Cold-Air-Networks/mirror-llvm-project
2
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 7

[IR] Remove ProfileCount Abstraction

Browse source 
This only exists to differentiate between real and synthetic profiles.
Remove the abstraction now that we plan to fully remove synthetic
profiles.

Reviewers: mtrofin, david-xl

Reviewed By: mtrofin

Pull Request: https://github.com/llvm/llvm-project/pull/204770
This commit is contained in:
Aiden Grossman 2026-06-23 20:26:14 +00:00 committed by GitHub
parent 65a7ccf656
commit 37a71b650a
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
24 changed files with 63 additions and 125 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

4
llvm/include/llvm/Analysis/ProfileSummaryInfo.h
View file

@ -325,9 +325,7 @@ private:
template <typename FuncT>
std::optional<uint64_t> getEntryCount(const FuncT *F) const {
if (!F->getEntryCount().has_value())
return std::nullopt;
return F->getEntryCount()->getCount();
return F->getEntryCount();
}
};

26
llvm/include/llvm/IR/Function.h
View file

@ -291,41 +291,19 @@ public:
}
}
enum ProfileCountType { PCT_Real, PCT_Synthetic };
/// Class to represent profile counts.
///
/// This class represents both real and synthetic profile counts.
class ProfileCount {
private:
uint64_t Count = 0;
ProfileCountType PCT = PCT_Real;
public:
ProfileCount(uint64_t Count, ProfileCountType PCT)
: Count(Count), PCT(PCT) {}
uint64_t getCount() const { return Count; }
ProfileCountType getType() const { return PCT; }
bool isSynthetic() const { return PCT == PCT_Synthetic; }
};
/// Set the entry count for this function.
///
/// Entry count is the number of times this function was executed based on
/// pgo data. \p Imports points to a set of GUIDs that needs to
/// be imported by the function for sample PGO, to enable the same inlines as
/// the profiled optimized binary.
void setEntryCount(ProfileCount Count,
const DenseSet<GlobalValue::GUID> *Imports = nullptr);
/// A convenience wrapper for setting entry count
void setEntryCount(uint64_t Count, ProfileCountType Type = PCT_Real,
void setEntryCount(uint64_t Count,
const DenseSet<GlobalValue::GUID> *Imports = nullptr);
/// Get the entry count for this function.
///
/// Entry count is the number of times the function was executed.
std::optional<ProfileCount> getEntryCount() const;
std::optional<uint64_t> getEntryCount() const;
/// Return true if the function is annotated with profile data.
///

9
llvm/include/llvm/Transforms/Utils/SampleProfileLoaderBaseImpl.h
View file

@ -47,7 +47,6 @@
namespace llvm {
using namespace sampleprof;
using namespace sampleprofutil;
using ProfileCount = Function::ProfileCount;
namespace vfs {
class FileSystem;
@ -1069,9 +1068,7 @@ void SampleProfileLoaderBaseImpl<BT>::initWeightPropagation(
// Sets the GUIDs that are inlined in the profiled binary. This is used
// for ThinLink to make correct liveness analysis, and also make the IR
// match the profiled binary before annotation.
getFunction(F).setEntryCount(
ProfileCount(Samples->getHeadSamples() + 1, Function::PCT_Real),
&InlinedGUIDs);
getFunction(F).setEntryCount(Samples->getHeadSamples() + 1, &InlinedGUIDs);
if (!SampleProfileUseProfi) {
// Compute dominance and loop info needed for propagation.
@ -1101,9 +1098,7 @@ void SampleProfileLoaderBaseImpl<BT>::finalizeWeightPropagation(
if (SampleProfileUseProfi) {
const BasicBlockT *EntryBB = getEntryBB(&F);
if (BlockWeights[EntryBB] > 0) {
getFunction(F).setEntryCount(
ProfileCount(BlockWeights[EntryBB], Function::PCT_Real),
&InlinedGUIDs);
getFunction(F).setEntryCount(BlockWeights[EntryBB], &InlinedGUIDs);
}
}
}

