blob: 08ba65d9483e09763e6b38623ef66058c50e3475 [file] [log] [blame]
//===----------------- LoopRotationUtils.cpp -----------------------------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
//
// This file provides utilities to convert a loop into a loop with bottom test.
//
//===----------------------------------------------------------------------===//
#include "llvm/Transforms/Utils/LoopRotationUtils.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/Analysis/AssumptionCache.h"
#include "llvm/Analysis/CodeMetrics.h"
#include "llvm/Analysis/DomTreeUpdater.h"
#include "llvm/Analysis/InstructionSimplify.h"
#include "llvm/Analysis/LoopInfo.h"
#include "llvm/Analysis/MemorySSA.h"
#include "llvm/Analysis/MemorySSAUpdater.h"
#include "llvm/Analysis/ScalarEvolution.h"
#include "llvm/Analysis/ValueTracking.h"
#include "llvm/IR/CFG.h"
#include "llvm/IR/DebugInfo.h"
#include "llvm/IR/Dominators.h"
#include "llvm/IR/IntrinsicInst.h"
#include "llvm/IR/MDBuilder.h"
#include "llvm/IR/ProfDataUtils.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Transforms/Utils/BasicBlockUtils.h"
#include "llvm/Transforms/Utils/Cloning.h"
#include "llvm/Transforms/Utils/Local.h"
#include "llvm/Transforms/Utils/SSAUpdater.h"
#include "llvm/Transforms/Utils/ValueMapper.h"
using namespace llvm;
#define DEBUG_TYPE "loop-rotate"
STATISTIC(NumNotRotatedDueToHeaderSize,
"Number of loops not rotated due to the header size");
STATISTIC(NumInstrsHoisted,
"Number of instructions hoisted into loop preheader");
STATISTIC(NumInstrsDuplicated,
"Number of instructions cloned into loop preheader");
STATISTIC(NumRotated, "Number of loops rotated");
static cl::opt<bool>
MultiRotate("loop-rotate-multi", cl::init(false), cl::Hidden,
cl::desc("Allow loop rotation multiple times in order to reach "
"a better latch exit"));
// Probability that a rotated loop has zero trip count / is never entered.
static constexpr uint32_t ZeroTripCountWeights[] = {1, 127};
namespace {
/// A simple loop rotation transformation.
class LoopRotate {
const unsigned MaxHeaderSize;
LoopInfo *LI;
const TargetTransformInfo *TTI;
AssumptionCache *AC;
DominatorTree *DT;
ScalarEvolution *SE;
MemorySSAUpdater *MSSAU;
const SimplifyQuery &SQ;
bool RotationOnly;
bool IsUtilMode;
bool PrepareForLTO;
public:
LoopRotate(unsigned MaxHeaderSize, LoopInfo *LI,
const TargetTransformInfo *TTI, AssumptionCache *AC,
DominatorTree *DT, ScalarEvolution *SE, MemorySSAUpdater *MSSAU,
const SimplifyQuery &SQ, bool RotationOnly, bool IsUtilMode,
bool PrepareForLTO)
: MaxHeaderSize(MaxHeaderSize), LI(LI), TTI(TTI), AC(AC), DT(DT), SE(SE),
MSSAU(MSSAU), SQ(SQ), RotationOnly(RotationOnly),
IsUtilMode(IsUtilMode), PrepareForLTO(PrepareForLTO) {}
bool processLoop(Loop *L);
private:
bool rotateLoop(Loop *L, bool SimplifiedLatch);
bool simplifyLoopLatch(Loop *L);
};
} // end anonymous namespace
/// Insert (K, V) pair into the ValueToValueMap, and verify the key did not
/// previously exist in the map, and the value was inserted.
static void InsertNewValueIntoMap(ValueToValueMapTy &VM, Value *K, Value *V) {
bool Inserted = VM.insert({K, V}).second;
assert(Inserted);
(void)Inserted;
}
/// RewriteUsesOfClonedInstructions - We just cloned the instructions from the
/// old header into the preheader. If there were uses of the values produced by
/// these instruction that were outside of the loop, we have to insert PHI nodes
/// to merge the two values. Do this now.
static void RewriteUsesOfClonedInstructions(BasicBlock *OrigHeader,
BasicBlock *OrigPreheader,
ValueToValueMapTy &ValueMap,
ScalarEvolution *SE,
SmallVectorImpl<PHINode*> *InsertedPHIs) {
// Remove PHI node entries that are no longer live.
BasicBlock::iterator I, E = OrigHeader->end();
for (I = OrigHeader->begin(); PHINode *PN = dyn_cast<PHINode>(I); ++I)
PN->removeIncomingValue(PN->getBasicBlockIndex(OrigPreheader));
// Now fix up users of the instructions in OrigHeader, inserting PHI nodes
// as necessary.
SSAUpdater SSA(InsertedPHIs);
for (I = OrigHeader->begin(); I != E; ++I) {
Value *OrigHeaderVal = &*I;
// If there are no uses of the value (e.g. because it returns void), there
// is nothing to rewrite.
if (OrigHeaderVal->use_empty())
continue;
Value *OrigPreHeaderVal = ValueMap.lookup(OrigHeaderVal);
// The value now exits in two versions: the initial value in the preheader
// and the loop "next" value in the original header.
SSA.Initialize(OrigHeaderVal->getType(), OrigHeaderVal->getName());
// Force re-computation of OrigHeaderVal, as some users now need to use the
// new PHI node.
if (SE)
SE->forgetValue(OrigHeaderVal);
SSA.AddAvailableValue(OrigHeader, OrigHeaderVal);
SSA.AddAvailableValue(OrigPreheader, OrigPreHeaderVal);
// Visit each use of the OrigHeader instruction.
for (Use &U : llvm::make_early_inc_range(OrigHeaderVal->uses())) {
// SSAUpdater can't handle a non-PHI use in the same block as an
// earlier def. We can easily handle those cases manually.
