/// Tests whether a function is "malloc-like".
///
/// A function is "malloc-like" if it returns either null or a pointer that
/// doesn't alias any other pointer visible to the caller.
static bool isFunctionMallocLike(Function *F, const SCCNodeSet &SCCNodes) {
SmallSetVector<Value *, 8> FlowsToReturn;
for (Function::iterator I = F->begin(), E = F->end(); I != E; ++I)
if (ReturnInst *Ret = dyn_cast<ReturnInst>(I->getTerminator()))
FlowsToReturn.insert(Ret->getReturnValue());
for (unsigned i = 0; i != FlowsToReturn.size(); ++i) {
Value *RetVal = FlowsToReturn[i];
if (Constant *C = dyn_cast<Constant>(RetVal)) {
if (!C->isNullValue() && !isa<UndefValue>(C))
return false;
continue;
}
if (isa<Argument>(RetVal))
return false;
if (Instruction *RVI = dyn_cast<Instruction>(RetVal))
switch (RVI->getOpcode()) {
// Extend the analysis by looking upwards.
case Instruction::BitCast:
case Instruction::GetElementPtr:
case Instruction::AddrSpaceCast:
FlowsToReturn.insert(RVI->getOperand(0));
continue;
case Instruction::Select: {
SelectInst *SI = cast<SelectInst>(RVI);
FlowsToReturn.insert(SI->getTrueValue());
FlowsToReturn.insert(SI->getFalseValue());
continue;
}
case Instruction::PHI: {
PHINode *PN = cast<PHINode>(RVI);
for (Value *IncValue : PN->incoming_values())
FlowsToReturn.insert(IncValue);
continue;
}
// Check whether the pointer came from an allocation.
case Instruction::Alloca:
break;
case Instruction::Call:
case Instruction::Invoke: {
CallSite CS(RVI);
if (CS.paramHasAttr(0, Attribute::NoAlias))
break;
if (CS.getCalledFunction() && SCCNodes.count(CS.getCalledFunction()))
break;
} // fall-through
default:
return false; // Did not come from an allocation.
}
if (PointerMayBeCaptured(RetVal, false, /*StoreCaptures=*/false))
return false;
}
return true;
}
// Analyze interleaved accesses and collect them into interleaved load and
// store groups.
//
// When generating code for an interleaved load group, we effectively hoist all
// loads in the group to the location of the first load in program order. When
// generating code for an interleaved store group, we sink all stores to the
// location of the last store. This code motion can change the order of load
// and store instructions and may break dependences.
//
// The code generation strategy mentioned above ensures that we won't violate
// any write-after-read (WAR) dependences.
//
// E.g., for the WAR dependence: a = A[i]; // (1)
// A[i] = b; // (2)
//
// The store group of (2) is always inserted at or below (2), and the load
// group of (1) is always inserted at or above (1). Thus, the instructions will
// never be reordered. All other dependences are checked to ensure the
// correctness of the instruction reordering.
//
// The algorithm visits all memory accesses in the loop in bottom-up program
// order. Program order is established by traversing the blocks in the loop in
// reverse postorder when collecting the accesses.
//
// We visit the memory accesses in bottom-up order because it can simplify the
// construction of store groups in the presence of write-after-write (WAW)
// dependences.
//
// E.g., for the WAW dependence: A[i] = a; // (1)
// A[i] = b; // (2)
// A[i + 1] = c; // (3)
//
// We will first create a store group with (3) and (2). (1) can't be added to
// this group because it and (2) are dependent. However, (1) can be grouped
// with other accesses that may precede it in program order. Note that a
// bottom-up order does not imply that WAW dependences should not be checked.
void InterleavedAccessInfo::analyzeInterleaving(
bool EnablePredicatedInterleavedMemAccesses) {
LLVM_DEBUG(dbgs() << "LV: Analyzing interleaved accesses...\n");
const ValueToValueMap &Strides = LAI->getSymbolicStrides();
// Holds all accesses with a constant stride.
