blob: 587b58ff562207a6b5e563b265bfd55fdc20da34 [file] [log] [blame]
//===-- LICM.cpp - Loop Invariant Code Motion Pass ------------------------===//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
// This pass performs loop invariant code motion, attempting to remove as much
// code from the body of a loop as possible. It does this by either hoisting
// code into the preheader block, or by sinking code to the exit blocks if it is
// safe. This pass also promotes must-aliased memory locations in the loop to
// live in registers, thus hoisting and sinking "invariant" loads and stores.
// Hoisting operations out of loops is a canonicalization transform. It
// enables and simplifies subsequent optimizations in the middle-end.
// Rematerialization of hoisted instructions to reduce register pressure is the
// responsibility of the back-end, which has more accurate information about
// register pressure and also handles other optimizations than LICM that
// increase live-ranges.
// This pass uses alias analysis for two purposes:
// 1. Moving loop invariant loads and calls out of loops. If we can determine
// that a load or call inside of a loop never aliases anything stored to,
// we can hoist it or sink it like any other instruction.
// 2. Scalar Promotion of Memory - If there is a store instruction inside of
// the loop, we try to move the store to happen AFTER the loop instead of
// inside of the loop. This can only happen if a few conditions are true:
// A. The pointer stored through is loop invariant
// B. There are no stores or loads in the loop which _may_ alias the
// pointer. There are no calls in the loop which mod/ref the pointer.
// If these conditions are true, we can promote the loads and stores in the
// loop of the pointer to use a temporary alloca'd variable. We then use
// the SSAUpdater to construct the appropriate SSA form for the value.
#include "llvm/Transforms/Scalar/LICM.h"
#include "llvm/ADT/SetOperations.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/Analysis/AliasAnalysis.h"
#include "llvm/Analysis/AliasSetTracker.h"
#include "llvm/Analysis/BasicAliasAnalysis.h"
#include "llvm/Analysis/BlockFrequencyInfo.h"
#include "llvm/Analysis/CaptureTracking.h"
#include "llvm/Analysis/ConstantFolding.h"
#include "llvm/Analysis/GlobalsModRef.h"
#include "llvm/Analysis/GuardUtils.h"
#include "llvm/Analysis/LazyBlockFrequencyInfo.h"
#include "llvm/Analysis/Loads.h"
#include "llvm/Analysis/LoopInfo.h"
#include "llvm/Analysis/LoopIterator.h"
#include "llvm/Analysis/LoopPass.h"
#include "llvm/Analysis/MemoryBuiltins.h"
#include "llvm/Analysis/MemorySSA.h"
#include "llvm/Analysis/MemorySSAUpdater.h"
#include "llvm/Analysis/MustExecute.h"
#include "llvm/Analysis/OptimizationRemarkEmitter.h"
#include "llvm/Analysis/ScalarEvolution.h"
#include "llvm/Analysis/ScalarEvolutionAliasAnalysis.h"
#include "llvm/Analysis/TargetLibraryInfo.h"
#include "llvm/Analysis/ValueTracking.h"
#include "llvm/IR/CFG.h"
#include "llvm/IR/Constants.h"
#include "llvm/IR/DataLayout.h"
#include "llvm/IR/DebugInfoMetadata.h"
#include "llvm/IR/DerivedTypes.h"
#include "llvm/IR/Dominators.h"
#include "llvm/IR/Instructions.h"
#include "llvm/IR/IntrinsicInst.h"
#include "llvm/IR/LLVMContext.h"
#include "llvm/IR/Metadata.h"
#include "llvm/IR/PatternMatch.h"
#include "llvm/IR/PredIteratorCache.h"
#include "llvm/InitializePasses.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Transforms/Scalar.h"
#include "llvm/Transforms/Scalar/LoopPassManager.h"
#include "llvm/Transforms/Utils/AssumeBundleBuilder.h"
#include "llvm/Transforms/Utils/BasicBlockUtils.h"
#include "llvm/Transforms/Utils/Local.h"
#include "llvm/Transforms/Utils/LoopUtils.h"
#include "llvm/Transforms/Utils/SSAUpdater.h"
#include <algorithm>
#include <utility>
using namespace llvm;
#define DEBUG_TYPE "licm"
STATISTIC(NumCreatedBlocks, "Number of blocks created");
STATISTIC(NumClonedBranches, "Number of branches cloned");
STATISTIC(NumSunk, "Number of instructions sunk out of loop");
STATISTIC(NumHoisted, "Number of instructions hoisted out of loop");
STATISTIC(NumMovedLoads, "Number of load insts hoisted or sunk");
STATISTIC(NumMovedCalls, "Number of call insts hoisted or sunk");
STATISTIC(NumPromoted, "Number of memory locations promoted to registers");
/// Memory promotion is enabled by default.
static cl::opt<bool>
DisablePromotion("disable-licm-promotion", cl::Hidden, cl::init(false),
cl::desc("Disable memory promotion in LICM pass"));
static cl::opt<bool> ControlFlowHoisting(
"licm-control-flow-hoisting", cl::Hidden, cl::init(false),
cl::desc("Enable control flow (and PHI) hoisting in LICM"));
static cl::opt<unsigned> HoistSinkColdnessThreshold(
"licm-coldness-threshold", cl::Hidden, cl::init(4),
cl::desc("Relative coldness Threshold of hoisting/sinking destination "
"block for LICM to be considered beneficial"));
static cl::opt<uint32_t> MaxNumUsesTraversed(
"licm-max-num-uses-traversed", cl::Hidden, cl::init(8),
cl::desc("Max num uses visited for identifying load "
"invariance in loop using invariant start (default = 8)"));
// Default value of zero implies we use the regular alias set tracker mechanism
// instead of the cross product using AA to identify aliasing of the memory
// location we are interested in.
static cl::opt<int>
LICMN2Theshold("licm-n2-threshold", cl::Hidden, cl::init(0),
cl::desc("How many instruction to cross product using AA"));
// Experimental option to allow imprecision in LICM in pathological cases, in
// exchange for faster compile. This is to be removed if MemorySSA starts to
// address the same issue. This flag applies only when LICM uses MemorySSA
// instead on AliasSetTracker. LICM calls MemorySSAWalker's
// getClobberingMemoryAccess, up to the value of the Cap, getting perfect
// accuracy. Afterwards, LICM will call into MemorySSA's getDefiningAccess,
// which may not be precise, since optimizeUses is capped. The result is
// correct, but we may not get as "far up" as possible to get which access is
// clobbering the one queried.
cl::opt<unsigned> llvm::SetLicmMssaOptCap(
"licm-mssa-optimization-cap", cl::init(100), cl::Hidden,
cl::desc("Enable imprecision in LICM in pathological cases, in exchange "
"for faster compile. Caps the MemorySSA clobbering calls."));
// Experimentally, memory promotion carries less importance than sinking and
// hoisting. Limit when we do promotion when using MemorySSA, in order to save
// compile time.
cl::opt<unsigned> llvm::SetLicmMssaNoAccForPromotionCap(
"licm-mssa-max-acc-promotion", cl::init(250), cl::Hidden,
cl::desc("[LICM & MemorySSA] When MSSA in LICM is disabled, this has no "
"effect. When MSSA in LICM is enabled, then this is the maximum "
"number of accesses allowed to be present in a loop in order to "
"enable memory promotion."));
static bool inSubLoop(BasicBlock *BB, Loop *CurLoop, LoopInfo *LI);
static bool isNotUsedOrFreeInLoop(const Instruction &I, const Loop *CurLoop,
const LoopSafetyInfo *SafetyInfo,
TargetTransformInfo *TTI, bool &FreeInLoop);
static void hoist(Instruction &I, const DominatorTree *DT, const Loop *CurLoop,
BasicBlock *Dest, ICFLoopSafetyInfo *SafetyInfo,
MemorySSAUpdater *MSSAU, ScalarEvolution *SE,
OptimizationRemarkEmitter *ORE);
static bool sink(Instruction &I, LoopInfo *LI, DominatorTree *DT,
BlockFrequencyInfo *BFI, const Loop *CurLoop,
ICFLoopSafetyInfo *SafetyInfo, MemorySSAUpdater *MSSAU,
OptimizationRemarkEmitter *ORE);
static bool isSafeToExecuteUnconditionally(Instruction &Inst,
const DominatorTree *DT,
const TargetLibraryInfo *TLI,
const Loop *CurLoop,
const LoopSafetyInfo *SafetyInfo,
OptimizationRemarkEmitter *ORE,
const Instruction *CtxI = nullptr);
static bool pointerInvalidatedByLoop(MemoryLocation MemLoc,
AliasSetTracker *CurAST, Loop *CurLoop,
AAResults *AA);
static bool pointerInvalidatedByLoopWithMSSA(MemorySSA *MSSA, MemoryUse *MU,
Loop *CurLoop, Instruction &I,
SinkAndHoistLICMFlags &Flags);
static bool pointerInvalidatedByBlockWithMSSA(BasicBlock &BB, MemorySSA &MSSA,
MemoryUse &MU);
static Instruction *cloneInstructionInExitBlock(
Instruction &I, BasicBlock &ExitBlock, PHINode &PN, const LoopInfo *LI,
const LoopSafetyInfo *SafetyInfo, MemorySSAUpdater *MSSAU);
static void eraseInstruction(Instruction &I, ICFLoopSafetyInfo &SafetyInfo,
AliasSetTracker *AST, MemorySSAUpdater *MSSAU);
static void moveInstructionBefore(Instruction &I, Instruction &Dest,
ICFLoopSafetyInfo &SafetyInfo,
MemorySSAUpdater *MSSAU, ScalarEvolution *SE);
static void foreachMemoryAccess(MemorySSA *MSSA, Loop *L,
function_ref<void(Instruction *)> Fn);
static SmallVector<SmallSetVector<Value *, 8>, 0>
collectPromotionCandidates(MemorySSA *MSSA, AliasAnalysis *AA, Loop *L,
SmallVectorImpl<Instruction *> &MaybePromotable);
namespace {
struct LoopInvariantCodeMotion {
bool runOnLoop(Loop *L, AAResults *AA, LoopInfo *LI, DominatorTree *DT,
BlockFrequencyInfo *BFI, TargetLibraryInfo *TLI,
TargetTransformInfo *TTI, ScalarEvolution *SE, MemorySSA *MSSA,
OptimizationRemarkEmitter *ORE);
LoopInvariantCodeMotion(unsigned LicmMssaOptCap,
unsigned LicmMssaNoAccForPromotionCap)
: LicmMssaOptCap(LicmMssaOptCap),
LicmMssaNoAccForPromotionCap(LicmMssaNoAccForPromotionCap) {}
unsigned LicmMssaOptCap;
unsigned LicmMssaNoAccForPromotionCap;
collectAliasInfoForLoop(Loop *L, LoopInfo *LI, AAResults *AA);
struct LegacyLICMPass : public LoopPass {
static char ID; // Pass identification, replacement for typeid
unsigned LicmMssaOptCap = SetLicmMssaOptCap,
unsigned LicmMssaNoAccForPromotionCap = SetLicmMssaNoAccForPromotionCap)
: LoopPass(ID), LICM(LicmMssaOptCap, LicmMssaNoAccForPromotionCap) {
bool runOnLoop(Loop *L, LPPassManager &LPM) override {
if (skipLoop(L))
return false;
LLVM_DEBUG(dbgs() << "Perform LICM on Loop with header at block "
<< L->getHeader()->getNameOrAsOperand() << "\n");
auto *SE = getAnalysisIfAvailable<ScalarEvolutionWrapperPass>();
MemorySSA *MSSA = EnableMSSALoopDependency
? (&getAnalysis<MemorySSAWrapperPass>().getMSSA())
: nullptr;
bool hasProfileData = L->getHeader()->getParent()->hasProfileData();
BlockFrequencyInfo *BFI =
hasProfileData ? &getAnalysis<LazyBlockFrequencyInfoPass>().getBFI()
: nullptr;
// For the old PM, we can't use OptimizationRemarkEmitter as an analysis
// pass. Function analyses need to be preserved across loop transformations
// but ORE cannot be preserved (see comment before the pass definition).
