`, `TraitSelector`, and `TraitProperty` as well as conversion and helper functions in `llvm/lib/Frontend/OpenMP/OMPContext.{h,cpp}`. The above enum classes are used in the parser, sema, and the AST attribute. The latter is not a collection of multiple primitive variant arguments that contain encodings via numbers and strings but instead a tree that mirrors the `match` clause (see `struct OpenMPTraitInfo`). The changes to the parser make it more forgiving when wrong syntax is read and they also resulted in more specialized diagnostics. The tests are updated and the core issues are detected as before. Here and elsewhere this patch tries to be generic, thus we do not distinguish what selector set, selector, or property is parsed except if they do behave exceptionally, as for example `user={condition(EXPR)}` does. The sema logic changed in two ways: First, the OMPDeclareVariantAttr representation changed, as mentioned above, and the sema was adjusted to work with the new `OpenMPTraitInfo`. Second, the matching and scoring logic moved into `OMPContext.{h,cpp}`. It is implemented on a flat representation of the `match` clause that is not tied to clang. `OpenMPTraitInfo` provides a method to generate this flat structure (see `struct VariantMatchInfo`) by computing integer score values and boolean user conditions from the `clang::Expr` we keep for them. The OpenMP context is now an explicit object (see `struct OMPContext`). This is in anticipation of construct traits that need to be tracked. The OpenMP context, as well as the `VariantMatchInfo`, are basically made up of a set of active or respectively required traits, e.g., 'host', and an ordered container of constructs which allows duplication. Matching and scoring is kept as generic as possible to allow easy extension in the future. --- Test changes: The messages checked in `OpenMP/declare_variant_messages.{c,cpp}` have been auto generated to match the new warnings and notes of the parser. The "subset" checks were reversed causing the wrong version to be picked. The tests have been adjusted to correct this. We do not print scores if the user did not provide one. We print spaces to make lists in the `match` clause more legible. Reviewers: kiranchandramohan, ABataev, RaviNarayanaswamy, gtbercea, grokos, sdmitriev, JonChesterfield, hfinkel, fghanim Subscribers: merge_guards_bot, rampitec, mgorny, hiraditya, aheejin, fedor.sergeev, simoncook, bollu, guansong, dexonsmith, jfb, s.egerton, llvm-commits, cfe-commits Tags: #clang, #llvm Differential Revision: https://reviews.llvm.org/D71830
2019-12-20 10:42:12 +08:00
TraitSet
llvm::omp::getOpenMPContextTraitSetForSelector(TraitSelector Selector) {
switch (Selector) {
#define OMP_TRAIT_SELECTOR(Enum, TraitSetEnum, Str, ReqProp) \
case TraitSelector::Enum: \
return TraitSet::TraitSetEnum;
#include "llvm/Frontend/OpenMP/OMPKinds.def"
}
llvm_unreachable("Unknown trait selector!");
[OpenMP][Part 2] Use reusable OpenMP context/traits handling This patch implements an almost complete handling of OpenMP contexts/traits such that we can reuse most of the logic in Flang through the OMPContext.{h,cpp} in llvm/Frontend/OpenMP. All but construct SIMD specifiers, e.g., inbranch, and the device ISA selector are define in `llvm/lib/Frontend/OpenMP/OMPKinds.def`. From these definitions we generate the enum classes `TraitSet`, `TraitSelector`, and `TraitProperty` as well as conversion and helper functions in `llvm/lib/Frontend/OpenMP/OMPContext.