2
llvm/lib/Analysis/BlockFrequencyInfoImpl.cpp
View file

@ -593,7 +593,7 @@ BlockFrequencyInfoImplBase::getProfileCountFromFreq(const Function &F,
if (!EntryCount)
return std::nullopt;
// Use 128 bit APInt to do the arithmetic to avoid overflow.
APInt BlockCount(128, EntryCount->getCount());
APInt BlockCount(128, *EntryCount);
APInt BlockFreq(128, Freq.getFrequency());
APInt EntryFreq(128, getEntryFreq().getFrequency());
BlockCount *= BlockFreq;

9
llvm/lib/Analysis/InlineCost.cpp
View file

@ -928,7 +928,7 @@ class InlineCostCallAnalyzer final : public CallAnalyzer {
// Make sure we have a nonzero entry count.
auto EntryCount = F.getEntryCount();
if (!EntryCount || !EntryCount->getCount())
if (!EntryCount || *EntryCount == 0)
return false;
BlockFrequencyInfo *CalleeBFI = &(GetBFI(F));
@ -1017,10 +1017,9 @@ class InlineCostCallAnalyzer final : public CallAnalyzer {
// Compute the cycle savings per call.
auto EntryProfileCount = F.getEntryCount();
assert(EntryProfileCount && EntryProfileCount->getCount());
auto EntryCount = EntryProfileCount->getCount();
CycleSavings += EntryCount / 2;
CycleSavings = CycleSavings.udiv(EntryCount);
assert(EntryProfileCount && *EntryProfileCount);
CycleSavings += *EntryProfileCount / 2;
CycleSavings = CycleSavings.udiv(*EntryProfileCount);
// Compute the total savings for the call site.
auto *CallerBB = CandidateCall.getParent();

2
llvm/lib/Analysis/ProfileSummaryInfo.cpp
View file

@ -114,7 +114,7 @@ bool ProfileSummaryInfo::isFunctionEntryCold(const Function *F) const {
// FIXME: The heuristic used below for determining coldness is based on
// preliminary SPEC tuning for inliner. This will eventually be a
// convenience method that calls isHotCount.
return FunctionCount && isColdCount(FunctionCount->getCount());
return FunctionCount && isColdCount(*FunctionCount);
}
/// Compute the hot and cold thresholds.

3
llvm/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
View file

@ -1608,8 +1608,7 @@ void AsmPrinter::emitBBAddrMapSection(const MachineFunction &MF) {
if (Features.FuncEntryCount) {
OutStreamer->AddComment("function entry count");
auto MaybeEntryCount = MF.getFunction().getEntryCount();
OutStreamer->emitULEB128IntValue(
MaybeEntryCount ? MaybeEntryCount->getCount() : 0);
OutStreamer->emitULEB128IntValue(MaybeEntryCount ? *MaybeEntryCount : 0);
}
const MachineBlockFrequencyInfo *MBFI =
Features.BBFreq

1
llvm/lib/CodeGen/MIRSampleProfile.cpp
View file

@ -38,7 +38,6 @@
using namespace llvm;
using namespace sampleprof;
using namespace llvm::sampleprofutil;
using ProfileCount = Function::ProfileCount;
#define DEBUG_TYPE "fs-profile-loader"

4
llvm/lib/CodeGen/MachineFunction.cpp
View file

@ -1669,9 +1669,7 @@ template <>
std::optional<uint64_t>
ProfileSummaryInfo::getEntryCount<llvm::MachineFunction>(
const llvm::MachineFunction *F) const {
if (!F->getFunction().getEntryCount().has_value())
return std::nullopt;
return F->getFunction().getEntryCount()->getCount();
return F->getFunction().getEntryCount();
}
#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)