OptimizationRemarkEmitter ORE(L->getHeader()->getParent());
return LICM.runOnLoop(
L, &getAnalysis<AAResultsWrapperPass>().getAAResults(),
&getAnalysis<DominatorTreeWrapperPass>().getDomTree(), BFI,
SE ? &SE->getSE() : nullptr, MSSA, &ORE);
/// This transformation requires natural loop information & requires that
/// loop preheaders be inserted into the CFG...
void getAnalysisUsage(AnalysisUsage &AU) const override {
if (EnableMSSALoopDependency) {
LoopInvariantCodeMotion LICM;
} // namespace
PreservedAnalyses LICMPass::run(Loop &L, LoopAnalysisManager &AM,
LoopStandardAnalysisResults &AR, LPMUpdater &) {
// For the new PM, we also can't use OptimizationRemarkEmitter as an analysis
// pass. Function analyses need to be preserved across loop transformations
// but ORE cannot be preserved (see comment before the pass definition).
OptimizationRemarkEmitter ORE(L.getHeader()->getParent());
LoopInvariantCodeMotion LICM(LicmMssaOptCap, LicmMssaNoAccForPromotionCap);
if (!LICM.runOnLoop(&L, &AR.AA, &AR.LI, &AR.DT, AR.BFI, &AR.TLI, &AR.TTI,
return PreservedAnalyses::all();
auto PA = getLoopPassPreservedAnalyses();
if (AR.MSSA)
return PA;
char LegacyLICMPass::ID = 0;
INITIALIZE_PASS_BEGIN(LegacyLICMPass, "licm", "Loop Invariant Code Motion",
false, false)
INITIALIZE_PASS_END(LegacyLICMPass, "licm", "Loop Invariant Code Motion", false,
Pass *llvm::createLICMPass() { return new LegacyLICMPass(); }
Pass *llvm::createLICMPass(unsigned LicmMssaOptCap,
unsigned LicmMssaNoAccForPromotionCap) {
return new LegacyLICMPass(LicmMssaOptCap, LicmMssaNoAccForPromotionCap);
llvm::SinkAndHoistLICMFlags::SinkAndHoistLICMFlags(bool IsSink, Loop *L,
MemorySSA *MSSA)
: SinkAndHoistLICMFlags(SetLicmMssaOptCap, SetLicmMssaNoAccForPromotionCap,
IsSink, L, MSSA) {}
unsigned LicmMssaOptCap, unsigned LicmMssaNoAccForPromotionCap, bool IsSink,
Loop *L, MemorySSA *MSSA)
: LicmMssaOptCap(LicmMssaOptCap),
IsSink(IsSink) {
assert(((L != nullptr) == (MSSA != nullptr)) &&
"Unexpected values for SinkAndHoistLICMFlags");
if (!MSSA)
unsigned AccessCapCount = 0;
for (auto *BB : L->getBlocks())
if (const auto *Accesses = MSSA->getBlockAccesses(BB))
for (const auto &MA : *Accesses) {
if (AccessCapCount > LicmMssaNoAccForPromotionCap) {
NoOfMemAccTooLarge = true;
/// Hoist expressions out of the specified loop. Note, alias info for inner
/// loop is not preserved so it is not a good idea to run LICM multiple
/// times on one loop.
bool LoopInvariantCodeMotion::runOnLoop(
Loop *L, AAResults *AA, LoopInfo *LI, DominatorTree *DT,
BlockFrequencyInfo *BFI, TargetLibraryInfo *TLI, TargetTransformInfo *TTI,
ScalarEvolution *SE, MemorySSA *MSSA, OptimizationRemarkEmitter *ORE) {
bool Changed = false;
assert(L->isLCSSAForm(*DT) && "Loop is not in LCSSA form.");
// If this loop has metadata indicating that LICM is not to be performed then
// just exit.
if (hasDisableLICMTransformsHint(L)) {
return false;
std::unique_ptr<AliasSetTracker> CurAST;
std::unique_ptr<MemorySSAUpdater> MSSAU;
std::unique_ptr<SinkAndHoistLICMFlags> Flags;
// Don't sink stores from loops with coroutine suspend instructions.
// LICM would sink instructions into the default destination of
// the coroutine switch. The default destination of the switch is to
// handle the case where the coroutine is suspended, by which point the
// coroutine frame may have been destroyed. No instruction can be sunk there.
// FIXME: This would unfortunately hurt the performance of coroutines, however
// there is currently no general solution for this. Similar issues could also
// potentially happen in other passes where instructions are being moved
// across that edge.
bool HasCoroSuspendInst = llvm::any_of(L->getBlocks(), [](BasicBlock *BB) {
return llvm::any_of(*BB, [](Instruction &I) {
IntrinsicInst *II = dyn_cast<IntrinsicInst>(&I);
return II && II->getIntrinsicID() == Intrinsic::coro_suspend;
if (!MSSA) {
LLVM_DEBUG(dbgs() << "LICM: Using Alias Set Tracker.\n");
CurAST = collectAliasInfoForLoop(L, LI, AA);
Flags = std::make_unique<SinkAndHoistLICMFlags>(
LicmMssaOptCap, LicmMssaNoAccForPromotionCap, /*IsSink=*/true);
} else {
LLVM_DEBUG(dbgs() << "LICM: Using MemorySSA.\n");
MSSAU = std::make_unique<MemorySSAUpdater>(MSSA);
Flags = std::make_unique<SinkAndHoistLICMFlags>(
LicmMssaOptCap, LicmMssaNoAccForPromotionCap, /*IsSink=*/true, L, MSSA);
// Get the preheader block to move instructions into...
BasicBlock *Preheader = L->getLoopPreheader();
// Compute loop safety information.
ICFLoopSafetyInfo SafetyInfo;
// We want to visit all of the instructions in this loop... that are not parts
// of our subloops (they have already had their invariants hoisted out of
// their loop, into this loop, so there is no need to process the BODIES of
// the subloops).
// Traverse the body of the loop in depth first order on the dominator tree so
// that we are guaranteed to see definitions before we see uses. This allows
// us to sink instructions in one pass, without iteration. After sinking
// instructions, we perform another pass to hoist them out of the loop.
if (L->hasDedicatedExits())
Changed |=
sinkRegion(DT->getNode(L->getHeader()), AA, LI, DT, BFI, TLI, TTI, L,
CurAST.get(), MSSAU.get(), &SafetyInfo, *Flags.get(), ORE);
if (Preheader)
Changed |= hoistRegion(DT->getNode(L->getHeader()), AA, LI, DT, BFI, TLI, L,
CurAST.get(), MSSAU.get(), SE, &SafetyInfo,
*Flags.get(), ORE);
// Now that all loop invariants have been removed from the loop, promote any
// memory references to scalars that we can.
// Don't sink stores from loops without dedicated block exits. Exits
// containing indirect branches are not transformed by loop simplify,
// make sure we catch that. An additional load may be generated in the
// preheader for SSA updater, so also avoid sinking when no preheader
// is available.
if (!DisablePromotion && Preheader && L->hasDedicatedExits() &&
!Flags->tooManyMemoryAccesses() && !HasCoroSuspendInst) {
// Figure out the loop exits and their insertion points
SmallVector<BasicBlock *, 8> ExitBlocks;
// We can't insert into a catchswitch.
bool HasCatchSwitch = llvm::any_of(ExitBlocks, [](BasicBlock *Exit) {
return isa<CatchSwitchInst>(Exit->getTerminator());
if (!HasCatchSwitch) {
SmallVector<Instruction *, 8> InsertPts;
SmallVector<MemoryAccess *, 8> MSSAInsertPts;
if (MSSAU)
for (BasicBlock *ExitBlock : ExitBlocks) {
if (MSSAU)
PredIteratorCache PIC;
bool Promoted = false;
if (CurAST.get()) {
// Loop over all of the alias sets in the tracker object.
for (AliasSet &AS : *CurAST) {
// We can promote this alias set if it has a store, if it is a "Must"
// alias set, if the pointer is loop invariant, and if we are not
// eliminating any volatile loads or stores.
if (AS.isForwardingAliasSet() || !AS.isMod() || !AS.isMustAlias() ||
!AS.empty() &&
"Must alias set should have at least one pointer element in it!");
SmallSetVector<Value *, 8> PointerMustAliases;
for (const auto &ASI : AS)
Promoted |= promoteLoopAccessesToScalars(
PointerMustAliases, ExitBlocks, InsertPts, MSSAInsertPts, PIC, LI,
DT, TLI, L, CurAST.get(), MSSAU.get(), &SafetyInfo, ORE);
} else {
SmallVector<Instruction *, 16> MaybePromotable;
foreachMemoryAccess(MSSA, L, [&](Instruction *I) {
// Promoting one set of accesses may make the pointers for another set
// loop invariant, so run this in a loop (with the MaybePromotable set
// decreasing in size over time).
bool LocalPromoted;
do {
LocalPromoted = false;
for (const SmallSetVector<Value *, 8> &PointerMustAliases :
collectPromotionCandidates(MSSA, AA, L, MaybePromotable)) {
LocalPromoted |= promoteLoopAccessesToScalars(
PointerMustAliases, ExitBlocks, InsertPts, MSSAInsertPts, PIC,
LI, DT, TLI, L, /*AST*/nullptr, MSSAU.get(), &SafetyInfo, ORE);
Promoted |= LocalPromoted;
} while (LocalPromoted);
// Once we have promoted values across the loop body we have to
// recursively reform LCSSA as any nested loop may now have values defined
// within the loop used in the outer loop.
// FIXME: This is really heavy handed. It would be a bit better to use an
// SSAUpdater strategy during promotion that was LCSSA aware and reformed
// it as it went.
if (Promoted)
formLCSSARecursively(*L, *DT, LI, SE);
Changed |= Promoted;
// Check that neither this loop nor its parent have had LCSSA broken. LICM is
// specifically moving instructions across the loop boundary and so it is
// especially in need of sanity checking here.
assert(L->isLCSSAForm(*DT) && "Loop not left in LCSSA form after LICM!");
assert((L->isOutermost() || L->getParentLoop()->isLCSSAForm(*DT)) &&
"Parent loop not left in LCSSA form after LICM!");
if (MSSAU.get() && VerifyMemorySSA)
if (Changed && SE)
return Changed;
/// Walk the specified region of the CFG (defined by all blocks dominated by
/// the specified block, and that are in the current loop) in reverse depth
/// first order w.r.t the DominatorTree. This allows us to visit uses before
/// definitions, allowing us to sink a loop body in one pass without iteration.
bool llvm::sinkRegion(DomTreeNode *N, AAResults *AA, LoopInfo *LI,
DominatorTree *DT, BlockFrequencyInfo *BFI,
TargetLibraryInfo *TLI, TargetTransformInfo *TTI,
Loop *CurLoop, AliasSetTracker *CurAST,
MemorySSAUpdater *MSSAU, ICFLoopSafetyInfo *SafetyInfo,
SinkAndHoistLICMFlags &Flags,
OptimizationRemarkEmitter *ORE) {
// Verify inputs.
assert(N != nullptr && AA != nullptr && LI != nullptr && DT != nullptr &&
CurLoop != nullptr && SafetyInfo != nullptr &&
"Unexpected input to sinkRegion.");
assert(((CurAST != nullptr) ^ (MSSAU != nullptr)) &&
"Either AliasSetTracker or MemorySSA should be initialized.");
// We want to visit children before parents. We will enque all the parents
// before their children in the worklist and process the worklist in reverse
// order.