{h,cpp}`. The above enum classes are used in the parser, sema, and the AST attribute. The latter is not a collection of multiple primitive variant arguments that contain encodings via numbers and strings but instead a tree that mirrors the `match` clause (see `struct OpenMPTraitInfo`). The changes to the parser make it more forgiving when wrong syntax is read and they also resulted in more specialized diagnostics. The tests are updated and the core issues are detected as before. Here and elsewhere this patch tries to be generic, thus we do not distinguish what selector set, selector, or property is parsed except if they do behave exceptionally, as for example `user={condition(EXPR)}` does. The sema logic changed in two ways: First, the OMPDeclareVariantAttr representation changed, as mentioned above, and the sema was adjusted to work with the new `OpenMPTraitInfo`. Second, the matching and scoring logic moved into `OMPContext.{h,cpp}`. It is implemented on a flat representation of the `match` clause that is not tied to clang. `OpenMPTraitInfo` provides a method to generate this flat structure (see `struct VariantMatchInfo`) by computing integer score values and boolean user conditions from the `clang::Expr` we keep for them. The OpenMP context is now an explicit object (see `struct OMPContext`). This is in anticipation of construct traits that need to be tracked. The OpenMP context, as well as the `VariantMatchInfo`, are basically made up of a set of active or respectively required traits, e.g., 'host', and an ordered container of constructs which allows duplication. Matching and scoring is kept as generic as possible to allow easy extension in the future. --- Test changes: The messages checked in `OpenMP/declare_variant_messages.{c,cpp}` have been auto generated to match the new warnings and notes of the parser. The "subset" checks were reversed causing the wrong version to be picked. The tests have been adjusted to correct this. We do not print scores if the user did not provide one. We print spaces to make lists in the `match` clause more legible. Reviewers: kiranchandramohan, ABataev, RaviNarayanaswamy, gtbercea, grokos, sdmitriev, JonChesterfield, hfinkel, fghanim Subscribers: merge_guards_bot, rampitec, mgorny, hiraditya, aheejin, fedor.sergeev, simoncook, bollu, guansong, dexonsmith, jfb, s.egerton, llvm-commits, cfe-commits Tags: #clang, #llvm Differential Revision: https://reviews.llvm.org/D71830
2019-12-20 10:42:12 +08:00
}
TraitSet
llvm::omp::getOpenMPContextTraitSetForProperty(TraitProperty Property) {
switch (Property) {
#define OMP_TRAIT_PROPERTY(Enum, TraitSetEnum, TraitSelectorEnum, Str) \
case TraitProperty::Enum: \
return TraitSet::TraitSetEnum;
#include "llvm/Frontend/OpenMP/OMPKinds.def"
}
llvm_unreachable("Unknown trait set!");
}
StringRef llvm::omp::getOpenMPContextTraitSetName(TraitSet Kind) {
switch (Kind) {
#define OMP_TRAIT_SET(Enum, Str) \
case TraitSet::Enum: \
return Str;
#include "llvm/Frontend/OpenMP/OMPKinds.def"
}
llvm_unreachable("Unknown trait set!");
}
TraitSelector llvm::omp::getOpenMPContextTraitSelectorKind(StringRef S) {
return StringSwitch<TraitSelector>(S)
#define OMP_TRAIT_SELECTOR(Enum, TraitSetEnum, Str, ReqProp) \
.Case(Str, TraitSelector::Enum)
#include "llvm/Frontend/OpenMP/OMPKinds.def"
.Default(TraitSelector::invalid);
}
TraitSelector
llvm::omp::getOpenMPContextTraitSelectorForProperty(TraitProperty Property) {
switch (Property) {
#define OMP_TRAIT_PROPERTY(Enum, TraitSetEnum, TraitSelectorEnum, Str) \
case TraitProperty::Enum: \
return TraitSelector::TraitSelectorEnum;
#include "llvm/Frontend/OpenMP/OMPKinds.def"
}
llvm_unreachable("Unknown trait set!");
}
StringRef llvm::omp::getOpenMPContextTraitSelectorName(TraitSelector Kind) {
switch (Kind) {
#define OMP_TRAIT_SELECTOR(Enum, TraitSetEnum, Str, ReqProp) \
case TraitSelector::Enum: \
return Str;
#include "llvm/Frontend/OpenMP/OMPKinds.def"
}
llvm_unreachable("Unknown trait selector!");
}
TraitProperty llvm::omp::getOpenMPContextTraitPropertyKind(
TraitSet Set, TraitSelector Selector, StringRef S) {
// Special handling for `device={isa(...)}` as we accept anything here. It is
// up to the target to decide if the feature is available.