Instruction *UserInst = cast<Instruction>(U.getUser());
if (!isa<PHINode>(UserInst)) {
BasicBlock *UserBB = UserInst->getParent();
// The original users in the OrigHeader are already using the
// original definitions.
if (UserBB == OrigHeader)
continue;
// Users in the OrigPreHeader need to use the value to which the
// original definitions are mapped.
if (UserBB == OrigPreheader) {
U = OrigPreHeaderVal;
continue;
}
}
// Anything else can be handled by SSAUpdater.
SSA.RewriteUse(U);
}
// Replace MetadataAsValue(ValueAsMetadata(OrigHeaderVal)) uses in debug
// intrinsics.
SmallVector<DbgValueInst *, 1> DbgValues;
SmallVector<DbgVariableRecord *, 1> DbgVariableRecords;
llvm::findDbgValues(DbgValues, OrigHeaderVal, &DbgVariableRecords);
for (auto &DbgValue : DbgValues) {
// The original users in the OrigHeader are already using the original
// definitions.
BasicBlock *UserBB = DbgValue->getParent();
if (UserBB == OrigHeader)
continue;
// Users in the OrigPreHeader need to use the value to which the
// original definitions are mapped and anything else can be handled by
// the SSAUpdater. To avoid adding PHINodes, check if the value is
// available in UserBB, if not substitute undef.
Value *NewVal;
if (UserBB == OrigPreheader)
NewVal = OrigPreHeaderVal;
else if (SSA.HasValueForBlock(UserBB))
NewVal = SSA.GetValueInMiddleOfBlock(UserBB);
else
NewVal = UndefValue::get(OrigHeaderVal->getType());
DbgValue->replaceVariableLocationOp(OrigHeaderVal, NewVal);
}
// RemoveDIs: duplicate implementation for non-instruction debug-info
// storage in DbgVariableRecords.
for (DbgVariableRecord *DVR : DbgVariableRecords) {
// The original users in the OrigHeader are already using the original
// definitions.
BasicBlock *UserBB = DVR->getMarker()->getParent();
if (UserBB == OrigHeader)
continue;
// Users in the OrigPreHeader need to use the value to which the
// original definitions are mapped and anything else can be handled by
// the SSAUpdater. To avoid adding PHINodes, check if the value is
// available in UserBB, if not substitute undef.
Value *NewVal;
if (UserBB == OrigPreheader)
NewVal = OrigPreHeaderVal;
else if (SSA.HasValueForBlock(UserBB))
NewVal = SSA.GetValueInMiddleOfBlock(UserBB);
else
NewVal = UndefValue::get(OrigHeaderVal->getType());
DVR->replaceVariableLocationOp(OrigHeaderVal, NewVal);
}
}
}
// Assuming both header and latch are exiting, look for a phi which is only
// used outside the loop (via a LCSSA phi) in the exit from the header.
// This means that rotating the loop can remove the phi.
static bool profitableToRotateLoopExitingLatch(Loop *L) {
BasicBlock *Header = L->getHeader();
BranchInst *BI = dyn_cast<BranchInst>(Header->getTerminator());
assert(BI && BI->isConditional() && "need header with conditional exit");
BasicBlock *HeaderExit = BI->getSuccessor(0);
if (L->contains(HeaderExit))
HeaderExit = BI->getSuccessor(1);
for (auto &Phi : Header->phis()) {
// Look for uses of this phi in the loop/via exits other than the header.
if (llvm::any_of(Phi.users(), [HeaderExit](const User *U) {
return cast<Instruction>(U)->getParent() != HeaderExit;
}))
continue;
return true;
}
return false;
}
// Check that latch exit is deoptimizing (which means - very unlikely to happen)
// and there is another exit from the loop which is non-deoptimizing.
// If we rotate latch to that exit our loop has a better chance of being fully
// canonical.
//
// It can give false positives in some rare cases.
static bool canRotateDeoptimizingLatchExit(Loop *L) {
BasicBlock *Latch = L->getLoopLatch();
assert(Latch && "need latch");
BranchInst *BI = dyn_cast<BranchInst>(Latch->getTerminator());
// Need normal exiting latch.
if (!BI || !BI->isConditional())
return false;
BasicBlock *Exit = BI->getSuccessor(1);
if (L->contains(Exit))
Exit = BI->getSuccessor(0);
// Latch exit is non-deoptimizing, no need to rotate.
if (!Exit->getPostdominatingDeoptimizeCall())
return false;
SmallVector<BasicBlock *, 4> Exits;
L->getUniqueExitBlocks(Exits);
if (!Exits.empty()) {
// There is at least one non-deoptimizing exit.
//
// Note, that BasicBlock::getPostdominatingDeoptimizeCall is not exact,
// as it can conservatively return false for deoptimizing exits with
// complex enough control flow down to deoptimize call.
//
// That means here we can report success for a case where
// all exits are deoptimizing but one of them has complex enough
// control flow (e.g. with loops).
