PassBuilderPipelines.cpp

Go to the documentation of this file.

//===- Construction of pass pipelines -------------------------------------===//
//
// 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 provides the implementation of the PassBuilder based on our
/// static pass registry as well as related functionality. It also provides
/// helpers to aid in analyzing, debugging, and testing passes and pass
/// pipelines.
///
//===----------------------------------------------------------------------===//
 
#include "PassesOptions.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/Analysis/AliasAnalysis.h"
#include "llvm/Analysis/BasicAliasAnalysis.h"
#include "llvm/Analysis/CGSCCPassManager.h"
#include "llvm/Analysis/FunctionPropertiesAnalysis.h"
#include "llvm/Analysis/GlobalsModRef.h"
#include "llvm/Analysis/InlineAdvisor.h"
#include "llvm/Analysis/InstCount.h"
#include "llvm/Analysis/ProfileSummaryInfo.h"
#include "llvm/Analysis/ScopedNoAliasAA.h"
#include "llvm/Analysis/TypeBasedAliasAnalysis.h"
#include "llvm/IR/PassManager.h"
#include "llvm/IR/Verifier.h"
#include "llvm/Pass.h"
#include "llvm/Passes/OptimizationLevel.h"
#include "llvm/Passes/PassBuilder.h"
#include "llvm/Passes/TriggerCrashPasses.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/ErrorHandling.h"
#include "llvm/Support/PGOOptions.h"
#include "llvm/Target/TargetMachine.h"
#include "llvm/Transforms/AggressiveInstCombine/AggressiveInstCombine.h"
#include "llvm/Transforms/Coroutines/CoroAnnotationElide.h"
#include "llvm/Transforms/Coroutines/CoroCleanup.h"
#include "llvm/Transforms/Coroutines/CoroConditionalWrapper.h"
#include "llvm/Transforms/Coroutines/CoroEarly.h"
#include "llvm/Transforms/Coroutines/CoroElide.h"
#include "llvm/Transforms/Coroutines/CoroSplit.h"
#include "llvm/Transforms/HipStdPar/HipStdPar.h"
#include "llvm/Transforms/IPO/AlwaysInliner.h"
#include "llvm/Transforms/IPO/Annotation2Metadata.h"
#include "llvm/Transforms/IPO/ArgumentPromotion.h"
#include "llvm/Transforms/IPO/Attributor.h"
#include "llvm/Transforms/IPO/CalledValuePropagation.h"
#include "llvm/Transforms/IPO/ConstantMerge.h"
#include "llvm/Transforms/IPO/CrossDSOCFI.h"
#include "llvm/Transforms/IPO/DeadArgumentElimination.h"
#include "llvm/Transforms/IPO/ElimAvailExtern.h"
#include "llvm/Transforms/IPO/EmbedBitcodePass.h"
#include "llvm/Transforms/IPO/ExpandVariadics.h"
#include "llvm/Transforms/IPO/FatLTOCleanup.h"
#include "llvm/Transforms/IPO/ForceFunctionAttrs.h"
#include "llvm/Transforms/IPO/FunctionAttrs.h"
#include "llvm/Transforms/IPO/GlobalDCE.h"
#include "llvm/Transforms/IPO/GlobalOpt.h"
#include "llvm/Transforms/IPO/GlobalSplit.h"
#include "llvm/Transforms/IPO/HotColdSplitting.h"
#include "llvm/Transforms/IPO/InferFunctionAttrs.h"
#include "llvm/Transforms/IPO/Inliner.h"
#include "llvm/Transforms/IPO/Instrumentor.h"
#include "llvm/Transforms/IPO/LowerTypeTests.h"
#include "llvm/Transforms/IPO/MemProfContextDisambiguation.h"
#include "llvm/Transforms/IPO/MergeFunctions.h"
#include "llvm/Transforms/IPO/ModuleInliner.h"
#include "llvm/Transforms/IPO/OpenMPOpt.h"
#include "llvm/Transforms/IPO/PartialInlining.h"
#include "llvm/Transforms/IPO/SCCP.h"
#include "llvm/Transforms/IPO/SampleProfile.h"
#include "llvm/Transforms/IPO/SampleProfileProbe.h"
#include "llvm/Transforms/IPO/WholeProgramDevirt.h"
#include "llvm/Transforms/InstCombine/InstCombine.h"
#include "llvm/Transforms/Instrumentation/AllocToken.h"
#include "llvm/Transforms/Instrumentation/CGProfile.h"
#include "llvm/Transforms/Instrumentation/ControlHeightReduction.h"
#include "llvm/Transforms/Instrumentation/InstrProfiling.h"
#include "llvm/Transforms/Instrumentation/MemProfInstrumentation.h"
#include "llvm/Transforms/Instrumentation/MemProfUse.h"
#include "llvm/Transforms/Instrumentation/PGOCtxProfFlattening.h"
#include "llvm/Transforms/Instrumentation/PGOCtxProfLowering.h"
#include "llvm/Transforms/Instrumentation/PGOForceFunctionAttrs.h"
#include "llvm/Transforms/Instrumentation/PGOInstrumentation.h"
#include "llvm/Transforms/Scalar/ADCE.h"
#include "llvm/Transforms/Scalar/AlignmentFromAssumptions.h"
#include "llvm/Transforms/Scalar/AnnotationRemarks.h"
#include "llvm/Transforms/Scalar/BDCE.h"
#include "llvm/Transforms/Scalar/CallSiteSplitting.h"
#include "llvm/Transforms/Scalar/ConstraintElimination.h"
#include "llvm/Transforms/Scalar/CorrelatedValuePropagation.h"
#include "llvm/Transforms/Scalar/DFAJumpThreading.h"
#include "llvm/Transforms/Scalar/DeadStoreElimination.h"
#include "llvm/Transforms/Scalar/DivRemPairs.h"
#include "llvm/Transforms/Scalar/DropUnnecessaryAssumes.h"
#include "llvm/Transforms/Scalar/EarlyCSE.h"
#include "llvm/Transforms/Scalar/ExpandMemCmp.h"
#include "llvm/Transforms/Scalar/Float2Int.h"
#include "llvm/Transforms/Scalar/GVN.h"
#include "llvm/Transforms/Scalar/GVNHoist.h"
#include "llvm/Transforms/Scalar/GVNSink.h"
#include "llvm/Transforms/Scalar/IndVarSimplify.h"
#include "llvm/Transforms/Scalar/InferAlignment.h"
#include "llvm/Transforms/Scalar/InstSimplifyPass.h"
#include "llvm/Transforms/Scalar/JumpTableToSwitch.h"
#include "llvm/Transforms/Scalar/JumpThreading.h"
#include "llvm/Transforms/Scalar/LICM.h"
#include "llvm/Transforms/Scalar/LoopDeletion.h"
#include "llvm/Transforms/Scalar/LoopDistribute.h"
#include "llvm/Transforms/Scalar/LoopFlatten.h"
#include "llvm/Transforms/Scalar/LoopFuse.h"
#include "llvm/Transforms/Scalar/LoopIdiomRecognize.h"
#include "llvm/Transforms/Scalar/LoopInstSimplify.h"
#include "llvm/Transforms/Scalar/LoopInterchange.h"
#include "llvm/Transforms/Scalar/LoopLoadElimination.h"
#include "llvm/Transforms/Scalar/LoopPassManager.h"
#include "llvm/Transforms/Scalar/LoopRotation.h"
#include "llvm/Transforms/Scalar/LoopSimplifyCFG.h"
#include "llvm/Transforms/Scalar/LoopSink.h"
#include "llvm/Transforms/Scalar/LoopUnrollAndJamPass.h"
#include "llvm/Transforms/Scalar/LoopUnrollPass.h"
#include "llvm/Transforms/Scalar/LoopVersioningLICM.h"
#include "llvm/Transforms/Scalar/LowerConstantIntrinsics.h"
#include "llvm/Transforms/Scalar/LowerExpectIntrinsic.h"
#include "llvm/Transforms/Scalar/LowerMatrixIntrinsics.h"
#include "llvm/Transforms/Scalar/MemCpyOptimizer.h"
#include "llvm/Transforms/Scalar/MergeICmps.h"
#include "llvm/Transforms/Scalar/MergedLoadStoreMotion.h"
#include "llvm/Transforms/Scalar/NewGVN.h"
#include "llvm/Transforms/Scalar/Reassociate.h"
#include "llvm/Transforms/Scalar/SCCP.h"
#include "llvm/Transforms/Scalar/SROA.h"
#include "llvm/Transforms/Scalar/SimpleLoopUnswitch.h"
#include "llvm/Transforms/Scalar/SimplifyCFG.h"
#include "llvm/Transforms/Scalar/SpeculativeExecution.h"
#include "llvm/Transforms/Scalar/TailRecursionElimination.h"
#include "llvm/Transforms/Scalar/WarnMissedTransforms.h"
#include "llvm/Transforms/Utils/AddDiscriminators.h"
#include "llvm/Transforms/Utils/AssignGUID.h"
#include "llvm/Transforms/Utils/AssumeBundleBuilder.h"
#include "llvm/Transforms/Utils/CanonicalizeAliases.h"
#include "llvm/Transforms/Utils/CountVisits.h"
#include "llvm/Transforms/Utils/EntryExitInstrumenter.h"
#include "llvm/Transforms/Utils/ExtraPassManager.h"
#include "llvm/Transforms/Utils/InjectTLIMappings.h"
#include "llvm/Transforms/Utils/LibCallsShrinkWrap.h"
#include "llvm/Transforms/Utils/LowerCommentStringPass.h"
#include "llvm/Transforms/Utils/Mem2Reg.h"
#include "llvm/Transforms/Utils/MoveAutoInit.h"
#include "llvm/Transforms/Utils/NameAnonGlobals.h"
#include "llvm/Transforms/Utils/RelLookupTableConverter.h"
#include "llvm/Transforms/Utils/SimplifyCFGOptions.h"
#include "llvm/Transforms/Vectorize/LoopVectorize.h"
#include "llvm/Transforms/Vectorize/SLPVectorizer.h"
#include "llvm/Transforms/Vectorize/VectorCombine.h"
 
using namespace llvm;
 
namespace llvm {
extern cl::opt<std::string> UseCtxProfile;
 
extern cl::opt<bool> EnableMemProfContextDisambiguation;
} // namespace llvm
 
PipelineTuningOptions::PipelineTuningOptions() {
  const PassesOptions &Opts = PassesOptions::Global;
  LoopInterleaving = true;
  LoopVectorization = true;
  SLPVectorization = false;
  LoopUnrolling = true;
  LoopInterchange = Opts.enable_loopinterchange;
  LoopFusion = false;
  ForgetAllSCEVInLoopUnroll = getForgetSCEVInLoopUnroll();
  LicmMssaOptCap = getLicmMssaOptCap();
  LicmMssaNoAccForPromotionCap = getLicmMssaNoAccForPromotionCap();
  CallGraphProfile = true;
  UnifiedLTO = false;
  MergeFunctions = Opts.enable_merge_functions;
  InlinerThreshold = -1;
  EagerlyInvalidateAnalyses = Opts.eagerly_invalidate_analyses;
  DevirtualizeSpeculatively = Opts.enable_devirtualize_speculatively;
}
 
namespace llvm {
extern cl::opt<unsigned> MaxDevirtIterations;
} // namespace llvm
 
void PassBuilder::invokePeepholeEPCallbacks(FunctionPassManager &FPM,
                                            OptimizationLevel Level) {
  for (auto &C : PeepholeEPCallbacks)
    C(FPM, Level);
}
void PassBuilder::invokeLateLoopOptimizationsEPCallbacks(
    LoopPassManager &LPM, OptimizationLevel Level) {
  for (auto &C : LateLoopOptimizationsEPCallbacks)
    C(LPM, Level);
}
void PassBuilder::invokeLoopOptimizerEndEPCallbacks(LoopPassManager &LPM,
                                                    OptimizationLevel Level) {
  for (auto &C : LoopOptimizerEndEPCallbacks)
    C(LPM, Level);
}
void PassBuilder::invokeScalarOptimizerLateEPCallbacks(
    FunctionPassManager &FPM, OptimizationLevel Level) {
  for (auto &C : ScalarOptimizerLateEPCallbacks)
    C(FPM, Level);
}
void PassBuilder::invokeCGSCCOptimizerLateEPCallbacks(CGSCCPassManager &CGPM,
                                                      OptimizationLevel Level) {
  for (auto &C : CGSCCOptimizerLateEPCallbacks)
    C(CGPM, Level);
}
void PassBuilder::invokeVectorizerStartEPCallbacks(FunctionPassManager &FPM,
                                                   OptimizationLevel Level) {
  for (auto &C : VectorizerStartEPCallbacks)
    C(FPM, Level);
}
void PassBuilder::invokeVectorizerEndEPCallbacks(FunctionPassManager &FPM,
                                                 OptimizationLevel Level) {
  for (auto &C : VectorizerEndEPCallbacks)
    C(FPM, Level);
}
void PassBuilder::invokeOptimizerEarlyEPCallbacks(ModulePassManager &MPM,
                                                  OptimizationLevel Level,
                                                  ThinOrFullLTOPhase Phase) {
  for (auto &C : OptimizerEarlyEPCallbacks)
    C(MPM, Level, Phase);
}
void PassBuilder::invokeOptimizerLastEPCallbacks(ModulePassManager &MPM,
                                                 OptimizationLevel Level,
                                                 ThinOrFullLTOPhase Phase) {
  for (auto &C : OptimizerLastEPCallbacks)
    C(MPM, Level, Phase);
}
void PassBuilder::invokeFullLinkTimeOptimizationEarlyEPCallbacks(
    ModulePassManager &MPM, OptimizationLevel Level) {
  for (auto &C : FullLinkTimeOptimizationEarlyEPCallbacks)
    C(MPM, Level);
}
void PassBuilder::invokeFullLinkTimeOptimizationLastEPCallbacks(
    ModulePassManager &MPM, OptimizationLevel Level) {
  for (auto &C : FullLinkTimeOptimizationLastEPCallbacks)
    C(MPM, Level);
}
void PassBuilder::invokeThinLinkTimeOptimizationEarlyEPCallbacks(
    ModulePassManager &MPM, OptimizationLevel Level) {
  for (auto &C : ThinLinkTimeOptimizationEarlyEPCallbacks)
    C(MPM, Level);
}
void PassBuilder::invokeThinLinkTimeOptimizationLastEPCallbacks(
    ModulePassManager &MPM, OptimizationLevel Level) {
  for (auto &C : ThinLinkTimeOptimizationLastEPCallbacks)
    C(MPM, Level);
}
void PassBuilder::invokePipelineStartEPCallbacks(ModulePassManager &MPM,
                                                 OptimizationLevel Level) {
  for (auto &C : PipelineStartEPCallbacks)
    C(MPM, Level);
}
void PassBuilder::invokePipelineEarlySimplificationEPCallbacks(
    ModulePassManager &MPM, OptimizationLevel Level, ThinOrFullLTOPhase Phase) {
  for (auto &C : PipelineEarlySimplificationEPCallbacks)
    C(MPM, Level, Phase);
}
 