if (Set == TraitSet::device && Selector == TraitSelector::device_isa)
return TraitProperty::device_isa___ANY;
#define OMP_TRAIT_PROPERTY(Enum, TraitSetEnum, TraitSelectorEnum, Str) \
[OpenMP][Part 2] Use reusable OpenMP context/traits handling This patch implements an almost complete handling of OpenMP contexts/traits such that we can reuse most of the logic in Flang through the OMPContext.{h,cpp} in llvm/Frontend/OpenMP. All but construct SIMD specifiers, e.g., inbranch, and the device ISA selector are define in `llvm/lib/Frontend/OpenMP/OMPKinds.def`. From these definitions we generate the enum classes `TraitSet`, `TraitSelector`, and `TraitProperty` as well as conversion and helper functions in `llvm/lib/Frontend/OpenMP/OMPContext.{h,cpp}`. The above enum classes are used in the parser, sema, and the AST attribute. The latter is not a collection of multiple primitive variant arguments that contain encodings via numbers and strings but instead a tree that mirrors the `match` clause (see `struct OpenMPTraitInfo`). The changes to the parser make it more forgiving when wrong syntax is read and they also resulted in more specialized diagnostics. The tests are updated and the core issues are detected as before. Here and elsewhere this patch tries to be generic, thus we do not distinguish what selector set, selector, or property is parsed except if they do behave exceptionally, as for example `user={condition(EXPR)}` does. The sema logic changed in two ways: First, the OMPDeclareVariantAttr representation changed, as mentioned above, and the sema was adjusted to work with the new `OpenMPTraitInfo`. Second, the matching and scoring logic moved into `OMPContext.{h,cpp}`. It is implemented on a flat representation of the `match` clause that is not tied to clang. `OpenMPTraitInfo` provides a method to generate this flat structure (see `struct VariantMatchInfo`) by computing integer score values and boolean user conditions from the `clang::Expr` we keep for them. The OpenMP context is now an explicit object (see `struct OMPContext`). This is in anticipation of construct traits that need to be tracked. The OpenMP context, as well as the `VariantMatchInfo`, are basically made up of a set of active or respectively required traits, e.g., 'host', and an ordered container of constructs which allows duplication. Matching and scoring is kept as generic as possible to allow easy extension in the future. --- Test changes: The messages checked in `OpenMP/declare_variant_messages.{c,cpp}` have been auto generated to match the new warnings and notes of the parser. The "subset" checks were reversed causing the wrong version to be picked. The tests have been adjusted to correct this. We do not print scores if the user did not provide one. We print spaces to make lists in the `match` clause more legible. Reviewers: kiranchandramohan, ABataev, RaviNarayanaswamy, gtbercea, grokos, sdmitriev, JonChesterfield, hfinkel, fghanim Subscribers: merge_guards_bot, rampitec, mgorny, hiraditya, aheejin, fedor.sergeev, simoncook, bollu, guansong, dexonsmith, jfb, s.egerton, llvm-commits, cfe-commits Tags: #clang, #llvm Differential Revision: https://reviews.llvm.org/D71830
2019-12-20 10:42:12 +08:00
if (Set == TraitSet::TraitSetEnum && Str == S) \
return TraitProperty::Enum;
#include "llvm/Frontend/OpenMP/OMPKinds.def"
return TraitProperty::invalid;
}
TraitProperty
llvm::omp::getOpenMPContextTraitPropertyForSelector(TraitSelector Selector) {
return StringSwitch<TraitProperty>(
getOpenMPContextTraitSelectorName(Selector))
#define OMP_TRAIT_PROPERTY(Enum, TraitSetEnum, TraitSelectorEnum, Str) \
.Case(Str, Selector == TraitSelector::TraitSelectorEnum \
? TraitProperty::Enum \
: TraitProperty::invalid)
#include "llvm/Frontend/OpenMP/OMPKinds.def"
.Default(TraitProperty::invalid);
}
StringRef llvm::omp::getOpenMPContextTraitPropertyName(TraitProperty Kind,
StringRef RawString) {
if (Kind == TraitProperty::device_isa___ANY)
return RawString;
switch (Kind) {
#define OMP_TRAIT_PROPERTY(Enum, TraitSetEnum, TraitSelectorEnum, Str) \
case TraitProperty::Enum: \
return Str;
#include "llvm/Frontend/OpenMP/OMPKinds.def"
}
llvm_unreachable("Unknown trait property!");
}
StringRef llvm::omp::getOpenMPContextTraitPropertyFullName(TraitProperty Kind) {
switch (Kind) {
#define OMP_TRAIT_PROPERTY(Enum, TraitSetEnum, TraitSelectorEnum, Str) \