SmallVector<DomTreeNode *, 16> Worklist = collectChildrenInLoop(N, CurLoop);
bool Changed = false;
for (DomTreeNode *DTN : reverse(Worklist)) {
BasicBlock *BB = DTN->getBlock();
// Only need to process the contents of this block if it is not part of a
// subloop (which would already have been processed).
if (inSubLoop(BB, CurLoop, LI))
for (BasicBlock::iterator II = BB->end(); II != BB->begin();) {
Instruction &I = *--II;
// If the instruction is dead, we would try to sink it because it isn't
// used in the loop, instead, just delete it.
if (isInstructionTriviallyDead(&I, TLI)) {
LLVM_DEBUG(dbgs() << "LICM deleting dead inst: " << I << '\n');
eraseInstruction(I, *SafetyInfo, CurAST, MSSAU);
Changed = true;
// Check to see if we can sink this instruction to the exit blocks
// of the loop. We can do this if the all users of the instruction are
// outside of the loop. In this case, it doesn't even matter if the
// operands of the instruction are loop invariant.
bool FreeInLoop = false;
if (!I.mayHaveSideEffects() &&
isNotUsedOrFreeInLoop(I, CurLoop, SafetyInfo, TTI, FreeInLoop) &&
canSinkOrHoistInst(I, AA, DT, CurLoop, CurAST, MSSAU, true, &Flags,
ORE)) {
if (sink(I, LI, DT, BFI, CurLoop, SafetyInfo, MSSAU, ORE)) {
if (!FreeInLoop) {
eraseInstruction(I, *SafetyInfo, CurAST, MSSAU);
Changed = true;
if (MSSAU && VerifyMemorySSA)
return Changed;
namespace {
// This is a helper class for hoistRegion to make it able to hoist control flow
// in order to be able to hoist phis. The way this works is that we initially
// start hoisting to the loop preheader, and when we see a loop invariant branch
// we make note of this. When we then come to hoist an instruction that's
// conditional on such a branch we duplicate the branch and the relevant control
// flow, then hoist the instruction into the block corresponding to its original
// block in the duplicated control flow.
class ControlFlowHoister {
// Information about the loop we are hoisting from
LoopInfo *LI;
DominatorTree *DT;
Loop *CurLoop;
MemorySSAUpdater *MSSAU;
// A map of blocks in the loop to the block their instructions will be hoisted
// to.
DenseMap<BasicBlock *, BasicBlock *> HoistDestinationMap;
// The branches that we can hoist, mapped to the block that marks a
// convergence point of their control flow.
DenseMap<BranchInst *, BasicBlock *> HoistableBranches;
ControlFlowHoister(LoopInfo *LI, DominatorTree *DT, Loop *CurLoop,
MemorySSAUpdater *MSSAU)
: LI(LI), DT(DT), CurLoop(CurLoop), MSSAU(MSSAU) {}
void registerPossiblyHoistableBranch(BranchInst *BI) {
// We can only hoist conditional branches with loop invariant operands.
if (!ControlFlowHoisting || !BI->isConditional() ||
// The branch destinations need to be in the loop, and we don't gain
// anything by duplicating conditional branches with duplicate successors,
// as it's essentially the same as an unconditional branch.
BasicBlock *TrueDest = BI->getSuccessor(0);
BasicBlock *FalseDest = BI->getSuccessor(1);
if (!CurLoop->contains(TrueDest) || !CurLoop->contains(FalseDest) ||
TrueDest == FalseDest)
// We can hoist BI if one branch destination is the successor of the other,
// or both have common successor which we check by seeing if the
// intersection of their successors is non-empty.
// TODO: This could be expanded to allowing branches where both ends
// eventually converge to a single block.
SmallPtrSet<BasicBlock *, 4> TrueDestSucc, FalseDestSucc;
TrueDestSucc.insert(succ_begin(TrueDest), succ_end(TrueDest));
FalseDestSucc.insert(succ_begin(FalseDest), succ_end(FalseDest));
BasicBlock *CommonSucc = nullptr;
if (TrueDestSucc.count(FalseDest)) {
CommonSucc = FalseDest;
} else if (FalseDestSucc.count(TrueDest)) {
CommonSucc = TrueDest;
} else {
set_intersect(TrueDestSucc, FalseDestSucc);
// If there's one common successor use that.
if (TrueDestSucc.size() == 1)
CommonSucc = *TrueDestSucc.begin();
// If there's more than one pick whichever appears first in the block list
// (we can't use the value returned by TrueDestSucc.begin() as it's
// unpredicatable which element gets returned).
else if (!TrueDestSucc.empty()) {
Function *F = TrueDest->getParent();
auto IsSucc = [&](BasicBlock &BB) { return TrueDestSucc.count(&BB); };
auto It = llvm::find_if(*F, IsSucc);
assert(It != F->end() && "Could not find successor in function");
CommonSucc = &*It;
// The common successor has to be dominated by the branch, as otherwise
// there will be some other path to the successor that will not be
// controlled by this branch so any phi we hoist would be controlled by the
// wrong condition. This also takes care of avoiding hoisting of loop back
// edges.
// TODO: In some cases this could be relaxed if the successor is dominated
// by another block that's been hoisted and we can guarantee that the
// control flow has been replicated exactly.
if (CommonSucc && DT->dominates(BI, CommonSucc))
HoistableBranches[BI] = CommonSucc;
bool canHoistPHI(PHINode *PN) {
// The phi must have loop invariant operands.
if (!ControlFlowHoisting || !CurLoop->hasLoopInvariantOperands(PN))
return false;
// We can hoist phis if the block they are in is the target of hoistable
// branches which cover all of the predecessors of the block.
SmallPtrSet<BasicBlock *, 8> PredecessorBlocks;
BasicBlock *BB = PN->getParent();
for (BasicBlock *PredBB : predecessors(BB))
// If we have less predecessor blocks than predecessors then the phi will
// have more than one incoming value for the same block which we can't
// handle.
// TODO: This could be handled be erasing some of the duplicate incoming
// values.
if (PredecessorBlocks.size() != pred_size(BB))
return false;
for (auto &Pair : HoistableBranches) {
if (Pair.second == BB) {
// Which blocks are predecessors via this branch depends on if the
// branch is triangle-like or diamond-like.
if (Pair.first->getSuccessor(0) == BB) {
} else if (Pair.first->getSuccessor(1) == BB) {
} else {
// PredecessorBlocks will now be empty if for every predecessor of BB we
// found a hoistable branch source.
return PredecessorBlocks.empty();
BasicBlock *getOrCreateHoistedBlock(BasicBlock *BB) {
if (!ControlFlowHoisting)
return CurLoop->getLoopPreheader();
// If BB has already been hoisted, return that
if (HoistDestinationMap.count(BB))
return HoistDestinationMap[BB];
// Check if this block is conditional based on a pending branch
auto HasBBAsSuccessor =
[&](DenseMap<BranchInst *, BasicBlock *>::value_type &Pair) {
return BB != Pair.second && (Pair.first->getSuccessor(0) == BB ||
Pair.first->getSuccessor(1) == BB);
auto It = llvm::find_if(HoistableBranches, HasBBAsSuccessor);
// If not involved in a pending branch, hoist to preheader
BasicBlock *InitialPreheader = CurLoop->getLoopPreheader();
if (It == HoistableBranches.end()) {
LLVM_DEBUG(dbgs() << "LICM using "
<< InitialPreheader->getNameOrAsOperand()
<< " as hoist destination for "
<< BB->getNameOrAsOperand() << "\n");
HoistDestinationMap[BB] = InitialPreheader;
return InitialPreheader;
BranchInst *BI = It->first;
assert(std::find_if(++It, HoistableBranches.end(), HasBBAsSuccessor) ==
HoistableBranches.end() &&
"BB is expected to be the target of at most one branch");
LLVMContext &C = BB->getContext();
BasicBlock *TrueDest = BI->getSuccessor(0);
BasicBlock *FalseDest = BI->getSuccessor(1);
BasicBlock *CommonSucc = HoistableBranches[BI];
BasicBlock *HoistTarget = getOrCreateHoistedBlock(BI->getParent());
// Create hoisted versions of blocks that currently don't have them
auto CreateHoistedBlock = [&](BasicBlock *Orig) {
if (HoistDestinationMap.count(Orig))
return HoistDestinationMap[Orig];
BasicBlock *New =
BasicBlock::Create(C, Orig->getName() + ".licm", Orig->getParent());
HoistDestinationMap[Orig] = New;
DT->addNewBlock(New, HoistTarget);
if (CurLoop->getParentLoop())
CurLoop->getParentLoop()->addBasicBlockToLoop(New, *LI);
LLVM_DEBUG(dbgs() << "LICM created " << New->getName()
<< " as hoist destination for " << Orig->getName()
<< "\n");
return New;
BasicBlock *HoistTrueDest = CreateHoistedBlock(TrueDest);
BasicBlock *HoistFalseDest = CreateHoistedBlock(FalseDest);
BasicBlock *HoistCommonSucc = CreateHoistedBlock(CommonSucc);
// Link up these blocks with branches.
if (!HoistCommonSucc->getTerminator()) {
// The new common successor we've generated will branch to whatever that
// hoist target branched to.
BasicBlock *TargetSucc = HoistTarget->getSingleSuccessor();
assert(TargetSucc && "Expected hoist target to have a single successor");
BranchInst::Create(TargetSucc, HoistCommonSucc);
if (!HoistTrueDest->getTerminator()) {
BranchInst::Create(HoistCommonSucc, HoistTrueDest);
if (!HoistFalseDest->getTerminator()) {
BranchInst::Create(HoistCommonSucc, HoistFalseDest);
// If BI is being cloned to what was originally the preheader then
// HoistCommonSucc will now be the new preheader.
if (HoistTarget == InitialPreheader) {
// Phis in the loop header now need to use the new preheader.
if (MSSAU)
HoistTarget->getSingleSuccessor(), HoistCommonSucc, {HoistTarget});
// The new preheader dominates the loop header.
DomTreeNode *PreheaderNode = DT->getNode(HoistCommonSucc);
DomTreeNode *HeaderNode = DT->getNode(CurLoop->getHeader());
DT->changeImmediateDominator(HeaderNode, PreheaderNode);
// The preheader hoist destination is now the new preheader, with the
// exception of the hoist destination of this branch.
for (auto &Pair : HoistDestinationMap)
if (Pair.second == InitialPreheader && Pair.first != BI->getParent())
Pair.second = HoistCommonSucc;
// Now finally clone BI.
BranchInst::Create(HoistTrueDest, HoistFalseDest, BI->getCondition()));
assert(CurLoop->getLoopPreheader() &&
"Hoisting blocks should not have destroyed preheader");
return HoistDestinationMap[BB];
} // namespace
// Hoisting/sinking instruction out of a loop isn't always beneficial. It's only
// only worthwhile if the destination block is actually colder than current
// block.
static bool worthSinkOrHoistInst(Instruction &I, BasicBlock *DstBlock,
OptimizationRemarkEmitter *ORE,
BlockFrequencyInfo *BFI) {
// Check block frequency only when runtime profile is available
// to avoid pathological cases. With static profile, lean towards
// hosting because it helps canonicalize the loop for vectorizer.
if (!DstBlock->getParent()->hasProfileData())
return true;
if (!HoistSinkColdnessThreshold || !BFI)
return true;
BasicBlock *SrcBlock = I.getParent();
if (BFI->getBlockFreq(DstBlock).getFrequency() / HoistSinkColdnessThreshold >
BFI->getBlockFreq(SrcBlock).getFrequency()) {
ORE->emit([&]() {
return OptimizationRemarkMissed(DEBUG_TYPE, "SinkHoistInst", &I)
<< "failed to sink or hoist instruction because containing block "
"has lower frequency than destination block";
return false;
return true;
/// Walk the specified region of the CFG (defined by all blocks dominated by
/// the specified block, and that are in the current loop) in depth first
/// order w.r.t the DominatorTree. This allows us to visit definitions before
/// uses, allowing us to hoist a loop body in one pass without iteration.
bool llvm::hoistRegion(DomTreeNode *N, AAResults *AA, LoopInfo *LI,
DominatorTree *DT, BlockFrequencyInfo *BFI,
TargetLibraryInfo *TLI, Loop *CurLoop,
AliasSetTracker *CurAST, MemorySSAUpdater *MSSAU,
ScalarEvolution *SE, ICFLoopSafetyInfo *SafetyInfo,
SinkAndHoistLICMFlags &Flags,
OptimizationRemarkEmitter *ORE) {
// Verify inputs.
assert(N != nullptr && AA != nullptr && LI != nullptr && DT != nullptr &&
CurLoop != nullptr && SafetyInfo != nullptr &&
"Unexpected input to hoistRegion.");
assert(((CurAST != nullptr) ^ (MSSAU != nullptr)) &&
"Either AliasSetTracker or MemorySSA should be initialized.");
ControlFlowHoister CFH(LI, DT, CurLoop, MSSAU);
// Keep track of instructions that have been hoisted, as they may need to be
// re-hoisted if they end up not dominating all of their uses.