// Get IR stats with InstCount before/after the optimization pipeline
static void instructionCountersPass(ModulePassManager &MPM,
                                    bool IsPreOptimization) {
  if (AreStatisticsEnabled()) {
    MPM.addPass(
        createModuleToFunctionPassAdaptor(InstCountPass(IsPreOptimization)));
    MPM.addPass(createModuleToFunctionPassAdaptor(
        FunctionPropertiesStatisticsPass(IsPreOptimization)));
  }
}
 
// Helper to add AnnotationRemarksPass.
static void addAnnotationRemarksPass(ModulePassManager &MPM) {
  MPM.addPass(createModuleToFunctionPassAdaptor(AnnotationRemarksPass()));
}
 
// Helper to check if the current compilation phase is preparing for LTO
static bool isLTOPreLink(ThinOrFullLTOPhase Phase) {
  return Phase == ThinOrFullLTOPhase::ThinLTOPreLink ||
         Phase == ThinOrFullLTOPhase::FullLTOPreLink;
}
 
// Helper to check if the current compilation phase is preparing for FullLTO
[[maybe_unused]] static bool isFullLTOPreLink(ThinOrFullLTOPhase Phase) {
  return Phase == ThinOrFullLTOPhase::FullLTOPreLink;
}
 
// Helper to check if the current compilation phase is preparing for ThinLTO
static bool isThinLTOPreLink(ThinOrFullLTOPhase Phase) {
  return Phase == ThinOrFullLTOPhase::ThinLTOPreLink;
}
 
// Helper to check if the current compilation phase is LTO backend
static bool isLTOPostLink(ThinOrFullLTOPhase Phase) {
  return Phase == ThinOrFullLTOPhase::ThinLTOPostLink ||
         Phase == ThinOrFullLTOPhase::FullLTOPostLink;
}
 
// Helper to check if the current compilation phase is FullLTO backend
static bool isFullLTOPostLink(ThinOrFullLTOPhase Phase) {
  return Phase == ThinOrFullLTOPhase::FullLTOPostLink;
}
 
// Helper to check if the current compilation phase is ThinLTO backend
static bool isThinLTOPostLink(ThinOrFullLTOPhase Phase) {
  return Phase == ThinOrFullLTOPhase::ThinLTOPostLink;
}
 
// Helper to wrap conditionally Coro passes.
static CoroConditionalWrapper buildCoroWrapper(ThinOrFullLTOPhase Phase) {
  // TODO: Skip passes according to Phase.
  ModulePassManager CoroPM;
  CoroPM.addPass(CoroEarlyPass());
  CGSCCPassManager CGPM;
  CGPM.addPass(CoroSplitPass());
  CoroPM.addPass(createModuleToPostOrderCGSCCPassAdaptor(std::move(CGPM)));
  CoroPM.addPass(CoroCleanupPass());
  CoroPM.addPass(GlobalDCEPass());
  return CoroConditionalWrapper(std::move(CoroPM));
}
 
static InlineParams getInlineParamsFromOptLevel(OptimizationLevel Level) {
  return getInlineParamsFromOptLevel(static_cast<unsigned>(Level));
}
 
static void addModuleInlinerPass(ModulePassManager &MPM,
                                 const PassesOptions &Opts,
                                 OptimizationLevel Level,
                                 ThinOrFullLTOPhase Phase) {
  InlineParams IP = ::getInlineParamsFromOptLevel(Level);
  if (Opts.enable_module_inliner)
    MPM.addPass(ModuleInlinerPass(IP, Opts.enable_ml_inliner, Phase));
  else
    MPM.addPass(ModuleInlinerWrapperPass(
        IP,
        /* MandatoryFirst */ true,
        InlineContext{Phase, InlinePass::CGSCCInliner}));
}
 
// TODO: Investigate the cost/benefit of tail call elimination on debugging.
FunctionPassManager
PassBuilder::buildO1FunctionSimplificationPipeline(OptimizationLevel Level,
                                                   ThinOrFullLTOPhase Phase) {
 
  FunctionPassManager FPM;
 
  if (AreStatisticsEnabled())
    FPM.addPass(CountVisitsPass());
 
  // Form SSA out of local memory accesses after breaking apart aggregates into
  // scalars.
  FPM.addPass(SROAPass(SROAOptions::ModifyCFG));
 
  // Catch trivial redundancies
  FPM.addPass(EarlyCSEPass(true /* Enable mem-ssa. */));
 
  // Hoisting of scalars and load expressions.
  FPM.addPass(
      SimplifyCFGPass(SimplifyCFGOptions().convertSwitchRangeToICmp(true)));
  FPM.addPass(InstCombinePass());
 
  FPM.addPass(LibCallsShrinkWrapPass());
 
  invokePeepholeEPCallbacks(FPM, Level);
 
  FPM.addPass(
      SimplifyCFGPass(SimplifyCFGOptions().convertSwitchRangeToICmp(true)));
 
  // Form canonically associated expression trees, and simplify the trees using
  // basic mathematical properties. For example, this will form (nearly)
  // minimal multiplication trees.
  FPM.addPass(ReassociatePass());
 
  // Add the primary loop simplification pipeline.
  // FIXME: Currently this is split into two loop pass pipelines because we run
  // some function passes in between them. These can and should be removed
  // and/or replaced by scheduling the loop pass equivalents in the correct
  // positions. But those equivalent passes aren't powerful enough yet.
  // Specifically, `SimplifyCFGPass` and `InstCombinePass` are currently still
  // used. We have `LoopSimplifyCFGPass` which isn't yet powerful enough yet to
  // fully replace `SimplifyCFGPass`, and the closest to the other we have is
  // `LoopInstSimplify`.
  LoopPassManager LPM1, LPM2;
 
  // Simplify the loop body. We do this initially to clean up after other loop
  // passes run, either when iterating on a loop or on inner loops with
  // implications on the outer loop.
  LPM1.addPass(LoopInstSimplifyPass());
  LPM1.addPass(LoopSimplifyCFGPass());
 
  // Try to remove as much code from the loop header as possible,
  // to reduce amount of IR that will have to be duplicated. However,
  // do not perform speculative hoisting the first time as LICM
  // will destroy metadata that may not need to be destroyed if run
  // after loop rotation.
  // TODO: Investigate promotion cap for O1.
  LPM1.addPass(LICMPass(PTO.LicmMssaOptCap, PTO.LicmMssaNoAccForPromotionCap,
                        /*AllowSpeculation=*/false));
 
  LPM1.addPass(
      LoopRotatePass(/*EnableHeaderDuplication=*/true, isLTOPreLink(Phase)));
  // TODO: Investigate promotion cap for O1.
  LPM1.addPass(LICMPass(PTO.LicmMssaOptCap, PTO.LicmMssaNoAccForPromotionCap,
                        /*AllowSpeculation=*/true));
  LPM1.addPass(SimpleLoopUnswitchPass());
  if (Opts.enable_loop_flatten)
    LPM1.addPass(LoopFlattenPass());
 
  LPM2.addPass(LoopIdiomRecognizePass());
  LPM2.addPass(IndVarSimplifyPass());
 
  invokeLateLoopOptimizationsEPCallbacks(LPM2, Level);
 
  LPM2.addPass(LoopDeletionPass());
 
  // Do not enable unrolling in PreLinkThinLTO phase during sample PGO
  // because it changes IR to makes profile annotation in back compile
  // inaccurate. The normal unroller doesn't pay attention to forced full unroll
  // attributes so we need to make sure and allow the full unroll pass to pay
  // attention to it.
  if (!isThinLTOPreLink(Phase) || !PGOOpt ||
      PGOOpt->Action != PGOOptions::SampleUse)
    LPM2.addPass(LoopFullUnrollPass(static_cast<int>(Level),
                                    /* OnlyWhenForced= */ !PTO.LoopUnrolling,
                                    PTO.ForgetAllSCEVInLoopUnroll));
 
  invokeLoopOptimizerEndEPCallbacks(LPM2, Level);
 
  FPM.addPass(createFunctionToLoopPassAdaptor(std::move(LPM1),
                                              /*UseMemorySSA=*/true));
  FPM.addPass(
      SimplifyCFGPass(SimplifyCFGOptions().convertSwitchRangeToICmp(true)));
  FPM.addPass(InstCombinePass());
  // The loop passes in LPM2 (LoopFullUnrollPass) do not preserve MemorySSA.
  // *All* loop passes must preserve it, in order to be able to use it.
  FPM.addPass(createFunctionToLoopPassAdaptor(std::move(LPM2),
                                              /*UseMemorySSA=*/false));
 
  // Delete small array after loop unroll.
  FPM.addPass(SROAPass(SROAOptions::ModifyCFG));
 
  // Specially optimize memory movement as it doesn't look like dataflow in SSA.
  FPM.addPass(MemCpyOptPass());
 
  // Sparse conditional constant propagation.
  // FIXME: It isn't clear why we do this *after* loop passes rather than
  // before...
  FPM.addPass(SCCPPass());
 
  // Delete dead bit computations (instcombine runs after to fold away the dead
  // computations, and then ADCE will run later to exploit any new DCE
  // opportunities that creates).
  FPM.addPass(BDCEPass());
 
  // Run instcombine after redundancy and dead bit elimination to exploit
  // opportunities opened up by them.
  FPM.addPass(InstCombinePass());
  invokePeepholeEPCallbacks(FPM, Level);
 
  FPM.addPass(CoroElidePass());
 
  invokeScalarOptimizerLateEPCallbacks(FPM, Level);
 
  // Finally, do an expensive DCE pass to catch all the dead code exposed by
  // the simplifications and basic cleanup after all the simplifications.
  // TODO: Investigate if this is too expensive.
  FPM.addPass(ADCEPass());
  FPM.addPass(
      SimplifyCFGPass(SimplifyCFGOptions().convertSwitchRangeToICmp(true)));
  FPM.addPass(InstCombinePass());
  invokePeepholeEPCallbacks(FPM, Level);
 
  return FPM;
}
 
FunctionPassManager
PassBuilder::buildFunctionSimplificationPipeline(OptimizationLevel Level,
                                                 ThinOrFullLTOPhase Phase) {
  assert(Level != OptimizationLevel::O0 && "Must request optimizations!");
 
  // The O1 pipeline has a separate pipeline creation function to simplify
  // construction readability.
  if (Level == OptimizationLevel::O1)
    return buildO1FunctionSimplificationPipeline(Level, Phase);
 
  FunctionPassManager FPM;
 
  if (AreStatisticsEnabled())
    FPM.addPass(CountVisitsPass());
 
  // Form SSA out of local memory accesses after breaking apart aggregates into
  // scalars.
  FPM.addPass(SROAPass(SROAOptions::ModifyCFG));
 
  // Catch trivial redundancies
  FPM.addPass(EarlyCSEPass(true /* Enable mem-ssa. */));
  if (EnableKnowledgeRetention)
    FPM.addPass(AssumeSimplifyPass());
 
  // Hoisting of scalars and load expressions.
  if (Opts.enable_gvn_hoist)
    FPM.addPass(GVNHoistPass());
 
  // Global value numbering based sinking.
  if (Opts.enable_gvn_sink) {
    FPM.addPass(GVNSinkPass());
    FPM.addPass(
        SimplifyCFGPass(SimplifyCFGOptions().convertSwitchRangeToICmp(true)));
  }
 
  // Speculative execution if the target has divergent branches; otherwise nop.
  FPM.addPass(SpeculativeExecutionPass(/* OnlyIfDivergentTarget =*/true));
 
  // Optimize based on known information about branches, and cleanup afterward.
  FPM.addPass(JumpThreadingPass());
  FPM.addPass(CorrelatedValuePropagationPass());
 
  // Jump table to switch conversion.
  if (Opts.enable_jump_table_to_switch)
    FPM.addPass(JumpTableToSwitchPass(/*InLTO=*/isLTOPostLink(Phase)));
 
  FPM.addPass(
      SimplifyCFGPass(SimplifyCFGOptions().convertSwitchRangeToICmp(true)));
  FPM.addPass(InstCombinePass());
  FPM.addPass(AggressiveInstCombinePass());
  FPM.addPass(LibCallsShrinkWrapPass());
 
  invokePeepholeEPCallbacks(FPM, Level);
 
  // For PGO use pipeline, try to optimize memory intrinsics such as memcpy
  // using the size value profile. Don't perform this when optimizing for size.
  if (PGOOpt && PGOOpt->Action == PGOOptions::IRUse)
    FPM.addPass(PGOMemOPSizeOpt());
 
  FPM.addPass(TailCallElimPass(/*UpdateFunctionEntryCount=*/
                               isInstrumentedPGOUse()));
  FPM.addPass(
      SimplifyCFGPass(SimplifyCFGOptions().convertSwitchRangeToICmp(true)));
 
  // Form canonically associated expression trees, and simplify the trees using
  // basic mathematical properties. For example, this will form (nearly)
  // minimal multiplication trees.
  FPM.addPass(ReassociatePass());
 
  if (Opts.enable_constraint_elimination)
    FPM.addPass(ConstraintEliminationPass());
 
  // Add the primary loop simplification pipeline.
  // FIXME: Currently this is split into two loop pass pipelines because we run
  // some function passes in between them. These can and should be removed
  // and/or replaced by scheduling the loop pass equivalents in the correct
  // positions. But those equivalent passes aren't powerful enough yet.
  // Specifically, `SimplifyCFGPass` and `InstCombinePass` are currently still
  // used. We have `LoopSimplifyCFGPass` which isn't yet powerful enough yet to
  // fully replace `SimplifyCFGPass`, and the closest to the other we have is
  // `LoopInstSimplify`.
  LoopPassManager LPM1, LPM2;
 
  // Simplify the loop body. We do this initially to clean up after other loop
  // passes run, either when iterating on a loop or on inner loops with
  // implications on the outer loop.
  LPM1.addPass(LoopInstSimplifyPass());
  LPM1.addPass(LoopSimplifyCFGPass());
 