//
// That should be a very rare case and false positives for this function
// have compile-time effect only.
return any_of(Exits, [](const BasicBlock *BB) {
return !BB->getPostdominatingDeoptimizeCall();
});
}
return false;
}
static void updateBranchWeights(BranchInst &PreHeaderBI, BranchInst &LoopBI,
bool HasConditionalPreHeader,
bool SuccsSwapped) {
MDNode *WeightMD = getBranchWeightMDNode(PreHeaderBI);
if (WeightMD == nullptr)
return;
// LoopBI should currently be a clone of PreHeaderBI with the same
// metadata. But we double check to make sure we don't have a degenerate case
// where instsimplify changed the instructions.
if (WeightMD != getBranchWeightMDNode(LoopBI))
return;
SmallVector<uint32_t, 2> Weights;
extractFromBranchWeightMD32(WeightMD, Weights);
if (Weights.size() != 2)
return;
uint32_t OrigLoopExitWeight = Weights[0];
uint32_t OrigLoopBackedgeWeight = Weights[1];
if (SuccsSwapped)
std::swap(OrigLoopExitWeight, OrigLoopBackedgeWeight);
// Update branch weights. Consider the following edge-counts:
//
// | |-------- |
// V V | V
// Br i1 ... | Br i1 ...
// | | | | |
// x| y| | becomes: | y0| |-----
// V V | | V V |
// Exit Loop | | Loop |
// | | | Br i1 ... |
// ----- | | | |
// x0| x1| y1 | |
// V V ----
// Exit
//
// The following must hold:
// - x == x0 + x1 # counts to "exit" must stay the same.
// - y0 == x - x0 == x1 # how often loop was entered at all.
// - y1 == y - y0 # How often loop was repeated (after first iter.).
//
// We cannot generally deduce how often we had a zero-trip count loop so we
// have to make a guess for how to distribute x among the new x0 and x1.
uint32_t ExitWeight0; // aka x0
uint32_t ExitWeight1; // aka x1
uint32_t EnterWeight; // aka y0
uint32_t LoopBackWeight; // aka y1
if (OrigLoopExitWeight > 0 && OrigLoopBackedgeWeight > 0) {
ExitWeight0 = 0;
if (HasConditionalPreHeader) {
// Here we cannot know how many 0-trip count loops we have, so we guess:
if (OrigLoopBackedgeWeight >= OrigLoopExitWeight) {
// If the loop count is bigger than the exit count then we set
// probabilities as if 0-trip count nearly never happens.
ExitWeight0 = ZeroTripCountWeights[0];
// Scale up counts if necessary so we can match `ZeroTripCountWeights`
// for the `ExitWeight0`:`ExitWeight1` (aka `x0`:`x1` ratio`) ratio.
while (OrigLoopExitWeight < ZeroTripCountWeights[1] + ExitWeight0) {
// ... but don't overflow.
uint32_t const HighBit = uint32_t{1} << (sizeof(uint32_t) * 8 - 1);
if ((OrigLoopBackedgeWeight & HighBit) != 0 ||
(OrigLoopExitWeight & HighBit) != 0)
break;
OrigLoopBackedgeWeight <<= 1;
OrigLoopExitWeight <<= 1;
}
} else {
// If there's a higher exit-count than backedge-count then we set
// probabilities as if there are only 0-trip and 1-trip cases.
ExitWeight0 = OrigLoopExitWeight - OrigLoopBackedgeWeight;
}
} else {
// Theoretically, if the loop body must be executed at least once, the
// backedge count must be not less than exit count. However the branch
// weight collected by sampling-based PGO may be not very accurate due to
// sampling. Therefore this workaround is required here to avoid underflow
// of unsigned in following update of branch weight.
if (OrigLoopExitWeight > OrigLoopBackedgeWeight)
OrigLoopBackedgeWeight = OrigLoopExitWeight;
}
assert(OrigLoopExitWeight >= ExitWeight0 && "Bad branch weight");
ExitWeight1 = OrigLoopExitWeight - ExitWeight0;
EnterWeight = ExitWeight1;
assert(OrigLoopBackedgeWeight >= EnterWeight && "Bad branch weight");
LoopBackWeight = OrigLoopBackedgeWeight - EnterWeight;
} else if (OrigLoopExitWeight == 0) {
if (OrigLoopBackedgeWeight == 0) {
// degenerate case... keep everything zero...
ExitWeight0 = 0;
ExitWeight1 = 0;
EnterWeight = 0;
LoopBackWeight = 0;
} else {
// Special case "LoopExitWeight == 0" weights which behaves like an
// endless where we don't want loop-enttry (y0) to be the same as
// loop-exit (x1).
ExitWeight0 = 0;
ExitWeight1 = 0;
EnterWeight = 1;
LoopBackWeight = OrigLoopBackedgeWeight;
}
} else {
// loop is never entered.
assert(OrigLoopBackedgeWeight == 0 && "remaining case is backedge zero");
ExitWeight0 = 1;
ExitWeight1 = 1;
EnterWeight = 0;
LoopBackWeight = 0;
}
const uint32_t LoopBIWeights[] = {
SuccsSwapped ? LoopBackWeight : ExitWeight1,
SuccsSwapped ? ExitWeight1 : LoopBackWeight,
};
setBranchWeights(LoopBI, LoopBIWeights);
if (HasConditionalPreHeader) {
const uint32_t PreHeaderBIWeights[] = {
SuccsSwapped ? EnterWeight : ExitWeight0,
SuccsSwapped ? ExitWeight0 : EnterWeight,
};
setBranchWeights(PreHeaderBI, PreHeaderBIWeights);
}
}
/// Rotate loop LP. Return true if the loop is rotated.
///
/// \param SimplifiedLatch is true if the latch was just folded into the final
/// loop exit. In this case we may want to rotate even though the new latch is
/// now an exiting branch. This rotation would have happened had the latch not
/// been simplified. However, if SimplifiedLatch is false, then we avoid
/// rotating loops in which the latch exits to avoid excessive or endless
/// rotation. LoopRotate should be repeatable and converge to a canonical
/// form. This property is satisfied because simplifying the loop latch can only
/// happen once across multiple invocations of the LoopRotate pass.