MapVector<Instruction *, StrideDescriptor> AccessStrideInfo;
collectConstStrideAccesses(AccessStrideInfo, Strides);
if (AccessStrideInfo.empty())
return;
// Collect the dependences in the loop.
collectDependences();
// Holds all interleaved store groups temporarily.
SmallSetVector<InterleaveGroup *, 4> StoreGroups;
// Holds all interleaved load groups temporarily.
SmallSetVector<InterleaveGroup *, 4> LoadGroups;
// Search in bottom-up program order for pairs of accesses (A and B) that can
// form interleaved load or store groups. In the algorithm below, access A
// precedes access B in program order. We initialize a group for B in the
// outer loop of the algorithm, and then in the inner loop, we attempt to
// insert each A into B's group if:
//
// 1. A and B have the same stride,
// 2. A and B have the same memory object size, and
// 3. A belongs in B's group according to its distance from B.
//
// Special care is taken to ensure group formation will not break any
// dependences.
for (auto BI = AccessStrideInfo.rbegin(), E = AccessStrideInfo.rend();
BI != E; ++BI) {
Instruction *B = BI->first;
StrideDescriptor DesB = BI->second;
// Initialize a group for B if it has an allowable stride. Even if we don't
// create a group for B, we continue with the bottom-up algorithm to ensure
// we don't break any of B's dependences.
InterleaveGroup *Group = nullptr;
if (isStrided(DesB.Stride) &&
(!isPredicated(B->getParent()) || EnablePredicatedInterleavedMemAccesses)) {
Group = getInterleaveGroup(B);
if (!Group) {
LLVM_DEBUG(dbgs() << "LV: Creating an interleave group with:" << *B
<< '\n');
Group = createInterleaveGroup(B, DesB.Stride, DesB.Align);
}
if (B->mayWriteToMemory())
StoreGroups.insert(Group);
else
LoadGroups.insert(Group);
}
for (auto AI = std::next(BI); AI != E; ++AI) {
Instruction *A = AI->first;
StrideDescriptor DesA = AI->second;
// Our code motion strategy implies that we can't have dependences
// between accesses in an interleaved group and other accesses located
// between the first and last member of the group. Note that this also
// means that a group can't have more than one member at a given offset.
// The accesses in a group can have dependences with other accesses, but
// we must ensure we don't extend the boundaries of the group such that
// we encompass those dependent accesses.
//
// For example, assume we have the sequence of accesses shown below in a
// stride-2 loop:
//
// (1, 2) is a group | A[i] = a; // (1)
// | A[i-1] = b; // (2) |
// A[i-3] = c; // (3)
// A[i] = d; // (4) | (2, 4) is not a group
//
// Because accesses (2) and (3) are dependent, we can group (2) with (1)
// but not with (4). If we did, the dependent access (3) would be within
// the boundaries of the (2, 4) group.
if (!canReorderMemAccessesForInterleavedGroups(&*AI, &*BI)) {
// If a dependence exists and A is already in a group, we know that A
// must be a store since A precedes B and WAR dependences are allowed.
// Thus, A would be sunk below B. We release A's group to prevent this
// illegal code motion. A will then be free to form another group with
// instructions that precede it.
if (isInterleaved(A)) {
InterleaveGroup *StoreGroup = getInterleaveGroup(A);
StoreGroups.remove(StoreGroup);
releaseGroup(StoreGroup);
}
// If a dependence exists and A is not already in a group (or it was
// and we just released it), B might be hoisted above A (if B is a
// load) or another store might be sunk below A (if B is a store). In
// either case, we can't add additional instructions to B's group. B
// will only form a group with instructions that it precedes.
break;
}
// At this point, we've checked for illegal code motion. If either A or B
// isn't strided, there's nothing left to do.
if (!isStrided(DesA.Stride) || !isStrided(DesB.Stride))
continue;
// Ignore A if it's already in a group or isn't the same kind of memory
// operation as B.