  // Try to remove as much code from the loop header as possible,
  // to reduce amount of IR that will have to be duplicated. However,
  // do not perform speculative hoisting the first time as LICM
  // will destroy metadata that may not need to be destroyed if run
  // after loop rotation.
  // TODO: Investigate promotion cap for O1.
  LPM1.addPass(LICMPass(PTO.LicmMssaOptCap, PTO.LicmMssaNoAccForPromotionCap,
                        /*AllowSpeculation=*/false));
 
  LPM1.addPass(
      LoopRotatePass(/*EnableHeaderDuplication=*/true, isLTOPreLink(Phase)));
  // TODO: Investigate promotion cap for O1.
  LPM1.addPass(LICMPass(PTO.LicmMssaOptCap, PTO.LicmMssaNoAccForPromotionCap,
                        /*AllowSpeculation=*/true));
  LPM1.addPass(
      SimpleLoopUnswitchPass(/* NonTrivial */ Level == OptimizationLevel::O3));
  if (Opts.enable_loop_flatten)
    LPM1.addPass(LoopFlattenPass());
 
  LPM2.addPass(LoopIdiomRecognizePass());
  LPM2.addPass(IndVarSimplifyPass());
 
  {
    ExtraLoopPassManager<ShouldRunExtraSimpleLoopUnswitch> ExtraPasses;
    ExtraPasses.addPass(SimpleLoopUnswitchPass(/* NonTrivial */ Level ==
                                               OptimizationLevel::O3));
    LPM2.addPass(std::move(ExtraPasses));
  }
 
  invokeLateLoopOptimizationsEPCallbacks(LPM2, Level);
 
  LPM2.addPass(LoopDeletionPass());
 
  // Do not enable unrolling in PreLinkThinLTO phase during sample PGO
  // because it changes IR to makes profile annotation in back compile
  // inaccurate. The normal unroller doesn't pay attention to forced full unroll
  // attributes so we need to make sure and allow the full unroll pass to pay
  // attention to it.
  if (!isThinLTOPreLink(Phase) || !PGOOpt ||
      PGOOpt->Action != PGOOptions::SampleUse)
    LPM2.addPass(LoopFullUnrollPass(static_cast<int>(Level),
                                    /* OnlyWhenForced= */ !PTO.LoopUnrolling,
                                    PTO.ForgetAllSCEVInLoopUnroll));
 
  invokeLoopOptimizerEndEPCallbacks(LPM2, Level);
 
  FPM.addPass(createFunctionToLoopPassAdaptor(std::move(LPM1),
                                              /*UseMemorySSA=*/true));
  FPM.addPass(
      SimplifyCFGPass(SimplifyCFGOptions().convertSwitchRangeToICmp(true)));
  FPM.addPass(InstCombinePass());
  // The loop passes in LPM2 (LoopIdiomRecognizePass, IndVarSimplifyPass,
  // LoopDeletionPass and LoopFullUnrollPass) do not preserve MemorySSA.
  // *All* loop passes must preserve it, in order to be able to use it.
  FPM.addPass(createFunctionToLoopPassAdaptor(std::move(LPM2),
                                              /*UseMemorySSA=*/false));
 
  // Delete small array after loop unroll.
  FPM.addPass(SROAPass(SROAOptions::ModifyCFG));
 
  // Try vectorization/scalarization transforms that are both improvements
  // themselves and can allow further folds with GVN and InstCombine.
  FPM.addPass(VectorCombinePass(/*TryEarlyFoldsOnly=*/true));
 
  // Eliminate redundancies.
  FPM.addPass(MergedLoadStoreMotionPass());
  if (Opts.enable_newgvn)
    FPM.addPass(NewGVNPass());
  else
    FPM.addPass(GVNPass());
 
  // Sparse conditional constant propagation.
  // FIXME: It isn't clear why we do this *after* loop passes rather than
  // before...
  FPM.addPass(SCCPPass());
 
  // Delete dead bit computations (instcombine runs after to fold away the dead
  // computations, and then ADCE will run later to exploit any new DCE
  // opportunities that creates).
  FPM.addPass(BDCEPass());
 
  // Run instcombine after redundancy and dead bit elimination to exploit
  // opportunities opened up by them.
  FPM.addPass(InstCombinePass());
  invokePeepholeEPCallbacks(FPM, Level);
 
  // Re-consider control flow based optimizations after redundancy elimination,
  // redo DCE, etc.
  if (Opts.enable_dfa_jump_thread)
    FPM.addPass(DFAJumpThreadingPass());
 
  FPM.addPass(JumpThreadingPass());
  FPM.addPass(CorrelatedValuePropagationPass());
 
  // Finally, do an expensive DCE pass to catch all the dead code exposed by
  // the simplifications and basic cleanup after all the simplifications.
  // TODO: Investigate if this is too expensive.
  FPM.addPass(ADCEPass());
 
  // Specially optimize memory movement as it doesn't look like dataflow in SSA.
  FPM.addPass(MemCpyOptPass());
 
  FPM.addPass(DSEPass());
  FPM.addPass(MoveAutoInitPass());
 
  FPM.addPass(createFunctionToLoopPassAdaptor(
      LICMPass(PTO.LicmMssaOptCap, PTO.LicmMssaNoAccForPromotionCap,
               /*AllowSpeculation=*/true),
      /*UseMemorySSA=*/true));
 
  FPM.addPass(CoroElidePass());
 
  invokeScalarOptimizerLateEPCallbacks(FPM, Level);
 
  FPM.addPass(SimplifyCFGPass(SimplifyCFGOptions()
                                  .convertSwitchRangeToICmp(true)
                                  .convertSwitchToArithmetic(true)
                                  .hoistCommonInsts(true)
                                  .sinkCommonInsts(true)));
  FPM.addPass(InstCombinePass());
  invokePeepholeEPCallbacks(FPM, Level);
 
  return FPM;
}
 
void PassBuilder::addRequiredLTOPreLinkPasses(ModulePassManager &MPM) {
  MPM.addPass(CanonicalizeAliasesPass());
  MPM.addPass(NameAnonGlobalPass());
  MPM.addPass(AssignGUIDPass());
}
 
void PassBuilder::addPreInlinerPasses(ModulePassManager &MPM,
                                      OptimizationLevel Level,
                                      ThinOrFullLTOPhase LTOPhase) {
  assert(Level != OptimizationLevel::O0 && "Not expecting O0 here!");
  if (Opts.disable_preinline)
    return;
  InlineParams IP;
 
  IP.DefaultThreshold = Opts.preinline_threshold;
 
  // FIXME: The hint threshold has the same value used by the regular inliner
  // when not optimzing for size. This should probably be lowered after
  // performance testing.
  // FIXME: this comment is cargo culted from the old pass manager, revisit).
  IP.HintThreshold = 325;
  IP.OptSizeHintThreshold = Opts.preinline_threshold;
  ModuleInlinerWrapperPass MIWP(
      IP, /* MandatoryFirst */ true,
      InlineContext{LTOPhase, InlinePass::EarlyInliner});
  CGSCCPassManager &CGPipeline = MIWP.getPM();
 
  FunctionPassManager FPM;
  FPM.addPass(SROAPass(SROAOptions::ModifyCFG));
  FPM.addPass(EarlyCSEPass()); // Catch trivial redundancies.
  FPM.addPass(SimplifyCFGPass(SimplifyCFGOptions().convertSwitchRangeToICmp(
      true)));                    // Merge & remove basic blocks.
  FPM.addPass(InstCombinePass()); // Combine silly sequences.
  invokePeepholeEPCallbacks(FPM, Level);
 
  CGPipeline.addPass(createCGSCCToFunctionPassAdaptor(
      std::move(FPM), PTO.EagerlyInvalidateAnalyses));
 
  MPM.addPass(std::move(MIWP));
 
  // Delete anything that is now dead to make sure that we don't instrument
  // dead code. Instrumentation can end up keeping dead code around and
  // dramatically increase code size.
  MPM.addPass(GlobalDCEPass());
}
 
void PassBuilder::addPostPGOLoopRotation(ModulePassManager &MPM,
                                         OptimizationLevel Level) {
  if (Opts.enable_post_pgo_loop_rotation) {
    // Disable header duplication in loop rotation at -Oz.
    MPM.addPass(createModuleToFunctionPassAdaptor(
        createFunctionToLoopPassAdaptor(LoopRotatePass(),
                                        /*UseMemorySSA=*/false),
        PTO.EagerlyInvalidateAnalyses));
  }
}
 
void PassBuilder::addPGOInstrPasses(ModulePassManager &MPM,
                                    OptimizationLevel Level, bool RunProfileGen,
                                    bool IsCS, bool AtomicCounterUpdate,
                                    std::string ProfileFile,
                                    std::string ProfileRemappingFile) {
  assert(Level != OptimizationLevel::O0 && "Not expecting O0 here!");
 
  if (!RunProfileGen) {
    assert(!ProfileFile.empty() && "Profile use expecting a profile file!");
    MPM.addPass(
        PGOInstrumentationUse(ProfileFile, ProfileRemappingFile, IsCS, FS));
    // Cache ProfileSummaryAnalysis once to avoid the potential need to insert
    // RequireAnalysisPass for PSI before subsequent non-module passes.
    MPM.addPass(RequireAnalysisPass<ProfileSummaryAnalysis, Module>());
    return;
  }
 
  // Perform PGO instrumentation.
  MPM.addPass(PGOInstrumentationGen(IsCS ? PGOInstrumentationType::CSFDO
                                         : PGOInstrumentationType::FDO));
 
  addPostPGOLoopRotation(MPM, Level);
  // Add the profile lowering pass.
  InstrProfOptions Options;
  if (!ProfileFile.empty())
    Options.InstrProfileOutput = ProfileFile;
  // Do counter promotion at Level greater than O0.
  Options.DoCounterPromotion = true;
  Options.UseBFIInPromotion = IsCS;
  if (Opts.enable_sampled_instrumentation) {
    Options.Sampling = true;
    // With sampling, there is little beneifit to enable counter promotion.
    // But note that sampling does work with counter promotion.
    Options.DoCounterPromotion = false;
  }
  Options.Atomic = AtomicCounterUpdate;
  MPM.addPass(InstrProfilingLoweringPass(Options, IsCS));
}
 
void PassBuilder::addPGOInstrPassesForO0(ModulePassManager &MPM,
                                         bool RunProfileGen, bool IsCS,
                                         bool AtomicCounterUpdate,
                                         std::string ProfileFile,
                                         std::string ProfileRemappingFile) {
  if (!RunProfileGen) {
    assert(!ProfileFile.empty() && "Profile use expecting a profile file!");
    MPM.addPass(
        PGOInstrumentationUse(ProfileFile, ProfileRemappingFile, IsCS, FS));
    // Cache ProfileSummaryAnalysis once to avoid the potential need to insert
    // RequireAnalysisPass for PSI before subsequent non-module passes.
    MPM.addPass(RequireAnalysisPass<ProfileSummaryAnalysis, Module>());
    return;
  }
 
  // Perform PGO instrumentation.
  MPM.addPass(PGOInstrumentationGen(IsCS ? PGOInstrumentationType::CSFDO
                                         : PGOInstrumentationType::FDO));
  // Add the profile lowering pass.
  InstrProfOptions Options;
  if (!ProfileFile.empty())
    Options.InstrProfileOutput = ProfileFile;
  // Do not do counter promotion at O0.
  Options.DoCounterPromotion = false;
  Options.UseBFIInPromotion = IsCS;
  Options.Atomic = AtomicCounterUpdate;
  MPM.addPass(InstrProfilingLoweringPass(Options, IsCS));
}
 
ModuleInlinerWrapperPass
PassBuilder::buildInlinerPipeline(OptimizationLevel Level,
                                  ThinOrFullLTOPhase Phase) {
  InlineParams IP;
  if (PTO.InlinerThreshold == -1)
    IP = ::getInlineParamsFromOptLevel(Level);
  else
    IP = getInlineParams(PTO.InlinerThreshold);
  // For PreLinkThinLTO + SamplePGO or PreLinkFullLTO + SamplePGO,
  // set hot-caller threshold to 0 to disable hot
  // callsite inline (as much as possible [1]) because it makes
  // profile annotation in the backend inaccurate.
  //
  // [1] Note the cost of a function could be below zero due to erased
  // prologue / epilogue.
  if (isLTOPreLink(Phase) && PGOOpt && PGOOpt->Action == PGOOptions::SampleUse)
    IP.HotCallSiteThreshold = 0;
 
  if (PGOOpt)
    IP.EnableDeferral = Opts.enable_npm_pgo_inline_deferral;
 
  ModuleInlinerWrapperPass MIWP(IP, Opts.mandatory_inlining_first,
                                InlineContext{Phase, InlinePass::CGSCCInliner},
                                Opts.enable_ml_inliner, MaxDevirtIterations);
 
  // Require the GlobalsAA analysis for the module so we can query it within
  // the CGSCC pipeline.
  if (Opts.enable_global_analyses) {
    MIWP.addModulePass(RequireAnalysisPass<GlobalsAA, Module>());
    // Invalidate AAManager so it can be recreated and pick up the newly
    // available GlobalsAA.
    MIWP.addModulePass(
        createModuleToFunctionPassAdaptor(InvalidateAnalysisPass<AAManager>()));
  }
 
  // Require the ProfileSummaryAnalysis for the module so we can query it within
  // the inliner pass.
  MIWP.addModulePass(RequireAnalysisPass<ProfileSummaryAnalysis, Module>());
 
  // Now begin the main postorder CGSCC pipeline.
  // FIXME: The current CGSCC pipeline has its origins in the legacy pass
  // manager and trying to emulate its precise behavior. Much of this doesn't
  // make a lot of sense and we should revisit the core CGSCC structure.
  CGSCCPassManager &MainCGPipeline = MIWP.getPM();
 
  // Note: historically, the PruneEH pass was run first to deduce nounwind and
  // generally clean up exception handling overhead. It isn't clear this is
  // valuable as the inliner doesn't currently care whether it is inlining an
  // invoke or a call.
 
  if (Opts.attributor_enable & AttributorRunOption::CGSCC)
    MainCGPipeline.addPass(AttributorCGSCCPass());
  else if (Opts.attributor_enable & AttributorRunOption::CGSCC_LIGHT)
    MainCGPipeline.addPass(AttributorLightCGSCCPass());
 
  // Deduce function attributes. We do another run of this after the function
  // simplification pipeline, so this only needs to run when it could affect the
  // function simplification pipeline, which is only the case with recursive
  // functions.
  MainCGPipeline.addPass(PostOrderFunctionAttrsPass(/*SkipNonRecursive*/ true));
 