///
/// If -loop-rotate-multi is enabled we can do multiple rotations in one go
/// so to reach a suitable (non-deoptimizing) exit.
bool LoopRotate::rotateLoop(Loop *L, bool SimplifiedLatch) {
// If the loop has only one block then there is not much to rotate.
if (L->getBlocks().size() == 1)
return false;
bool Rotated = false;
do {
BasicBlock *OrigHeader = L->getHeader();
BasicBlock *OrigLatch = L->getLoopLatch();
BranchInst *BI = dyn_cast<BranchInst>(OrigHeader->getTerminator());
if (!BI || BI->isUnconditional())
return Rotated;
// If the loop header is not one of the loop exiting blocks then
// either this loop is already rotated or it is not
// suitable for loop rotation transformations.
if (!L->isLoopExiting(OrigHeader))
return Rotated;
// If the loop latch already contains a branch that leaves the loop then the
// loop is already rotated.
if (!OrigLatch)
return Rotated;
// Rotate if either the loop latch does *not* exit the loop, or if the loop
// latch was just simplified. Or if we think it will be profitable.
if (L->isLoopExiting(OrigLatch) && !SimplifiedLatch && IsUtilMode == false &&
!profitableToRotateLoopExitingLatch(L) &&
!canRotateDeoptimizingLatchExit(L))
return Rotated;
// Check size of original header and reject loop if it is very big or we can't
// duplicate blocks inside it.
{
SmallPtrSet<const Value *, 32> EphValues;
CodeMetrics::collectEphemeralValues(L, AC, EphValues);
CodeMetrics Metrics;
Metrics.analyzeBasicBlock(OrigHeader, *TTI, EphValues, PrepareForLTO);
if (Metrics.notDuplicatable) {
LLVM_DEBUG(
dbgs() << "LoopRotation: NOT rotating - contains non-duplicatable"
<< " instructions: ";
L->dump());
return Rotated;
}
if (Metrics.convergent) {
LLVM_DEBUG(dbgs() << "LoopRotation: NOT rotating - contains convergent "
"instructions: ";
L->dump());
return Rotated;
}
if (!Metrics.NumInsts.isValid()) {
LLVM_DEBUG(dbgs() << "LoopRotation: NOT rotating - contains instructions"
" with invalid cost: ";
L->dump());
return Rotated;
}
if (Metrics.NumInsts > MaxHeaderSize) {
LLVM_DEBUG(dbgs() << "LoopRotation: NOT rotating - contains "
<< Metrics.NumInsts
<< " instructions, which is more than the threshold ("
<< MaxHeaderSize << " instructions): ";
L->dump());
++NumNotRotatedDueToHeaderSize;
return Rotated;
}
// When preparing for LTO, avoid rotating loops with calls that could be
// inlined during the LTO stage.
if (PrepareForLTO && Metrics.NumInlineCandidates > 0)
return Rotated;
}
// Now, this loop is suitable for rotation.
BasicBlock *OrigPreheader = L->getLoopPreheader();
// If the loop could not be converted to canonical form, it must have an
// indirectbr in it, just give up.
if (!OrigPreheader || !L->hasDedicatedExits())
return Rotated;
// Anything ScalarEvolution may know about this loop or the PHI nodes
// in its header will soon be invalidated. We should also invalidate
// all outer loops because insertion and deletion of blocks that happens
// during the rotation may violate invariants related to backedge taken
// infos in them.
if (SE) {
SE->forgetTopmostLoop(L);
// We may hoist some instructions out of loop. In case if they were cached
// as "loop variant" or "loop computable", these caches must be dropped.
// We also may fold basic blocks, so cached block dispositions also need
// to be dropped.
SE->forgetBlockAndLoopDispositions();
}
LLVM_DEBUG(dbgs() << "LoopRotation: rotating "; L->dump());
if (MSSAU && VerifyMemorySSA)
MSSAU->getMemorySSA()->verifyMemorySSA();
// Find new Loop header. NewHeader is a Header's one and only successor
// that is inside loop. Header's other successor is outside the
// loop. Otherwise loop is not suitable for rotation.
BasicBlock *Exit = BI->getSuccessor(0);
BasicBlock *NewHeader = BI->getSuccessor(1);
bool BISuccsSwapped = L->contains(Exit);
if (BISuccsSwapped)
std::swap(Exit, NewHeader);
assert(NewHeader && "Unable to determine new loop header");
assert(L->contains(NewHeader) && !L->contains(Exit) &&
"Unable to determine loop header and exit blocks");
// This code assumes that the new header has exactly one predecessor.
// Remove any single-entry PHI nodes in it.
assert(NewHeader->getSinglePredecessor() &&
"New header doesn't have one pred!");
FoldSingleEntryPHINodes(NewHeader);
// Begin by walking OrigHeader and populating ValueMap with an entry for
// each Instruction.
BasicBlock::iterator I = OrigHeader->begin(), E = OrigHeader->end();
ValueToValueMapTy ValueMap, ValueMapMSSA;
// For PHI nodes, the value available in OldPreHeader is just the
// incoming value from OldPreHeader.
for (; PHINode *PN = dyn_cast<PHINode>(I); ++I)
InsertNewValueIntoMap(ValueMap, PN,
PN->getIncomingValueForBlock(OrigPreheader));
// For the rest of the instructions, either hoist to the OrigPreheader if
// possible or create a clone in the OldPreHeader if not.