SmallVector<Instruction *, 16> HoistedInstructions;
// For PHI hoisting to work we need to hoist blocks before their successors.
// We can do this by iterating through the blocks in the loop in reverse
// post-order.
LoopBlocksRPO Worklist(CurLoop);
bool Changed = false;
for (BasicBlock *BB : Worklist) {
// Only need to process the contents of this block if it is not part of a
// subloop (which would already have been processed).
if (inSubLoop(BB, CurLoop, LI))
for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E;) {
Instruction &I = *II++;
// Try constant folding this instruction. If all the operands are
// constants, it is technically hoistable, but it would be better to
// just fold it.
if (Constant *C = ConstantFoldInstruction(
&I, I.getModule()->getDataLayout(), TLI)) {
LLVM_DEBUG(dbgs() << "LICM folding inst: " << I << " --> " << *C
<< '\n');
if (CurAST)
CurAST->copyValue(&I, C);
// FIXME MSSA: Such replacements may make accesses unoptimized (D51960).
if (isInstructionTriviallyDead(&I, TLI))
eraseInstruction(I, *SafetyInfo, CurAST, MSSAU);
Changed = true;
// Try hoisting the instruction out to the preheader. We can only do
// this if all of the operands of the instruction are loop invariant and
// if it is safe to hoist the instruction. We also check block frequency
// to make sure instruction only gets hoisted into colder blocks.
// TODO: It may be safe to hoist if we are hoisting to a conditional block
// and we have accurately duplicated the control flow from the loop header
// to that block.
if (CurLoop->hasLoopInvariantOperands(&I) &&
canSinkOrHoistInst(I, AA, DT, CurLoop, CurAST, MSSAU, true, &Flags,
ORE) &&
worthSinkOrHoistInst(I, CurLoop->getLoopPreheader(), ORE, BFI) &&
I, DT, TLI, CurLoop, SafetyInfo, ORE,
CurLoop->getLoopPreheader()->getTerminator())) {
hoist(I, DT, CurLoop, CFH.getOrCreateHoistedBlock(BB), SafetyInfo,
Changed = true;
// Attempt to remove floating point division out of the loop by
// converting it to a reciprocal multiplication.
if (I.getOpcode() == Instruction::FDiv && I.hasAllowReciprocal() &&
CurLoop->isLoopInvariant(I.getOperand(1))) {
auto Divisor = I.getOperand(1);
auto One = llvm::ConstantFP::get(Divisor->getType(), 1.0);
auto ReciprocalDivisor = BinaryOperator::CreateFDiv(One, Divisor);
SafetyInfo->insertInstructionTo(ReciprocalDivisor, I.getParent());
auto Product =
BinaryOperator::CreateFMul(I.getOperand(0), ReciprocalDivisor);
SafetyInfo->insertInstructionTo(Product, I.getParent());
eraseInstruction(I, *SafetyInfo, CurAST, MSSAU);
hoist(*ReciprocalDivisor, DT, CurLoop, CFH.getOrCreateHoistedBlock(BB),
SafetyInfo, MSSAU, SE, ORE);
Changed = true;
auto IsInvariantStart = [&](Instruction &I) {
using namespace PatternMatch;
return I.use_empty() &&
match(&I, m_Intrinsic<Intrinsic::invariant_start>());
auto MustExecuteWithoutWritesBefore = [&](Instruction &I) {
return SafetyInfo->isGuaranteedToExecute(I, DT, CurLoop) &&
SafetyInfo->doesNotWriteMemoryBefore(I, CurLoop);
if ((IsInvariantStart(I) || isGuard(&I)) &&
CurLoop->hasLoopInvariantOperands(&I) &&
MustExecuteWithoutWritesBefore(I)) {
hoist(I, DT, CurLoop, CFH.getOrCreateHoistedBlock(BB), SafetyInfo,
Changed = true;
if (PHINode *PN = dyn_cast<PHINode>(&I)) {
if (CFH.canHoistPHI(PN)) {
// Redirect incoming blocks first to ensure that we create hoisted
// versions of those blocks before we hoist the phi.
for (unsigned int i = 0; i < PN->getNumIncomingValues(); ++i)
i, CFH.getOrCreateHoistedBlock(PN->getIncomingBlock(i)));
hoist(*PN, DT, CurLoop, CFH.getOrCreateHoistedBlock(BB), SafetyInfo,
assert(DT->dominates(PN, BB) && "Conditional PHIs not expected");
Changed = true;
// Remember possibly hoistable branches so we can actually hoist them
// later if needed.
if (BranchInst *BI = dyn_cast<BranchInst>(&I))
// If we hoisted instructions to a conditional block they may not dominate
// their uses that weren't hoisted (such as phis where some operands are not
// loop invariant). If so make them unconditional by moving them to their
// immediate dominator. We iterate through the instructions in reverse order
// which ensures that when we rehoist an instruction we rehoist its operands,
// and also keep track of where in the block we are rehoisting to to make sure
// that we rehoist instructions before the instructions that use them.
Instruction *HoistPoint = nullptr;
if (ControlFlowHoisting) {
for (Instruction *I : reverse(HoistedInstructions)) {
if (!llvm::all_of(I->uses(),
[&](Use &U) { return DT->dominates(I, U); })) {
BasicBlock *Dominator =
if (!HoistPoint || !DT->dominates(HoistPoint->getParent(), Dominator)) {
if (HoistPoint)
assert(DT->dominates(Dominator, HoistPoint->getParent()) &&
"New hoist point expected to dominate old hoist point");
HoistPoint = Dominator->getTerminator();
LLVM_DEBUG(dbgs() << "LICM rehoisting to "
<< HoistPoint->getParent()->getNameOrAsOperand()
<< ": " << *I << "\n");
moveInstructionBefore(*I, *HoistPoint, *SafetyInfo, MSSAU, SE);
HoistPoint = I;
Changed = true;
if (MSSAU && VerifyMemorySSA)
// Now that we've finished hoisting make sure that LI and DT are still
// valid.
if (Changed) {
assert(DT->verify(DominatorTree::VerificationLevel::Fast) &&
"Dominator tree verification failed");
return Changed;
// Return true if LI is invariant within scope of the loop. LI is invariant if
// CurLoop is dominated by an invariant.start representing the same memory
// location and size as the memory location LI loads from, and also the
// invariant.start has no uses.
static bool isLoadInvariantInLoop(LoadInst *LI, DominatorTree *DT,
Loop *CurLoop) {
Value *Addr = LI->getOperand(0);
const DataLayout &DL = LI->getModule()->getDataLayout();
const TypeSize LocSizeInBits = DL.getTypeSizeInBits(LI->getType());
// It is not currently possible for clang to generate an invariant.start
// intrinsic with scalable vector types because we don't support thread local
// sizeless types and we don't permit sizeless types in structs or classes.
// Furthermore, even if support is added for this in future the intrinsic
// itself is defined to have a size of -1 for variable sized objects. This
// makes it impossible to verify if the intrinsic envelops our region of
// interest. For example, both <vscale x 32 x i8> and <vscale x 16 x i8>
// types would have a -1 parameter, but the former is clearly double the size
// of the latter.
if (LocSizeInBits.isScalable())
return false;
// if the type is i8 addrspace(x)*, we know this is the type of
// llvm.invariant.start operand
auto *PtrInt8Ty = PointerType::get(Type::getInt8Ty(LI->getContext()),
unsigned BitcastsVisited = 0;
// Look through bitcasts until we reach the i8* type (this is invariant.start
// operand type).
while (Addr->getType() != PtrInt8Ty) {
auto *BC = dyn_cast<BitCastInst>(Addr);
// Avoid traversing high number of bitcast uses.
if (++BitcastsVisited > MaxNumUsesTraversed || !BC)
return false;
Addr = BC->getOperand(0);
unsigned UsesVisited = 0;
// Traverse all uses of the load operand value, to see if invariant.start is
// one of the uses, and whether it dominates the load instruction.
for (auto *U : Addr->users()) {
// Avoid traversing for Load operand with high number of users.
if (++UsesVisited > MaxNumUsesTraversed)
return false;
IntrinsicInst *II = dyn_cast<IntrinsicInst>(U);
// If there are escaping uses of invariant.start instruction, the load maybe
// non-invariant.
if (!II || II->getIntrinsicID() != Intrinsic::invariant_start ||
ConstantInt *InvariantSize = cast<ConstantInt>(II->getArgOperand(0));
// The intrinsic supports having a -1 argument for variable sized objects
// so we should check for that here.
if (InvariantSize->isNegative())
uint64_t InvariantSizeInBits = InvariantSize->getSExtValue() * 8;
// Confirm the invariant.start location size contains the load operand size
// in bits. Also, the invariant.start should dominate the load, and we
// should not hoist the load out of a loop that contains this dominating
// invariant.start.
if (LocSizeInBits.getFixedSize() <= InvariantSizeInBits &&
DT->properlyDominates(II->getParent(), CurLoop->getHeader()))
return true;
return false;
namespace {
/// Return true if-and-only-if we know how to (mechanically) both hoist and
/// sink a given instruction out of a loop. Does not address legality
/// concerns such as aliasing or speculation safety.
bool isHoistableAndSinkableInst(Instruction &I) {
// Only these instructions are hoistable/sinkable.
return (isa<LoadInst>(I) || isa<StoreInst>(I) || isa<CallInst>(I) ||
isa<FenceInst>(I) || isa<CastInst>(I) || isa<UnaryOperator>(I) ||
isa<BinaryOperator>(I) || isa<SelectInst>(I) ||
isa<GetElementPtrInst>(I) || isa<CmpInst>(I) ||
isa<InsertElementInst>(I) || isa<ExtractElementInst>(I) ||
isa<ShuffleVectorInst>(I) || isa<ExtractValueInst>(I) ||
isa<InsertValueInst>(I) || isa<FreezeInst>(I));
/// Return true if all of the alias sets within this AST are known not to
/// contain a Mod, or if MSSA knows thare are no MemoryDefs in the loop.
bool isReadOnly(AliasSetTracker *CurAST, const MemorySSAUpdater *MSSAU,
const Loop *L) {
if (CurAST) {
for (AliasSet &AS : *CurAST) {
if (!AS.isForwardingAliasSet() && AS.isMod()) {
return false;
return true;
} else { /*MSSAU*/
for (auto *BB : L->getBlocks())
if (MSSAU->getMemorySSA()->getBlockDefs(BB))
return false;
return true;
/// Return true if I is the only Instruction with a MemoryAccess in L.
bool isOnlyMemoryAccess(const Instruction *I, const Loop *L,
const MemorySSAUpdater *MSSAU) {
for (auto *BB : L->getBlocks())
if (auto *Accs = MSSAU->getMemorySSA()->getBlockAccesses(BB)) {
int NotAPhi = 0;
for (const auto &Acc : *Accs) {
if (isa<MemoryPhi>(&Acc))
const auto *MUD = cast<MemoryUseOrDef>(&Acc);
if (MUD->getMemoryInst() != I || NotAPhi++ == 1)
return false;
return true;
bool llvm::canSinkOrHoistInst(Instruction &I, AAResults *AA, DominatorTree *DT,
Loop *CurLoop, AliasSetTracker *CurAST,
MemorySSAUpdater *MSSAU,
bool TargetExecutesOncePerLoop,
SinkAndHoistLICMFlags *Flags,
OptimizationRemarkEmitter *ORE) {
assert(((CurAST != nullptr) ^ (MSSAU != nullptr)) &&
"Either AliasSetTracker or MemorySSA should be initialized.");
// If we don't understand the instruction, bail early.
if (!isHoistableAndSinkableInst(I))
return false;
MemorySSA *MSSA = MSSAU ? MSSAU->getMemorySSA() : nullptr;
if (MSSA)
assert(Flags != nullptr && "Flags cannot be null.");
// Loads have extra constraints we have to verify before we can hoist them.
if (LoadInst *LI = dyn_cast<LoadInst>(&I)) {
if (!LI->isUnordered())
return false; // Don't sink/hoist volatile or ordered atomic loads!