  // When at O3 add argument promotion to the pass pipeline.
  // FIXME: It isn't at all clear why this should be limited to O3.
  if (Level == OptimizationLevel::O3)
    MainCGPipeline.addPass(ArgumentPromotionPass());
 
  // Try to perform OpenMP specific optimizations. This is a (quick!) no-op if
  // there are no OpenMP runtime calls present in the module.
  if (Level == OptimizationLevel::O2 || Level == OptimizationLevel::O3)
    MainCGPipeline.addPass(OpenMPOptCGSCCPass(Phase));
 
  invokeCGSCCOptimizerLateEPCallbacks(MainCGPipeline, Level);
 
  // Add the core function simplification pipeline nested inside the
  // CGSCC walk.
  MainCGPipeline.addPass(createCGSCCToFunctionPassAdaptor(
      buildFunctionSimplificationPipeline(Level, Phase),
      PTO.EagerlyInvalidateAnalyses, /*NoRerun=*/true));
 
  // Finally, deduce any function attributes based on the fully simplified
  // function.
  MainCGPipeline.addPass(PostOrderFunctionAttrsPass());
 
  // Mark that the function is fully simplified and that it shouldn't be
  // simplified again if we somehow revisit it due to CGSCC mutations unless
  // it's been modified since.
  MainCGPipeline.addPass(createCGSCCToFunctionPassAdaptor(
      RequireAnalysisPass<ShouldNotRunFunctionPassesAnalysis, Function>()));
 
  if (!isThinLTOPreLink(Phase)) {
    MainCGPipeline.addPass(CoroSplitPass(Level != OptimizationLevel::O0));
    MainCGPipeline.addPass(CoroAnnotationElidePass());
  }
 
  // Make sure we don't affect potential future NoRerun CGSCC adaptors.
  MIWP.addLateModulePass(createModuleToFunctionPassAdaptor(
      InvalidateAnalysisPass<ShouldNotRunFunctionPassesAnalysis>()));
 
  return MIWP;
}
 
ModulePassManager
PassBuilder::buildModuleInlinerPipeline(OptimizationLevel Level,
                                        ThinOrFullLTOPhase Phase) {
  ModulePassManager MPM;
 
  InlineParams IP = ::getInlineParamsFromOptLevel(Level);
  // For PreLinkThinLTO + SamplePGO or PreLinkFullLTO + SamplePGO,
  // set hot-caller threshold to 0 to disable hot
  // callsite inline (as much as possible [1]) because it makes
  // profile annotation in the backend inaccurate.
  //
  // [1] Note the cost of a function could be below zero due to erased
  // prologue / epilogue.
  if (isLTOPreLink(Phase) && PGOOpt && PGOOpt->Action == PGOOptions::SampleUse)
    IP.HotCallSiteThreshold = 0;
 
  if (PGOOpt)
    IP.EnableDeferral = Opts.enable_npm_pgo_inline_deferral;
 
  // The inline deferral logic is used to avoid losing some
  // inlining chance in future. It is helpful in SCC inliner, in which
  // inlining is processed in bottom-up order.
  // While in module inliner, the inlining order is a priority-based order
  // by default. The inline deferral is unnecessary there. So we disable the
  // inline deferral logic in module inliner.
  IP.EnableDeferral = false;
 
  MPM.addPass(ModuleInlinerPass(IP, Opts.enable_ml_inliner, Phase));
  if (!UseCtxProfile.empty() && Phase == ThinOrFullLTOPhase::ThinLTOPostLink) {
    MPM.addPass(GlobalOptPass());
    MPM.addPass(GlobalDCEPass());
    MPM.addPass(AssignGUIDPass());
    MPM.addPass(PGOCtxProfFlatteningPass(/*IsPreThinlink=*/false));
  }
 
  MPM.addPass(createModuleToFunctionPassAdaptor(
      buildFunctionSimplificationPipeline(Level, Phase),
      PTO.EagerlyInvalidateAnalyses));
 
  if (!isThinLTOPreLink(Phase)) {
    MPM.addPass(createModuleToPostOrderCGSCCPassAdaptor(
        CoroSplitPass(Level != OptimizationLevel::O0)));
    MPM.addPass(
        createModuleToPostOrderCGSCCPassAdaptor(CoroAnnotationElidePass()));
  }
 
  return MPM;
}
 
ModulePassManager
PassBuilder::buildModuleSimplificationPipeline(OptimizationLevel Level,
                                               ThinOrFullLTOPhase Phase) {
  assert(Level != OptimizationLevel::O0 &&
         "Should not be used for O0 pipeline");
 
  assert(!isFullLTOPostLink(Phase) &&
         "FullLTOPostLink shouldn't call buildModuleSimplificationPipeline!");
 
  ModulePassManager MPM;
 
  // Place pseudo probe instrumentation as the first pass of the pipeline to
  // minimize the impact of optimization changes.
  if (PGOOpt && PGOOpt->PseudoProbeForProfiling && !isThinLTOPostLink(Phase))
    MPM.addPass(SampleProfileProbePass(TM));
 
  bool HasSampleProfile = PGOOpt && (PGOOpt->Action == PGOOptions::SampleUse);
 
  // In ThinLTO mode, when flattened profile is used, all the available
  // profile information will be annotated in PreLink phase so there is
  // no need to load the profile again in PostLink.
  bool LoadSampleProfile = HasSampleProfile && !(Opts.flattened_profile_used &&
                                                 isThinLTOPostLink(Phase));
 
  // During the ThinLTO backend phase we perform early indirect call promotion
  // here, before globalopt. Otherwise imported available_externally functions
  // look unreferenced and are removed. If we are going to load the sample
  // profile then defer until later.
  // TODO: See if we can move later and consolidate with the location where
  // we perform ICP when we are loading a sample profile.
  // TODO: We pass HasSampleProfile (whether there was a sample profile file
  // passed to the compile) to the SamplePGO flag of ICP. This is used to
  // determine whether the new direct calls are annotated with prof metadata.
  // Ideally this should be determined from whether the IR is annotated with
  // sample profile, and not whether the a sample profile was provided on the
  // command line. E.g. for flattened profiles where we will not be reloading
  // the sample profile in the ThinLTO backend, we ideally shouldn't have to
  // provide the sample profile file.
  if (isThinLTOPostLink(Phase) && !LoadSampleProfile)
    MPM.addPass(PGOIndirectCallPromotion(true /* InLTO */, HasSampleProfile));
 
  // Create an early function pass manager to cleanup the output of the
  // frontend. Not necessary with LTO post link pipelines since the pre link
  // pipeline already cleaned up the frontend output.
  if (!isThinLTOPostLink(Phase)) {
    // Do basic inference of function attributes from known properties of system
    // libraries and other oracles.
    MPM.addPass(InferFunctionAttrsPass());
    MPM.addPass(CoroEarlyPass());
 
    FunctionPassManager EarlyFPM;
    EarlyFPM.addPass(EntryExitInstrumenterPass(/*PostInlining=*/false));
    // Lower llvm.expect to metadata before attempting transforms.
    // Compare/branch metadata may alter the behavior of passes like
    // SimplifyCFG.
    EarlyFPM.addPass(LowerExpectIntrinsicPass());
    EarlyFPM.addPass(SimplifyCFGPass());
    EarlyFPM.addPass(SROAPass(SROAOptions::ModifyCFG));
    EarlyFPM.addPass(EarlyCSEPass());
    if (Level == OptimizationLevel::O3)
      EarlyFPM.addPass(CallSiteSplittingPass());
    MPM.addPass(createModuleToFunctionPassAdaptor(
        std::move(EarlyFPM), PTO.EagerlyInvalidateAnalyses));
  }
 
  if (LoadSampleProfile) {
    // Annotate sample profile right after early FPM to ensure freshness of
    // the debug info.
    MPM.addPass(SampleProfileLoaderPass(
        PGOOpt->ProfileFile, PGOOpt->ProfileRemappingFile, Phase, FS));
    // Cache ProfileSummaryAnalysis once to avoid the potential need to insert
    // RequireAnalysisPass for PSI before subsequent non-module passes.
    MPM.addPass(RequireAnalysisPass<ProfileSummaryAnalysis, Module>());
    // Do not invoke ICP in the LTOPrelink phase as it makes it hard
    // for the profile annotation to be accurate in the LTO backend.
    if (!isLTOPreLink(Phase))
      // We perform early indirect call promotion here, before globalopt.
      // This is important for the ThinLTO backend phase because otherwise
      // imported available_externally functions look unreferenced and are
      // removed.
      MPM.addPass(
          PGOIndirectCallPromotion(true /* IsInLTO */, true /* SamplePGO */));
  }
 
  // Try to perform OpenMP specific optimizations on the module. This is a
  // (quick!) no-op if there are no OpenMP runtime calls present in the module.
  MPM.addPass(OpenMPOptPass(Phase));
 
  if (Opts.attributor_enable & AttributorRunOption::MODULE)
    MPM.addPass(AttributorPass());
  else if (Opts.attributor_enable & AttributorRunOption::MODULE_LIGHT)
    MPM.addPass(AttributorLightPass());
 
  // Lower type metadata and the type.test intrinsic in the ThinLTO
  // post link pipeline after ICP. This is to enable usage of the type
  // tests in ICP sequences.
  if (isThinLTOPostLink(Phase))
    MPM.addPass(DropTypeTestsPass());
 
  invokePipelineEarlySimplificationEPCallbacks(MPM, Level, Phase);
 
  // Interprocedural constant propagation now that basic cleanup has occurred
  // and prior to optimizing globals.
  // FIXME: This position in the pipeline hasn't been carefully considered in
  // years, it should be re-analyzed.
  MPM.addPass(
      IPSCCPPass(IPSCCPOptions(/*AllowFuncSpec=*/!isLTOPreLink(Phase))));
 
  // Attach metadata to indirect call sites indicating the set of functions
  // they may target at run-time. This should follow IPSCCP.
  MPM.addPass(CalledValuePropagationPass());
 
  // Optimize globals to try and fold them into constants.
  MPM.addPass(GlobalOptPass());
 
  // Create a small function pass pipeline to cleanup after all the global
  // optimizations.
  FunctionPassManager GlobalCleanupPM;
  // FIXME: Should this instead by a run of SROA?
  GlobalCleanupPM.addPass(PromotePass());
  GlobalCleanupPM.addPass(InstCombinePass());
  invokePeepholeEPCallbacks(GlobalCleanupPM, Level);
  GlobalCleanupPM.addPass(
      SimplifyCFGPass(SimplifyCFGOptions().convertSwitchRangeToICmp(true)));
  MPM.addPass(createModuleToFunctionPassAdaptor(std::move(GlobalCleanupPM),
                                                PTO.EagerlyInvalidateAnalyses));
 
  // We already asserted this happens in non-FullLTOPostLink earlier.
  const bool IsPreLink = !isThinLTOPostLink(Phase);
  // Enable contextual profiling instrumentation.
  const bool IsCtxProfGen =
      IsPreLink && PGOCtxProfLoweringPass::isCtxIRPGOInstrEnabled();
  const bool IsPGOPreLink = !IsCtxProfGen && PGOOpt && IsPreLink;
  const bool IsPGOInstrGen =
      IsPGOPreLink && PGOOpt->Action == PGOOptions::IRInstr;
  const bool IsPGOInstrUse =
      IsPGOPreLink && PGOOpt->Action == PGOOptions::IRUse;
  const bool IsMemprofUse = IsPGOPreLink && !PGOOpt->MemoryProfile.empty();
  // We don't want to mix pgo ctx gen and pgo gen; we also don't currently
  // enable ctx profiling from the frontend.
  assert(!(IsPGOInstrGen && PGOCtxProfLoweringPass::isCtxIRPGOInstrEnabled()) &&
         "Enabling both instrumented PGO and contextual instrumentation is not "
         "supported.");
  const bool IsCtxProfUse = !UseCtxProfile.empty() && isThinLTOPreLink(Phase);
 
  assert((Opts.instrument_cold_function_only_path.empty() ||
          isPGOInstrumentColdFunctionOnly()) &&
         "--instrument-cold-function-only-path is provided but "
         "--pgo-instrument-cold-function-only is not enabled");
  const bool IsColdFuncOnlyInstrGen =
      isPGOInstrumentColdFunctionOnly() && IsPGOPreLink &&
      !Opts.instrument_cold_function_only_path.empty();
 
  if (IsPGOInstrGen || IsPGOInstrUse || IsMemprofUse || IsCtxProfGen ||
      IsCtxProfUse || IsColdFuncOnlyInstrGen)
    addPreInlinerPasses(MPM, Level, Phase);
 
  // Add all the requested passes for instrumentation PGO, if requested.
  if (IsPGOInstrGen || IsPGOInstrUse) {
    addPGOInstrPasses(MPM, Level,
                      /*RunProfileGen=*/IsPGOInstrGen,
                      /*IsCS=*/false, PGOOpt->AtomicCounterUpdate,
                      PGOOpt->ProfileFile, PGOOpt->ProfileRemappingFile);
  } else if (IsCtxProfGen || IsCtxProfUse) {
    MPM.addPass(PGOInstrumentationGen(PGOInstrumentationType::CTXPROF));
    // In pre-link, we just want the instrumented IR. We use the contextual
    // profile in the post-thinlink phase.
    // The instrumentation will be removed in post-thinlink after IPO.
    if (IsCtxProfUse) {
      MPM.addPass(AssignGUIDPass());
      MPM.addPass(PGOCtxProfFlatteningPass(/*IsPreThinlink=*/true));
      return MPM;
    }
    // Block further inlining in the instrumented ctxprof case. This avoids
    // confusingly collecting profiles for the same GUID corresponding to
    // different variants of the function. We could do like PGO and identify
    // functions by a (GUID, Hash) tuple, but since the ctxprof "use" waits for
    // thinlto to happen before performing any further optimizations, it's
    // unnecessary to collect profiles for non-prevailing copies.
    MPM.addPass(NoinlineNonPrevailing());
    addPostPGOLoopRotation(MPM, Level);
    MPM.addPass(AssignGUIDPass());
    MPM.addPass(PGOCtxProfLoweringPass());
  } else if (IsColdFuncOnlyInstrGen) {
    addPGOInstrPasses(MPM, Level, /* RunProfileGen */ true, /* IsCS */ false,
                      /* AtomicCounterUpdate */ false,
                      Opts.instrument_cold_function_only_path.str(),
                      /* ProfileRemappingFile */ "");
  }
 
  if (IsPGOInstrGen || IsPGOInstrUse || IsCtxProfGen)
    MPM.addPass(PGOIndirectCallPromotion(false, false));
 
  if (IsPGOPreLink && PGOOpt->CSAction == PGOOptions::CSIRInstr)
    MPM.addPass(PGOInstrumentationGenCreateVar(
        PGOOpt->CSProfileGenFile, Opts.enable_sampled_instrumentation));
 
  if (IsMemprofUse)
    MPM.addPass(MemProfUsePass(PGOOpt->MemoryProfile, FS));
 
  if (PGOOpt && (PGOOpt->Action == PGOOptions::IRUse ||
                 PGOOpt->Action == PGOOptions::SampleUse))
    MPM.addPass(PGOForceFunctionAttrsPass(PGOOpt->ColdOptType));
 