Instruction *LoopEntryBranch = OrigPreheader->getTerminator();
// Record all debug intrinsics preceding LoopEntryBranch to avoid
// duplication.
using DbgIntrinsicHash =
std::pair<std::pair<hash_code, DILocalVariable *>, DIExpression *>;
auto makeHash = [](auto *D) -> DbgIntrinsicHash {
auto VarLocOps = D->location_ops();
return {{hash_combine_range(VarLocOps.begin(), VarLocOps.end()),
D->getVariable()},
D->getExpression()};
};
SmallDenseSet<DbgIntrinsicHash, 8> DbgIntrinsics;
for (Instruction &I : llvm::drop_begin(llvm::reverse(*OrigPreheader))) {
if (auto *DII = dyn_cast<DbgVariableIntrinsic>(&I)) {
DbgIntrinsics.insert(makeHash(DII));
// Until RemoveDIs supports dbg.declares in DbgVariableRecord format,
// we'll need to collect DbgVariableRecords attached to any other debug
// intrinsics.
for (const DbgVariableRecord &DVR :
filterDbgVars(DII->getDbgRecordRange()))
DbgIntrinsics.insert(makeHash(&DVR));
} else {
break;
}
}
// Build DbgVariableRecord hashes for DbgVariableRecords attached to the
// terminator, which isn't considered in the loop above.
for (const DbgVariableRecord &DVR :
filterDbgVars(OrigPreheader->getTerminator()->getDbgRecordRange()))
DbgIntrinsics.insert(makeHash(&DVR));
// Remember the local noalias scope declarations in the header. After the
// rotation, they must be duplicated and the scope must be cloned. This
// avoids unwanted interaction across iterations.
SmallVector<NoAliasScopeDeclInst *, 6> NoAliasDeclInstructions;
for (Instruction &I : *OrigHeader)
if (auto *Decl = dyn_cast<NoAliasScopeDeclInst>(&I))
NoAliasDeclInstructions.push_back(Decl);
Module *M = OrigHeader->getModule();
// Track the next DbgRecord to clone. If we have a sequence where an
// instruction is hoisted instead of being cloned:
// DbgRecord blah
// %foo = add i32 0, 0
// DbgRecord xyzzy
// %bar = call i32 @foobar()
// where %foo is hoisted, then the DbgRecord "blah" will be seen twice, once
// attached to %foo, then when %foo his hoisted it will "fall down" onto the
// function call:
// DbgRecord blah
// DbgRecord xyzzy
// %bar = call i32 @foobar()
// causing it to appear attached to the call too.
//
// To avoid this, cloneDebugInfoFrom takes an optional "start cloning from
// here" position to account for this behaviour. We point it at any
// DbgRecords on the next instruction, here labelled xyzzy, before we hoist
// %foo. Later, we only only clone DbgRecords from that position (xyzzy)
// onwards, which avoids cloning DbgRecord "blah" multiple times. (Stored as
// a range because it gives us a natural way of testing whether
// there were DbgRecords on the next instruction before we hoisted things).
iterator_range<DbgRecord::self_iterator> NextDbgInsts =
(I != E) ? I->getDbgRecordRange() : DbgMarker::getEmptyDbgRecordRange();
while (I != E) {
Instruction *Inst = &*I++;
// If the instruction's operands are invariant and it doesn't read or write
// memory, then it is safe to hoist. Doing this doesn't change the order of
// execution in the preheader, but does prevent the instruction from
// executing in each iteration of the loop. This means it is safe to hoist
// something that might trap, but isn't safe to hoist something that reads
// memory (without proving that the loop doesn't write).
if (L->hasLoopInvariantOperands(Inst) && !Inst->mayReadFromMemory() &&
!Inst->mayWriteToMemory() && !Inst->isTerminator() &&
!isa<DbgInfoIntrinsic>(Inst) && !isa<AllocaInst>(Inst) &&
// It is not safe to hoist the value of these instructions in
// coroutines, as the addresses of otherwise eligible variables (e.g.
// thread-local variables and errno) may change if the coroutine is
// resumed in a different thread.Therefore, we disable this
// optimization for correctness. However, this may block other correct
// optimizations.
// FIXME: This should be reverted once we have a better model for
// memory access in coroutines.
!Inst->getFunction()->isPresplitCoroutine()) {
if (LoopEntryBranch->getParent()->IsNewDbgInfoFormat &&
!NextDbgInsts.empty()) {
auto DbgValueRange =
LoopEntryBranch->cloneDebugInfoFrom(Inst, NextDbgInsts.begin());
RemapDbgRecordRange(M, DbgValueRange, ValueMap,
RF_NoModuleLevelChanges | RF_IgnoreMissingLocals);
// Erase anything we've seen before.
for (DbgVariableRecord &DVR :
make_early_inc_range(filterDbgVars(DbgValueRange)))
if (DbgIntrinsics.count(makeHash(&DVR)))
DVR.eraseFromParent();
}
NextDbgInsts = I->getDbgRecordRange();
Inst->moveBefore(LoopEntryBranch);
++NumInstrsHoisted;
continue;
}
// Otherwise, create a duplicate of the instruction.
Instruction *C = Inst->clone();
C->insertBefore(LoopEntryBranch);
++NumInstrsDuplicated;
if (LoopEntryBranch->getParent()->IsNewDbgInfoFormat &&
!NextDbgInsts.empty()) {
auto Range = C->cloneDebugInfoFrom(Inst, NextDbgInsts.begin());
RemapDbgRecordRange(M, Range, ValueMap,
RF_NoModuleLevelChanges | RF_IgnoreMissingLocals);
NextDbgInsts = DbgMarker::getEmptyDbgRecordRange();
// Erase anything we've seen before.
for (DbgVariableRecord &DVR :
make_early_inc_range(filterDbgVars(Range)))
if (DbgIntrinsics.count(makeHash(&DVR)))
DVR.eraseFromParent();
}
// Eagerly remap the operands of the instruction.