// Loads from constant memory are always safe to move, even if they end up
// in the same alias set as something that ends up being modified.
if (AA->pointsToConstantMemory(LI->getOperand(0)))
return true;
if (LI->hasMetadata(LLVMContext::MD_invariant_load))
return true;
if (LI->isAtomic() && !TargetExecutesOncePerLoop)
return false; // Don't risk duplicating unordered loads
// This checks for an invariant.start dominating the load.
if (isLoadInvariantInLoop(LI, DT, CurLoop))
return true;
bool Invalidated;
if (CurAST)
Invalidated = pointerInvalidatedByLoop(MemoryLocation::get(LI), CurAST,
CurLoop, AA);
Invalidated = pointerInvalidatedByLoopWithMSSA(
MSSA, cast<MemoryUse>(MSSA->getMemoryAccess(LI)), CurLoop, I, *Flags);
// Check loop-invariant address because this may also be a sinkable load
// whose address is not necessarily loop-invariant.
if (ORE && Invalidated && CurLoop->isLoopInvariant(LI->getPointerOperand()))
ORE->emit([&]() {
return OptimizationRemarkMissed(
DEBUG_TYPE, "LoadWithLoopInvariantAddressInvalidated", LI)
<< "failed to move load with loop-invariant address "
"because the loop may invalidate its value";
return !Invalidated;
} else if (CallInst *CI = dyn_cast<CallInst>(&I)) {
// Don't sink or hoist dbg info; it's legal, but not useful.
if (isa<DbgInfoIntrinsic>(I))
return false;
// Don't sink calls which can throw.
if (CI->mayThrow())
return false;
// Convergent attribute has been used on operations that involve
// inter-thread communication which results are implicitly affected by the
// enclosing control flows. It is not safe to hoist or sink such operations
// across control flow.
if (CI->isConvergent())
return false;
using namespace PatternMatch;
if (match(CI, m_Intrinsic<Intrinsic::assume>()))
// Assumes don't actually alias anything or throw
return true;
if (match(CI, m_Intrinsic<Intrinsic::experimental_widenable_condition>()))
// Widenable conditions don't actually alias anything or throw
return true;
// Handle simple cases by querying alias analysis.
FunctionModRefBehavior Behavior = AA->getModRefBehavior(CI);
if (Behavior == FMRB_DoesNotAccessMemory)
return true;
if (AAResults::onlyReadsMemory(Behavior)) {
// A readonly argmemonly function only reads from memory pointed to by
// it's arguments with arbitrary offsets. If we can prove there are no
// writes to this memory in the loop, we can hoist or sink.
if (AAResults::onlyAccessesArgPointees(Behavior)) {
// TODO: expand to writeable arguments
for (Value *Op : CI->arg_operands())
if (Op->getType()->isPointerTy()) {
bool Invalidated;
if (CurAST)
Invalidated = pointerInvalidatedByLoop(
MemoryLocation::getBeforeOrAfter(Op), CurAST, CurLoop, AA);
Invalidated = pointerInvalidatedByLoopWithMSSA(
MSSA, cast<MemoryUse>(MSSA->getMemoryAccess(CI)), CurLoop, I,
if (Invalidated)
return false;
return true;
// If this call only reads from memory and there are no writes to memory
// in the loop, we can hoist or sink the call as appropriate.
if (isReadOnly(CurAST, MSSAU, CurLoop))
return true;
// FIXME: This should use mod/ref information to see if we can hoist or
// sink the call.
return false;
} else if (auto *FI = dyn_cast<FenceInst>(&I)) {
// Fences alias (most) everything to provide ordering. For the moment,
// just give up if there are any other memory operations in the loop.
if (CurAST) {
auto Begin = CurAST->begin();
assert(Begin != CurAST->end() && "must contain FI");
if (std::next(Begin) != CurAST->end())
// constant memory for instance, TODO: handle better
return false;
auto *UniqueI = Begin->getUniqueInstruction();
if (!UniqueI)
// other memory op, give up
return false;
(void)FI; // suppress unused variable warning
assert(UniqueI == FI && "AS must contain FI");
return true;
} else // MSSAU
return isOnlyMemoryAccess(FI, CurLoop, MSSAU);
} else if (auto *SI = dyn_cast<StoreInst>(&I)) {
if (!SI->isUnordered())
return false; // Don't sink/hoist volatile or ordered atomic store!
// We can only hoist a store that we can prove writes a value which is not
// read or overwritten within the loop. For those cases, we fallback to
// load store promotion instead. TODO: We can extend this to cases where
// there is exactly one write to the location and that write dominates an
// arbitrary number of reads in the loop.
if (CurAST) {
auto &AS = CurAST->getAliasSetFor(MemoryLocation::get(SI));
if (AS.isRef() || !AS.isMustAlias())
// Quick exit test, handled by the full path below as well.
return false;
auto *UniqueI = AS.getUniqueInstruction();
if (!UniqueI)
// other memory op, give up
return false;
assert(UniqueI == SI && "AS must contain SI");
return true;
} else { // MSSAU
if (isOnlyMemoryAccess(SI, CurLoop, MSSAU))
return true;
// If there are more accesses than the Promotion cap or no "quota" to
// check clobber, then give up as we're not walking a list that long.
if (Flags->tooManyMemoryAccesses() || Flags->tooManyClobberingCalls())
return false;
// If there are interfering Uses (i.e. their defining access is in the
// loop), or ordered loads (stored as Defs!), don't move this store.
// Could do better here, but this is conservatively correct.
// TODO: Cache set of Uses on the first walk in runOnLoop, update when
// moving accesses. Can also extend to dominating uses.
auto *SIMD = MSSA->getMemoryAccess(SI);
for (auto *BB : CurLoop->getBlocks())
if (auto *Accesses = MSSA->getBlockAccesses(BB)) {
for (const auto &MA : *Accesses)
if (const auto *MU = dyn_cast<MemoryUse>(&MA)) {
auto *MD = MU->getDefiningAccess();
if (!MSSA->isLiveOnEntryDef(MD) &&
return false;
// Disable hoisting past potentially interfering loads. Optimized
// Uses may point to an access outside the loop, as getClobbering
// checks the previous iteration when walking the backedge.
// FIXME: More precise: no Uses that alias SI.
if (!Flags->getIsSink() && !MSSA->dominates(SIMD, MU))
return false;
} else if (const auto *MD = dyn_cast<MemoryDef>(&MA)) {
if (auto *LI = dyn_cast<LoadInst>(MD->getMemoryInst())) {
(void)LI; // Silence warning.
assert(!LI->isUnordered() && "Expected unordered load");
return false;
// Any call, while it may not be clobbering SI, it may be a use.
if (auto *CI = dyn_cast<CallInst>(MD->getMemoryInst())) {
// Check if the call may read from the memory locattion written
// to by SI. Check CI's attributes and arguments; the number of
// such checks performed is limited above by NoOfMemAccTooLarge.
ModRefInfo MRI = AA->getModRefInfo(CI, MemoryLocation::get(SI));
if (isModOrRefSet(MRI))
return false;
auto *Source = MSSA->getSkipSelfWalker()->getClobberingMemoryAccess(SI);
// If there are no clobbering Defs in the loop, store is safe to hoist.
return MSSA->isLiveOnEntryDef(Source) ||
assert(!I.mayReadOrWriteMemory() && "unhandled aliasing");
// We've established mechanical ability and aliasing, it's up to the caller
// to check fault safety
return true;
/// Returns true if a PHINode is a trivially replaceable with an
/// Instruction.
/// This is true when all incoming values are that instruction.
/// This pattern occurs most often with LCSSA PHI nodes.
static bool isTriviallyReplaceablePHI(const PHINode &PN, const Instruction &I) {
for (const Value *IncValue : PN.incoming_values())
if (IncValue != &I)
return false;
return true;
/// Return true if the instruction is free in the loop.
static bool isFreeInLoop(const Instruction &I, const Loop *CurLoop,
const TargetTransformInfo *TTI) {
if (const GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(&I)) {
if (TTI->getUserCost(GEP, TargetTransformInfo::TCK_SizeAndLatency) !=
return false;
// For a GEP, we cannot simply use getUserCost because currently it
// optimistically assume that a GEP will fold into addressing mode
// regardless of its users.
const BasicBlock *BB = GEP->getParent();
for (const User *U : GEP->users()) {
const Instruction *UI = cast<Instruction>(U);
if (CurLoop->contains(UI) &&
(BB != UI->getParent() ||
(!isa<StoreInst>(UI) && !isa<LoadInst>(UI))))
return false;
return true;
} else
return TTI->getUserCost(&I, TargetTransformInfo::TCK_SizeAndLatency) ==
/// Return true if the only users of this instruction are outside of
/// the loop. If this is true, we can sink the instruction to the exit
/// blocks of the loop.
/// We also return true if the instruction could be folded away in lowering.
/// (e.g., a GEP can be folded into a load as an addressing mode in the loop).
static bool isNotUsedOrFreeInLoop(const Instruction &I, const Loop *CurLoop,
const LoopSafetyInfo *SafetyInfo,
TargetTransformInfo *TTI, bool &FreeInLoop) {
const auto &BlockColors = SafetyInfo->getBlockColors();
bool IsFree = isFreeInLoop(I, CurLoop, TTI);
for (const User *U : I.users()) {
const Instruction *UI = cast<Instruction>(U);
if (const PHINode *PN = dyn_cast<PHINode>(UI)) {
const BasicBlock *BB = PN->getParent();
// We cannot sink uses in catchswitches.
if (isa<CatchSwitchInst>(BB->getTerminator()))
return false;
// We need to sink a callsite to a unique funclet. Avoid sinking if the
// phi use is too muddled.
if (isa<CallInst>(I))
if (!BlockColors.empty() &&
BlockColors.find(const_cast<BasicBlock *>(BB))->second.size() != 1)
return false;
if (CurLoop->contains(UI)) {
if (IsFree) {
FreeInLoop = true;
return false;
return true;
static Instruction *cloneInstructionInExitBlock(
Instruction &I, BasicBlock &ExitBlock, PHINode &PN, const LoopInfo *LI,
const LoopSafetyInfo *SafetyInfo, MemorySSAUpdater *MSSAU) {
Instruction *New;
if (auto *CI = dyn_cast<CallInst>(&I)) {
const auto &BlockColors = SafetyInfo->getBlockColors();
// Sinking call-sites need to be handled differently from other
// instructions. The cloned call-site needs a funclet bundle operand
// appropriate for its location in the CFG.