case TraitProperty::Enum: \
return "(" #TraitSetEnum "," #TraitSelectorEnum "," Str ")";
#include "llvm/Frontend/OpenMP/OMPKinds.def"
}
llvm_unreachable("Unknown trait property!");
}
bool llvm::omp::isValidTraitSelectorForTraitSet(TraitSelector Selector,
TraitSet Set,
bool &AllowsTraitScore,
bool &RequiresProperty) {
AllowsTraitScore = Set != TraitSet::construct && Set != TraitSet::device;
switch (Selector) {
#define OMP_TRAIT_SELECTOR(Enum, TraitSetEnum, Str, ReqProp) \
case TraitSelector::Enum: \
RequiresProperty = ReqProp; \
return Set == TraitSet::TraitSetEnum;
#include "llvm/Frontend/OpenMP/OMPKinds.def"
}
llvm_unreachable("Unknown trait selector!");
}
bool llvm::omp::isValidTraitPropertyForTraitSetAndSelector(
TraitProperty Property, TraitSelector Selector, TraitSet Set) {
switch (Property) {
#define OMP_TRAIT_PROPERTY(Enum, TraitSetEnum, TraitSelectorEnum, Str) \
case TraitProperty::Enum: \
return Set == TraitSet::TraitSetEnum && \
Selector == TraitSelector::TraitSelectorEnum;
#include "llvm/Frontend/OpenMP/OMPKinds.def"
}
llvm_unreachable("Unknown trait property!");
}
[OpenMP][Part 2] Use reusable OpenMP context/traits handling This patch implements an almost complete handling of OpenMP contexts/traits such that we can reuse most of the logic in Flang through the OMPContext.{h,cpp} in llvm/Frontend/OpenMP. All but construct SIMD specifiers, e.g., inbranch, and the device ISA selector are define in `llvm/lib/Frontend/OpenMP/OMPKinds.def`. From these definitions we generate the enum classes `TraitSet`, `TraitSelector`, and `TraitProperty` as well as conversion and helper functions in `llvm/lib/Frontend/OpenMP/OMPContext.{h,cpp}`. The above enum classes are used in the parser, sema, and the AST attribute. The latter is not a collection of multiple primitive variant arguments that contain encodings via numbers and strings but instead a tree that mirrors the `match` clause (see `struct OpenMPTraitInfo`). The changes to the parser make it more forgiving when wrong syntax is read and they also resulted in more specialized diagnostics. The tests are updated and the core issues are detected as before. Here and elsewhere this patch tries to be generic, thus we do not distinguish what selector set, selector, or property is parsed except if they do behave exceptionally, as for example `user={condition(EXPR)}` does. The sema logic changed in two ways: First, the OMPDeclareVariantAttr representation changed, as mentioned above, and the sema was adjusted to work with the new `OpenMPTraitInfo`. Second, the matching and scoring logic moved into `OMPContext.{h,cpp}`. It is implemented on a flat representation of the `match` clause that is not tied to clang. `OpenMPTraitInfo` provides a method to generate this flat structure (see `struct VariantMatchInfo`) by computing integer score values and boolean user conditions from the `clang::Expr` we keep for them. The OpenMP context is now an explicit object (see `struct OMPContext`). This is in anticipation of construct traits that need to be tracked. The OpenMP context, as well as the `VariantMatchInfo`, are basically made up of a set of active or respectively required traits, e.g., 'host', and an ordered container of constructs which allows duplication. Matching and scoring is kept as generic as possible to allow easy extension in the future. --- Test changes: The messages checked in `OpenMP/declare_variant_messages.{c,cpp}` have been auto generated to match the new warnings and notes of the parser. The "subset" checks were reversed causing the wrong version to be picked. The tests have been adjusted to correct this. We do not print scores if the user did not provide one. We print spaces to make lists in the `match` clause more legible. Reviewers: kiranchandramohan, ABataev, RaviNarayanaswamy, gtbercea, grokos, sdmitriev, JonChesterfield, hfinkel, fghanim Subscribers: merge_guards_bot, rampitec, mgorny, hiraditya, aheejin, fedor.sergeev, simoncook, bollu, guansong, dexonsmith, jfb, s.egerton, llvm-commits, cfe-commits Tags: #clang, #llvm Differential Revision: https://reviews.llvm.org/D71830
2019-12-20 10:42:12 +08:00
std::string llvm::omp::listOpenMPContextTraitSets() {
std::string S;
#define OMP_TRAIT_SET(Enum, Str) \
if (StringRef(Str) != "invalid") \