  MPM.addPass(AlwaysInlinerPass(/*InsertLifetimeIntrinsics=*/true));
 
  if (Opts.enable_module_inliner)
    MPM.addPass(buildModuleInlinerPipeline(Level, Phase));
  else
    MPM.addPass(buildInlinerPipeline(Level, Phase));
 
  // Remove any dead arguments exposed by cleanups, constant folding globals,
  // and argument promotion.
  MPM.addPass(DeadArgumentEliminationPass());
 
  if (isThinLTOPostLink(Phase))
    MPM.addPass(SimplifyTypeTestsPass());
 
  if (!isThinLTOPreLink(Phase))
    MPM.addPass(CoroCleanupPass());
 
  // Optimize globals now that functions are fully simplified.
  MPM.addPass(GlobalOptPass());
  MPM.addPass(GlobalDCEPass());
 
  return MPM;
}
 
/// TODO: Should LTO cause any differences to this set of passes?
void PassBuilder::addVectorPasses(OptimizationLevel Level,
                                  FunctionPassManager &FPM,
                                  ThinOrFullLTOPhase LTOPhase) {
  FPM.addPass(LoopVectorizePass(
      LoopVectorizeOptions(!PTO.LoopInterleaving, !PTO.LoopVectorization)));
 
  // Drop dereferenceable assumes after vectorization, as they are no longer
  // needed and can inhibit further optimization.
  if (!isLTOPreLink(LTOPhase))
    FPM.addPass(DropUnnecessaryAssumesPass(/*DropDereferenceable=*/true));
 
  FPM.addPass(InferAlignmentPass());
  if (isFullLTOPostLink(LTOPhase)) {
    // The vectorizer may have significantly shortened a loop body; unroll
    // again. Unroll small loops to hide loop backedge latency and saturate any
    // parallel execution resources of an out-of-order processor. We also then
    // need to clean up redundancies and loop invariant code.
    // FIXME: It would be really good to use a loop-integrated instruction
    // combiner for cleanup here so that the unrolling and LICM can be pipelined
    // across the loop nests.
    // We do UnrollAndJam in a separate LPM to ensure it happens before unroll
    if (Opts.enable_unroll_and_jam && PTO.LoopUnrolling)
      FPM.addPass(createFunctionToLoopPassAdaptor(
          LoopUnrollAndJamPass(static_cast<int>(Level))));
    FPM.addPass(LoopUnrollPass(LoopUnrollOptions(
        static_cast<int>(Level), /*OnlyWhenForced=*/!PTO.LoopUnrolling,
        PTO.ForgetAllSCEVInLoopUnroll)));
    FPM.addPass(WarnMissedTransformationsPass());
    // Now that we are done with loop unrolling, be it either by LoopVectorizer,
    // or LoopUnroll passes, some variable-offset GEP's into alloca's could have
    // become constant-offset, thus enabling SROA and alloca promotion. Do so.
    // NOTE: we are very late in the pipeline, and we don't have any LICM
    // or SimplifyCFG passes scheduled after us, that would cleanup
    // the CFG mess this may created if allowed to modify CFG, so forbid that.
 
    // We also turn on struct to vector canonicalization here, which allows
    // converting allocas of homogeneous structs into vector allocas when the
    // allocas' users are all memory intrinsics. This allows promotion in some
    // cases because structs cannot promote to SSA values, but vectors can. We
    // only turn this on after memcpyopt runs because this might hinder
    // memcpyopt's optimizations if done before. Look at the documentation for
    // `tryCanonicalizeStructToVector` in SROA.cpp to see why.
    FPM.addPass(SROAPass(SROAOptions(SROAOptions::PreserveCFG,
                                     /*AggregateToVector=*/true)));
  }
 
  if (!isFullLTOPostLink(LTOPhase)) {
    // Eliminate loads by forwarding stores from the previous iteration to loads
    // of the current iteration.
    FPM.addPass(LoopLoadEliminationPass());
  }
  // Cleanup after the loop optimization passes.
  FPM.addPass(InstCombinePass());
 
  if (Level > OptimizationLevel::O1 && Opts.extra_vectorizer_passes) {
    ExtraFunctionPassManager<ShouldRunExtraVectorPasses> ExtraPasses;
    // At higher optimization levels, try to clean up any runtime overlap and
    // alignment checks inserted by the vectorizer. We want to track correlated
    // runtime checks for two inner loops in the same outer loop, fold any
    // common computations, hoist loop-invariant aspects out of any outer loop,
    // and unswitch the runtime checks if possible. Once hoisted, we may have
    // dead (or speculatable) control flows or more combining opportunities.
    ExtraPasses.addPass(EarlyCSEPass());
    ExtraPasses.addPass(CorrelatedValuePropagationPass());
    ExtraPasses.addPass(InstCombinePass());
    LoopPassManager LPM;
    LPM.addPass(LICMPass(PTO.LicmMssaOptCap, PTO.LicmMssaNoAccForPromotionCap,
                         /*AllowSpeculation=*/true));
    LPM.addPass(SimpleLoopUnswitchPass(/* NonTrivial */ Level ==
                                       OptimizationLevel::O3));
    ExtraPasses.addPass(
        createFunctionToLoopPassAdaptor(std::move(LPM), /*UseMemorySSA=*/true));
    ExtraPasses.addPass(
        SimplifyCFGPass(SimplifyCFGOptions().convertSwitchRangeToICmp(true)));
    ExtraPasses.addPass(InstCombinePass());
    FPM.addPass(std::move(ExtraPasses));
  }
 
  // Now that we've formed fast to execute loop structures, we do further
  // optimizations. These are run afterward as they might block doing complex
  // analyses and transforms such as what are needed for loop vectorization.
 
  // Cleanup after loop vectorization, etc. Simplification passes like CVP and
  // GVN, loop transforms, and others have already run, so it's now better to
  // convert to more optimized IR using more aggressive simplify CFG options.
  // The extra sinking transform can create larger basic blocks, so do this
  // before SLP vectorization.
  FPM.addPass(SimplifyCFGPass(SimplifyCFGOptions()
                                  .forwardSwitchCondToPhi(true)
                                  .convertSwitchRangeToICmp(true)
                                  .convertSwitchToArithmetic(true)
                                  .convertSwitchToLookupTable(true)
                                  .needCanonicalLoops(false)
                                  .hoistCommonInsts(true)
                                  .sinkCommonInsts(true)));
 
  if (isFullLTOPostLink(LTOPhase)) {
    FPM.addPass(SCCPPass());
    FPM.addPass(InstCombinePass());
    FPM.addPass(BDCEPass());
  }
 
  // Optimize parallel scalar instruction chains into SIMD instructions.
  if (PTO.SLPVectorization) {
    FPM.addPass(SLPVectorizerPass());
    if (Level >= OptimizationLevel::O2 && Opts.extra_vectorizer_passes) {
      FPM.addPass(EarlyCSEPass());
    }
  }
  // Enhance/cleanup vector code.
  FPM.addPass(VectorCombinePass());
 
  if (!isFullLTOPostLink(LTOPhase)) {
    FPM.addPass(InstCombinePass());
    // Unroll small loops to hide loop backedge latency and saturate any
    // parallel execution resources of an out-of-order processor. We also then
    // need to clean up redundancies and loop invariant code.
    // FIXME: It would be really good to use a loop-integrated instruction
    // combiner for cleanup here so that the unrolling and LICM can be pipelined
    // across the loop nests.
    // We do UnrollAndJam in a separate LPM to ensure it happens before unroll
    if (Opts.enable_unroll_and_jam && PTO.LoopUnrolling) {
      FPM.addPass(createFunctionToLoopPassAdaptor(
          LoopUnrollAndJamPass(static_cast<int>(Level))));
    }
    FPM.addPass(LoopUnrollPass(LoopUnrollOptions(
        static_cast<int>(Level), /*OnlyWhenForced=*/!PTO.LoopUnrolling,
        PTO.ForgetAllSCEVInLoopUnroll)));
    FPM.addPass(WarnMissedTransformationsPass());
    // Now that we are done with loop unrolling, be it either by LoopVectorizer,
    // or LoopUnroll passes, some variable-offset GEP's into alloca's could have
    // become constant-offset, thus enabling SROA and alloca promotion. Do so.
    // NOTE: we are very late in the pipeline, and we don't have any LICM
    // or SimplifyCFG passes scheduled after us, that would cleanup
    // the CFG mess this may created if allowed to modify CFG, so forbid that.
 
    // We also turn on struct to vector canonicalization here, which allows
    // converting allocas of homogeneous structs into vector allocas when the
    // allocas' users are all memory intrinsics. This allows promotion in some
    // cases because structs cannot promote to SSA values, but vectors can. We
    // only turn this on after memcpyopt runs because this might hinder
    // memcpyopt's optimizations if done before. Look at the documentation for
    // `tryCanonicalizeStructToVector` in SROA.cpp to see why.
    FPM.addPass(SROAPass(SROAOptions(SROAOptions::PreserveCFG,
                                     /*AggregateToVector=*/true)));
  }
 
  FPM.addPass(InferAlignmentPass());
  FPM.addPass(InstCombinePass());
 
  // This is needed for two reasons:
  //   1. It works around problems that instcombine introduces, such as sinking
  //      expensive FP divides into loops containing multiplications using the
  //      divide result.
  //   2. It helps to clean up some loop-invariant code created by the loop
  //      unroll pass when IsFullLTO=false.
  FPM.addPass(createFunctionToLoopPassAdaptor(
      LICMPass(PTO.LicmMssaOptCap, PTO.LicmMssaNoAccForPromotionCap,
               /*AllowSpeculation=*/true),
      /*UseMemorySSA=*/true));
 
  // Now that we've vectorized and unrolled loops, we may have more refined
  // alignment information, try to re-derive it here.
  FPM.addPass(AlignmentFromAssumptionsPass());
}
 
ModulePassManager
PassBuilder::buildModuleOptimizationPipeline(OptimizationLevel Level,
                                             ThinOrFullLTOPhase LTOPhase) {
  ModulePassManager MPM;
 
  // Run partial inlining pass to partially inline functions that have
  // large bodies.
  if (Opts.enable_partial_inlining)
    MPM.addPass(PartialInlinerPass());
 
  // Remove avail extern fns and globals definitions since we aren't compiling
  // an object file for later LTO. For LTO we want to preserve these so they
  // are eligible for inlining at link-time. Note if they are unreferenced they
  // will be removed by GlobalDCE later, so this only impacts referenced
  // available externally globals. Eventually they will be suppressed during
  // codegen, but eliminating here enables more opportunity for GlobalDCE as it
  // may make globals referenced by available external functions dead and saves
  // running remaining passes on the eliminated functions. These should be
  // preserved during prelinking for link-time inlining decisions.
  if (!isLTOPreLink(LTOPhase))
    MPM.addPass(EliminateAvailableExternallyPass());
 
  // Do RPO function attribute inference across the module to forward-propagate
  // attributes where applicable.
  // FIXME: Is this really an optimization rather than a canonicalization?
  MPM.addPass(ReversePostOrderFunctionAttrsPass());
 
  // Do a post inline PGO instrumentation and use pass. This is a context
  // sensitive PGO pass. We don't want to do this in LTOPreLink phrase as
  // cross-module inline has not been done yet. The context sensitive
  // instrumentation is after all the inlines are done.
  if (!isLTOPreLink(LTOPhase) && PGOOpt) {
    if (PGOOpt->CSAction == PGOOptions::CSIRInstr)
      addPGOInstrPasses(MPM, Level, /*RunProfileGen=*/true,
                        /*IsCS=*/true, PGOOpt->AtomicCounterUpdate,
                        PGOOpt->CSProfileGenFile, PGOOpt->ProfileRemappingFile);
    else if (PGOOpt->CSAction == PGOOptions::CSIRUse)
      addPGOInstrPasses(MPM, Level, /*RunProfileGen=*/false,
                        /*IsCS=*/true, PGOOpt->AtomicCounterUpdate,
                        PGOOpt->ProfileFile, PGOOpt->ProfileRemappingFile);
  }
 
  // Re-compute GlobalsAA here prior to function passes. This is particularly
  // useful as the above will have inlined, DCE'ed, and function-attr
  // propagated everything. We should at this point have a reasonably minimal
  // and richly annotated call graph. By computing aliasing and mod/ref
  // information for all local globals here, the late loop passes and notably
  // the vectorizer will be able to use them to help recognize vectorizable
  // memory operations.
  if (Opts.enable_global_analyses)
    MPM.addPass(RecomputeGlobalsAAPass());
 
  invokeOptimizerEarlyEPCallbacks(MPM, Level, LTOPhase);
 
  FunctionPassManager OptimizePM;
 
  // Only drop unnecessary assumes post-inline and post-link, as otherwise
  // additional uses of the affected value may be introduced through inlining
  // and CSE.
  if (!isLTOPreLink(LTOPhase))
    OptimizePM.addPass(DropUnnecessaryAssumesPass());
 
  // Scheduling LoopVersioningLICM when inlining is over, because after that
  // we may see more accurate aliasing. Reason to run this late is that too
  // early versioning may prevent further inlining due to increase of code
  // size. Other optimizations which runs later might get benefit of no-alias
  // assumption in clone loop.
  if (Opts.enable_loop_versioning_licm) {
    OptimizePM.addPass(
        createFunctionToLoopPassAdaptor(LoopVersioningLICMPass()));
    // LoopVersioningLICM pass might increase new LICM opportunities.
    OptimizePM.addPass(createFunctionToLoopPassAdaptor(
        LICMPass(PTO.LicmMssaOptCap, PTO.LicmMssaNoAccForPromotionCap,
                 /*AllowSpeculation=*/true),
        /*USeMemorySSA=*/true));
  }
 
  OptimizePM.addPass(Float2IntPass());
  // Defer until LTO post-link where some constants may become known.
  if (!isLTOPreLink(LTOPhase))
    OptimizePM.addPass(LowerConstantIntrinsicsPass());
 
  if (Opts.enable_matrix) {
    OptimizePM.addPass(LowerMatrixIntrinsicsPass());
    OptimizePM.addPass(EarlyCSEPass());
  }
 
  // CHR pass should only be applied with the profile information.
  // The check is to check the profile summary information in CHR.
  if (Opts.enable_chr && Level == OptimizationLevel::O3)
    OptimizePM.addPass(ControlHeightReductionPass());
 
  // FIXME: We need to run some loop optimizations to re-rotate loops after
  // simplifycfg and others undo their rotation.
 