RemapInstruction(C, ValueMap,
RF_NoModuleLevelChanges | RF_IgnoreMissingLocals);
// Avoid inserting the same intrinsic twice.
if (auto *DII = dyn_cast<DbgVariableIntrinsic>(C))
if (DbgIntrinsics.count(makeHash(DII))) {
C->eraseFromParent();
continue;
}
// With the operands remapped, see if the instruction constant folds or is
// otherwise simplifyable. This commonly occurs because the entry from PHI
// nodes allows icmps and other instructions to fold.
Value *V = simplifyInstruction(C, SQ);
if (V && LI->replacementPreservesLCSSAForm(C, V)) {
// If so, then delete the temporary instruction and stick the folded value
// in the map.
InsertNewValueIntoMap(ValueMap, Inst, V);
if (!C->mayHaveSideEffects()) {
C->eraseFromParent();
C = nullptr;
}
} else {
InsertNewValueIntoMap(ValueMap, Inst, C);
}
if (C) {
// Otherwise, stick the new instruction into the new block!
C->setName(Inst->getName());
if (auto *II = dyn_cast<AssumeInst>(C))
AC->registerAssumption(II);
// MemorySSA cares whether the cloned instruction was inserted or not, and
// not whether it can be remapped to a simplified value.
if (MSSAU)
InsertNewValueIntoMap(ValueMapMSSA, Inst, C);
}
}
if (!NoAliasDeclInstructions.empty()) {
// There are noalias scope declarations:
// (general):
// Original: OrigPre { OrigHeader NewHeader ... Latch }
// after: (OrigPre+OrigHeader') { NewHeader ... Latch OrigHeader }
//
// with D: llvm.experimental.noalias.scope.decl,
// U: !noalias or !alias.scope depending on D
// ... { D U1 U2 } can transform into:
// (0) : ... { D U1 U2 } // no relevant rotation for this part
// (1) : ... D' { U1 U2 D } // D is part of OrigHeader
// (2) : ... D' U1' { U2 D U1 } // D, U1 are part of OrigHeader
//
// We now want to transform:
// (1) -> : ... D' { D U1 U2 D'' }
// (2) -> : ... D' U1' { D U2 D'' U1'' }
// D: original llvm.experimental.noalias.scope.decl
// D', U1': duplicate with replaced scopes
// D'', U1'': different duplicate with replaced scopes
// This ensures a safe fallback to 'may_alias' introduced by the rotate,
// as U1'' and U1' scopes will not be compatible wrt to the local restrict
// Clone the llvm.experimental.noalias.decl again for the NewHeader.
BasicBlock::iterator NewHeaderInsertionPoint =
NewHeader->getFirstNonPHIIt();
for (NoAliasScopeDeclInst *NAD : NoAliasDeclInstructions) {
LLVM_DEBUG(dbgs() << " Cloning llvm.experimental.noalias.scope.decl:"
<< *NAD << "\n");
Instruction *NewNAD = NAD->clone();
NewNAD->insertBefore(*NewHeader, NewHeaderInsertionPoint);
}
// Scopes must now be duplicated, once for OrigHeader and once for
// OrigPreHeader'.
{
auto &Context = NewHeader->getContext();
SmallVector<MDNode *, 8> NoAliasDeclScopes;
for (NoAliasScopeDeclInst *NAD : NoAliasDeclInstructions)
NoAliasDeclScopes.push_back(NAD->getScopeList());
LLVM_DEBUG(dbgs() << " Updating OrigHeader scopes\n");
cloneAndAdaptNoAliasScopes(NoAliasDeclScopes, {OrigHeader}, Context,
"h.rot");
LLVM_DEBUG(OrigHeader->dump());
// Keep the compile time impact low by only adapting the inserted block
// of instructions in the OrigPreHeader. This might result in slightly
// more aliasing between these instructions and those that were already
// present, but it will be much faster when the original PreHeader is
// large.
LLVM_DEBUG(dbgs() << " Updating part of OrigPreheader scopes\n");
auto *FirstDecl =
cast<Instruction>(ValueMap[*NoAliasDeclInstructions.begin()]);
auto *LastInst = &OrigPreheader->back();
cloneAndAdaptNoAliasScopes(NoAliasDeclScopes, FirstDecl, LastInst,
Context, "pre.rot");
LLVM_DEBUG(OrigPreheader->dump());
LLVM_DEBUG(dbgs() << " Updated NewHeader:\n");
LLVM_DEBUG(NewHeader->dump());
}
}
// Along with all the other instructions, we just cloned OrigHeader's
// terminator into OrigPreHeader. Fix up the PHI nodes in each of OrigHeader's
// successors by duplicating their incoming values for OrigHeader.
for (BasicBlock *SuccBB : successors(OrigHeader))
for (BasicBlock::iterator BI = SuccBB->begin();
PHINode *PN = dyn_cast<PHINode>(BI); ++BI)
PN->addIncoming(PN->getIncomingValueForBlock(OrigHeader), OrigPreheader);
// Now that OrigPreHeader has a clone of OrigHeader's terminator, remove
// OrigPreHeader's old terminator (the original branch into the loop), and
// remove the corresponding incoming values from the PHI nodes in OrigHeader.
