//===- ModuleSummaryAnalysis.cpp - Module summary index builder -----------===// // // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. // See https://llvm.org/LICENSE.txt for license information. // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // // This pass builds a ModuleSummaryIndex object for the module, to be written // to bitcode or LLVM assembly. // //===----------------------------------------------------------------------===// #include "llvm/Analysis/ModuleSummaryAnalysis.h" #include "llvm/ADT/ArrayRef.h" #include "llvm/ADT/DenseSet.h" #include "llvm/ADT/MapVector.h" #include "llvm/ADT/STLExtras.h" #include "llvm/ADT/SetVector.h" #include "llvm/ADT/SmallPtrSet.h" #include "llvm/ADT/SmallVector.h" #include "llvm/ADT/StringRef.h" #include "llvm/Analysis/BlockFrequencyInfo.h" #include "llvm/Analysis/BranchProbabilityInfo.h" #include "llvm/Analysis/IndirectCallPromotionAnalysis.h" #include "llvm/Analysis/LoopInfo.h" #include "llvm/Analysis/MemoryProfileInfo.h" #include "llvm/Analysis/ProfileSummaryInfo.h" #include "llvm/Analysis/StackSafetyAnalysis.h" #include "llvm/Analysis/TypeMetadataUtils.h" #include "llvm/IR/Attributes.h" #include "llvm/IR/BasicBlock.h" #include "llvm/IR/Constant.h" #include "llvm/IR/Constants.h" #include "llvm/IR/Dominators.h" #include "llvm/IR/Function.h" #include "llvm/IR/GlobalAlias.h" #include "llvm/IR/GlobalValue.h" #include "llvm/IR/GlobalVariable.h" #include "llvm/IR/Instructions.h" #include "llvm/IR/IntrinsicInst.h" #include "llvm/IR/Metadata.h" #include "llvm/IR/Module.h" #include "llvm/IR/ModuleSummaryIndex.h" #include "llvm/IR/Use.h" #include "llvm/IR/User.h" #include "llvm/InitializePasses.h" #include "llvm/Object/ModuleSymbolTable.h" #include "llvm/Object/SymbolicFile.h" #include "llvm/Pass.h" #include "llvm/Support/Casting.h" #include "llvm/Support/CommandLine.h" #include "llvm/Support/FileSystem.h" #include #include #include #include using namespace llvm; using namespace llvm::memprof; #define DEBUG_TYPE "module-summary-analysis" // Option to force edges cold which will block importing when the // -import-cold-multiplier is set to 0. Useful for debugging. namespace llvm { FunctionSummary::ForceSummaryHotnessType ForceSummaryEdgesCold = FunctionSummary::FSHT_None; } // namespace llvm static cl::opt FSEC( "force-summary-edges-cold", cl::Hidden, cl::location(ForceSummaryEdgesCold), cl::desc("Force all edges in the function summary to cold"), cl::values(clEnumValN(FunctionSummary::FSHT_None, "none", "None."), clEnumValN(FunctionSummary::FSHT_AllNonCritical, "all-non-critical", "All non-critical edges."), clEnumValN(FunctionSummary::FSHT_All, "all", "All edges."))); static cl::opt ModuleSummaryDotFile( "module-summary-dot-file", cl::Hidden, cl::value_desc("filename"), cl::desc("File to emit dot graph of new summary into")); extern cl::opt ScalePartialSampleProfileWorkingSetSize; // Walk through the operands of a given User via worklist iteration and populate // the set of GlobalValue references encountered. Invoked either on an // Instruction or a GlobalVariable (which walks its initializer). // Return true if any of the operands contains blockaddress. This is important // to know when computing summary for global var, because if global variable // references basic block address we can't import it separately from function // containing that basic block. For simplicity we currently don't import such // global vars at all. When importing function we aren't interested if any // instruction in it takes an address of any basic block, because instruction // can only take an address of basic block located in the same function. static bool findRefEdges(ModuleSummaryIndex &Index, const User *CurUser, SetVector &RefEdges, SmallPtrSet &Visited) { bool HasBlockAddress = false; SmallVector Worklist; if (Visited.insert(CurUser).second) Worklist.push_back(CurUser); while (!Worklist.empty()) { const User *U = Worklist.pop_back_val(); const auto *CB = dyn_cast(U); for (const auto &OI : U->operands()) { const User *Operand = dyn_cast(OI); if (!Operand) continue; if (isa(Operand)) { HasBlockAddress = true; continue; } if (auto *GV = dyn_cast(Operand)) { // We have a reference to a global value. This should be added to // the reference set unless it is a callee. Callees are handled // specially by WriteFunction and are added to a separate list. if (!