  // Optimize the loop execution. These passes operate on entire loop nests
  // rather than on each loop in an inside-out manner, and so they are actually
  // function passes.
 
  invokeVectorizerStartEPCallbacks(OptimizePM, Level);
 
  LoopPassManager LPM;
  // First rotate loops that may have been un-rotated by prior passes.
  // Disable header duplication at -Oz.
  LPM.addPass(LoopRotatePass(/*EnableLoopHeaderDuplication=*/true,
                             isLTOPreLink(LTOPhase),
                             /*CheckExitCount=*/true));
  // Some loops may have become dead by now. Try to delete them.
  // FIXME: see discussion in https://reviews.llvm.org/D112851,
  //        this may need to be revisited once we run GVN before loop deletion
  //        in the simplification pipeline.
  LPM.addPass(LoopDeletionPass());
 
  if (PTO.LoopInterchange)
    LPM.addPass(LoopInterchangePass());
 
  OptimizePM.addPass(
      createFunctionToLoopPassAdaptor(std::move(LPM), /*UseMemorySSA=*/false));
 
  // FIXME: This may not be the right place in the pipeline.
  // We need to have the data to support the right place.
  if (PTO.LoopFusion)
    OptimizePM.addPass(LoopFusePass());
 
  // Distribute loops to allow partial vectorization.  I.e. isolate dependences
  // into separate loop that would otherwise inhibit vectorization.  This is
  // currently only performed for loops marked with the metadata
  // llvm.loop.distribute=true or when -enable-loop-distribute is specified.
  OptimizePM.addPass(LoopDistributePass());
 
  // Populates the VFABI attribute with the scalar-to-vector mappings
  // from the TargetLibraryInfo.
  OptimizePM.addPass(InjectTLIMappings());
 
  addVectorPasses(Level, OptimizePM, LTOPhase);
 
  invokeVectorizerEndEPCallbacks(OptimizePM, Level);
 
  // LoopSink pass sinks instructions hoisted by LICM, which serves as a
  // canonicalization pass that enables other optimizations. As a result,
  // LoopSink pass needs to be a very late IR pass to avoid undoing LICM
  // result too early.
  OptimizePM.addPass(LoopSinkPass());
 
  // And finally clean up LCSSA form before generating code.
  OptimizePM.addPass(InstSimplifyPass());
 
  // This hoists/decomposes div/rem ops. It should run after other sink/hoist
  // passes to avoid re-sinking, but before SimplifyCFG because it can allow
  // flattening of blocks.
  OptimizePM.addPass(DivRemPairsPass());
 
  // Merge adjacent icmps into memcmp, then expand memcmp to loads/compares.
  // TODO: move this furter up so that it can be optimized by GVN, etc.
  if (Opts.enable_mergeicmps)
    OptimizePM.addPass(MergeICmpsPass());
  OptimizePM.addPass(ExpandMemCmpPass());
 
  // Try to annotate calls that were created during optimization.
  OptimizePM.addPass(
      TailCallElimPass(/*UpdateFunctionEntryCount=*/isInstrumentedPGOUse()));
 
  // LoopSink (and other loop passes since the last simplifyCFG) might have
  // resulted in single-entry-single-exit or empty blocks. Clean up the CFG.
  OptimizePM.addPass(
      SimplifyCFGPass(SimplifyCFGOptions()
                          .convertSwitchRangeToICmp(true)
                          .convertSwitchToArithmetic(true)
                          .speculateUnpredictables(true)
                          .hoistLoadsStoresWithCondFaulting(true)));
 
  // Add the core optimizing pipeline.
  MPM.addPass(createModuleToFunctionPassAdaptor(std::move(OptimizePM),
                                                PTO.EagerlyInvalidateAnalyses));
 
  // AllocToken transforms heap allocation calls; this needs to run late after
  // other allocation call transformations (such as those in InstCombine).
  if (!isLTOPreLink(LTOPhase))
    MPM.addPass(AllocTokenPass());
 
  invokeOptimizerLastEPCallbacks(MPM, Level, LTOPhase);
 
  // Run the Instrumentor pass late.
  if (Opts.enable_instrumentor)
    MPM.addPass(InstrumentorPass(FS));
 
  // Split out cold code. Splitting is done late to avoid hiding context from
  // other optimizations and inadvertently regressing performance. The tradeoff
  // is that this has a higher code size cost than splitting early.
  if (Opts.hot_cold_split && !isLTOPreLink(LTOPhase))
    MPM.addPass(HotColdSplittingPass());
 
  // Now we need to do some global optimization transforms.
  // FIXME: It would seem like these should come first in the optimization
  // pipeline and maybe be the bottom of the canonicalization pipeline? Weird
  // ordering here.
  MPM.addPass(GlobalDCEPass());
  MPM.addPass(ConstantMergePass());
 
  // Merge functions if requested. It has a better chance to merge functions
  // after ConstantMerge folded jump tables.
  if (PTO.MergeFunctions)
    MPM.addPass(MergeFunctionsPass());
 
  if (PTO.CallGraphProfile && !isLTOPreLink(LTOPhase))
    MPM.addPass(CGProfilePass(isLTOPostLink(LTOPhase)));
 
  // RelLookupTableConverterPass runs later in LTO post-link pipeline.
  if (!isLTOPreLink(LTOPhase))
    MPM.addPass(RelLookupTableConverterPass());
 
  // Add devirtualization pass only when LTO is not enabled, as otherwise
  // the pass is already enabled in the LTO pipeline.
  if (PTO.DevirtualizeSpeculatively && LTOPhase == ThinOrFullLTOPhase::None) {
    // TODO: explore a better pipeline configuration that can improve
    // compilation time overhead.
    // FIXME: move this earlier (lots of pass ordering tests will need fixing)
    MPM.addPass(AssignGUIDPass());
    MPM.addPass(WholeProgramDevirtPass(
        /*ExportSummary*/ nullptr,
        /*ImportSummary*/ nullptr,
        /*DevirtSpeculatively*/ PTO.DevirtualizeSpeculatively));
    MPM.addPass(DropTypeTestsPass());
    // Given that the devirtualization creates more opportunities for inlining,
    // we run the Inliner again here to maximize the optimization gain we
    // get from devirtualization.
    // Also, we can't run devirtualization before inlining because the
    // devirtualization depends on the passes optimizing/eliminating vtable GVs
    // and those passes are only effective after inlining.
    addModuleInlinerPass(MPM, Opts, Level, ThinOrFullLTOPhase::None);
  }
 
  // Attach !implicit.ref metadata from all functions to copyright strings.
  MPM.addPass(LowerCommentStringPass());
 
  return MPM;
}
 
ModulePassManager
PassBuilder::buildPerModuleDefaultPipeline(OptimizationLevel Level,
                                           ThinOrFullLTOPhase Phase) {
  if (Level == OptimizationLevel::O0)
    return buildO0DefaultPipeline(Level, Phase);
 
  ModulePassManager MPM;
  instructionCountersPass(MPM, /* IsPreOptimization */ true);
  // Currently this pipeline is only invoked in an LTO pre link pass or when we
  // are not running LTO. If that changes the below checks may need updating.
  assert(isLTOPreLink(Phase) || Phase == ThinOrFullLTOPhase::None);
 
  // If we are invoking this in non-LTO mode, remove any MemProf related
  // attributes and metadata, as we don't know whether we are linking with
  // a library containing the necessary interfaces.
  if (Phase == ThinOrFullLTOPhase::None)
    MPM.addPass(MemProfRemoveInfo());
 
  // Convert @llvm.global.annotations to !annotation metadata.
  MPM.addPass(Annotation2MetadataPass());
 
  // Force any function attributes we want the rest of the pipeline to observe.
  MPM.addPass(ForceFunctionAttrsPass());
 
  if (Opts.opt_pipeline_trigger_crash)
    MPM.addPass(createModuleToFunctionPassAdaptor(TriggerCrashFunctionPass()));
 
  if (PGOOpt && PGOOpt->DebugInfoForProfiling)
    MPM.addPass(createModuleToFunctionPassAdaptor(AddDiscriminatorsPass()));
 
  // Apply module pipeline start EP callback.
  invokePipelineStartEPCallbacks(MPM, Level);
 
  // Add the core simplification pipeline.
  MPM.addPass(buildModuleSimplificationPipeline(Level, Phase));
 
  // Now add the optimization pipeline.
  MPM.addPass(buildModuleOptimizationPipeline(Level, Phase));
 
  if (PGOOpt && PGOOpt->PseudoProbeForProfiling &&
      PGOOpt->Action == PGOOptions::SampleUse)
    MPM.addPass(PseudoProbeUpdatePass());
 
  // Emit annotation remarks.
  addAnnotationRemarksPass(MPM);
 
  if (isLTOPreLink(Phase))
    addRequiredLTOPreLinkPasses(MPM);
 
  instructionCountersPass(MPM, /* IsPreOptimization */ false);
  return MPM;
}
 
ModulePassManager
PassBuilder::buildFatLTODefaultPipeline(OptimizationLevel Level, bool ThinLTO,
                                        bool EmitSummary, bool Verify) {
  ModulePassManager MPM;
 
  instructionCountersPass(MPM, /* IsPreOptimization */ true);
 
  if (ThinLTO)
    MPM.addPass(buildThinLTOPreLinkDefaultPipeline(Level));
  else
    MPM.addPass(buildLTOPreLinkDefaultPipeline(Level));
  // AssignGUIDPass attaches !guid metadata (MD_unique_id) to global objects,
  // triggering the bitcode writer to emit a METADATA_KIND_BLOCK. Standard LTO
  // bitcode emission runs VerifierPass by default, which registers metadata
  // kind IDs in LLVMContext. Running VerifierPass here before EmbedBitcodePass
  // to get the same behavior.
  if (Verify)
    MPM.addPass(VerifierPass());
  MPM.addPass(EmbedBitcodePass(ThinLTO, EmitSummary));
 
  // Perform any cleanups to the IR that aren't suitable for per TU compilation,
  // like removing CFI/WPD related instructions. Note, we reuse
  // DropTypeTestsPass to clean up type tests rather than duplicate that logic
  // in FatLtoCleanup.
  MPM.addPass(FatLtoCleanup());
 
  // If we're doing FatLTO w/ CFI enabled, we don't want the type tests in the
  // object code, only in the bitcode section, so drop it before we run
  // module optimization and generate machine code. If llvm.type.test() isn't in
  // the IR, this won't do anything.
  MPM.addPass(DropTypeTestsPass(lowertypetests::DropTestKind::All));
 
  // Use the ThinLTO post-link pipeline with sample profiling
  if (ThinLTO && PGOOpt && PGOOpt->Action == PGOOptions::SampleUse)
    MPM.addPass(buildThinLTODefaultPipeline(Level, /*ImportSummary=*/nullptr));
  else {
    // ModuleSimplification does not run the coroutine passes for
    // ThinLTOPreLink, so we need the coroutine passes to run for ThinLTO
    // builds, otherwise they will miscompile.
    if (ThinLTO) {
      // TODO: replace w/ buildCoroWrapper() when it takes phase and level into
      // consideration.
      CGSCCPassManager CGPM;
      CGPM.addPass(CoroSplitPass(Level != OptimizationLevel::O0));
      CGPM.addPass(CoroAnnotationElidePass());
      MPM.addPass(createModuleToPostOrderCGSCCPassAdaptor(std::move(CGPM)));
      MPM.addPass(CoroCleanupPass());
    }
 
    // otherwise, just use module optimization
    MPM.addPass(
        buildModuleOptimizationPipeline(Level, ThinOrFullLTOPhase::None));
    // Emit annotation remarks.
    addAnnotationRemarksPass(MPM);
  }
 
  instructionCountersPass(MPM, /* IsPreOptimization */ false);
 
  return MPM;
}
 
ModulePassManager
PassBuilder::buildThinLTOPreLinkDefaultPipeline(OptimizationLevel Level) {
  if (Level == OptimizationLevel::O0)
    return buildO0DefaultPipeline(Level, ThinOrFullLTOPhase::ThinLTOPreLink);
 
  ModulePassManager MPM;
 
  instructionCountersPass(MPM, /* IsPreOptimization */ true);
 
  // Convert @llvm.global.annotations to !annotation metadata.
  MPM.addPass(Annotation2MetadataPass());
 
  // Force any function attributes we want the rest of the pipeline to observe.
  MPM.addPass(ForceFunctionAttrsPass());
 
  if (PGOOpt && PGOOpt->DebugInfoForProfiling)
    MPM.addPass(createModuleToFunctionPassAdaptor(AddDiscriminatorsPass()));
 
  // Apply module pipeline start EP callback.
  invokePipelineStartEPCallbacks(MPM, Level);
 
  // If we are planning to perform ThinLTO later, we don't bloat the code with
  // unrolling/vectorization/... now. Just simplify the module as much as we
  // can.
  MPM.addPass(buildModuleSimplificationPipeline(
      Level, ThinOrFullLTOPhase::ThinLTOPreLink));
  // In pre-link, for ctx prof use, we stop here with an instrumented IR. We let
  // thinlto use the contextual info to perform imports; then use the contextual
  // profile in the post-thinlink phase.
  if (!UseCtxProfile.empty()) {
    addRequiredLTOPreLinkPasses(MPM);
    return MPM;
  }
 
  // Run partial inlining pass to partially inline functions that have
  // large bodies.
  // FIXME: It isn't clear whether this is really the right place to run this
  // in ThinLTO. Because there is another canonicalization and simplification
  // phase that will run after the thin link, running this here ends up with
  // less information than will be available later and it may grow functions in
  // ways that aren't beneficial.
  if (Opts.enable_partial_inlining)
    MPM.addPass(PartialInlinerPass());
 
  if (PGOOpt && PGOOpt->PseudoProbeForProfiling &&
      PGOOpt->Action == PGOOptions::SampleUse)
    MPM.addPass(PseudoProbeUpdatePass());
 