LoopEntryBranch->eraseFromParent();
OrigPreheader->flushTerminatorDbgRecords();
// Update MemorySSA before the rewrite call below changes the 1:1
// instruction:cloned_instruction_or_value mapping.
if (MSSAU) {
InsertNewValueIntoMap(ValueMapMSSA, OrigHeader, OrigPreheader);
MSSAU->updateForClonedBlockIntoPred(OrigHeader, OrigPreheader,
ValueMapMSSA);
}
SmallVector<PHINode*, 2> InsertedPHIs;
// If there were any uses of instructions in the duplicated block outside the
// loop, update them, inserting PHI nodes as required
RewriteUsesOfClonedInstructions(OrigHeader, OrigPreheader, ValueMap, SE,
&InsertedPHIs);
// Attach dbg.value intrinsics to the new phis if that phi uses a value that
// previously had debug metadata attached. This keeps the debug info
// up-to-date in the loop body.
if (!InsertedPHIs.empty())
insertDebugValuesForPHIs(OrigHeader, InsertedPHIs);
// NewHeader is now the header of the loop.
L->moveToHeader(NewHeader);
assert(L->getHeader() == NewHeader && "Latch block is our new header");
// Inform DT about changes to the CFG.
if (DT) {
// The OrigPreheader branches to the NewHeader and Exit now. Then, inform
// the DT about the removed edge to the OrigHeader (that got removed).
SmallVector<DominatorTree::UpdateType, 3> Updates;
Updates.push_back({DominatorTree::Insert, OrigPreheader, Exit});
Updates.push_back({DominatorTree::Insert, OrigPreheader, NewHeader});
Updates.push_back({DominatorTree::Delete, OrigPreheader, OrigHeader});
if (MSSAU) {
MSSAU->applyUpdates(Updates, *DT, /*UpdateDT=*/true);
if (VerifyMemorySSA)
MSSAU->getMemorySSA()->verifyMemorySSA();
} else {
DT->applyUpdates(Updates);
}
}
// At this point, we've finished our major CFG changes. As part of cloning
// the loop into the preheader we've simplified instructions and the
// duplicated conditional branch may now be branching on a constant. If it is
// branching on a constant and if that constant means that we enter the loop,
// then we fold away the cond branch to an uncond branch. This simplifies the
// loop in cases important for nested loops, and it also means we don't have
// to split as many edges.
BranchInst *PHBI = cast<BranchInst>(OrigPreheader->getTerminator());
assert(PHBI->isConditional() && "Should be clone of BI condbr!");
const Value *Cond = PHBI->getCondition();
const bool HasConditionalPreHeader =
!isa<ConstantInt>(Cond) ||
PHBI->getSuccessor(cast<ConstantInt>(Cond)->isZero()) != NewHeader;
updateBranchWeights(*PHBI, *BI, HasConditionalPreHeader, BISuccsSwapped);
if (HasConditionalPreHeader) {
// The conditional branch can't be folded, handle the general case.
// Split edges as necessary to preserve LoopSimplify form.
// Right now OrigPreHeader has two successors, NewHeader and ExitBlock, and
// thus is not a preheader anymore.
// Split the edge to form a real preheader.
BasicBlock *NewPH = SplitCriticalEdge(
OrigPreheader, NewHeader,
CriticalEdgeSplittingOptions(DT, LI, MSSAU).setPreserveLCSSA());
NewPH->setName(NewHeader->getName() + ".lr.ph");
// Preserve canonical loop form, which means that 'Exit' should have only
// one predecessor. Note that Exit could be an exit block for multiple
// nested loops, causing both of the edges to now be critical and need to
// be split.
SmallVector<BasicBlock *, 4> ExitPreds(predecessors(Exit));
bool SplitLatchEdge = false;
for (BasicBlock *ExitPred : ExitPreds) {
// We only need to split loop exit edges.
Loop *PredLoop = LI->getLoopFor(ExitPred);
if (!PredLoop || PredLoop->contains(Exit) ||
isa<IndirectBrInst>(ExitPred->getTerminator()))
continue;
SplitLatchEdge |= L->getLoopLatch() == ExitPred;
BasicBlock *ExitSplit = SplitCriticalEdge(
ExitPred, Exit,
CriticalEdgeSplittingOptions(DT, LI, MSSAU).setPreserveLCSSA());
ExitSplit->moveBefore(Exit);
}
assert(SplitLatchEdge &&
"Despite splitting all preds, failed to split latch exit?");
(void)SplitLatchEdge;
} else {
// We can fold the conditional branch in the preheader, this makes things
// simpler. The first step is to remove the extra edge to the Exit block.
Exit->removePredecessor(OrigPreheader, true /*preserve LCSSA*/);
BranchInst *NewBI = BranchInst::Create(NewHeader, PHBI->getIterator());
NewBI->setDebugLoc(PHBI->getDebugLoc());
PHBI->eraseFromParent();
// With our CFG finalized, update DomTree if it is available.
if (DT) DT->deleteEdge(OrigPreheader, Exit);
// Update MSSA too, if available.
if (MSSAU)
MSSAU->removeEdge(OrigPreheader, Exit);
}
assert(L->getLoopPreheader() && "Invalid loop preheader after loop rotation");
assert(L->getLoopLatch() && "Invalid loop latch after loop rotation");
if (MSSAU && VerifyMemorySSA)
MSSAU->getMemorySSA()->verifyMemorySSA();
// Now that the CFG and DomTree are in a consistent state again, try to merge
// the OrigHeader block into OrigLatch. This will succeed if they are
// connected by an unconditional branch. This is just a cleanup so the
// emitted code isn't too gross in this common case.
DomTreeUpdater DTU(DT, DomTreeUpdater::UpdateStrategy::Eager);
BasicBlock *PredBB = OrigHeader->getUniquePredecessor();
bool DidMerge = MergeBlockIntoPredecessor(OrigHeader, &DTU, LI, MSSAU);
if (DidMerge)
RemoveRedundantDbgInstrs(PredBB);
if (MSSAU && VerifyMemorySSA)
MSSAU->getMemorySSA()->verifyMemorySSA();
LLVM_DEBUG(dbgs() << "LoopRotation: into "; L->dump());
++NumRotated;
Rotated = true;
SimplifiedLatch = false;
// Check that new latch is a deoptimizing exit and then repeat rotation if possible.
