From e017cd98983168b9b6683d62ff18365617273042 Mon Sep 17 00:00:00 2001 From: Johannes Doerfert Date: Fri, 13 Mar 2020 19:39:09 -0500 Subject: [PATCH] [OpenMP][NFC] Reduce instantiation time with different data structure See rational here: https://reviews.llvm.org/D71847#1922648 Reviewed By: rnk Differential Revision: https://reviews.llvm.org/D76170 --- include/llvm/Frontend/OpenMP/OMPContext.h | 10 +++--- include/llvm/Frontend/OpenMP/OMPKinds.def | 7 ++++ lib/Frontend/OpenMP/OMPContext.cpp | 39 +++++++++++++++-------- 3 files changed, 38 insertions(+), 18 deletions(-) diff --git a/include/llvm/Frontend/OpenMP/OMPContext.h b/include/llvm/Frontend/OpenMP/OMPContext.h index 960c557f55d..0a9d9c277d9 100644 --- a/include/llvm/Frontend/OpenMP/OMPContext.h +++ b/include/llvm/Frontend/OpenMP/OMPContext.h @@ -16,6 +16,7 @@ #define LLVM_OPENMP_CONTEXT_H #include "llvm/ADT/APSInt.h" +#include "llvm/ADT/BitVector.h" #include "llvm/ADT/SetVector.h" #include "llvm/ADT/SmallSet.h" #include "llvm/ADT/Triple.h" @@ -43,6 +44,7 @@ enum class TraitSelector { /// IDs for all OpenMP context trait properties (host/gpu/bsc/llvm/...) enum class TraitProperty { #define OMP_TRAIT_PROPERTY(Enum, ...) Enum, +#define OMP_LAST_TRAIT_PROPERTY(Enum) Last = Enum #include "llvm/Frontend/OpenMP/OMPKinds.def" }; @@ -122,12 +124,12 @@ struct VariantMatchInfo { void addTrait(TraitSet Set, TraitProperty Property, APInt *Score = nullptr) { if (Score) ScoreMap[Property] = *Score; - RequiredTraits.insert(Property); + RequiredTraits.set(unsigned(Property)); if (Set == TraitSet::construct) ConstructTraits.push_back(Property); } - SmallSet RequiredTraits; + BitVector RequiredTraits = BitVector(unsigned(TraitProperty::Last) + 1); SmallVector ConstructTraits; SmallDenseMap ScoreMap; }; @@ -142,12 +144,12 @@ struct OMPContext { addTrait(getOpenMPContextTraitSetForProperty(Property), Property); } void addTrait(TraitSet Set, TraitProperty Property) { - ActiveTraits.insert(Property); + ActiveTraits.set(unsigned(Property)); if (Set == TraitSet::construct) ConstructTraits.push_back(Property); } - SmallSet ActiveTraits; + BitVector ActiveTraits = BitVector(unsigned(TraitProperty::Last) + 1); SmallVector ConstructTraits; }; diff --git a/include/llvm/Frontend/OpenMP/OMPKinds.def b/include/llvm/Frontend/OpenMP/OMPKinds.def index 10ac5a759e7..5d26f07a7f5 100644 --- a/include/llvm/Frontend/OpenMP/OMPKinds.def +++ b/include/llvm/Frontend/OpenMP/OMPKinds.def @@ -427,6 +427,9 @@ __OMP_PROC_BIND_KIND(unknown, 7) #ifndef OMP_TRAIT_PROPERTY #define OMP_TRAIT_PROPERTY(Enum, TraitSetEnum, TraitSelectorEnum, Str) #endif +#ifndef OMP_LAST_TRAIT_PROPERTY +#define OMP_LAST_TRAIT_PROPERTY(Enum) +#endif #define __OMP_TRAIT_SET(Name) OMP_TRAIT_SET(Name, #Name) #define __OMP_TRAIT_SELECTOR(TraitSet, Name, RequiresProperty) \ @@ -534,10 +537,14 @@ __OMP_REQUIRES_TRAIT(reverse_offload) __OMP_REQUIRES_TRAIT(dynamic_allocators) __OMP_REQUIRES_TRAIT(atomic_default_mem_order) +OMP_LAST_TRAIT_PROPERTY( + implementation_atomic_default_mem_order_atomic_default_mem_order) + #undef __OMP_TRAIT_SELECTOR_AND_PROPERTY #undef OMP_TRAIT_SELECTOR #undef __OMP_TRAIT_SELECTOR #undef OMP_TRAIT_PROPERTY +#undef OMP_LAST_TRAIT_PROPERTY #undef __OMP_TRAIT_PROPERTY #undef __OMP_REQUIRES_TRAIT #undef OMP_REQUIRES_TRAIT diff --git a/lib/Frontend/OpenMP/OMPContext.cpp b/lib/Frontend/OpenMP/OMPContext.cpp index 62e5d1d8c46..748caa645a3 100644 --- a/lib/Frontend/OpenMP/OMPContext.cpp +++ b/lib/Frontend/OpenMP/OMPContext.cpp @@ -26,8 +26,9 @@ using namespace omp; OMPContext::OMPContext(bool IsDeviceCompilation, Triple TargetTriple) { // Add the appropriate device kind trait based on the triple and the // IsDeviceCompilation flag. - ActiveTraits.insert(IsDeviceCompilation ? TraitProperty::device_kind_nohost - : TraitProperty::device_kind_host); + ActiveTraits.set(unsigned(IsDeviceCompilation + ? TraitProperty::device_kind_nohost + : TraitProperty::device_kind_host)); switch (TargetTriple.getArch()) { case Triple::arm: case Triple::armeb: @@ -43,12 +44,12 @@ OMPContext::OMPContext(bool IsDeviceCompilation, Triple TargetTriple) { case Triple::ppc64le: case Triple::x86: case Triple::x86_64: - ActiveTraits.insert(TraitProperty::device_kind_cpu); + ActiveTraits.set(unsigned(TraitProperty::device_kind_cpu)); break; case Triple::amdgcn: case Triple::nvptx: case Triple::nvptx64: - ActiveTraits.insert(TraitProperty::device_kind_gpu); + ActiveTraits.set(unsigned(TraitProperty::device_kind_gpu)); break; default: break; @@ -58,7 +59,7 @@ OMPContext::OMPContext(bool IsDeviceCompilation, Triple TargetTriple) { #define OMP_TRAIT_PROPERTY(Enum, TraitSetEnum, TraitSelectorEnum, Str) \ if (TraitSelector::TraitSelectorEnum == TraitSelector::device_arch) \ if (TargetTriple.getArch() == TargetTriple.getArchTypeForLLVMName(Str)) \ - ActiveTraits.insert(TraitProperty::Enum); + ActiveTraits.set(unsigned(TraitProperty::Enum)); #include "llvm/Frontend/OpenMP/OMPKinds.def" // TODO: What exactly do we want to see as device ISA trait? @@ -67,20 +68,22 @@ OMPContext::OMPContext(bool IsDeviceCompilation, Triple TargetTriple) { // LLVM is the "OpenMP vendor" but we could also interpret vendor as the // target vendor. - ActiveTraits.insert(TraitProperty::implementation_vendor_llvm); + ActiveTraits.set(unsigned(TraitProperty::implementation_vendor_llvm)); // The user condition true is accepted but not false. - ActiveTraits.insert(TraitProperty::user_condition_true); + ActiveTraits.set(unsigned(TraitProperty::user_condition_true)); // This is for sure some device. - ActiveTraits.insert(TraitProperty::device_kind_any); + ActiveTraits.set(unsigned(TraitProperty::device_kind_any)); LLVM_DEBUG({ dbgs() << "[" << DEBUG_TYPE << "] New OpenMP context with the following properties:\n"; - for (auto &Property : ActiveTraits) + for (const auto &SetBitsIt : ActiveTraits.set_bits()) { + TraitProperty Property = TraitProperty(SetBitsIt); dbgs() << "\t " << getOpenMPContextTraitPropertyFullName(Property) << "\n"; + } }); } @@ -122,17 +125,24 @@ static bool isStrictSubset(const VariantMatchInfo &VMI0, // 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. - return set_is_strict_subset(VMI0.RequiredTraits, VMI1.RequiredTraits) && - isSubset(VMI0.ConstructTraits, VMI1.ConstructTraits); + if (VMI0.RequiredTraits.count() >= VMI1.RequiredTraits.count()) + return false; + for (const auto &SetBitsIt : VMI0.RequiredTraits.set_bits()) + if (!VMI1.RequiredTraits.test(SetBitsIt)) + return false; + if (!isSubset(VMI0.ConstructTraits, VMI1.ConstructTraits)) + return false; + return true; } static int isVariantApplicableInContextHelper( const VariantMatchInfo &VMI, const OMPContext &Ctx, SmallVectorImpl *ConstructMatches) { - for (TraitProperty Property : VMI.RequiredTraits) { + for (const auto &SetBitsIt : VMI.RequiredTraits.set_bits()) { + TraitProperty Property = TraitProperty(SetBitsIt); - bool IsActiveTrait = Ctx.ActiveTraits.count(Property); + bool IsActiveTrait = Ctx.ActiveTraits.test(unsigned(Property)); if (!IsActiveTrait) { LLVM_DEBUG(dbgs() << "[" << DEBUG_TYPE << "] Property " << getOpenMPContextTraitPropertyName(Property) @@ -181,7 +191,8 @@ static APInt getVariantMatchScore(const VariantMatchInfo &VMI, APInt Score(64, 1); unsigned NoConstructTraits = VMI.ConstructTraits.size(); - for (TraitProperty Property : VMI.RequiredTraits) { + for (const auto &SetBitsIt : VMI.RequiredTraits.set_bits()) { + TraitProperty Property = TraitProperty(SetBitsIt); // If there is a user score attached, use it. if (VMI.ScoreMap.count(Property)) { const APInt &UserScore = VMI.ScoreMap.lookup(Property);