22
llvm/lib/IR/Function.cpp
View file

@ -56,7 +56,6 @@
#include <string>
using namespace llvm;
using ProfileCount = Function::ProfileCount;
// Explicit instantiations of SymbolTableListTraits since some of the methods
// are not in the public header file...
@ -1079,29 +1078,18 @@ void Function::setValueSubclassDataBit(unsigned Bit, bool On) {
setValueSubclassData(getSubclassDataFromValue() & ~(1 << Bit));
}
void Function::setEntryCount(ProfileCount Count,
void Function::setEntryCount(uint64_t Count,
const DenseSet<GlobalValue::GUID> *S) {
#if !defined(NDEBUG)
auto PrevCount = getEntryCount();
assert(!PrevCount || PrevCount->getType() == Count.getType());
#endif
auto ImportGUIDs = getImportGUIDs();
if (S == nullptr && ImportGUIDs.size())
S = &ImportGUIDs;
MDBuilder MDB(getContext());
setMetadata(
LLVMContext::MD_prof,
MDB.createFunctionEntryCount(Count.getCount(), Count.isSynthetic(), S));
setMetadata(LLVMContext::MD_prof,
MDB.createFunctionEntryCount(Count, false, S));
}
void Function::setEntryCount(uint64_t Count, Function::ProfileCountType Type,
const DenseSet<GlobalValue::GUID> *Imports) {
setEntryCount(ProfileCount(Count, Type), Imports);
}
std::optional<ProfileCount> Function::getEntryCount() const {
std::optional<uint64_t> Function::getEntryCount() const {
MDNode *MD = getMetadata(LLVMContext::MD_prof);
if (MD && MD->getOperand(0))
if (MDString *MDS = dyn_cast<MDString>(MD->getOperand(0))) {
@ -1113,7 +1101,7 @@ std::optional<ProfileCount> Function::getEntryCount() const {
// Treat this the same as unknown.
if (Count == static_cast<uint64_t>(-1))
return std::nullopt;
return ProfileCount(Count, PCT_Real);
return Count;
}
return std::nullopt;
}

6
llvm/lib/IR/ProfDataUtils.cpp
View file

@ -282,15 +282,13 @@ void llvm::setExplicitlyUnknownBranchWeightsIfProfiled(Instruction &I,
F = F ? F : I.getFunction();
assert(F && "Either pass a instruction attached to a Function, or explicitly "
"pass the Function that it will be attached to");
if (std::optional<Function::ProfileCount> EC = F->getEntryCount();
EC && EC->getCount() > 0)
if (std::optional<uint64_t> EC = F->getEntryCount(); EC && *EC > 0)
setExplicitlyUnknownBranchWeights(I, PassName);
}
MDNode *llvm::getExplicitlyUnknownBranchWeightsIfProfiled(Function &F,
StringRef PassName) {
if (std::optional<Function::ProfileCount> EC = F.getEntryCount();
!EC || EC->getCount() == 0)
if (std::optional<uint64_t> EC = F.getEntryCount(); !EC || *EC == 0)
return nullptr;
MDBuilder MDB(F.getContext());
return MDNode::get(

4
llvm/lib/Transforms/AggressiveInstCombine/AggressiveInstCombine.cpp
View file

@ -1953,8 +1953,8 @@ void StrNCmpInliner::inlineCompare(Value *LHS, StringRef RHS, uint64_t N,
Function *F = CI->getFunction();
assert(F && "Instruction does not belong to a function!");
std::optional<Function::ProfileCount> EC = F->getEntryCount();
if (EC && EC->getCount() > 0)
std::optional<uint64_t> EC = F->getEntryCount();
if (EC && *EC > 0)
setExplicitlyUnknownBranchWeights(*CondBrInst, DEBUG_TYPE);
} else {
B.CreateBr(BBNE);

9
llvm/lib/Transforms/IPO/FunctionSpecialization.cpp
View file

@ -795,14 +795,13 @@ bool FunctionSpecializer::run() {
std::optional<uint64_t> Count =
BFI.getBlockProfileCount(Call->getParent());
if (Count && !ProfcheckDisableMetadataFixes) {
std::optional<llvm::Function::ProfileCount> MaybeCloneCount =
Clone->getEntryCount();
std::optional<uint64_t> MaybeCloneCount = Clone->getEntryCount();
if (MaybeCloneCount) {
uint64_t CallCount = *Count + MaybeCloneCount->getCount();
uint64_t CallCount = *Count + *MaybeCloneCount;
Clone->setEntryCount(CallCount);
if (std::optional<llvm::Function::ProfileCount> MaybeOriginalCount =
if (std::optional<uint64_t> MaybeOriginalCount =
S.F->getEntryCount()) {
uint64_t OriginalCount = MaybeOriginalCount->getCount();
uint64_t OriginalCount = *MaybeOriginalCount;
if (OriginalCount >= *Count) {
S.F->setEntryCount(OriginalCount - *Count);
} else {