  // Handle Optimizer{Early,Last}EPCallbacks added by clang on PreLink. Actual
  // optimization is going to be done in PostLink stage, but clang can't add
  // callbacks there in case of in-process ThinLTO called by linker.
  invokeOptimizerEarlyEPCallbacks(MPM, Level,
                                  /*Phase=*/ThinOrFullLTOPhase::ThinLTOPreLink);
  invokeOptimizerLastEPCallbacks(MPM, Level,
                                 /*Phase=*/ThinOrFullLTOPhase::ThinLTOPreLink);
 
  // Emit annotation remarks.
  addAnnotationRemarksPass(MPM);
 
  // Attach !implicit.ref metadata from all functions to copyright strings.
  MPM.addPass(LowerCommentStringPass());
 
  addRequiredLTOPreLinkPasses(MPM);
 
  instructionCountersPass(MPM, /* IsPreOptimization */ false);
 
  return MPM;
}
 
ModulePassManager PassBuilder::buildThinLTODefaultPipeline(
    OptimizationLevel Level, const ModuleSummaryIndex *ImportSummary) {
  ModulePassManager MPM;
 
  instructionCountersPass(MPM, /* IsPreOptimization */ true);
 
  invokeThinLinkTimeOptimizationEarlyEPCallbacks(MPM, Level);
 
  // If we are invoking this without a summary index noting that we are linking
  // with a library containing the necessary APIs, remove any MemProf related
  // attributes and metadata.
  if (!ImportSummary || !ImportSummary->withSupportsHotColdNew())
    MPM.addPass(MemProfRemoveInfo());
 
  if (ImportSummary) {
    // For ThinLTO we must apply the context disambiguation decisions early, to
    // ensure we can correctly match the callsites to summary data.
    if (EnableMemProfContextDisambiguation)
      MPM.addPass(MemProfContextDisambiguation(
          ImportSummary, PGOOpt && PGOOpt->Action == PGOOptions::SampleUse));
 
    // These passes import type identifier resolutions for whole-program
    // devirtualization and CFI. They must run early because other passes may
    // disturb the specific instruction patterns that these passes look for,
    // creating dependencies on resolutions that may not appear in the summary.
    //
    // For example, GVN may transform the pattern assume(type.test) appearing in
    // two basic blocks into assume(phi(type.test, type.test)), which would
    // transform a dependency on a WPD resolution into a dependency on a type
    // identifier resolution for CFI.
    //
    // Also, WPD has access to more precise information than ICP and can
    // devirtualize more effectively, so it should operate on the IR first.
    //
    // The WPD and LowerTypeTest passes need to run at -O0 to lower type
    // metadata and intrinsics.
    MPM.addPass(WholeProgramDevirtPass(nullptr, ImportSummary));
    MPM.addPass(LowerTypeTestsPass(nullptr, ImportSummary));
  }
 
  if (Level == OptimizationLevel::O0) {
    // Run a second time to clean up any type tests left behind by WPD for use
    // in ICP.
    MPM.addPass(DropTypeTestsPass());
    MPM.addPass(buildCoroWrapper(ThinOrFullLTOPhase::ThinLTOPostLink));
 
    // AllocToken transforms heap allocation calls; this needs to run late after
    // other allocation call transformations (such as those in InstCombine).
    MPM.addPass(AllocTokenPass());
 
    // Drop available_externally and unreferenced globals. This is necessary
    // with ThinLTO in order to avoid leaving undefined references to dead
    // globals in the object file.
    MPM.addPass(EliminateAvailableExternallyPass());
    MPM.addPass(GlobalDCEPass());
 
    invokeThinLinkTimeOptimizationLastEPCallbacks(MPM, Level);
 
    return MPM;
  }
  if (!UseCtxProfile.empty()) {
    MPM.addPass(
        buildModuleInlinerPipeline(Level, ThinOrFullLTOPhase::ThinLTOPostLink));
  } else {
    // Add the core simplification pipeline.
    MPM.addPass(buildModuleSimplificationPipeline(
        Level, ThinOrFullLTOPhase::ThinLTOPostLink));
  }
  // Now add the optimization pipeline.
  MPM.addPass(buildModuleOptimizationPipeline(
      Level, ThinOrFullLTOPhase::ThinLTOPostLink));
 
  invokeThinLinkTimeOptimizationLastEPCallbacks(MPM, Level);
 
  // Emit annotation remarks.
  addAnnotationRemarksPass(MPM);
 
  instructionCountersPass(MPM, /* IsPreOptimization */ false);
 
  return MPM;
}
 
ModulePassManager
PassBuilder::buildLTOPreLinkDefaultPipeline(OptimizationLevel Level) {
  // FIXME: We should use a customized pre-link pipeline!
  return buildPerModuleDefaultPipeline(Level,
                                       ThinOrFullLTOPhase::FullLTOPreLink);
}
 
ModulePassManager
PassBuilder::buildLTODefaultPipeline(OptimizationLevel Level,
                                     ModuleSummaryIndex *ExportSummary) {
  ModulePassManager MPM;
 
  instructionCountersPass(MPM, /* IsPreOptimization */ true);
 
  invokeFullLinkTimeOptimizationEarlyEPCallbacks(MPM, Level);
 
  // If we are invoking this without a summary index noting that we are linking
  // with a library containing the necessary APIs, remove any MemProf related
  // attributes and metadata.
  if (!ExportSummary || !ExportSummary->withSupportsHotColdNew())
    MPM.addPass(MemProfRemoveInfo());
 
  // Create a function that performs CFI checks for cross-DSO calls with targets
  // in the current module.
  MPM.addPass(CrossDSOCFIPass());
 
  if (Level == OptimizationLevel::O0) {
    // The WPD and LowerTypeTest passes need to run at -O0 to lower type
    // metadata and intrinsics.
    MPM.addPass(WholeProgramDevirtPass(ExportSummary, nullptr));
    MPM.addPass(LowerTypeTestsPass(ExportSummary, nullptr));
    // Run a second time to clean up any type tests left behind by WPD for use
    // in ICP.
    MPM.addPass(DropTypeTestsPass());
 
    MPM.addPass(buildCoroWrapper(ThinOrFullLTOPhase::FullLTOPostLink));
 
    // AllocToken transforms heap allocation calls; this needs to run late after
    // other allocation call transformations (such as those in InstCombine).
    MPM.addPass(AllocTokenPass());
 
    invokeFullLinkTimeOptimizationLastEPCallbacks(MPM, Level);
 
    // Emit annotation remarks.
    addAnnotationRemarksPass(MPM);
 
    return MPM;
  }
 
  if (PGOOpt && PGOOpt->Action == PGOOptions::SampleUse) {
    // Load sample profile before running the LTO optimization pipeline.
    MPM.addPass(SampleProfileLoaderPass(PGOOpt->ProfileFile,
                                        PGOOpt->ProfileRemappingFile,
                                        ThinOrFullLTOPhase::FullLTOPostLink));
    // Cache ProfileSummaryAnalysis once to avoid the potential need to insert
    // RequireAnalysisPass for PSI before subsequent non-module passes.
    MPM.addPass(RequireAnalysisPass<ProfileSummaryAnalysis, Module>());
  }
 
  // Try to run OpenMP optimizations, quick no-op if no OpenMP metadata present.
  MPM.addPass(OpenMPOptPass(ThinOrFullLTOPhase::FullLTOPostLink));
 
  // Remove unused virtual tables to improve the quality of code generated by
  // whole-program devirtualization and bitset lowering.
  MPM.addPass(GlobalDCEPass(/*InLTOPostLink=*/true));
 
  // Do basic inference of function attributes from known properties of system
  // libraries and other oracles.
  MPM.addPass(InferFunctionAttrsPass());
 
  if (Level >= OptimizationLevel::O2) {
    MPM.addPass(createModuleToFunctionPassAdaptor(
        CallSiteSplittingPass(), PTO.EagerlyInvalidateAnalyses));
 
    // Indirect call promotion. This should promote all the targets that are
    // left by the earlier promotion pass that promotes intra-module targets.
    // This two-step promotion is to save the compile time. For LTO, it should
    // produce the same result as if we only do promotion here.
    MPM.addPass(PGOIndirectCallPromotion(
        true /* InLTO */, PGOOpt && PGOOpt->Action == PGOOptions::SampleUse));
 
    // Promoting by-reference arguments to by-value exposes more constants to
    // IPSCCP.
    CGSCCPassManager CGPM;
    CGPM.addPass(PostOrderFunctionAttrsPass());
    CGPM.addPass(ArgumentPromotionPass());
    CGPM.addPass(
        createCGSCCToFunctionPassAdaptor(SROAPass(SROAOptions::ModifyCFG)));
    MPM.addPass(createModuleToPostOrderCGSCCPassAdaptor(std::move(CGPM)));
 
    // Propagate constants at call sites into the functions they call.  This
    // opens opportunities for globalopt (and inlining) by substituting function
    // pointers passed as arguments to direct uses of functions.
    MPM.addPass(IPSCCPPass(IPSCCPOptions(/*AllowFuncSpec=*/true)));
 
    // Attach metadata to indirect call sites indicating the set of functions
    // they may target at run-time. This should follow IPSCCP.
    MPM.addPass(CalledValuePropagationPass());
  }
 
  // Do RPO function attribute inference across the module to forward-propagate
  // attributes where applicable.
  // FIXME: Is this really an optimization rather than a canonicalization?
  MPM.addPass(ReversePostOrderFunctionAttrsPass());
 
  // Use in-range annotations on GEP indices to split globals where beneficial.
  MPM.addPass(GlobalSplitPass());
 
  // Run whole program optimization of virtual call when the list of callees
  // is fixed.
  MPM.addPass(WholeProgramDevirtPass(ExportSummary, nullptr));
 
  MPM.addPass(NoRecurseLTOInferencePass());
  // Stop here at -O1.
  if (Level == OptimizationLevel::O1) {
    MPM.addPass(createModuleToFunctionPassAdaptor(
        LowerConstantIntrinsicsPass(), PTO.EagerlyInvalidateAnalyses));
 
    // The LowerTypeTestsPass needs to run to lower type metadata and the
    // type.test intrinsics. The pass does nothing if CFI is disabled.
    MPM.addPass(LowerTypeTestsPass(ExportSummary, nullptr));
    // Run a second time to clean up any type tests left behind by WPD for use
    // in ICP (which is performed earlier than this in the regular LTO
    // pipeline).
    MPM.addPass(DropTypeTestsPass());
 
    MPM.addPass(buildCoroWrapper(ThinOrFullLTOPhase::FullLTOPostLink));
 
    // AllocToken transforms heap allocation calls; this needs to run late after
    // other allocation call transformations (such as those in InstCombine).
    MPM.addPass(AllocTokenPass());
 
    invokeFullLinkTimeOptimizationLastEPCallbacks(MPM, Level);
 
    // Emit annotation remarks.
    addAnnotationRemarksPass(MPM);
 
    instructionCountersPass(MPM, /* IsPreOptimization */ false);
 
    return MPM;
  }
 
  // TODO: Skip to match buildCoroWrapper.
  MPM.addPass(CoroEarlyPass());
 
  // Optimize globals to try and fold them into constants.
  MPM.addPass(GlobalOptPass());
 
  // Promote any localized globals to SSA registers.
  MPM.addPass(createModuleToFunctionPassAdaptor(PromotePass()));
 
  // Linking modules together can lead to duplicate global constant, only
  // keep one copy of each constant.
  MPM.addPass(ConstantMergePass());
 
  // Remove unused arguments from functions.
  MPM.addPass(DeadArgumentEliminationPass());
 
  // Reduce the code after globalopt and ipsccp.  Both can open up significant
  // simplification opportunities, and both can propagate functions through
  // function pointers.  When this happens, we often have to resolve varargs
  // calls, etc, so let instcombine do this.
  FunctionPassManager PeepholeFPM;
  PeepholeFPM.addPass(InstCombinePass());
  if (Level >= OptimizationLevel::O2)
    PeepholeFPM.addPass(AggressiveInstCombinePass());
  invokePeepholeEPCallbacks(PeepholeFPM, Level);
 
  MPM.addPass(createModuleToFunctionPassAdaptor(std::move(PeepholeFPM),
                                                PTO.EagerlyInvalidateAnalyses));
 
  // Lower variadic functions for supported targets prior to inlining.
  MPM.addPass(ExpandVariadicsPass(ExpandVariadicsMode::Optimize));
 
  // Note: historically, the PruneEH pass was run first to deduce nounwind and
  // generally clean up exception handling overhead. It isn't clear this is
  // valuable as the inliner doesn't currently care whether it is inlining an
  // invoke or a call.
  // Run the inliner now.
  addModuleInlinerPass(MPM, Opts, Level, ThinOrFullLTOPhase::FullLTOPostLink);
 
  // Perform context disambiguation after inlining, since that would reduce the
  // amount of additional cloning required to distinguish the allocation
  // contexts.
  if (EnableMemProfContextDisambiguation)
    MPM.addPass(MemProfContextDisambiguation(
        /*Summary=*/nullptr,
        PGOOpt && PGOOpt->Action == PGOOptions::SampleUse));
 
  // Optimize globals again after we ran the inliner.
  MPM.addPass(GlobalOptPass());
 
  // Run the OpenMPOpt pass again after global optimizations.
  MPM.addPass(OpenMPOptPass(ThinOrFullLTOPhase::FullLTOPostLink));
 
  // Garbage collect dead functions.
  MPM.addPass(GlobalDCEPass(/*InLTOPostLink=*/true));
 
  // If we didn't decide to inline a function, check to see if we can
  // transform it to pass arguments by value instead of by reference.
  CGSCCPassManager CGPM;
  CGPM.addPass(ArgumentPromotionPass());
  CGPM.addPass(CoroSplitPass(Level != OptimizationLevel::O0));
  CGPM.addPass(CoroAnnotationElidePass());
  invokeCGSCCOptimizerLateEPCallbacks(CGPM, Level);
  MPM.addPass(createModuleToPostOrderCGSCCPassAdaptor(std::move(CGPM)));
 
  FunctionPassManager FPM;
  // The IPO Passes may leave cruft around. Clean up after them.
  FPM.addPass(InstCombinePass());
  invokePeepholeEPCallbacks(FPM, Level);
 
  if (Opts.enable_constraint_elimination)
    FPM.addPass(ConstraintEliminationPass());
 
  FPM.addPass(JumpThreadingPass());
 