[OpenMP][Part 2] Use reusable OpenMP context/traits handling This patch implements an almost complete handling of OpenMP contexts/traits such that we can reuse most of the logic in Flang through the OMPContext.{h,cpp} in llvm/Frontend/OpenMP. All but construct SIMD specifiers, e.g., inbranch, and the device ISA selector are define in `llvm/lib/Frontend/OpenMP/OMPKinds.def`. From these definitions we generate the enum classes `TraitSet`, `TraitSelector`, and `TraitProperty` as well as conversion and helper functions in `llvm/lib/Frontend/OpenMP/OMPContext.{h,cpp}`. The above enum classes are used in the parser, sema, and the AST attribute. The latter is not a collection of multiple primitive variant arguments that contain encodings via numbers and strings but instead a tree that mirrors the `match` clause (see `struct OpenMPTraitInfo`). The changes to the parser make it more forgiving when wrong syntax is read and they also resulted in more specialized diagnostics. The tests are updated and the core issues are detected as before. Here and elsewhere this patch tries to be generic, thus we do not distinguish what selector set, selector, or property is parsed except if they do behave exceptionally, as for example `user={condition(EXPR)}` does. The sema logic changed in two ways: First, the OMPDeclareVariantAttr representation changed, as mentioned above, and the sema was adjusted to work with the new `OpenMPTraitInfo`. Second, the matching and scoring logic moved into `OMPContext.{h,cpp}`. It is implemented on a flat representation of the `match` clause that is not tied to clang. `OpenMPTraitInfo` provides a method to generate this flat structure (see `struct VariantMatchInfo`) by computing integer score values and boolean user conditions from the `clang::Expr` we keep for them. The OpenMP context is now an explicit object (see `struct OMPContext`). This is in anticipation of construct traits that need to be tracked. The OpenMP context, as well as the `VariantMatchInfo`, are basically made up of a set of active or respectively required traits, e.g., 'host', and an ordered container of constructs which allows duplication. Matching and scoring is kept as generic as possible to allow easy extension in the future. --- Test changes: The messages checked in `OpenMP/declare_variant_messages.{c,cpp}` have been auto generated to match the new warnings and notes of the parser. The "subset" checks were reversed causing the wrong version to be picked. The tests have been adjusted to correct this. We do not print scores if the user did not provide one. We print spaces to make lists in the `match` clause more legible. Reviewers: kiranchandramohan, ABataev, RaviNarayanaswamy, gtbercea, grokos, sdmitriev, JonChesterfield, hfinkel, fghanim Subscribers: merge_guards_bot, rampitec, mgorny, hiraditya, aheejin, fedor.sergeev, simoncook, bollu, guansong, dexonsmith, jfb, s.egerton, llvm-commits, cfe-commits Tags: #clang, #llvm Differential Revision: https://reviews.llvm.org/D71830
2019-12-20 10:42:12 +08:00
S.append("'").append(Str).append("'").append(" ");
#include "llvm/Frontend/OpenMP/OMPKinds.def"
S.pop_back();
return S;
}
std::string llvm::omp::listOpenMPContextTraitSelectors(TraitSet Set) {
std::string S;
#define OMP_TRAIT_SELECTOR(Enum, TraitSetEnum, Str, ReqProp) \
if (TraitSet::TraitSetEnum == Set && StringRef(Str) != "Invalid") \
[OpenMP][Part 2] Use reusable OpenMP context/traits handling This patch implements an almost complete handling of OpenMP contexts/traits such that we can reuse most of the logic in Flang through the OMPContext.{h,cpp} in llvm/Frontend/OpenMP. All but construct SIMD specifiers, e.g., inbranch, and the device ISA selector are define in `llvm/lib/Frontend/OpenMP/OMPKinds.def`. From these definitions we generate the enum classes `TraitSet`, `TraitSelector`, and `TraitProperty` as well as conversion and helper functions in `llvm/lib/Frontend/OpenMP/OMPContext.