// Note that mayReadFromMemory() isn't mutually exclusive to
// mayWriteToMemory in the case of atomic loads. We shouldn't see those
// here, canVectorizeMemory() should have returned false - except for the
// case we asked for optimization remarks.
if (isInterleaved(A) ||
(A->mayReadFromMemory() != B->mayReadFromMemory()) ||
(A->mayWriteToMemory() != B->mayWriteToMemory()))
continue;
// Check rules 1 and 2. Ignore A if its stride or size is different from
// that of B.
if (DesA.Stride != DesB.Stride || DesA.Size != DesB.Size)
continue;
// Ignore A if the memory object of A and B don't belong to the same
// address space
if (getLoadStoreAddressSpace(A) != getLoadStoreAddressSpace(B))
continue;
// Calculate the distance from A to B.
const SCEVConstant *DistToB = dyn_cast<SCEVConstant>(
PSE.getSE()->getMinusSCEV(DesA.Scev, DesB.Scev));
if (!DistToB)
continue;
int64_t DistanceToB = DistToB->getAPInt().getSExtValue();
// Check rule 3. Ignore A if its distance to B is not a multiple of the
// size.
if (DistanceToB % static_cast<int64_t>(DesB.Size))
continue;
// All members of a predicated interleave-group must have the same predicate,
// and currently must reside in the same BB.
BasicBlock *BlockA = A->getParent();
BasicBlock *BlockB = B->getParent();
if ((isPredicated(BlockA) || isPredicated(BlockB)) &&
(!EnablePredicatedInterleavedMemAccesses || BlockA != BlockB))
continue;
// The index of A is the index of B plus A's distance to B in multiples
// of the size.
int IndexA =
Group->getIndex(B) + DistanceToB / static_cast<int64_t>(DesB.Size);
// Try to insert A into B's group.
if (Group->insertMember(A, IndexA, DesA.Align)) {
LLVM_DEBUG(dbgs() << "LV: Inserted:" << *A << '\n'
<< " into the interleave group with" << *B
<< '\n');
InterleaveGroupMap[A] = Group;
// Set the first load in program order as the insert position.
if (A->mayReadFromMemory())
Group->setInsertPos(A);
}
} // Iteration over A accesses.
} // Iteration over B accesses.
// Remove interleaved store groups with gaps.
for (InterleaveGroup *Group : StoreGroups)
if (Group->getNumMembers() != Group->getFactor()) {
LLVM_DEBUG(
dbgs() << "LV: Invalidate candidate interleaved store group due "
"to gaps.\n");
releaseGroup(Group);
}
// Remove interleaved groups with gaps (currently only loads) whose memory
// accesses may wrap around. We have to revisit the getPtrStride analysis,
// this time with ShouldCheckWrap=true, since collectConstStrideAccesses does
// not check wrapping (see documentation there).
// FORNOW we use Assume=false;
// TODO: Change to Assume=true but making sure we don't exceed the threshold
// of runtime SCEV assumptions checks (thereby potentially failing to
// vectorize altogether).
// Additional optional optimizations:
// TODO: If we are peeling the loop and we know that the first pointer doesn't
// wrap then we can deduce that all pointers in the group don't wrap.
// This means that we can forcefully peel the loop in order to only have to
// check the first pointer for no-wrap. When we'll change to use Assume=true
// we'll only need at most one runtime check per interleaved group.
for (InterleaveGroup *Group : LoadGroups) {
// Case 1: A full group. Can Skip the checks; For full groups, if the wide
// load would wrap around the address space we would do a memory access at
// nullptr even without the transformation.
if (Group->getNumMembers() == Group->getFactor())
continue;
// Case 2: If first and last members of the group don't wrap this implies
// that all the pointers in the group don't wrap.
// So we check only group member 0 (which is always guaranteed to exist),
// and group member Factor - 1; If the latter doesn't exist we rely on
// peeling (if it is a non-reveresed accsess -- see Case 3).