6
llvm/lib/Transforms/IPO/PartialInlining.cpp
View file

@ -1355,8 +1355,7 @@ bool PartialInlinerImpl::tryPartialInline(FunctionCloner &Cloner) {
if (CalleeEntryCount)
computeCallsiteToProfCountMap(Cloner.ClonedFunc, CallSiteToProfCountMap);
uint64_t CalleeEntryCountV =
(CalleeEntryCount ? CalleeEntryCount->getCount() : 0);
uint64_t CalleeEntryCountV = (CalleeEntryCount ? *CalleeEntryCount : 0);
bool AnyInline = false;
for (User *User : Users) {
@ -1408,8 +1407,7 @@ bool PartialInlinerImpl::tryPartialInline(FunctionCloner &Cloner) {
if (AnyInline) {
Cloner.IsFunctionInlined = true;
if (CalleeEntryCount)
Cloner.OrigFunc->setEntryCount(Function::ProfileCount(
CalleeEntryCountV, CalleeEntryCount->getType()));
Cloner.OrigFunc->setEntryCount(CalleeEntryCountV);
OptimizationRemarkEmitter OrigFuncORE(Cloner.OrigFunc);
OrigFuncORE.emit([&]() {
return OptimizationRemark(DEBUG_TYPE, "PartiallyInlined", Cloner.OrigFunc)

3
llvm/lib/Transforms/IPO/SampleProfile.cpp
View file

@ -93,7 +93,6 @@
using namespace llvm;
using namespace sampleprof;
using namespace llvm::sampleprofutil;
using ProfileCount = Function::ProfileCount;
#define DEBUG_TYPE "sample-profile"
#define CSINLINE_DEBUG DEBUG_TYPE "-inline"
@ -2291,7 +2290,7 @@ bool SampleProfileLoader::runOnFunction(Function &F,
// Initialize entry count when the function has no existing entry
// count value.
if (!F.getEntryCount())
F.setEntryCount(ProfileCount(initialEntryCount, Function::PCT_Real));
F.setEntryCount(initialEntryCount);
auto &FAM = AM.getResult<FunctionAnalysisManagerModuleProxy>(*F.getParent())
.getManager();
ORE = &FAM.getResult<OptimizationRemarkEmitterAnalysis>(F);

4
llvm/lib/Transforms/IPO/WholeProgramDevirt.cpp
View file

@ -1581,12 +1581,12 @@ void DevirtModule::applyICallBranchFunnel(VTableSlotInfo &SlotInfo,
auto EC = BFI.getBlockFreq(&F.getEntryBlock());
auto CC = F.getEntryCount();
double CallCount = 0.0;
if (EC.getFrequency() != 0 && CC && CC->getCount() != 0) {
if (EC.getFrequency() != 0 && CC && *CC != 0) {
double CallFreq =
static_cast<double>(
BFI.getBlockFreq(CB.getParent()).getFrequency()) /
EC.getFrequency();
CallCount = CallFreq * CC->getCount();
CallCount = CallFreq * *CC;
}
FunctionEntryCounts[&JT] += CallCount;
}