{h,cpp}`. The above enum classes are used in the parser, sema, and the AST attribute. The latter is not a collection of multiple primitive variant arguments that contain encodings via numbers and strings but instead a tree that mirrors the `match` clause (see `struct OpenMPTraitInfo`). The changes to the parser make it more forgiving when wrong syntax is read and they also resulted in more specialized diagnostics. The tests are updated and the core issues are detected as before. Here and elsewhere this patch tries to be generic, thus we do not distinguish what selector set, selector, or property is parsed except if they do behave exceptionally, as for example `user={condition(EXPR)}` does. The sema logic changed in two ways: First, the OMPDeclareVariantAttr representation changed, as mentioned above, and the sema was adjusted to work with the new `OpenMPTraitInfo`. Second, the matching and scoring logic moved into `OMPContext.{h,cpp}`. It is implemented on a flat representation of the `match` clause that is not tied to clang. `OpenMPTraitInfo` provides a method to generate this flat structure (see `struct VariantMatchInfo`) by computing integer score values and boolean user conditions from the `clang::Expr` we keep for them. The OpenMP context is now an explicit object (see `struct OMPContext`). This is in anticipation of construct traits that need to be tracked. The OpenMP context, as well as the `VariantMatchInfo`, are basically made up of a set of active or respectively required traits, e.g., 'host', and an ordered container of constructs which allows duplication. Matching and scoring is kept as generic as possible to allow easy extension in the future. --- Test changes: The messages checked in `OpenMP/declare_variant_messages.{c,cpp}` have been auto generated to match the new warnings and notes of the parser. The "subset" checks were reversed causing the wrong version to be picked. The tests have been adjusted to correct this. We do not print scores if the user did not provide one. We print spaces to make lists in the `match` clause more legible. Reviewers: kiranchandramohan, ABataev, RaviNarayanaswamy, gtbercea, grokos, sdmitriev, JonChesterfield, hfinkel, fghanim Subscribers: merge_guards_bot, rampitec, mgorny, hiraditya, aheejin, fedor.sergeev, simoncook, bollu, guansong, dexonsmith, jfb, s.egerton, llvm-commits, cfe-commits Tags: #clang, #llvm Differential Revision: https://reviews.llvm.org/D71830
2019-12-20 10:42:12 +08:00
S.append("'").append(Str).append("'").append(" ");
#include "llvm/Frontend/OpenMP/OMPKinds.def"
S.pop_back();
return S;
}
std::string
llvm::omp::listOpenMPContextTraitProperties(TraitSet Set,
TraitSelector Selector) {
std::string S;
#define OMP_TRAIT_PROPERTY(Enum, TraitSetEnum, TraitSelectorEnum, Str) \
if (TraitSet::TraitSetEnum == Set && \
TraitSelector::TraitSelectorEnum == Selector && \
StringRef(Str) != "invalid") \
[OpenMP][Part 2] Use reusable OpenMP context/traits handling This patch implements an almost complete handling of OpenMP contexts/traits such that we can reuse most of the logic in Flang through the OMPContext.{h,cpp} in llvm/Frontend/OpenMP. All but construct SIMD specifiers, e.g., inbranch, and the device ISA selector are define in `llvm/lib/Frontend/OpenMP/OMPKinds.def`. From these definitions we generate the enum classes `TraitSet`, `TraitSelector`, and `TraitProperty` as well as conversion and helper functions in `llvm/lib/Frontend/OpenMP/OMPContext.{h,cpp}`. The above enum classes are used in the parser, sema, and the AST attribute. The latter is not a collection of multiple primitive variant arguments that contain encodings via numbers and strings but instead a tree that mirrors the `match` clause (see `struct OpenMPTraitInfo`). The changes to the parser make it more forgiving when wrong syntax is read and they also resulted in more specialized diagnostics. The tests are updated and the core issues are detected as before. Here and elsewhere this patch tries to be generic, thus we do not distinguish what selector set, selector, or property is parsed except if they do behave exceptionally, as for example `user={condition(EXPR)}` does. The sema logic changed in two ways: First, the OMPDeclareVariantAttr representation changed, as mentioned above, and the sema was adjusted to work with the new `OpenMPTraitInfo`. Second, the matching and scoring logic moved into `OMPContext.