(CB && CB->isCallee(&OI))) RefEdges.insert(Index.getOrInsertValueInfo(GV)); continue; } if (Visited.insert(Operand).second) Worklist.push_back(Operand); } } return HasBlockAddress; } static CalleeInfo::HotnessType getHotness(uint64_t ProfileCount, ProfileSummaryInfo *PSI) { if (!PSI) return CalleeInfo::HotnessType::Unknown; if (PSI->isHotCount(ProfileCount)) return CalleeInfo::HotnessType::Hot; if (PSI->isColdCount(ProfileCount)) return CalleeInfo::HotnessType::Cold; return CalleeInfo::HotnessType::None; } static bool isNonRenamableLocal(const GlobalValue &GV) { return GV.hasSection() && GV.hasLocalLinkage(); } /// Determine whether this call has all constant integer arguments (excluding /// "this") and summarize it to VCalls or ConstVCalls as appropriate. static void addVCallToSet(DevirtCallSite Call, GlobalValue::GUID Guid, SetVector &VCalls, SetVector &ConstVCalls) { std::vector Args; // Start from the second argument to skip the "this" pointer. for (auto &Arg : drop_begin(Call.CB.args())) { auto *CI = dyn_cast(Arg); if (!CI || CI->getBitWidth() > 64) { VCalls.insert({Guid, Call.Offset}); return; } Args.push_back(CI->getZExtValue()); } ConstVCalls.insert({{Guid, Call.Offset}, std::move(Args)}); } /// If this intrinsic call requires that we add information to the function /// summary, do so via the non-constant reference arguments. static void addIntrinsicToSummary( const CallInst *CI, SetVector &TypeTests, SetVector &TypeTestAssumeVCalls, SetVector &TypeCheckedLoadVCalls, SetVector &TypeTestAssumeConstVCalls, SetVector &TypeCheckedLoadConstVCalls, DominatorTree &DT) { switch (CI->getCalledFunction()->getIntrinsicID()) { case Intrinsic::type_test: case Intrinsic::public_type_test: { auto *TypeMDVal = cast(CI->getArgOperand(1)); auto *TypeId = dyn_cast(TypeMDVal->getMetadata()); if (!TypeId) break; GlobalValue::GUID Guid = GlobalValue::getGUID(TypeId->getString()); // Produce a summary from type.test intrinsics. We only summarize type.test // intrinsics that are used other than by an llvm.assume intrinsic. // Intrinsics that are assumed are relevant only to the devirtualization // pass, not the type test lowering pass. bool HasNonAssumeUses = llvm::any_of(CI->uses(), [](const Use &CIU) { return !isa(CIU.getUser()); }); if (HasNonAssumeUses) TypeTests.insert(Guid); SmallVector DevirtCalls; SmallVector Assumes; findDevirtualizableCallsForTypeTest(DevirtCalls, Assumes, CI, DT); for (auto &Call : DevirtCalls) addVCallToSet(Call, Guid, TypeTestAssumeVCalls, TypeTestAssumeConstVCalls); break; } case Intrinsic::type_checked_load_relative: case Intrinsic::type_checked_load: { auto *TypeMDVal = cast(CI->getArgOperand(2)); auto *TypeId = dyn_cast(TypeMDVal->getMetadata()); if (!TypeId) break; GlobalValue::GUID Guid = GlobalValue::getGUID(TypeId->getString()); SmallVector DevirtCalls; SmallVector LoadedPtrs; SmallVector Preds; bool HasNonCallUses = false; findDevirtualizableCallsForTypeCheckedLoad(DevirtCalls, LoadedPtrs, Preds, HasNonCallUses, CI, DT); // Any non-call uses of the result of llvm.type.checked.load will // prevent us from optimizing away the llvm.type.test. if (HasNonCallUses) TypeTests.insert(Guid); for (auto &Call : DevirtCalls) addVCallToSet(Call, Guid, TypeCheckedLoadVCalls, TypeCheckedLoadConstVCalls); break; } default: break; } } static bool isNonVolatileLoad(const Instruction *I) { if (const auto *LI = dyn_cast(I)) return !LI->isVolatile(); return false; } static bool isNonVolatileStore(const Instruction *I) { if (const auto *SI = dyn_cast(I)) return !SI->isVolatile(); return false; } // Returns true if the function definition must be unreachable. // // Note if this helper function returns true, `F` is guaranteed // to be unreachable; if it returns false, `F` might still // be unreachable but not covered by this helper function. static bool mustBeUnreachableFunction(const Function &F) { // A function must be unreachable if its entry block ends with an // 'unreachable'. assert(!F.isDeclaration()); return isa(F.getEntryBlock().getTerminator()); } static void computeFunctionSummary( ModuleSummaryIndex &Index, const Module &M, const Function &F, BlockFrequencyInfo *BFI, ProfileSummaryInfo *PSI, DominatorTree &DT, bool HasLocalsInUsedOrAsm, DenseSet &CantBePromoted, bool IsThinLTO, std::function GetSSICallback) { // Summary not currently supported for anonymous functions, they should // have been named. assert(F.hasName()); unsigned NumInsts = 0; // Map from callee ValueId to profile count. Used to accumulate profile // counts for all static calls to a given callee. MapVector, std::vector>> CallGraphEdges; SetVector RefEdges, LoadRefEdges, StoreRefEdges; SetVector TypeTests; SetVector TypeTestAssumeVCalls, TypeCheckedLoadVCalls; SetVector TypeTestAssumeConstVCalls, TypeCheckedLoadConstVCalls; ICallPromotionAnalysis ICallAnalysis; SmallPtrSet Visited; // Add personality function, prefix data and prologue data to function's ref // list. findRefEdges(Index, &F, RefEdges, Visited); std::vector NonVolatileLoads; std::vector NonVolatileStores; std::vector