12
llvm/lib/Transforms/Instrumentation/PGOInstrumentation.cpp
View file

@ -128,7 +128,6 @@
#include <vector>
using namespace llvm;
using ProfileCount = Function::ProfileCount;
using VPCandidateInfo = ValueProfileCollector::CandidateInfo;
#define DEBUG_TYPE "pgo-instrumentation"
@ -1675,7 +1674,7 @@ void PGOUseFunc::populateCounters() {
// Fix the obviously inconsistent entry count.
if (FuncMaxCount > 0 && FuncEntryCount == 0)
FuncEntryCount = 1;
F.setEntryCount(ProfileCount(FuncEntryCount, Function::PCT_Real));
F.setEntryCount(FuncEntryCount);
markFunctionAttributes(FuncEntryCount, FuncMaxCount);
LLVM_DEBUG(FuncInfo.dumpInfo("after reading profile."));
@ -1943,7 +1942,7 @@ static bool skipPGOGen(const Function &F) {
return true;
if (PGOInstrumentColdFunctionOnly) {
if (auto EntryCount = F.getEntryCount())
return EntryCount->getCount() > PGOColdInstrumentEntryThreshold;
return *EntryCount > PGOColdInstrumentEntryThreshold;
return !PGOTreatUnknownAsCold;
}
return false;
@ -2031,8 +2030,7 @@ static void fixFuncEntryCount(PGOUseFunc &Func, LoopInfo &LI,
BlockFrequencyInfo NBFI(F, NBPI, LI);
#ifndef NDEBUG
auto BFIEntryCount = F.getEntryCount();
assert(BFIEntryCount && (BFIEntryCount->getCount() > 0) &&
"Invalid BFI Entrycount");
assert(BFIEntryCount && (*BFIEntryCount > 0) && "Invalid BFI Entrycount");
#endif
auto SumCount = APFloat::getZero(APFloat::IEEEdouble());
auto SumBFICount = APFloat::getZero(APFloat::IEEEdouble());
@ -2063,7 +2061,7 @@ static void fixFuncEntryCount(PGOUseFunc &Func, LoopInfo &LI,
if (NewEntryCount == 0)
NewEntryCount = 1;
if (NewEntryCount != FuncEntryCount) {
F.setEntryCount(ProfileCount(NewEntryCount, Function::PCT_Real));
F.setEntryCount(NewEntryCount);
LLVM_DEBUG(dbgs() << "FixFuncEntryCount: in " << F.getName()
<< ", entry_count " << FuncEntryCount << " --> "
<< NewEntryCount << "\n");
@ -2255,7 +2253,7 @@ static bool annotateAllFunctions(
if (!Func.readCounters(AllZeros, PseudoKind))
continue;
if (AllZeros) {
F.setEntryCount(ProfileCount(0, Function::PCT_Real));
F.setEntryCount(0);
if (Func.getProgramMaxCount() != 0)
ColdFunctions.push_back(&F);
continue;

8
llvm/lib/Transforms/Scalar/TailRecursionElimination.cpp
View file

@ -442,7 +442,7 @@ class TailRecursionEliminator {
: F(F), TTI(TTI), AA(AA), ORE(ORE), DTU(DTU), BFI(BFI),
OrigEntryBBFreq(
BFI ? BFI->getBlockFreq(&F.getEntryBlock()).getFrequency() : 0U),
OrigEntryCount(F.getEntryCount() ? F.getEntryCount()->getCount() : 0) {
OrigEntryCount(F.getEntryCount() ? *F.getEntryCount() : 0) {
if (BFI) {
// The assert is meant as API documentation for the caller.
assert((OrigEntryCount != 0 && OrigEntryBBFreq != 0) &&
@ -771,7 +771,7 @@ bool TailRecursionEliminator::eliminateCall(CallInst *CI) {
static_cast<double>(OrigEntryBBFreq);
auto ToSubtract =
static_cast<uint64_t>(std::round(RelativeBBFreq * OrigEntryCount));
auto OldEntryCount = F.getEntryCount()->getCount();
auto OldEntryCount = *F.getEntryCount();
if (OldEntryCount <= ToSubtract) {
LLVM_DEBUG(
errs() << "[TRE] The entrycount attributable to the recursive call, "
@ -779,7 +779,7 @@ bool TailRecursionEliminator::eliminateCall(CallInst *CI) {
<< ", should be strictly lower than the function entry count, "
<< OldEntryCount << "\n");
} else {
F.setEntryCount(OldEntryCount - ToSubtract, F.getEntryCount()->getType());
F.setEntryCount(OldEntryCount - ToSubtract);
}
}
return true;
@ -992,7 +992,7 @@ PreservedAnalyses TailCallElimPass::run(Function &F,
// the lines asking for the cached result, should they be nullptr (which, in
// the case of the PDT, is likely), updates to the trees would be missed.
auto *BFI = (!ForceDisableBFI && UpdateFunctionEntryCount &&
F.getEntryCount().has_value() && F.getEntryCount()->getCount())
F.getEntryCount().has_value() && *F.getEntryCount())
? &AM.getResult<BlockFrequencyAnalysis>(F)
: nullptr;
auto &ORE = AM.getResult<OptimizationRemarkEmitterAnalysis>(F);