{h,cpp}`. It is implemented on a flat representation of the `match` clause that is not tied to clang. `OpenMPTraitInfo` provides a method to generate this flat structure (see `struct VariantMatchInfo`) by computing integer score values and boolean user conditions from the `clang::Expr` we keep for them. The OpenMP context is now an explicit object (see `struct OMPContext`). This is in anticipation of construct traits that need to be tracked. The OpenMP context, as well as the `VariantMatchInfo`, are basically made up of a set of active or respectively required traits, e.g., 'host', and an ordered container of constructs which allows duplication. Matching and scoring is kept as generic as possible to allow easy extension in the future. --- Test changes: The messages checked in `OpenMP/declare_variant_messages.{c,cpp}` have been auto generated to match the new warnings and notes of the parser. The "subset" checks were reversed causing the wrong version to be picked. The tests have been adjusted to correct this. We do not print scores if the user did not provide one. We print spaces to make lists in the `match` clause more legible. Reviewers: kiranchandramohan, ABataev, RaviNarayanaswamy, gtbercea, grokos, sdmitriev, JonChesterfield, hfinkel, fghanim Subscribers: merge_guards_bot, rampitec, mgorny, hiraditya, aheejin, fedor.sergeev, simoncook, bollu, guansong, dexonsmith, jfb, s.egerton, llvm-commits, cfe-commits Tags: #clang, #llvm Differential Revision: https://reviews.llvm.org/D71830
2019-12-20 10:42:12 +08:00
S.append("'").append(Str).append("'").append(" ");
#include "llvm/Frontend/OpenMP/OMPKinds.def"
if (S.empty())
return "<none>";
[OpenMP][Part 2] Use reusable OpenMP context/traits handling This patch implements an almost complete handling of OpenMP contexts/traits such that we can reuse most of the logic in Flang through the OMPContext.{h,cpp} in llvm/Frontend/OpenMP. All but construct SIMD specifiers, e.g., inbranch, and the device ISA selector are define in `llvm/lib/Frontend/OpenMP/OMPKinds.def`. From these definitions we generate the enum classes `TraitSet`, `TraitSelector`, and `TraitProperty` as well as conversion and helper functions in `llvm/lib/Frontend/OpenMP/OMPContext.{h,cpp}`. The above enum classes are used in the parser, sema, and the AST attribute. The latter is not a collection of multiple primitive variant arguments that contain encodings via numbers and strings but instead a tree that mirrors the `match` clause (see `struct OpenMPTraitInfo`). The changes to the parser make it more forgiving when wrong syntax is read and they also resulted in more specialized diagnostics. The tests are updated and the core issues are detected as before. Here and elsewhere this patch tries to be generic, thus we do not distinguish what selector set, selector, or property is parsed except if they do behave exceptionally, as for example `user={condition(EXPR)}` does. The sema logic changed in two ways: First, the OMPDeclareVariantAttr representation changed, as mentioned above, and the sema was adjusted to work with the new `OpenMPTraitInfo`. Second, the matching and scoring logic moved into `OMPContext.{h,cpp}`. It is implemented on a flat representation of the `match` clause that is not tied to clang. `OpenMPTraitInfo` provides a method to generate this flat structure (see `struct VariantMatchInfo`) by computing integer score values and boolean user conditions from the `clang::Expr` we keep for them. The OpenMP context is now an explicit object (see `struct OMPContext`). This is in anticipation of construct traits that need to be tracked. The OpenMP context, as well as the `VariantMatchInfo`, are basically made up of a set of active or respectively required traits, e.g., 'host', and an ordered container of constructs which allows duplication. Matching and scoring is kept as generic as possible to allow easy extension in the future. --- Test changes: The messages checked in `OpenMP/declare_variant_messages.{c,cpp}` have been auto generated to match the new warnings and notes of the parser. The "subset" checks were reversed causing the wrong version to be picked. The tests have been adjusted to correct this. We do not print scores if the user did not provide one. We print spaces to make lists in the `match` clause more legible. Reviewers: kiranchandramohan, ABataev, RaviNarayanaswamy, gtbercea, grokos, sdmitriev, JonChesterfield, hfinkel, fghanim Subscribers: merge_guards_bot, rampitec, mgorny, hiraditya, aheejin, fedor.sergeev, simoncook, bollu, guansong, dexonsmith, jfb, s.egerton, llvm-commits, cfe-commits Tags: #clang, #llvm Differential Revision: https://reviews.llvm.org/D71830
2019-12-20 10:42:12 +08:00
S.pop_back();
return S;
}