Value *FirstMemberPtr = getLoadStorePointerOperand(Group->getMember(0));
if (!getPtrStride(PSE, FirstMemberPtr, TheLoop, Strides, /*Assume=*/false,
/*ShouldCheckWrap=*/true)) {
LLVM_DEBUG(
dbgs() << "LV: Invalidate candidate interleaved group due to "
"first group member potentially pointer-wrapping.\n");
releaseGroup(Group);
continue;
}
Instruction *LastMember = Group->getMember(Group->getFactor() - 1);
if (LastMember) {
Value *LastMemberPtr = getLoadStorePointerOperand(LastMember);
if (!getPtrStride(PSE, LastMemberPtr, TheLoop, Strides, /*Assume=*/false,
/*ShouldCheckWrap=*/true)) {
LLVM_DEBUG(
dbgs() << "LV: Invalidate candidate interleaved group due to "
"last group member potentially pointer-wrapping.\n");
releaseGroup(Group);
}
} else {
// Case 3: A non-reversed interleaved load group with gaps: We need
// to execute at least one scalar epilogue iteration. This will ensure
// we don't speculatively access memory out-of-bounds. We only need
// to look for a member at index factor - 1, since every group must have
// a member at index zero.
if (Group->isReverse()) {
LLVM_DEBUG(
dbgs() << "LV: Invalidate candidate interleaved group due to "
"a reverse access with gaps.\n");
releaseGroup(Group);
continue;
}
LLVM_DEBUG(
dbgs() << "LV: Interleaved group requires epilogue iteration.\n");
RequiresScalarEpilogue = true;
}
}
}
/// UpdateSuccessorsPHIs - After FromBB is tail duplicated into its predecessor
/// blocks, the successors have gained new predecessors. Update the PHI
/// instructions in them accordingly.
void
TailDuplicatePass::UpdateSuccessorsPHIs(MachineBasicBlock *FromBB, bool isDead,
SmallVector<MachineBasicBlock*, 8> &TDBBs,
SmallSetVector<MachineBasicBlock*,8> &Succs) {
for (SmallSetVector<MachineBasicBlock*, 8>::iterator SI = Succs.begin(),
SE = Succs.end(); SI != SE; ++SI) {
MachineBasicBlock *SuccBB = *SI;
for (MachineBasicBlock::iterator II = SuccBB->begin(), EE = SuccBB->end();
II != EE; ++II) {
if (!II->isPHI())
break;
unsigned Idx = 0;
for (unsigned i = 1, e = II->getNumOperands(); i != e; i += 2) {
MachineOperand &MO = II->getOperand(i+1);
if (MO.getMBB() == FromBB) {
Idx = i;
break;
}
}
assert(Idx != 0);
MachineOperand &MO0 = II->getOperand(Idx);
unsigned Reg = MO0.getReg();
if (isDead) {
// Folded into the previous BB.
// There could be duplicate phi source entries. FIXME: Should sdisel
// or earlier pass fixed this?
for (unsigned i = II->getNumOperands()-2; i != Idx; i -= 2) {
MachineOperand &MO = II->getOperand(i+1);
if (MO.getMBB() == FromBB) {
II->RemoveOperand(i+1);
II->RemoveOperand(i);
}
}
} else
Idx = 0;
// If Idx is set, the operands at Idx and Idx+1 must be removed.
// We reuse the location to avoid expensive RemoveOperand calls.