4
llvm/lib/Transforms/Utils/CodeExtractor.cpp
View file

@ -66,7 +66,6 @@
using namespace llvm;
using namespace llvm::PatternMatch;
using ProfileCount = Function::ProfileCount;
#define DEBUG_TYPE "code-extractor"
@ -1092,8 +1091,7 @@ Function *CodeExtractor::constructFunctionDeclaration(
if (BFI) {
auto Count = BFI->getProfileCountFromFreq(EntryFreq);
if (Count.has_value())
newFunction->setEntryCount(
ProfileCount(*Count, Function::PCT_Real)); // FIXME
newFunction->setEntryCount(*Count);
}
return newFunction;

32
llvm/lib/Transforms/Utils/InlineFunction.cpp
View file

@ -86,7 +86,6 @@
using namespace llvm;
using namespace llvm::memprof;
using ProfileCount = Function::ProfileCount;
static cl::opt<bool>
EnableNoAliasConversion("enable-noalias-to-md-conversion", cl::init(true),
@ -2162,15 +2161,14 @@ static void updateCallerBFI(BasicBlock *CallSiteBlock,
/// Update the branch metadata for cloned call instructions.
static void updateCallProfile(Function *Callee, const ValueToValueMapTy &VMap,
const ProfileCount &CalleeEntryCount,
const uint64_t &CalleeEntryCount,
const CallBase &TheCall, ProfileSummaryInfo *PSI,
BlockFrequencyInfo *CallerBFI) {
if (CalleeEntryCount.isSynthetic() || CalleeEntryCount.getCount() < 1)
if (CalleeEntryCount < 1)
return;
auto CallSiteCount =
PSI ? PSI->getProfileCount(TheCall, CallerBFI) : std::nullopt;
int64_t CallCount =
std::min(CallSiteCount.value_or(0), CalleeEntryCount.getCount());
int64_t CallCount = std::min(CallSiteCount.value_or(0), CalleeEntryCount);
updateProfileCallee(Callee, -CallCount, &VMap);
}
@ -2181,14 +2179,12 @@ void llvm::updateProfileCallee(
if (!CalleeCount)
return;
const uint64_t PriorEntryCount = CalleeCount->getCount();
// Since CallSiteCount is an estimate, it could exceed the original callee
// count and has to be set to 0 so guard against underflow.
const uint64_t NewEntryCount =
(EntryDelta < 0 && static_cast<uint64_t>(-EntryDelta) > PriorEntryCount)
(EntryDelta < 0 && static_cast<uint64_t>(-EntryDelta) > *CalleeCount)
? 0
: PriorEntryCount + EntryDelta;
: *CalleeCount + EntryDelta;
auto updateVTableProfWeight = [](CallBase *CB, const uint64_t NewEntryCount,
const uint64_t PriorEntryCount) {
@ -2199,18 +2195,18 @@ void llvm::updateProfileCallee(
// During inlining ?
if (VMap) {
uint64_t CloneEntryCount = PriorEntryCount - NewEntryCount;
uint64_t CloneEntryCount = *CalleeCount - NewEntryCount;
for (auto Entry : *VMap) {
if (isa<CallInst>(Entry.first))
if (auto *CI = dyn_cast_or_null<CallInst>(Entry.second)) {
CI->updateProfWeight(CloneEntryCount, PriorEntryCount);
updateVTableProfWeight(CI, CloneEntryCount, PriorEntryCount);
CI->updateProfWeight(CloneEntryCount, *CalleeCount);
updateVTableProfWeight(CI, CloneEntryCount, *CalleeCount);
}
if (isa<InvokeInst>(Entry.first))
if (auto *II = dyn_cast_or_null<InvokeInst>(Entry.second)) {
II->updateProfWeight(CloneEntryCount, PriorEntryCount);
updateVTableProfWeight(II, CloneEntryCount, PriorEntryCount);
II->updateProfWeight(CloneEntryCount, *CalleeCount);
updateVTableProfWeight(II, CloneEntryCount, *CalleeCount);
}
}
}
@ -2223,12 +2219,12 @@ void llvm::updateProfileCallee(
if (!VMap || VMap->count(&BB))
for (Instruction &I : BB) {
if (CallInst *CI = dyn_cast<CallInst>(&I)) {
CI->updateProfWeight(NewEntryCount, PriorEntryCount);
updateVTableProfWeight(CI, NewEntryCount, PriorEntryCount);
CI->updateProfWeight(NewEntryCount, *CalleeCount);
updateVTableProfWeight(CI, NewEntryCount, *CalleeCount);
}
if (InvokeInst *II = dyn_cast<InvokeInst>(&I)) {
II->updateProfWeight(NewEntryCount, PriorEntryCount);
updateVTableProfWeight(II, NewEntryCount, PriorEntryCount);
II->updateProfWeight(NewEntryCount, *CalleeCount);
updateVTableProfWeight(II, NewEntryCount, *CalleeCount);
}
}
}