SmallVector<OperandBundleDef, 1> OpBundles;
for (unsigned BundleIdx = 0, BundleEnd = CI->getNumOperandBundles();
BundleIdx != BundleEnd; ++BundleIdx) {
OperandBundleUse Bundle = CI->getOperandBundleAt(BundleIdx);
if (Bundle.getTagID() == LLVMContext::OB_funclet)
if (!BlockColors.empty()) {
const ColorVector &CV = BlockColors.find(&ExitBlock)->second;
assert(CV.size() == 1 && "non-unique color for exit block!");
BasicBlock *BBColor = CV.front();
Instruction *EHPad = BBColor->getFirstNonPHI();
if (EHPad->isEHPad())
OpBundles.emplace_back("funclet", EHPad);
New = CallInst::Create(CI, OpBundles);
} else {
New = I.clone();
ExitBlock.getInstList().insert(ExitBlock.getFirstInsertionPt(), New);
if (!I.getName().empty())
New->setName(I.getName() + ".le");
if (MSSAU && MSSAU->getMemorySSA()->getMemoryAccess(&I)) {
// Create a new MemoryAccess and let MemorySSA set its defining access.
MemoryAccess *NewMemAcc = MSSAU->createMemoryAccessInBB(
New, nullptr, New->getParent(), MemorySSA::Beginning);
if (NewMemAcc) {
if (auto *MemDef = dyn_cast<MemoryDef>(NewMemAcc))
MSSAU->insertDef(MemDef, /*RenameUses=*/true);
else {
auto *MemUse = cast<MemoryUse>(NewMemAcc);
MSSAU->insertUse(MemUse, /*RenameUses=*/true);
// Build LCSSA PHI nodes for any in-loop operands (if legal). Note that
// this is particularly cheap because we can rip off the PHI node that we're
// replacing for the number and blocks of the predecessors.
// OPT: If this shows up in a profile, we can instead finish sinking all
// invariant instructions, and then walk their operands to re-establish
// LCSSA. That will eliminate creating PHI nodes just to nuke them when
// sinking bottom-up.
for (Use &Op : New->operands())
if (LI->wouldBeOutOfLoopUseRequiringLCSSA(Op.get(), PN.getParent())) {
auto *OInst = cast<Instruction>(Op.get());
PHINode *OpPN =
PHINode::Create(OInst->getType(), PN.getNumIncomingValues(),
OInst->getName() + ".lcssa", &ExitBlock.front());
for (unsigned i = 0, e = PN.getNumIncomingValues(); i != e; ++i)
OpPN->addIncoming(OInst, PN.getIncomingBlock(i));
Op = OpPN;
return New;
static void eraseInstruction(Instruction &I, ICFLoopSafetyInfo &SafetyInfo,
AliasSetTracker *AST, MemorySSAUpdater *MSSAU) {
if (AST)
if (MSSAU)
static void moveInstructionBefore(Instruction &I, Instruction &Dest,
ICFLoopSafetyInfo &SafetyInfo,
MemorySSAUpdater *MSSAU,
ScalarEvolution *SE) {
SafetyInfo.insertInstructionTo(&I, Dest.getParent());
if (MSSAU)
if (MemoryUseOrDef *OldMemAcc = cast_or_null<MemoryUseOrDef>(
MSSAU->moveToPlace(OldMemAcc, Dest.getParent(),
if (SE)
static Instruction *sinkThroughTriviallyReplaceablePHI(
PHINode *TPN, Instruction *I, LoopInfo *LI,
SmallDenseMap<BasicBlock *, Instruction *, 32> &SunkCopies,
const LoopSafetyInfo *SafetyInfo, const Loop *CurLoop,
MemorySSAUpdater *MSSAU) {
assert(isTriviallyReplaceablePHI(*TPN, *I) &&
"Expect only trivially replaceable PHI");
BasicBlock *ExitBlock = TPN->getParent();
Instruction *New;
auto It = SunkCopies.find(ExitBlock);
if (It != SunkCopies.end())
New = It->second;
New = SunkCopies[ExitBlock] = cloneInstructionInExitBlock(
*I, *ExitBlock, *TPN, LI, SafetyInfo, MSSAU);
return New;
static bool canSplitPredecessors(PHINode *PN, LoopSafetyInfo *SafetyInfo) {
BasicBlock *BB = PN->getParent();
if (!BB->canSplitPredecessors())
return false;
// It's not impossible to split EHPad blocks, but if BlockColors already exist
// it require updating BlockColors for all offspring blocks accordingly. By
// skipping such corner case, we can make updating BlockColors after splitting
// predecessor fairly simple.
if (!SafetyInfo->getBlockColors().empty() && BB->getFirstNonPHI()->isEHPad())
return false;
for (pred_iterator PI = pred_begin(BB), E = pred_end(BB); PI != E; ++PI) {
BasicBlock *BBPred = *PI;
if (isa<IndirectBrInst>(BBPred->getTerminator()) ||
return false;
return true;
static void splitPredecessorsOfLoopExit(PHINode *PN, DominatorTree *DT,
LoopInfo *LI, const Loop *CurLoop,
LoopSafetyInfo *SafetyInfo,
MemorySSAUpdater *MSSAU) {
#ifndef NDEBUG
SmallVector<BasicBlock *, 32> ExitBlocks;
SmallPtrSet<BasicBlock *, 32> ExitBlockSet(ExitBlocks.begin(),
BasicBlock *ExitBB = PN->getParent();
assert(ExitBlockSet.count(ExitBB) && "Expect the PHI is in an exit block.");
// Split predecessors of the loop exit to make instructions in the loop are
// exposed to exit blocks through trivially replaceable PHIs while keeping the
// loop in the canonical form where each predecessor of each exit block should
// be contained within the loop. For example, this will convert the loop below
// from
// LB1:
// %v1 =
// br %LE, %LB2
// LB2:
// %v2 =
// br %LE, %LB1
// LE:
// %p = phi [%v1, %LB1], [%v2, %LB2] <-- non-trivially replaceable
// to
// LB1:
// %v1 =
// br %LE.split, %LB2
// LB2:
// %v2 =
// br %LE.split2, %LB1
// LE.split:
// %p1 = phi [%v1, %LB1] <-- trivially replaceable
// br %LE
// LE.split2:
// %p2 = phi [%v2, %LB2] <-- trivially replaceable
// br %LE
// LE:
// %p = phi [%p1, %LE.split], [%p2, %LE.split2]
const auto &BlockColors = SafetyInfo->getBlockColors();
SmallSetVector<BasicBlock *, 8> PredBBs(pred_begin(ExitBB), pred_end(ExitBB));
while (!PredBBs.empty()) {
BasicBlock *PredBB = *PredBBs.begin();
assert(CurLoop->contains(PredBB) &&
"Expect all predecessors are in the loop");
if (PN->getBasicBlockIndex(PredBB) >= 0) {
BasicBlock *NewPred = SplitBlockPredecessors(
ExitBB, PredBB, ".split.loop.exit", DT, LI, MSSAU, true);
// Since we do not allow splitting EH-block with BlockColors in
// canSplitPredecessors(), we can simply assign predecessor's color to
// the new block.
if (!BlockColors.empty())
// Grab a reference to the ColorVector to be inserted before getting the
// reference to the vector we are copying because inserting the new
// element in BlockColors might cause the map to be reallocated.
SafetyInfo->copyColors(NewPred, PredBB);
/// When an instruction is found to only be used outside of the loop, this
/// function moves it to the exit blocks and patches up SSA form as needed.
/// This method is guaranteed to remove the original instruction from its
/// position, and may either delete it or move it to outside of the loop.
static bool sink(Instruction &I, LoopInfo *LI, DominatorTree *DT,
BlockFrequencyInfo *BFI, const Loop *CurLoop,
ICFLoopSafetyInfo *SafetyInfo, MemorySSAUpdater *MSSAU,
OptimizationRemarkEmitter *ORE) {
bool Changed = false;
LLVM_DEBUG(dbgs() << "LICM sinking instruction: " << I << "\n");
// Iterate over users to be ready for actual sinking. Replace users via
// unreachable blocks with undef and make all user PHIs trivially replaceable.
SmallPtrSet<Instruction *, 8> VisitedUsers;
for (Value::user_iterator UI = I.user_begin(), UE = I.user_end(); UI != UE;) {
auto *User = cast<Instruction>(*UI);
Use &U = UI.getUse();
if (VisitedUsers.count(User) || CurLoop->contains(User))
if (!DT->isReachableFromEntry(User->getParent())) {
U = UndefValue::get(I.getType());
Changed = true;
// The user must be a PHI node.
PHINode *PN = cast<PHINode>(User);
// Surprisingly, instructions can be used outside of loops without any
// exits. This can only happen in PHI nodes if the incoming block is
// unreachable.
BasicBlock *BB = PN->getIncomingBlock(U);
if (!DT->isReachableFromEntry(BB)) {
U = UndefValue::get(I.getType());
Changed = true;
if (isTriviallyReplaceablePHI(*PN, I))
if (!canSplitPredecessors(PN, SafetyInfo))
return Changed;
// Split predecessors of the PHI so that we can make users trivially
// replaceable.
splitPredecessorsOfLoopExit(PN, DT, LI, CurLoop, SafetyInfo, MSSAU);
// Should rebuild the iterators, as they may be invalidated by
// splitPredecessorsOfLoopExit().
UI = I.user_begin();
UE = I.user_end();
if (VisitedUsers.empty())
return Changed;
ORE->emit([&]() {
return OptimizationRemark(DEBUG_TYPE, "InstSunk", &I)
<< "sinking " << ore::NV("Inst", &I);
if (isa<LoadInst>(I))
else if (isa<CallInst>(I))
#ifndef NDEBUG
SmallVector<BasicBlock *, 32> ExitBlocks;
SmallPtrSet<BasicBlock *, 32> ExitBlockSet(ExitBlocks.begin(),
// Clones of this instruction. Don't create more than one per exit block!
SmallDenseMap<BasicBlock *, Instruction *, 32> SunkCopies;
// If this instruction is only used outside of the loop, then all users are
// PHI nodes in exit blocks due to LCSSA form. Just RAUW them with clones of
// the instruction.
// First check if I is worth sinking for all uses. Sink only when it is worth
// across all uses.
SmallSetVector<User*, 8> Users(I.user_begin(), I.user_end());
SmallVector<PHINode *, 8> ExitPNs;
for (auto *UI : Users) {
auto *User = cast<Instruction>(UI);
if (CurLoop->contains(User))
PHINode *PN = cast<PHINode>(User);
assert(ExitBlockSet.count(PN->getParent()) &&
"The LCSSA PHI is not in an exit block!");
if (!worthSinkOrHoistInst(I, PN->getParent(), ORE, BFI)) {
return Changed;
for (auto *PN : ExitPNs) {
// The PHI must be trivially replaceable.
Instruction *New = sinkThroughTriviallyReplaceablePHI(
PN, &I, LI, SunkCopies, SafetyInfo, CurLoop, MSSAU);
eraseInstruction(*PN, *SafetyInfo, nullptr, nullptr);
Changed = true;
return Changed;
/// When an instruction is found to only use loop invariant operands that
/// is safe to hoist, this instruction is called to do the dirty work.
static void hoist(Instruction &I, const DominatorTree *DT, const Loop *CurLoop,
BasicBlock *Dest, ICFLoopSafetyInfo *SafetyInfo,
MemorySSAUpdater *MSSAU, ScalarEvolution *SE,
OptimizationRemarkEmitter *ORE) {
LLVM_DEBUG(dbgs() << "LICM hoisting to " << Dest->getNameOrAsOperand() << ": "
<< I << "\n");
ORE->emit([&]() {
return OptimizationRemark(DEBUG_TYPE, "Hoisted", &I) << "hoisting "
<< ore::NV("Inst", &I);
// Metadata can be dependent on conditions we are hoisting above.
// Conservatively strip all metadata on the instruction unless we were
// guaranteed to execute I if we entered the loop, in which case the metadata
// is valid in the loop preheader.
if (I.hasMetadataOtherThanDebugLoc() &&
// The check on hasMetadataOtherThanDebugLoc is to prevent us from burning
// time in isGuaranteedToExecute if we don't actually have anything to
// drop. It is a compile time optimization, not required for correctness.