// Deoptimizing latch exit is not a generally typical case, so we just loop over.
// TODO: if it becomes a performance bottleneck extend rotation algorithm
// to handle multiple rotations in one go.
} while (MultiRotate && canRotateDeoptimizingLatchExit(L));
return true;
}
/// Determine whether the instructions in this range may be safely and cheaply
/// speculated. This is not an important enough situation to develop complex
/// heuristics. We handle a single arithmetic instruction along with any type
/// conversions.
static bool shouldSpeculateInstrs(BasicBlock::iterator Begin,
BasicBlock::iterator End, Loop *L) {
bool seenIncrement = false;
bool MultiExitLoop = false;
if (!L->getExitingBlock())
MultiExitLoop = true;
for (BasicBlock::iterator I = Begin; I != End; ++I) {
if (!isSafeToSpeculativelyExecute(&*I))
return false;
if (isa<DbgInfoIntrinsic>(I))
continue;
switch (I->getOpcode()) {
default:
return false;
case Instruction::GetElementPtr:
// GEPs are cheap if all indices are constant.
if (!cast<GEPOperator>(I)->hasAllConstantIndices())
return false;
// fall-thru to increment case
[[fallthrough]];
case Instruction::Add:
case Instruction::Sub:
case Instruction::And:
case Instruction::Or:
case Instruction::Xor:
case Instruction::Shl:
case Instruction::LShr:
case Instruction::AShr: {
Value *IVOpnd =
!isa<Constant>(I->getOperand(0))
? I->getOperand(0)
: !isa<Constant>(I->getOperand(1)) ? I->getOperand(1) : nullptr;
if (!IVOpnd)
return false;
// If increment operand is used outside of the loop, this speculation
// could cause extra live range interference.
if (MultiExitLoop) {
for (User *UseI : IVOpnd->users()) {
auto *UserInst = cast<Instruction>(UseI);
if (!L->contains(UserInst))
return false;
}
}
if (seenIncrement)
return false;
seenIncrement = true;
break;
}
case Instruction::Trunc:
case Instruction::ZExt:
case Instruction::SExt:
// ignore type conversions
break;
}
}
return true;
}
/// Fold the loop tail into the loop exit by speculating the loop tail
/// instructions. Typically, this is a single post-increment. In the case of a
/// simple 2-block loop, hoisting the increment can be much better than
/// duplicating the entire loop header. In the case of loops with early exits,
/// rotation will not work anyway, but simplifyLoopLatch will put the loop in
/// canonical form so downstream passes can handle it.
///
/// I don't believe this invalidates SCEV.
bool LoopRotate::simplifyLoopLatch(Loop *L) {
BasicBlock *Latch = L->getLoopLatch();
if (!Latch || Latch->hasAddressTaken())
return false;
BranchInst *Jmp = dyn_cast<BranchInst>(Latch->getTerminator());
if (!Jmp || !Jmp->isUnconditional())
return false;
BasicBlock *LastExit = Latch->getSinglePredecessor();
if (!LastExit || !L->isLoopExiting(LastExit))
return false;
BranchInst *BI = dyn_cast<BranchInst>(LastExit->getTerminator());
if (!BI)
return false;
if (!shouldSpeculateInstrs(Latch->begin(), Jmp->getIterator(), L))
return false;
LLVM_DEBUG(dbgs() << "Folding loop latch " << Latch->getName() << " into "
<< LastExit->getName() << "\n");
DomTreeUpdater DTU(DT, DomTreeUpdater::UpdateStrategy::Eager);
MergeBlockIntoPredecessor(Latch, &DTU, LI, MSSAU, nullptr,
/*PredecessorWithTwoSuccessors=*/true);
if (SE) {
// Merging blocks may remove blocks reference in the block disposition cache. Clear the cache.
SE->forgetBlockAndLoopDispositions();
}
if (MSSAU && VerifyMemorySSA)
MSSAU->getMemorySSA()->verifyMemorySSA();
return true;
}
/// Rotate \c L, and return true if any modification was made.
bool LoopRotate::processLoop(Loop *L) {
// Save the loop metadata.
MDNode *LoopMD = L->getLoopID();
bool SimplifiedLatch = false;
// Simplify the loop latch before attempting to rotate the header
// upward. Rotation may not be needed if the loop tail can be folded into the
// loop exit.
if (!RotationOnly)
SimplifiedLatch = simplifyLoopLatch(L);
bool MadeChange = rotateLoop(L, SimplifiedLatch);
assert((!MadeChange || L->isLoopExiting(L->getLoopLatch())) &&
"Loop latch should be exiting after loop-rotate.");
// Restore the loop metadata.
// NB! We presume LoopRotation DOESN'T ADD its own metadata.
if ((MadeChange || SimplifiedLatch) && LoopMD)
L->setLoopID(LoopMD);
return MadeChange || SimplifiedLatch;
}
/// The utility to convert a loop into a loop with bottom test.
bool llvm::LoopRotation(Loop *L, LoopInfo *LI, const TargetTransformInfo *TTI,
AssumptionCache *AC, DominatorTree *DT,
ScalarEvolution *SE, MemorySSAUpdater *MSSAU,
const SimplifyQuery &SQ, bool RotationOnly = true,
unsigned Threshold = unsigned(-1),
bool IsUtilMode = true, bool PrepareForLTO) {
LoopRotate LR(Threshold, LI, TTI, AC, DT, SE, MSSAU, SQ, RotationOnly,
IsUtilMode, PrepareForLTO);
return LR.processLoop(L);
}