5
llvm/lib/Transforms/Utils/LowerMemIntrinsics.cpp
View file

@ -72,9 +72,8 @@ struct LoopExpansionInfo {
std::optional<uint64_t> getAverageMemOpLoopTripCount(const MemIntrinsic &I) {
if (ProfcheckDisableMetadataFixes)
return std::nullopt;
if (std::optional<Function::ProfileCount> EC =
I.getFunction()->getEntryCount();
!EC || !EC->getCount())
if (std::optional<uint64_t> EC = I.getFunction()->getEntryCount();
!EC || *EC == 0)
return std::nullopt;
if (const auto Len = I.getLengthInBytes())
return Len->getZExtValue();

4
llvm/lib/Transforms/Utils/ProfileVerify.cpp
View file

@ -120,7 +120,7 @@ bool ProfileInjector::inject() {
F.setEntryCount(DefaultFunctionEntryCount);
// If there is an entry count that's 0, then don't bother injecting. We won't
// verify these either.
if (F.getEntryCount()->getCount() == 0)
if (*F.getEntryCount() == 0)
return false;
bool Changed = false;
// Cycle through the weights list. If we didn't, tests with more than (say)
@ -248,7 +248,7 @@ PreservedAnalyses ProfileVerifierPass::run(Function &F,
emitProfileError("function entry count missing (set to 0 if cold)", F);
return PreservedAnalyses::all();
}
} else if (EntryCount->getCount() == 0) {
} else if (*EntryCount == 0) {
return PreservedAnalyses::all();
}
for (const auto &BB : F) {

2
llvm/lib/Transforms/Vectorize/VPlan.cpp
View file

@ -1796,7 +1796,7 @@ void LoopVectorizationPlanner::updateLoopMetadataAndProfileInfo(
unsigned AverageVectorTripCount = 0;
unsigned RemainderAverageTripCount = 0;
auto EC = VectorLoop->getLoopPreheader()->getParent()->getEntryCount();
auto IsProfiled = EC && EC->getCount();
auto IsProfiled = EC && *EC != 0;
if (!OrigAverageTripCount) {
if (!IsProfiled)
return;

7
llvm/unittests/IR/MetadataTest.cpp
View file

@ -5215,14 +5215,13 @@ TEST_F(FunctionAttachmentTest, Verifier) {
EXPECT_FALSE(verifyFunction(*F));
}
TEST_F(FunctionAttachmentTest, RealEntryCount) {
TEST_F(FunctionAttachmentTest, EntryCount) {
Function *F = getFunction("foo");
EXPECT_FALSE(F->getEntryCount().has_value());
F->setEntryCount(12304, Function::PCT_Real);
F->setEntryCount(12304);
auto Count = F->getEntryCount();
EXPECT_TRUE(Count.has_value());
EXPECT_EQ(12304u, Count->getCount());
EXPECT_EQ(Function::PCT_Real, Count->getType());
EXPECT_EQ(12304u, *Count);
}
TEST_F(FunctionAttachmentTest, SubprogramAttachment) {