!SafetyInfo->isGuaranteedToExecute(I, DT, CurLoop))
if (isa<PHINode>(I))
// Move the new node to the end of the phi list in the destination block.
moveInstructionBefore(I, *Dest->getFirstNonPHI(), *SafetyInfo, MSSAU, SE);
// Move the new node to the destination block, before its terminator.
moveInstructionBefore(I, *Dest->getTerminator(), *SafetyInfo, MSSAU, SE);
if (isa<LoadInst>(I))
else if (isa<CallInst>(I))
/// Only sink or hoist an instruction if it is not a trapping instruction,
/// or if the instruction is known not to trap when moved to the preheader.
/// or if it is a trapping instruction and is guaranteed to execute.
static bool isSafeToExecuteUnconditionally(Instruction &Inst,
const DominatorTree *DT,
const TargetLibraryInfo *TLI,
const Loop *CurLoop,
const LoopSafetyInfo *SafetyInfo,
OptimizationRemarkEmitter *ORE,
const Instruction *CtxI) {
if (isSafeToSpeculativelyExecute(&Inst, CtxI, DT, TLI))
return true;
bool GuaranteedToExecute =
SafetyInfo->isGuaranteedToExecute(Inst, DT, CurLoop);
if (!GuaranteedToExecute) {
auto *LI = dyn_cast<LoadInst>(&Inst);
if (LI && CurLoop->isLoopInvariant(LI->getPointerOperand()))
ORE->emit([&]() {
return OptimizationRemarkMissed(
DEBUG_TYPE, "LoadWithLoopInvariantAddressCondExecuted", LI)
<< "failed to hoist load with loop-invariant address "
"because load is conditionally executed";
return GuaranteedToExecute;
namespace {
class LoopPromoter : public LoadAndStorePromoter {
Value *SomePtr; // Designated pointer to store to.
const SmallSetVector<Value *, 8> &PointerMustAliases;
SmallVectorImpl<BasicBlock *> &LoopExitBlocks;
SmallVectorImpl<Instruction *> &LoopInsertPts;
SmallVectorImpl<MemoryAccess *> &MSSAInsertPts;
PredIteratorCache &PredCache;
AliasSetTracker *AST;
MemorySSAUpdater *MSSAU;
LoopInfo &LI;
DebugLoc DL;
int Alignment;
bool UnorderedAtomic;
AAMDNodes AATags;
ICFLoopSafetyInfo &SafetyInfo;
// We're about to add a use of V in a loop exit block. Insert an LCSSA phi
// (if legal) if doing so would add an out-of-loop use to an instruction
// defined in-loop.
Value *maybeInsertLCSSAPHI(Value *V, BasicBlock *BB) const {
if (!LI.wouldBeOutOfLoopUseRequiringLCSSA(V, BB))
return V;
Instruction *I = cast<Instruction>(V);
// We need to create an LCSSA PHI node for the incoming value and
// store that.
PHINode *PN = PHINode::Create(I->getType(), PredCache.size(BB),
I->getName() + ".lcssa", &BB->front());
for (BasicBlock *Pred : PredCache.get(BB))
PN->addIncoming(I, Pred);
return PN;
LoopPromoter(Value *SP, ArrayRef<const Instruction *> Insts, SSAUpdater &S,
const SmallSetVector<Value *, 8> &PMA,
SmallVectorImpl<BasicBlock *> &LEB,
SmallVectorImpl<Instruction *> &LIP,
SmallVectorImpl<MemoryAccess *> &MSSAIP, PredIteratorCache &PIC,
AliasSetTracker *ast, MemorySSAUpdater *MSSAU, LoopInfo &li,
DebugLoc dl, int alignment, bool UnorderedAtomic,
const AAMDNodes &AATags, ICFLoopSafetyInfo &SafetyInfo)
: LoadAndStorePromoter(Insts, S), SomePtr(SP), PointerMustAliases(PMA),
LoopExitBlocks(LEB), LoopInsertPts(LIP), MSSAInsertPts(MSSAIP),
PredCache(PIC), AST(ast), MSSAU(MSSAU), LI(li), DL(std::move(dl)),
Alignment(alignment), UnorderedAtomic(UnorderedAtomic), AATags(AATags),
SafetyInfo(SafetyInfo) {}
bool isInstInList(Instruction *I,
const SmallVectorImpl<Instruction *> &) const override {
Value *Ptr;
if (LoadInst *LI = dyn_cast<LoadInst>(I))
Ptr = LI->getOperand(0);
Ptr = cast<StoreInst>(I)->getPointerOperand();
return PointerMustAliases.count(Ptr);
void doExtraRewritesBeforeFinalDeletion() override {
// Insert stores after in the loop exit blocks. Each exit block gets a
// store of the live-out values that feed them. Since we've already told
// the SSA updater about the defs in the loop and the preheader
// definition, it is all set and we can start using it.
for (unsigned i = 0, e = LoopExitBlocks.size(); i != e; ++i) {
BasicBlock *ExitBlock = LoopExitBlocks[i];
Value *LiveInValue = SSA.GetValueInMiddleOfBlock(ExitBlock);
LiveInValue = maybeInsertLCSSAPHI(LiveInValue, ExitBlock);
Value *Ptr = maybeInsertLCSSAPHI(SomePtr, ExitBlock);
Instruction *InsertPos = LoopInsertPts[i];
StoreInst *NewSI = new StoreInst(LiveInValue, Ptr, InsertPos);
if (UnorderedAtomic)
if (AATags)
if (MSSAU) {
MemoryAccess *MSSAInsertPoint = MSSAInsertPts[i];
MemoryAccess *NewMemAcc;
if (!MSSAInsertPoint) {
NewMemAcc = MSSAU->createMemoryAccessInBB(
NewSI, nullptr, NewSI->getParent(), MemorySSA::Beginning);
} else {
NewMemAcc =
MSSAU->createMemoryAccessAfter(NewSI, nullptr, MSSAInsertPoint);
MSSAInsertPts[i] = NewMemAcc;
MSSAU->insertDef(cast<MemoryDef>(NewMemAcc), true);
// FIXME: true for safety, false may still be correct.
void replaceLoadWithValue(LoadInst *LI, Value *V) const override {
// Update alias analysis.
if (AST)
AST->copyValue(LI, V);
void instructionDeleted(Instruction *I) const override {
if (AST)
if (MSSAU)
/// Return true iff we can prove that a caller of this function can not inspect
/// the contents of the provided object in a well defined program.
bool isKnownNonEscaping(Value *Object, const TargetLibraryInfo *TLI) {
if (isa<AllocaInst>(Object))
// Since the alloca goes out of scope, we know the caller can't retain a
// reference to it and be well defined. Thus, we don't need to check for
// capture.
return true;
// For all other objects we need to know that the caller can't possibly
// have gotten a reference to the object. There are two components of
// that:
// 1) Object can't be escaped by this function. This is what
// PointerMayBeCaptured checks.
// 2) Object can't have been captured at definition site. For this, we
// need to know the return value is noalias. At the moment, we use a
// weaker condition and handle only AllocLikeFunctions (which are
// known to be noalias). TODO
return isAllocLikeFn(Object, TLI) &&
!PointerMayBeCaptured(Object, true, true);
} // namespace
/// Try to promote memory values to scalars by sinking stores out of the
/// loop and moving loads to before the loop. We do this by looping over
/// the stores in the loop, looking for stores to Must pointers which are
/// loop invariant.
bool llvm::promoteLoopAccessesToScalars(
const SmallSetVector<Value *, 8> &PointerMustAliases,
SmallVectorImpl<BasicBlock *> &ExitBlocks,
SmallVectorImpl<Instruction *> &InsertPts,
SmallVectorImpl<MemoryAccess *> &MSSAInsertPts, PredIteratorCache &PIC,
LoopInfo *LI, DominatorTree *DT, const TargetLibraryInfo *TLI,
Loop *CurLoop, AliasSetTracker *CurAST, MemorySSAUpdater *MSSAU,
ICFLoopSafetyInfo *SafetyInfo, OptimizationRemarkEmitter *ORE) {
// Verify inputs.
assert(LI != nullptr && DT != nullptr && CurLoop != nullptr &&
SafetyInfo != nullptr &&
"Unexpected Input to promoteLoopAccessesToScalars");
Value *SomePtr = *PointerMustAliases.begin();
BasicBlock *Preheader = CurLoop->getLoopPreheader();
// It is not safe to promote a load/store from the loop if the load/store is
// conditional. For example, turning:
// for () { if (c) *P += 1; }
// into:
// tmp = *P; for () { if (c) tmp +=1; } *P = tmp;
// is not safe, because *P may only be valid to access if 'c' is true.
// The safety property divides into two parts:
// p1) The memory may not be dereferenceable on entry to the loop. In this
// case, we can't insert the required load in the preheader.
// p2) The memory model does not allow us to insert a store along any dynamic
// path which did not originally have one.
// If at least one store is guaranteed to execute, both properties are
// satisfied, and promotion is legal.
// This, however, is not a necessary condition. Even if no store/load is
// guaranteed to execute, we can still establish these properties.
// We can establish (p1) by proving that hoisting the load into the preheader
// is safe (i.e. proving dereferenceability on all paths through the loop). We
// can use any access within the alias set to prove dereferenceability,
// since they're all must alias.
// There are two ways establish (p2):
// a) Prove the location is thread-local. In this case the memory model
// requirement does not apply, and stores are safe to insert.
// b) Prove a store dominates every exit block. In this case, if an exit
// blocks is reached, the original dynamic path would have taken us through
// the store, so inserting a store into the exit block is safe. Note that this
// is different from the store being guaranteed to execute. For instance,
// if an exception is thrown on the first iteration of the loop, the original
// store is never executed, but the exit blocks are not executed either.
bool DereferenceableInPH = false;
bool SafeToInsertStore = false;
SmallVector<Instruction *, 64> LoopUses;
// We start with an alignment of one and try to find instructions that allow
// us to prove better alignment.
Align Alignment;
// Keep track of which types of access we see
bool SawUnorderedAtomic = false;
bool SawNotAtomic = false;
AAMDNodes AATags;
const DataLayout &MDL = Preheader->getModule()->getDataLayout();
bool IsKnownThreadLocalObject = false;
if (SafetyInfo->anyBlockMayThrow()) {
// If a loop can throw, we have to insert a store along each unwind edge.
// That said, we can't actually make the unwind edge explicit. Therefore,
// we have to prove that the store is dead along the unwind edge. We do
// this by proving that the caller can't have a reference to the object
// after return and thus can't possibly load from the object.
Value *Object = getUnderlyingObject(SomePtr);
if (!isKnownNonEscaping(Object, TLI))
return false;
// Subtlety: Alloca's aren't visible to callers, but *are* potentially
// visible to other threads if captured and used during their lifetimes.
IsKnownThreadLocalObject = !isa<AllocaInst>(Object);
// Check that all of the pointers in the alias set have the same type. We
// cannot (yet) promote a memory location that is loaded and stored in
// different sizes. While we are at it, collect alignment and AA info.
for (Value *ASIV : PointerMustAliases) {
// Check that all of the pointers in the alias set have the same type. We
// cannot (yet) promote a memory location that is loaded and stored in
// different sizes.
if (SomePtr->getType() != ASIV->getType())
return false;
for (User *U : ASIV->users()) {
// Ignore instructions that are outside the loop.
Instruction *UI = dyn_cast<Instruction>(U);
if (!UI || !CurLoop->contains(UI))
// If there is an non-load/store instruction in the loop, we can't promote
// it.
if (LoadInst *Load = dyn_cast<LoadInst>(UI)) {
if (!Load->isUnordered())
return false;
SawUnorderedAtomic |= Load->isAtomic();
SawNotAtomic |= !Load->isAtomic();
Align InstAlignment = Load->getAlign();
// Note that proving a load safe to speculate requires proving
// sufficient alignment at the target location. Proving it guaranteed
// to execute does as well. Thus we can increase our guaranteed
// alignment as well.
if (!DereferenceableInPH || (InstAlignment > Alignment))
if (isSafeToExecuteUnconditionally(*Load, DT, TLI, CurLoop,
SafetyInfo, ORE,
Preheader->getTerminator())) {
DereferenceableInPH = true;
Alignment = std::max(Alignment, InstAlignment);
} else if (const StoreInst *Store = dyn_cast<StoreInst>(UI)) {
// Stores *of* the pointer are not interesting, only stores *to* the
// pointer.
if (UI->getOperand(1) != ASIV)
if (!Store->isUnordered())
return false;
SawUnorderedAtomic |= Store->isAtomic();
SawNotAtomic |= !Store->isAtomic();
// If the store is guaranteed to execute, both properties are satisfied.