Callsites; std::vector Allocs; #ifndef NDEBUG DenseSet CallsThatMayHaveMemprofSummary; #endif bool HasInlineAsmMaybeReferencingInternal = false; bool HasIndirBranchToBlockAddress = false; bool HasUnknownCall = false; bool MayThrow = false; for (const BasicBlock &BB : F) { // We don't allow inlining of function with indirect branch to blockaddress. // If the blockaddress escapes the function, e.g., via a global variable, // inlining may lead to an invalid cross-function reference. So we shouldn't // import such function either. if (BB.hasAddressTaken()) { for (User *U : BlockAddress::get(const_cast(&BB))->users()) if (!isa(*U)) { HasIndirBranchToBlockAddress = true; break; } } for (const Instruction &I : BB) { if (I.isDebugOrPseudoInst()) continue; ++NumInsts; // Regular LTO module doesn't participate in ThinLTO import, // so no reference from it can be read/writeonly, since this // would require importing variable as local copy if (IsThinLTO) { if (isNonVolatileLoad(&I)) { // Postpone processing of non-volatile load instructions // See comments below Visited.insert(&I); NonVolatileLoads.push_back(&I); continue; } else if (isNonVolatileStore(&I)) { Visited.insert(&I); NonVolatileStores.push_back(&I); // All references from second operand of store (destination address) // can be considered write-only if they're not referenced by any // non-store instruction. References from first operand of store // (stored value) can't be treated either as read- or as write-only // so we add them to RefEdges as we do with all other instructions // except non-volatile load. Value *Stored = I.getOperand(0); if (auto *GV = dyn_cast(Stored)) // findRefEdges will try to examine GV operands, so instead // of calling it we should add GV to RefEdges directly. RefEdges.insert(Index.getOrInsertValueInfo(GV)); else if (auto *U = dyn_cast(Stored)) findRefEdges(Index, U, RefEdges, Visited); continue; } } findRefEdges(Index, &I, RefEdges, Visited); const auto *CB = dyn_cast(&I); if (!CB) { if (I.mayThrow()) MayThrow = true; continue; } const auto *CI = dyn_cast(&I); // Since we don't know exactly which local values are referenced in inline // assembly, conservatively mark the function as possibly referencing // a local value from inline assembly to ensure we don't export a // reference (which would require renaming and promotion of the // referenced value). if (HasLocalsInUsedOrAsm && CI && CI->isInlineAsm()) HasInlineAsmMaybeReferencingInternal = true; auto *CalledValue = CB->getCalledOperand(); auto *CalledFunction = CB->getCalledFunction(); if (CalledValue && !CalledFunction) { CalledValue = CalledValue->stripPointerCasts(); // Stripping pointer casts can reveal a called function. CalledFunction = dyn_cast(CalledValue); } // Check if this is an alias to a function. If so, get the // called aliasee for the checks below. if (auto *GA = dyn_cast(CalledValue)) { assert(!CalledFunction && "Expected null called function in callsite for alias"); CalledFunction = dyn_cast(GA->getAliaseeObject()); } // Check if this is a direct call to a known function or a known // intrinsic, or an indirect call with profile data. if (CalledFunction) { if (CI && CalledFunction->isIntrinsic()) { addIntrinsicToSummary( CI, TypeTests, TypeTestAssumeVCalls, TypeCheckedLoadVCalls, TypeTestAssumeConstVCalls, TypeCheckedLoadConstVCalls, DT); continue; } // We should have named any anonymous globals assert(CalledFunction->hasName()); auto ScaledCount = PSI->getProfileCount(*CB, BFI); auto Hotness = ScaledCount ? getHotness(*ScaledCount, PSI) : CalleeInfo::HotnessType::Unknown; if (ForceSummaryEdgesCold != FunctionSummary::FSHT_None) Hotness = CalleeInfo::HotnessType::Cold; // Use the original CalledValue, in case it was an alias. We want // to record the call edge to the alias in that case. Eventually // an alias summary will be created to associate the alias and // aliasee. auto &ValueInfo = CallGraphEdges[Index.getOrInsertValueInfo( cast(CalledValue))]; ValueInfo.updateHotness(Hotness); // Add the relative block frequency to CalleeInfo if there is no profile // information. if (BFI != nullptr && Hotness == CalleeInfo::HotnessType::Unknown) { uint64_t BBFreq = BFI->getBlockFreq(&BB).getFrequency(); uint64_t EntryFreq = BFI->getEntryFreq(); ValueInfo.updateRelBlockFreq(BBFreq, EntryFreq); } } else { HasUnknownCall = true; // Skip inline assembly calls. if (CI && CI->isInlineAsm()) continue; // Skip direct calls. if (!CalledValue || isa(CalledValue)) continue; // Check if the instruction has a callees metadata. If so, add callees // to CallGraphEdges to reflect the references from the metadata, and // to enable importing for subsequent indirect call promotion and // inlining. if (auto *MD = I.getMetadata(LLVMContext::MD_callees)) { for (const auto &Op : MD->operands()) { Function *Callee = mdconst::extract_or_null(Op); if (Callee) CallGraphEdges[Index.getOrInsertValueInfo(Callee)]; } } uint32_t NumVals, NumCandidates; uint64_t TotalCount; auto CandidateProfileData = ICallAnalysis.getPromotionCandidatesForInstruction( &I, NumVals, TotalCount, NumCandidates); for (const auto &Candidate : CandidateProfileData) CallGraphEdges[Index.getOrInsertValueInfo(Candidate.Value)] .updateHotness(getHotness(Candidate.Count, PSI)); } // Summarize memprof related metadata. This is only needed for ThinLTO. if (!IsThinLTO) continue; // TODO: Skip indirect calls for now. Need to handle these better, likely // by creating multiple Callsites, one per target, then speculatively // devirtualize while applying clone info in the ThinLTO backends. This // will also be important because we will have a different set of clone // versions per target. This handling needs to match that in the ThinLTO // backend so we handle things consistently for matching of callsite // summaries to instructions. if (!CalledFunction) continue; // Ensure we keep this analysis in sync with the handling in the ThinLTO // backend (see MemProfContextDisambiguation::applyImport). Save this call // so that we can skip it in checking the reverse case later. assert(mayHaveMemprofSummary(CB)); #ifndef NDEBUG CallsThatMayHaveMemprofSummary.insert(CB); #endif // Compute the list of stack ids first (so we can trim them from the stack // ids on any MIBs). CallStack InstCallsite( I.getMetadata(LLVMContext::MD_callsite)); auto *MemProfMD = I.getMetadata(LLVMContext::MD_memprof); if (MemProfMD) { std::vector MIBs; for (auto &MDOp : MemProfMD->operands()) { auto *MIBMD = cast(MDOp); MDNode *StackNode = getMIBStackNode(MIBMD); assert(StackNode); SmallVector StackIdIndices; CallStack StackContext(StackNode); // Collapse out any on the allocation call (inlining). for (auto ContextIter = StackContext.beginAfterSharedPrefix(InstCallsite); ContextIter != StackContext.end(); ++ContextIter) { unsigned StackIdIdx = Index.addOrGetStackIdIndex(*ContextIter); // If this is a direct recursion, simply skip the duplicate // entries. If this is mutual recursion, handling is left to // the LTO link analysis client. if (StackIdIndices.empty() || StackIdIndices.back() != StackIdIdx) StackIdIndices.push_back(StackIdIdx); } MIBs.push_back( MIBInfo(getMIBAllocType(MIBMD), std::move(StackIdIndices))); } Allocs.push_back(AllocInfo(std::move(MIBs))); } else if (!InstCallsite.empty()) { SmallVector StackIdIndices; for (auto StackId : InstCallsite) StackIdIndices.push_back(Index.addOrGetStackIdIndex(StackId)); // Use the original CalledValue, in case it was an alias. We want // to record the call edge to the alias in that case. Eventually // an alias summary will be created to associate the alias and // aliasee. auto CalleeValueInfo = Index.getOrInsertValueInfo(cast(CalledValue)); Callsites.push_back({CalleeValueInfo, StackIdIndices}); } } } if (PSI->hasPartialSampleProfile() && ScalePartialSampleProfileWorkingSetSize) Index.addBlockCount(F.size()); std::vector Refs; if (IsThinLTO) { auto AddRefEdges = [&](const std::vector &Instrs, SetVector &Edges, SmallPtrSet &Cache) { for (const auto *I : Instrs) { Cache.erase(I); findRefEdges(Index, I, Edges, Cache); } }; // By now we processed all instructions in a function, except // non-volatile loads and non-volatile value stores. Let's find // ref edges for both of instruction sets AddRefEdges(NonVolatileLoads, LoadRefEdges, Visited); // We can add some values to the Visited set when processing load // instructions which are also used by stores in NonVolatileStores. // For example this can happen if we have following code: // // store %Derived* @foo, %Derived** bitcast (%Base** @bar to %Derived**) // %42 = load %Derived*, %Derived** bitcast (%Base** @bar to %Derived**) // // After processing loads we'll add bitcast to the Visited set, and if // we use the same set while processing stores, we'll never see store // to @bar and @bar will be mistakenly treated as readonly. SmallPtrSet StoreCache; AddRefEdges(NonVolatileStores, StoreRefEdges, StoreCache); // If both load and store instruction reference the same variable // we won't be able to optimize it. Add all such reference edges // to RefEdges set. for (const auto &VI : StoreRefEdges) if (LoadRefEdges.remove(VI)) RefEdges.insert(VI); unsigned RefCnt = RefEdges.size(); // All new reference edges inserted in two loops below are either // read or write only. They will be grouped in the end of RefEdges // vector, so we can use a single integer value to identify them. for (const