  // Do a post inline PGO instrumentation and use pass. This is a context
  // sensitive PGO pass.
  if (PGOOpt) {
    if (PGOOpt->CSAction == PGOOptions::CSIRInstr)
      addPGOInstrPasses(MPM, Level, /*RunProfileGen=*/true,
                        /*IsCS=*/true, PGOOpt->AtomicCounterUpdate,
                        PGOOpt->CSProfileGenFile, PGOOpt->ProfileRemappingFile);
    else if (PGOOpt->CSAction == PGOOptions::CSIRUse)
      addPGOInstrPasses(MPM, Level, /*RunProfileGen=*/false,
                        /*IsCS=*/true, PGOOpt->AtomicCounterUpdate,
                        PGOOpt->ProfileFile, PGOOpt->ProfileRemappingFile);
  }
 
  // Break up allocas
  FPM.addPass(SROAPass(SROAOptions::ModifyCFG));
 
  // LTO provides additional opportunities for tailcall elimination due to
  // link-time inlining, and visibility of nocapture attribute.
  FPM.addPass(
      TailCallElimPass(/*UpdateFunctionEntryCount=*/isInstrumentedPGOUse()));
 
  // Run a few AA driver optimizations here and now to cleanup the code.
  MPM.addPass(createModuleToFunctionPassAdaptor(std::move(FPM),
                                                PTO.EagerlyInvalidateAnalyses));
 
  MPM.addPass(
      createModuleToPostOrderCGSCCPassAdaptor(PostOrderFunctionAttrsPass()));
 
  // Require the GlobalsAA analysis for the module so we can query it within
  // MainFPM.
  if (Opts.enable_global_analyses) {
    MPM.addPass(RequireAnalysisPass<GlobalsAA, Module>());
    // Invalidate AAManager so it can be recreated and pick up the newly
    // available GlobalsAA.
    MPM.addPass(
        createModuleToFunctionPassAdaptor(InvalidateAnalysisPass<AAManager>()));
  }
 
  FunctionPassManager MainFPM;
  MainFPM.addPass(createFunctionToLoopPassAdaptor(
      LICMPass(PTO.LicmMssaOptCap, PTO.LicmMssaNoAccForPromotionCap,
               /*AllowSpeculation=*/true),
      /*USeMemorySSA=*/true));
 
  if (Opts.enable_newgvn)
    MainFPM.addPass(NewGVNPass());
  else
    MainFPM.addPass(GVNPass());
 
  // Remove dead memcpy()'s.
  MainFPM.addPass(MemCpyOptPass());
 
  // Nuke dead stores.
  MainFPM.addPass(DSEPass());
  MainFPM.addPass(MoveAutoInitPass());
  MainFPM.addPass(MergedLoadStoreMotionPass());
 
  MainFPM.addPass(LowerConstantIntrinsicsPass());
 
  invokeVectorizerStartEPCallbacks(MainFPM, Level);
 
  LoopPassManager LPM;
  if (Opts.enable_loop_flatten && Level >= OptimizationLevel::O2)
    LPM.addPass(LoopFlattenPass());
  LPM.addPass(IndVarSimplifyPass());
  LPM.addPass(LoopDeletionPass());
  // FIXME: Add loop interchange.
 
  // Unroll small loops and perform peeling.
  LPM.addPass(LoopFullUnrollPass(static_cast<int>(Level),
                                 /* OnlyWhenForced= */ !PTO.LoopUnrolling,
                                 PTO.ForgetAllSCEVInLoopUnroll));
  // The loop passes in LPM (LoopFullUnrollPass) do not preserve MemorySSA.
  // *All* loop passes must preserve it, in order to be able to use it.
  MainFPM.addPass(
      createFunctionToLoopPassAdaptor(std::move(LPM), /*UseMemorySSA=*/false));
 
  MainFPM.addPass(LoopDistributePass());
 
  addVectorPasses(Level, MainFPM, ThinOrFullLTOPhase::FullLTOPostLink);
 
  invokeVectorizerEndEPCallbacks(MainFPM, Level);
 
  // Run the OpenMPOpt CGSCC pass again late.
  MPM.addPass(createModuleToPostOrderCGSCCPassAdaptor(
      OpenMPOptCGSCCPass(ThinOrFullLTOPhase::FullLTOPostLink)));
 
  invokePeepholeEPCallbacks(MainFPM, Level);
  MainFPM.addPass(JumpThreadingPass());
  MPM.addPass(createModuleToFunctionPassAdaptor(std::move(MainFPM),
                                                PTO.EagerlyInvalidateAnalyses));
 
  // Lower type metadata and the type.test intrinsic. This pass supports
  // clang's control flow integrity mechanisms (-fsanitize=cfi*) and needs
  // to be run at link time if CFI is enabled. This pass does nothing if
  // CFI is disabled.
  MPM.addPass(LowerTypeTestsPass(ExportSummary, nullptr));
  // Run a second time to clean up any type tests left behind by WPD for use
  // in ICP (which is performed earlier than this in the regular LTO pipeline).
  MPM.addPass(DropTypeTestsPass());
 
  // Enable splitting late in the FullLTO post-link pipeline.
  if (Opts.hot_cold_split)
    MPM.addPass(HotColdSplittingPass());
 
  // Add late LTO optimization passes.
  FunctionPassManager LateFPM;
 
  // LoopSink pass sinks instructions hoisted by LICM, which serves as a
  // canonicalization pass that enables other optimizations. As a result,
  // LoopSink pass needs to be a very late IR pass to avoid undoing LICM
  // result too early.
  LateFPM.addPass(LoopSinkPass());
 
  // This hoists/decomposes div/rem ops. It should run after other sink/hoist
  // passes to avoid re-sinking, but before SimplifyCFG because it can allow
  // flattening of blocks.
  LateFPM.addPass(DivRemPairsPass());
 
  // Delete basic blocks, which optimization passes may have killed.
  LateFPM.addPass(SimplifyCFGPass(SimplifyCFGOptions()
                                      .convertSwitchRangeToICmp(true)
                                      .convertSwitchToArithmetic(true)
                                      .hoistCommonInsts(true)
                                      .speculateUnpredictables(true)));
  MPM.addPass(createModuleToFunctionPassAdaptor(std::move(LateFPM)));
 
  // Drop bodies of available eternally objects to improve GlobalDCE.
  MPM.addPass(EliminateAvailableExternallyPass());
 
  // Now that we have optimized the program, discard unreachable functions.
  MPM.addPass(GlobalDCEPass(/*InLTOPostLink=*/true));
 
  if (PTO.MergeFunctions)
    MPM.addPass(MergeFunctionsPass());
 
  MPM.addPass(RelLookupTableConverterPass());
 
  if (PTO.CallGraphProfile)
    MPM.addPass(CGProfilePass(/*InLTOPostLink=*/true));
 
  MPM.addPass(CoroCleanupPass());
 
  // AllocToken transforms heap allocation calls; this needs to run late after
  // other allocation call transformations (such as those in InstCombine).
  MPM.addPass(AllocTokenPass());
 
  invokeFullLinkTimeOptimizationLastEPCallbacks(MPM, Level);
 
  // Emit annotation remarks.
  addAnnotationRemarksPass(MPM);
 
  instructionCountersPass(MPM, /* IsPreOptimization */ false);
 
  return MPM;
}
 
ModulePassManager
PassBuilder::buildO0DefaultPipeline(OptimizationLevel Level,
                                    ThinOrFullLTOPhase Phase) {
  assert(Level == OptimizationLevel::O0 &&
         "buildO0DefaultPipeline should only be used with O0");
 
  ModulePassManager MPM;
 
  instructionCountersPass(MPM, /* IsPreOptimization */ true);
 
  // Perform pseudo probe instrumentation in O0 mode. This is for the
  // consistency between different build modes. For example, a LTO build can be
  // mixed with an O0 prelink and an O2 postlink. Loading a sample profile in
  // the postlink will require pseudo probe instrumentation in the prelink.
  if (PGOOpt && PGOOpt->PseudoProbeForProfiling)
    MPM.addPass(SampleProfileProbePass(TM));
 
  if (PGOOpt && (PGOOpt->Action == PGOOptions::IRInstr ||
                 PGOOpt->Action == PGOOptions::IRUse))
    addPGOInstrPassesForO0(
        MPM,
        /*RunProfileGen=*/(PGOOpt->Action == PGOOptions::IRInstr),
        /*IsCS=*/false, PGOOpt->AtomicCounterUpdate, PGOOpt->ProfileFile,
        PGOOpt->ProfileRemappingFile);
 
  // Instrument function entry and exit before all inlining.
  MPM.addPass(createModuleToFunctionPassAdaptor(
      EntryExitInstrumenterPass(/*PostInlining=*/false)));
 
  invokePipelineStartEPCallbacks(MPM, Level);
 
  if (PGOOpt && PGOOpt->DebugInfoForProfiling)
    MPM.addPass(createModuleToFunctionPassAdaptor(AddDiscriminatorsPass()));
 
  if (PGOOpt && PGOOpt->Action == PGOOptions::SampleUse) {
    // Explicitly disable sample loader inlining and use flattened profile in O0
    // pipeline.
    MPM.addPass(SampleProfileLoaderPass(PGOOpt->ProfileFile,
                                        PGOOpt->ProfileRemappingFile,
                                        ThinOrFullLTOPhase::None, FS,
                                        /*DisableSampleProfileInlining=*/true,
                                        /*UseFlattenedProfile=*/true));
    // Cache ProfileSummaryAnalysis once to avoid the potential need to insert
    // RequireAnalysisPass for PSI before subsequent non-module passes.
    MPM.addPass(RequireAnalysisPass<ProfileSummaryAnalysis, Module>());
  }
 
  invokePipelineEarlySimplificationEPCallbacks(MPM, Level, Phase);
 
  // Build a minimal pipeline based on the semantics required by LLVM,
  // which is just that always inlining occurs. Further, disable generating
  // lifetime intrinsics to avoid enabling further optimizations during
  // code generation.
  MPM.addPass(AlwaysInlinerPass(
      /*InsertLifetimeIntrinsics=*/false));
 
  if (PTO.MergeFunctions)
    MPM.addPass(MergeFunctionsPass());
 
  if (Opts.enable_matrix)
    MPM.addPass(
        createModuleToFunctionPassAdaptor(LowerMatrixIntrinsicsPass(true)));
 
  if (!CGSCCOptimizerLateEPCallbacks.empty()) {
    CGSCCPassManager CGPM;
    invokeCGSCCOptimizerLateEPCallbacks(CGPM, Level);
    if (!CGPM.isEmpty())
      MPM.addPass(createModuleToPostOrderCGSCCPassAdaptor(std::move(CGPM)));
  }
  if (!LateLoopOptimizationsEPCallbacks.empty()) {
    LoopPassManager LPM;
    invokeLateLoopOptimizationsEPCallbacks(LPM, Level);
    if (!LPM.isEmpty()) {
      MPM.addPass(createModuleToFunctionPassAdaptor(
          createFunctionToLoopPassAdaptor(std::move(LPM))));
    }
  }
  if (!LoopOptimizerEndEPCallbacks.empty()) {
    LoopPassManager LPM;
    invokeLoopOptimizerEndEPCallbacks(LPM, Level);
    if (!LPM.isEmpty()) {
      MPM.addPass(createModuleToFunctionPassAdaptor(
          createFunctionToLoopPassAdaptor(std::move(LPM))));
    }
  }
  if (!ScalarOptimizerLateEPCallbacks.empty()) {
    FunctionPassManager FPM;
    invokeScalarOptimizerLateEPCallbacks(FPM, Level);
    if (!FPM.isEmpty())
      MPM.addPass(createModuleToFunctionPassAdaptor(std::move(FPM)));
  }
 
  invokeOptimizerEarlyEPCallbacks(MPM, Level, Phase);
 
  if (!VectorizerStartEPCallbacks.empty()) {
    FunctionPassManager FPM;
    invokeVectorizerStartEPCallbacks(FPM, Level);
    if (!FPM.isEmpty())
      MPM.addPass(createModuleToFunctionPassAdaptor(std::move(FPM)));
  }
 
  if (!VectorizerEndEPCallbacks.empty()) {
    FunctionPassManager FPM;
    invokeVectorizerEndEPCallbacks(FPM, Level);
    if (!FPM.isEmpty())
      MPM.addPass(createModuleToFunctionPassAdaptor(std::move(FPM)));
  }
 
  MPM.addPass(buildCoroWrapper(Phase));
 
  // AllocToken transforms heap allocation calls; this needs to run late after
  // other allocation call transformations (such as those in InstCombine).
  if (!isLTOPreLink(Phase))
    MPM.addPass(AllocTokenPass());
 
  invokeOptimizerLastEPCallbacks(MPM, Level, Phase);
 
  if (Opts.enable_instrumentor)
    MPM.addPass(InstrumentorPass(FS));
 
  // Attach !implicit.ref metadata from all functions to copyright strings.
  MPM.addPass(LowerCommentStringPass());
 
  if (isLTOPreLink(Phase))
    addRequiredLTOPreLinkPasses(MPM);
 
  // Emit annotation remarks.
  addAnnotationRemarksPass(MPM);
 
  instructionCountersPass(MPM, /* IsPreOptimization */ false);
 
  return MPM;
}
 
AAManager PassBuilder::buildDefaultAAPipeline() {
  AAManager AA;
 
  // The order in which these are registered determines their priority when
  // being queried.
 
  // Add any target-specific alias analyses that should be run early.
  if (TM)
    TM->registerEarlyDefaultAliasAnalyses(AA);
 
  // First we register the basic alias analysis that provides the majority of
  // per-function local AA logic. This is a stateless, on-demand local set of
  // AA techniques.
  AA.registerFunctionAnalysis<BasicAA>();
 
  // Next we query fast, specialized alias analyses that wrap IR-embedded
  // information about aliasing.
  AA.registerFunctionAnalysis<ScopedNoAliasAA>();
  AA.registerFunctionAnalysis<TypeBasedAA>();
 
  // Add support for querying global aliasing information when available.
  // Because the `AAManager` is a function analysis and `GlobalsAA` is a module
  // analysis, all that the `AAManager` can do is query for any *cached*
  // results from `GlobalsAA` through a readonly proxy.
  if (Opts.enable_global_analyses)
    AA.registerModuleAnalysis<GlobalsAA>();
 
  // Add target-specific alias analyses.
  if (TM)
    TM->registerDefaultAliasAnalyses(AA);
 
  return AA;
}
 
bool PassBuilder::isInstrumentedPGOUse() const {
  return (PGOOpt && PGOOpt->Action == PGOOptions::IRUse) ||
         !UseCtxProfile.empty();
}