// We may want to check if a store is guaranteed to execute even if we
// already know that promotion is safe, since it may have higher
// alignment than any other guaranteed stores, in which case we can
// raise the alignment on the promoted store.
Align InstAlignment = Store->getAlign();
if (!DereferenceableInPH || !SafeToInsertStore ||
(InstAlignment > Alignment)) {
if (SafetyInfo->isGuaranteedToExecute(*UI, DT, CurLoop)) {
DereferenceableInPH = true;
SafeToInsertStore = true;
Alignment = std::max(Alignment, InstAlignment);
// If a store dominates all exit blocks, it is safe to sink.
// As explained above, if an exit block was executed, a dominating
// store must have been executed at least once, so we are not
// introducing stores on paths that did not have them.
// Note that this only looks at explicit exit blocks. If we ever
// start sinking stores into unwind edges (see above), this will break.
if (!SafeToInsertStore)
SafeToInsertStore = llvm::all_of(ExitBlocks, [&](BasicBlock *Exit) {
return DT->dominates(Store->getParent(), Exit);
// If the store is not guaranteed to execute, we may still get
// deref info through it.
if (!DereferenceableInPH) {
DereferenceableInPH = isDereferenceableAndAlignedPointer(
Store->getPointerOperand(), Store->getValueOperand()->getType(),
Store->getAlign(), MDL, Preheader->getTerminator(), DT, TLI);
} else
return false; // Not a load or store.
// Merge the AA tags.
if (LoopUses.empty()) {
// On the first load/store, just take its AA tags.
} else if (AATags) {
UI->getAAMetadata(AATags, /* Merge = */ true);
// If we found both an unordered atomic instruction and a non-atomic memory
// access, bail. We can't blindly promote non-atomic to atomic since we
// might not be able to lower the result. We can't downgrade since that
// would violate memory model. Also, align 0 is an error for atomics.
if (SawUnorderedAtomic && SawNotAtomic)
return false;
// If we're inserting an atomic load in the preheader, we must be able to
// lower it. We're only guaranteed to be able to lower naturally aligned
// atomics.
auto *SomePtrElemType = SomePtr->getType()->getPointerElementType();
if (SawUnorderedAtomic &&
Alignment < MDL.getTypeStoreSize(SomePtrElemType))
return false;
// If we couldn't prove we can hoist the load, bail.
if (!DereferenceableInPH)
return false;
// We know we can hoist the load, but don't have a guaranteed store.
// Check whether the location is thread-local. If it is, then we can insert
// stores along paths which originally didn't have them without violating the
// memory model.
if (!SafeToInsertStore) {
if (IsKnownThreadLocalObject)
SafeToInsertStore = true;
else {
Value *Object = getUnderlyingObject(SomePtr);
SafeToInsertStore =
(isAllocLikeFn(Object, TLI) || isa<AllocaInst>(Object)) &&
!PointerMayBeCaptured(Object, true, true);
// If we've still failed to prove we can sink the store, give up.
if (!SafeToInsertStore)
return false;
// Otherwise, this is safe to promote, lets do it!
LLVM_DEBUG(dbgs() << "LICM: Promoting value stored to in loop: " << *SomePtr
<< '\n');
ORE->emit([&]() {
return OptimizationRemark(DEBUG_TYPE, "PromoteLoopAccessesToScalar",
<< "Moving accesses to memory location out of the loop";
// Look at all the loop uses, and try to merge their locations.
std::vector<const DILocation *> LoopUsesLocs;
for (auto U : LoopUses)
auto DL = DebugLoc(DILocation::getMergedLocations(LoopUsesLocs));
// We use the SSAUpdater interface to insert phi nodes as required.
SmallVector<PHINode *, 16> NewPHIs;
SSAUpdater SSA(&NewPHIs);
LoopPromoter Promoter(SomePtr, LoopUses, SSA, PointerMustAliases, ExitBlocks,
InsertPts, MSSAInsertPts, PIC, CurAST, MSSAU, *LI, DL,
Alignment.value(), SawUnorderedAtomic, AATags,
// Set up the preheader to have a definition of the value. It is the live-out
// value from the preheader that uses in the loop will use.
LoadInst *PreheaderLoad = new LoadInst(
SomePtr->getType()->getPointerElementType(), SomePtr,
SomePtr->getName() + ".promoted", Preheader->getTerminator());
if (SawUnorderedAtomic)
if (AATags)
SSA.AddAvailableValue(Preheader, PreheaderLoad);
if (MSSAU) {
MemoryAccess *PreheaderLoadMemoryAccess = MSSAU->createMemoryAccessInBB(
PreheaderLoad, nullptr, PreheaderLoad->getParent(), MemorySSA::End);
MemoryUse *NewMemUse = cast<MemoryUse>(PreheaderLoadMemoryAccess);
MSSAU->insertUse(NewMemUse, /*RenameUses=*/true);
if (MSSAU && VerifyMemorySSA)
// Rewrite all the loads in the loop and remember all the definitions from
// stores in the loop.;
if (MSSAU && VerifyMemorySSA)
// If the SSAUpdater didn't use the load in the preheader, just zap it now.
if (PreheaderLoad->use_empty())
eraseInstruction(*PreheaderLoad, *SafetyInfo, CurAST, MSSAU);
return true;
static void foreachMemoryAccess(MemorySSA *MSSA, Loop *L,
function_ref<void(Instruction *)> Fn) {
for (const BasicBlock *BB : L->blocks())
if (const auto *Accesses = MSSA->getBlockAccesses(BB))
for (const auto &Access : *Accesses)
if (const auto *MUD = dyn_cast<MemoryUseOrDef>(&Access))
static SmallVector<SmallSetVector<Value *, 8>, 0>
collectPromotionCandidates(MemorySSA *MSSA, AliasAnalysis *AA, Loop *L,
SmallVectorImpl<Instruction *> &MaybePromotable) {
AliasSetTracker AST(*AA);
auto IsPotentiallyPromotable = [L](const Instruction *I) {
if (const auto *SI = dyn_cast<StoreInst>(I))
return L->isLoopInvariant(SI->getPointerOperand());
if (const auto *LI = dyn_cast<LoadInst>(I))
return L->isLoopInvariant(LI->getPointerOperand());
return false;
// Populate AST with potentially promotable accesses and remove them from
// MaybePromotable, so they will not be checked again on the next iteration.
SmallPtrSet<Value *, 16> AttemptingPromotion;
llvm::erase_if(MaybePromotable, [&](Instruction *I) {
if (IsPotentiallyPromotable(I)) {
return true;
return false;
// We're only interested in must-alias sets that contain a mod.
SmallVector<const AliasSet *, 8> Sets;
for (AliasSet &AS : AST)
if (!AS.isForwardingAliasSet() && AS.isMod() && AS.isMustAlias())
if (Sets.empty())
return {}; // Nothing to promote...
// Discard any sets for which there is an aliasing non-promotable access.
foreachMemoryAccess(MSSA, L, [&](Instruction *I) {
if (AttemptingPromotion.contains(I))
llvm::erase_if(Sets, [&](const AliasSet *AS) {
return AS->aliasesUnknownInst(I, *AA);
SmallVector<SmallSetVector<Value *, 8>, 0> Result;
for (const AliasSet *Set : Sets) {
SmallSetVector<Value *, 8> PointerMustAliases;
for (const auto &ASI : *Set)
return Result;
/// Returns an owning pointer to an alias set which incorporates aliasing info
/// from L and all subloops of L.
LoopInvariantCodeMotion::collectAliasInfoForLoop(Loop *L, LoopInfo *LI,
AAResults *AA) {
auto CurAST = std::make_unique<AliasSetTracker>(*AA);
// Add everything from all the sub loops.
for (Loop *InnerL : L->getSubLoops())
for (BasicBlock *BB : InnerL->blocks())
// And merge in this loop (without anything from inner loops).
for (BasicBlock *BB : L->blocks())
if (LI->getLoopFor(BB) == L)
return CurAST;
static bool pointerInvalidatedByLoop(MemoryLocation MemLoc,
AliasSetTracker *CurAST, Loop *CurLoop,
AAResults *AA) {
// First check to see if any of the basic blocks in CurLoop invalidate *V.
bool isInvalidatedAccordingToAST = CurAST->getAliasSetFor(MemLoc).isMod();
if (!isInvalidatedAccordingToAST || !LICMN2Theshold)
return isInvalidatedAccordingToAST;
// Check with a diagnostic analysis if we can refine the information above.
// This is to identify the limitations of using the AST.
// The alias set mechanism used by LICM has a major weakness in that it
// combines all things which may alias into a single set *before* asking
// modref questions. As a result, a single readonly call within a loop will
// collapse all loads and stores into a single alias set and report
// invalidation if the loop contains any store. For example, readonly calls
// with deopt states have this form and create a general alias set with all
// loads and stores. In order to get any LICM in loops containing possible
// deopt states we need a more precise invalidation of checking the mod ref
// info of each instruction within the loop and LI. This has a complexity of
// O(N^2), so currently, it is used only as a diagnostic tool since the
// default value of LICMN2Threshold is zero.
// Don't look at nested loops.
if (CurLoop->begin() != CurLoop->end())
return true;
int N = 0;
for (BasicBlock *BB : CurLoop->getBlocks())
for (Instruction &I : *BB) {
if (N >= LICMN2Theshold) {
LLVM_DEBUG(dbgs() << "Alasing N2 threshold exhausted for "
<< *(MemLoc.Ptr) << "\n");
return true;
auto Res = AA->getModRefInfo(&I, MemLoc);
if (isModSet(Res)) {
LLVM_DEBUG(dbgs() << "Aliasing failed on " << I << " for "
<< *(MemLoc.Ptr) << "\n");
return true;
LLVM_DEBUG(dbgs() << "Aliasing okay for " << *(MemLoc.Ptr) << "\n");
return false;
bool pointerInvalidatedByLoopWithMSSA(MemorySSA *MSSA, MemoryUse *MU,
Loop *CurLoop, Instruction &I,
SinkAndHoistLICMFlags &Flags) {
// For hoisting, use the walker to determine safety
if (!Flags.getIsSink()) {
MemoryAccess *Source;
// See declaration of SetLicmMssaOptCap for usage details.
if (Flags.tooManyClobberingCalls())
Source = MU->getDefiningAccess();
else {
Source = MSSA->getSkipSelfWalker()->getClobberingMemoryAccess(MU);
return !MSSA->isLiveOnEntryDef(Source) &&
// For sinking, we'd need to check all Defs below this use. The getClobbering
// call will look on the backedge of the loop, but will check aliasing with
// the instructions on the previous iteration.
// For example:
// for (i ... )
// load a[i] ( Use (LoE)
// store a[i] ( 1 = Def (2), with 2 = Phi for the loop.
// i++;
// The load sees no clobbering inside the loop, as the backedge alias check
// does phi translation, and will check aliasing against store a[i-1].
// However sinking the load outside the loop, below the store is incorrect.
// For now, only sink if there are no Defs in the loop, and the existing ones
// precede the use and are in the same block.
// FIXME: Increase precision: Safe to sink if Use post dominates the Def;
// needs PostDominatorTreeAnalysis.
// FIXME: More precise: no Defs that alias this Use.
if (Flags.tooManyMemoryAccesses())
return true;
for (auto *BB : CurLoop->getBlocks())
if (pointerInvalidatedByBlockWithMSSA(*BB, *