auto &VI : LoadRefEdges) RefEdges.insert(VI); unsigned FirstWORef = RefEdges.size(); for (const auto &VI : StoreRefEdges) RefEdges.insert(VI); Refs = RefEdges.takeVector(); for (; RefCnt < FirstWORef; ++RefCnt) Refs[RefCnt].setReadOnly(); for (; RefCnt < Refs.size(); ++RefCnt) Refs[RefCnt].setWriteOnly(); } else { Refs = RefEdges.takeVector(); } // Explicit add hot edges to enforce importing for designated GUIDs for // sample PGO, to enable the same inlines as the profiled optimized binary. for (auto &I : F.getImportGUIDs()) CallGraphEdges[Index.getOrInsertValueInfo(I)].updateHotness( ForceSummaryEdgesCold == FunctionSummary::FSHT_All ? CalleeInfo::HotnessType::Cold : CalleeInfo::HotnessType::Critical); #ifndef NDEBUG // Make sure that all calls we decided could not have memprof summaries get a // false value for mayHaveMemprofSummary, to ensure that this handling remains // in sync with the ThinLTO backend handling. if (IsThinLTO) { for (const BasicBlock &BB : F) { for (const Instruction &I : BB) { const auto *CB = dyn_cast(&I); if (!CB) continue; // We already checked these above. if (CallsThatMayHaveMemprofSummary.count(CB)) continue; assert(!mayHaveMemprofSummary(CB)); } } } #endif bool NonRenamableLocal = isNonRenamableLocal(F); bool NotEligibleForImport = NonRenamableLocal || HasInlineAsmMaybeReferencingInternal || HasIndirBranchToBlockAddress; GlobalValueSummary::GVFlags Flags( F.getLinkage(), F.getVisibility(), NotEligibleForImport, /* Live = */ false, F.isDSOLocal(), F.canBeOmittedFromSymbolTable()); FunctionSummary::FFlags FunFlags{ F.doesNotAccessMemory(), F.onlyReadsMemory() && !F.doesNotAccessMemory(), F.hasFnAttribute(Attribute::NoRecurse), F.returnDoesNotAlias(), // FIXME: refactor this to use the same code that inliner is using. // Don't try to import functions with noinline attribute. F.getAttributes().hasFnAttr(Attribute::NoInline), F.hasFnAttribute(Attribute::AlwaysInline), F.hasFnAttribute(Attribute::NoUnwind), MayThrow, HasUnknownCall, mustBeUnreachableFunction(F)}; std::vector ParamAccesses; if (auto *SSI = GetSSICallback(F)) ParamAccesses = SSI->getParamAccesses(Index); auto FuncSummary = std::make_unique( Flags, NumInsts, FunFlags, /*EntryCount=*/0, std::move(Refs), CallGraphEdges.takeVector(), TypeTests.takeVector(), TypeTestAssumeVCalls.takeVector(), TypeCheckedLoadVCalls.takeVector(), TypeTestAssumeConstVCalls.takeVector(), TypeCheckedLoadConstVCalls.takeVector(), std::move(ParamAccesses), std::move(Callsites), std::move(Allocs)); if (NonRenamableLocal) CantBePromoted.insert(F.getGUID()); Index.addGlobalValueSummary(F, std::move(FuncSummary)); } /// Find function pointers referenced within the given vtable initializer /// (or subset of an initializer) \p I. The starting offset of \p I within /// the vtable initializer is \p StartingOffset. Any discovered function /// pointers are added to \p VTableFuncs along with their cumulative offset /// within the initializer. static void findFuncPointers(const Constant *I, uint64_t StartingOffset, const Module &M, ModuleSummaryIndex &Index, VTableFuncList &VTableFuncs) { // First check if this is a function pointer. if (I->getType()->isPointerTy()) { auto C = I->stripPointerCasts(); auto A = dyn_cast(C); if (isa(C) || (A && isa(A->getAliasee()))) { auto GV = dyn_cast(C); assert(GV); // We can disregard __cxa_pure_virtual as a possible call target, as // calls to pure virtuals are UB. if (GV && GV->getName() != "__cxa_pure_virtual") VTableFuncs.push_back({Index.getOrInsertValueInfo(GV), StartingOffset}); return; } } // Walk through the elements in the constant struct or array and recursively // look for virtual function pointers. const DataLayout &DL = M.getDataLayout(); if (auto *C = dyn_cast(I)) { StructType *STy = dyn_cast(C->getType()); assert(STy); const StructLayout *SL = DL.getStructLayout(C->getType()); for (auto EI : llvm::enumerate(STy->elements())) { auto Offset = SL->getElementOffset(EI.index()); unsigned Op = SL->getElementContainingOffset(Offset); findFuncPointers(cast(I->getOperand(Op)), StartingOffset + Offset, M, Index, VTableFuncs); } } else if (auto *C = dyn_cast(I)) { ArrayType *ATy = C->getType(); Type *EltTy = ATy->getElementType(); uint64_t EltSize = DL.getTypeAllocSize(EltTy); for (unsigned i = 0, e = ATy->getNumElements(); i != e; ++i) { findFuncPointers(cast(I->getOperand(i)), StartingOffset + i * EltSize, M, Index, VTableFuncs); } } } // Identify the function pointers referenced by vtable definition \p V. static void computeVTableFuncs(ModuleSummaryIndex &Index, const GlobalVariable &V, const Module &M, VTableFuncList &VTableFuncs) { if (!V.isConstant()) return; findFuncPointers(V.getInitializer(), /*StartingOffset=*/0, M, Index, VTableFuncs); #ifndef NDEBUG // Validate that the VTableFuncs list is ordered by offset. uint64_t PrevOffset = 0; for (auto &P : VTableFuncs) { // The findVFuncPointers traversal should have encountered the // functions in offset order. We need to use ">=" since PrevOffset // starts at 0. assert(P.VTableOffset >= PrevOffset); PrevOffset = P.VTableOffset; } #endif } /// Record vtable definition \p V for each type metadata it references. static void recordTypeIdCompatibleVtableReferences(ModuleSummaryIndex &Index, const GlobalVariable &V, SmallVectorImpl &Types) { for (MDNode *Type : Types) { auto TypeID = Type->getOperand(1).get(); uint64_t Offset = cast( cast(Type->getOperand(0))->getValue()) ->getZExtValue(); if (auto *TypeId = dyn_cast(TypeID)) Index.getOrInsertTypeIdCompatibleVtableSummary(TypeId->getString()) .push_back({Offset, Index.getOrInsertValueInfo(&V)}); } } static void computeVariableSummary(ModuleSummaryIndex &Index, const GlobalVariable &V, DenseSet &CantBePromoted, const Module &M, SmallVectorImpl &Types) { SetVector RefEdges; SmallPtrSet Visited; bool HasBlockAddress = findRefEdges(Index, &V, RefEdges, Visited); bool NonRenamableLocal = isNonRenamableLocal(V); GlobalValueSummary::GVFlags Flags( V.getLinkage(), V.getVisibility(), NonRenamableLocal, /* Live = */ false, V.isDSOLocal(), V.canBeOmittedFromSymbolTable()); VTableFuncList VTableFuncs; // If splitting is not enabled, then we compute the summary information // necessary for index-based whole program devirtualization. if (!Index.enableSplitLTOUnit()) { Types.clear(); V.getMetadata(LLVMContext::MD_type, Types); if (!Types.empty()) { // Identify the function pointers referenced by this vtable definition. computeVTableFuncs(Index, V, M, VTableFuncs); // Record this vtable definition for each type metadata it references. recordTypeIdCompatibleVtableReferences(Index, V, Types); } } // Don't mark variables we won't be able to internalize as read/write-only. bool CanBeInternalized = !V.hasComdat() && !V.hasAppendingLinkage() && !V.isInterposable() && !V.hasAvailableExternallyLinkage() && !V.hasDLLExportStorageClass(); bool Constant = V.isConstant(); GlobalVarSummary::GVarFlags VarFlags(CanBeInternalized, Constant ? false : CanBeInternalized, Constant, V.getVCallVisibility()); auto GVarSummary = std::make_unique(Flags, VarFlags, RefEdges.takeVector()); if (NonRenamableLocal) CantBePromoted.insert(V.getGUID()); if (HasBlockAddress) GVarSummary->setNotEligibleToImport(); if (!VTableFuncs.empty()) GVarSummary->setVTableFuncs(VTableFuncs); Index.addGlobalValueSummary(V, std::move(GVarSummary)); } static void computeAliasSummary(ModuleSummaryIndex &Index, const GlobalAlias &A, DenseSet &CantBePromoted) { // Skip summary for indirect function aliases as summary for aliasee will not // be emitted. const GlobalObject *Aliasee = A.getAliaseeObject(); if (isa(Aliasee)) return; bool NonRenamableLocal = isNonRenamableLocal(A); GlobalValueSummary::GVFlags Flags( A.getLinkage(), A.getVisibility(), NonRenamableLocal, /* Live = */ false, A.isDSOLocal(), A.canBeOmittedFromSymbolTable()); auto AS = std::make_unique(Flags); auto AliaseeVI = Index.getValueInfo(Aliasee->getGUID()); assert(AliaseeVI && "Alias expects aliasee summary to be available"); assert(AliaseeVI.getSummaryList().size() == 1 && "Expected a single entry per aliasee in per-module index"); AS->setAliasee(AliaseeVI, AliaseeVI.getSummaryList()[0].get()); if (NonRenamableLocal) CantBePromoted.insert(A.getGUID()); Index.addGlobalValueSummary(A, std::move(AS)); } // Set LiveRoot flag on entries matching the given value name. static void setLiveRoot(ModuleSummaryIndex &Index, StringRef Name) { if (ValueInfo VI = Index.getValueInfo(GlobalValue::getGUID(Name))) for (const auto &Summary : VI.getSummaryList()) Summary->setLive(true); } ModuleSummaryIndex llvm::buildModuleSummaryIndex( const Module &M, std::function GetBFICallback, ProfileSummaryInfo *PSI, std::function GetSSICallback) { assert(PSI); bool EnableSplitLTOUnit = false; bool UnifiedLTO = false; if (auto *MD = mdconst::extract_or_null( M.getModuleFlag("EnableSplitLTOUnit"))) EnableSplitLTOUnit = MD->getZExtValue(); if (auto *MD = mdconst::extract_or_null(M.getModuleFlag("UnifiedLTO"))) UnifiedLTO = MD->getZExtValue(); ModuleSummaryIndex Index(/*HaveGVs=*/true, EnableSplitLTOUnit, UnifiedLTO); // Identify the local values in the llvm.used and llvm.compiler.used sets, // which should not be exported as they would then require renaming and // promotion, but we may have opaque uses e.g. in inline asm. We collect them // here because we use this information to mark functions containing inline // assembly calls as not importable. SmallPtrSet LocalsUsed; SmallVector Used; // First collect those in the llvm.used set. collectUsedGlobalVariables(M, Used, /*CompilerUsed=*/false); // Next collect those in the llvm.compiler.used set. collectUsedGlobalVariables(M, Used, /*CompilerUsed=*/true); DenseSet CantBePromoted; for (auto *V : Used) { if (V->hasLocalLinkage()) { LocalsUsed.insert(V); CantBePromoted.insert(V->getGUID()); } } bool HasLocalInlineAsmSymbol = false; if (!M.getModuleInlineAsm().empty()) { // Collect the local values defined by module level asm, and set up // summaries for these symbols so that they can be marked as NoRename, // to prevent export of any use of them in regular IR that would require // renaming within the module level asm. Note we don't need to create a // summary for weak or global defs, as they don't need to be flagged as // NoRename, and defs in module level asm can't be imported anyway. // Also, any values used but not defined within module level asm should // be listed on the llvm.used or llvm.compiler.used global and marked as // referenced from there. ModuleSymbolTable::CollectAsmSymbols( M, [&](StringRef Name, object::BasicSymbolRef::Flags Flags) { // Symbols not marked as Weak or Global are local definitions. if (Flags & (object::BasicSymbolRef::SF_Weak | object::BasicSymbolRef::SF_Global)) return; HasLocalInlineAsmSymbol = true; GlobalValue *GV = M.getNamedValue(Name); if (!GV) return; assert(GV->isDeclaration() && "Def in module asm already has definition"); GlobalValueSummary::GVFlags GVFlags( GlobalValue::InternalLinkage, GlobalValue::DefaultVisibility, /* NotEligibleToImport = */ true, /* Live = */ true, /* Local */ GV->isDSOLocal(), GV->canBeOmittedFromSymbolTable()); CantBePromoted.insert(GV->getGUID()); // Create the appropriate summary type. if (Function *F = dyn_cast(GV)) { std::unique_ptr Summary = std::make_unique( GVFlags, /*InstCount=*/0, FunctionSummary::FFlags{ F->hasFnAttribute(Attribute::ReadNone), F->hasFnAttribute(Attribute::ReadOnly), F->hasFnAttribute(Attribute::NoRecurse), F->returnDoesNotAlias(), /* NoInline = */ false, F->hasFnAttribute(Attribute::AlwaysInline), F->hasFnAttribute(Attribute::NoUnwind), /* MayThrow */ true, /* HasUnknownCall */ true, /* MustBeUnreachable */ false}, /*EntryCount=*/0, ArrayRef{}, ArrayRef{}, ArrayRef{}, ArrayRef{}, ArrayRef{}, ArrayRef{}, ArrayRef{}, ArrayRef{}, ArrayRef{}, ArrayRef{}); Index.addGlobalValueSummary(*GV, std::move(Summary)); } else { std::unique_ptr Summary = std::make_unique( GVFlags, GlobalVarSummary::GVarFlags( false, false, cast(GV)->isConstant(), GlobalObject::VCallVisibilityPublic), ArrayRef{}); Index.addGlobalValueSummary(*GV, std::move(Summary)); } }); } bool IsThinLTO = true; if (auto *MD = mdconst::extract_or_null(M.getModuleFlag("ThinLTO"))) IsThinLTO = MD->getZExtValue(); // Compute summaries for all functions defined in module, and save in the // index. for (const auto &F : M) { if (F.isDeclaration()) continue; DominatorTree DT(const_cast(F)); BlockFrequencyInfo *BFI = nullptr; std::unique_ptr BFIPtr; if (GetBFICallback) BFI = GetBFICallback(F); else if (F.hasProfileData()) { LoopInfo LI{DT}; BranchProbabilityInfo BPI{F, LI}; BFIPtr = std::make_unique(F, BPI, LI); BFI = BFIPtr.get(); } computeFunctionSummary(Index, M, F, BFI, PSI, DT, !LocalsUsed.empty() || HasLocalInlineAsmSymbol, CantBePromoted, IsThinLTO, GetSSICallback); } // Compute summaries for all variables defined in module, and save in the // index. SmallVector Types; for (const GlobalVariable &G : M.globals()) { if (G.isDeclaration()) continue; computeVariableSummary(Index, G, CantBePromoted, M, Types); } // Compute summaries for all aliases defined in module, and save in the // index. for (const GlobalAlias &A : M.aliases()) computeAliasSummary(Index, A, CantBePromoted); // Iterate through ifuncs, set their resolvers all alive. for (const GlobalIFunc &I : M.ifuncs()) { I.applyAlongResolverPath([&Index](const GlobalValue &GV) { Index.getGlobalValueSummary(GV)->setLive(true); }); } for (auto *V : LocalsUsed) { auto *Summary = Index.getGlobalValueSummary(*V); assert(Summary && "Missing summary for global value"); Summary->setNotEligibleToImport(); } // The linker doesn't know about these LLVM produced values, so we need // to flag them as live in the index to ensure index-based dead value // analysis treats them as live roots of the analysis. setLiveRoot(Index, "llvm.used"); setLiveRoot(Index, "llvm.compiler.used"); setLiveRoot(Index, "llvm.global_ctors"); setLiveRoot(Index, "llvm.global_dtors"); setLiveRoot(Index, "llvm.global.annotations"); for (auto &GlobalList : Index) { // Ignore entries for references that are undefined in the current module. if (GlobalList.second.SummaryList.empty()) continue; assert(GlobalList.second.SummaryList.size() == 1 && "Expected module's index to have one summary per GUID"); auto &Summary = GlobalList.second.SummaryList[0]; if (!IsThinLTO) { Summary->setNotEligibleToImport(); continue; } bool AllRefsCanBeExternallyReferenced = llvm::all_of(Summary->refs(), [&](const ValueInfo &VI) { return !CantBePromoted.count(VI.getGUID()); }); if (!AllRefsCanBeExternallyReferenced) { Summary->setNotEligibleToImport(); continue; } if (auto *FuncSummary = dyn_cast(Summary.get())) { bool AllCallsCanBeExternallyReferenced = llvm::all_of( FuncSummary->calls(), [&](const FunctionSummary::EdgeTy &Edge) { return !CantBePromoted.count(Edge.first.getGUID()); }); if (!AllCallsCanBeExternallyReferenced) Summary->setNotEligibleToImport(); } } if (!ModuleSummaryDotFile.empty()) { std::error_code EC; raw_fd_ostream OSDot(ModuleSummaryDotFile, EC, sys::fs::OpenFlags::OF_None); if (EC) report_fatal_error(Twine("Failed to open dot file ") + ModuleSummaryDotFile + ": " + EC.message() + "\n"); Index.exportToDot(OSDot, {}); } return Index; } AnalysisKey ModuleSummaryIndexAnalysis::Key; ModuleSummaryIndex ModuleSummaryIndexAnalysis::run(Module &M, ModuleAnalysisManager &AM) { ProfileSummaryInfo &PSI = AM.getResult(M); auto &FAM = AM.getResult(M).getManager(); bool NeedSSI = needsParamAccessSummary(M); return buildModuleSummaryIndex( M, [&FAM](const Function &F) { return &FAM.getResult( *const_cast(&F)); }, &PSI, [&FAM, NeedSSI](const Function &F) -> const StackSafetyInfo * { return NeedSSI ? &FAM.getResult( const_cast(F)) : nullptr; }); } char ModuleSummaryIndexWrapperPass::ID = 0; INITIALIZE_PASS_BEGIN(ModuleSummaryIndexWrapperPass, "module-summary-analysis", "Module Summary Analysis", false, true) INITIALIZE_PASS_DEPENDENCY(BlockFrequencyInfoWrapperPass) INITIALIZE_PASS_DEPENDENCY(ProfileSummaryInfoWrapperPass) INITIALIZE_PASS_DEPENDENCY(StackSafetyInfoWrapperPass) INITIALIZE_PASS_END(ModuleSummaryIndexWrapperPass, "module-summary-analysis", "Module Summary Analysis", false, true) ModulePass *llvm::createModuleSummaryIndexWrapperPass() { return new ModuleSummaryIndexWrapperPass(); } ModuleSummaryIndexWrapperPass::ModuleSummaryIndexWrapperPass() : ModulePass(ID) { initializeModuleSummaryIndexWrapperPassPass(*PassRegistry::getPassRegistry()); } bool ModuleSummaryIndexWrapperPass::runOnModule(Module &M) { auto *PSI = &getAnalysis().getPSI(); bool NeedSSI = needsParamAccessSummary(M); Index.emplace(buildModuleSummaryIndex( M, [this](const Function &F) { return &(this->getAnalysis( *const_cast(&F)) .getBFI()); }, PSI, [&](const Function &F) -> const StackSafetyInfo * { return NeedSSI ? &getAnalysis( const_cast(F)) .getResult() : nullptr; })); return false; } bool ModuleSummaryIndexWrapperPass::doFinalization(Module &M) { Index.reset(); return false; } void ModuleSummaryIndexWrapperPass::getAnalysisUsage(AnalysisUsage &AU) const { AU.setPreservesAll(); AU.addRequired(); AU.addRequired(); AU.addRequired(); } char ImmutableModuleSummaryIndexWrapperPass::ID = 0; ImmutableModuleSummaryIndexWrapperPass::ImmutableModuleSummaryIndexWrapperPass( const ModuleSummaryIndex *Index) : ImmutablePass(ID), Index(Index) { initializeImmutableModuleSummaryIndexWrapperPassPass( *PassRegistry::getPassRegistry()); } void ImmutableModuleSummaryIndexWrapperPass::getAnalysisUsage( AnalysisUsage &AU) const { AU.setPreservesAll(); } ImmutablePass *llvm::createImmutableModuleSummaryIndexWrapperPass( const ModuleSummaryIndex *Index) { return new ImmutableModuleSummaryIndexWrapperPass(Index); } INITIALIZE_PASS(ImmutableModuleSummaryIndexWrapperPass, "module-summary-info", "Module summary info", false, true) bool llvm::mayHaveMemprofSummary(const CallBase *CB) { if (!CB) return false; if (CB->isDebugOrPseudoInst()) return false; auto *CI = dyn_cast(CB); auto *CalledValue = CB->getCalledOperand(); auto *CalledFunction = CB->getCalledFunction(); if (CalledValue && !CalledFunction) { CalledValue = CalledValue->stripPointerCasts(); // Stripping pointer casts can reveal a called function. CalledFunction = dyn_cast(CalledValue); } // Check if this is an alias to a function. If so, get the // called aliasee for the checks below. if (auto *GA = dyn_cast(CalledValue)) { assert(!CalledFunction && "Expected null called function in callsite for alias"); CalledFunction = dyn_cast(GA->getAliaseeObject()); } // Check if this is a direct call to a known function or a known // intrinsic, or an indirect call with profile data. if (CalledFunction) { if (CI && CalledFunction->isIntrinsic()) return false; } else { // TODO: For now skip indirect calls. See comments in // computeFunctionSummary for what is needed to handle this. return false; } return true; }