DenseMap<unsigned,AvailableValsTy>::iterator LI=SSAUpdateVals.find(Reg);
if (LI != SSAUpdateVals.end()) {
// This register is defined in the tail block.
for (unsigned j = 0, ee = LI->second.size(); j != ee; ++j) {
MachineBasicBlock *SrcBB = LI->second[j].first;
unsigned SrcReg = LI->second[j].second;
if (Idx != 0) {
II->getOperand(Idx).setReg(SrcReg);
II->getOperand(Idx+1).setMBB(SrcBB);
Idx = 0;
} else {
II->addOperand(MachineOperand::CreateReg(SrcReg, false));
II->addOperand(MachineOperand::CreateMBB(SrcBB));
}
}
} else {
// Live in tail block, must also be live in predecessors.
for (unsigned j = 0, ee = TDBBs.size(); j != ee; ++j) {
MachineBasicBlock *SrcBB = TDBBs[j];
if (Idx != 0) {
II->getOperand(Idx).setReg(Reg);
II->getOperand(Idx+1).setMBB(SrcBB);
Idx = 0;
} else {
II->addOperand(MachineOperand::CreateReg(Reg, false));
II->addOperand(MachineOperand::CreateMBB(SrcBB));
}
}
}
if (Idx != 0) {
II->RemoveOperand(Idx+1);
II->RemoveOperand(Idx);
}
}
}
}
bool MachineCopyPropagation::CopyPropagateBlock(MachineBasicBlock &MBB) {
SmallSetVector<MachineInstr*, 8> MaybeDeadCopies; // Candidates for deletion
DenseMap<unsigned, MachineInstr*> AvailCopyMap; // Def -> available copies map
DenseMap<unsigned, MachineInstr*> CopyMap; // Def -> copies map
SourceMap SrcMap; // Src -> Def map
bool Changed = false;
for (MachineBasicBlock::iterator I = MBB.begin(), E = MBB.end(); I != E; ) {
MachineInstr *MI = &*I;
++I;
if (MI->isCopy()) {
unsigned Def = MI->getOperand(0).getReg();
unsigned Src = MI->getOperand(1).getReg();
if (TargetRegisterInfo::isVirtualRegister(Def) ||
TargetRegisterInfo::isVirtualRegister(Src))
report_fatal_error("MachineCopyPropagation should be run after"
" register allocation!");
DenseMap<unsigned, MachineInstr*>::iterator CI = AvailCopyMap.find(Src);
if (CI != AvailCopyMap.end()) {
MachineInstr *CopyMI = CI->second;
if (!MRI->isReserved(Def) &&
(!MRI->isReserved(Src) || NoInterveningSideEffect(CopyMI, MI)) &&
isNopCopy(CopyMI, Def, Src, TRI)) {
// The two copies cancel out and the source of the first copy
// hasn't been overridden, eliminate the second one. e.g.
// %ECX<def> = COPY %EAX<kill>
// ... nothing clobbered EAX.
// %EAX<def> = COPY %ECX
// =>
// %ECX<def> = COPY %EAX
//
// Also avoid eliminating a copy from reserved registers unless the
// definition is proven not clobbered. e.g.
// %RSP<def> = COPY %RAX
// CALL
// %RAX<def> = COPY %RSP
// Clear any kills of Def between CopyMI and MI. This extends the
// live range.
for (MachineBasicBlock::iterator I = CopyMI, E = MI; I != E; ++I)
I->clearRegisterKills(Def, TRI);
removeCopy(MI);
Changed = true;
++NumDeletes;
continue;
}
}
// If Src is defined by a previous copy, it cannot be eliminated.
for (MCRegAliasIterator AI(Src, TRI, true); AI.isValid(); ++AI) {
CI = CopyMap.find(*AI);
if (CI != CopyMap.end())
MaybeDeadCopies.remove(CI->second);
}
// Copy is now a candidate for deletion.
MaybeDeadCopies.insert(MI);
// If 'Src' is previously source of another copy, then this earlier copy's
// source is no longer available. e.g.
// %xmm9<def> = copy %xmm2
// ...
// %xmm2<def> = copy %xmm0
// ...
// %xmm2<def> = copy %xmm9
SourceNoLongerAvailable(Def, SrcMap, AvailCopyMap);
// Remember Def is defined by the copy.
// ... Make sure to clear the def maps of aliases first.
for (MCRegAliasIterator AI(Def, TRI, false); AI.isValid(); ++AI) {
CopyMap.erase(*AI);
AvailCopyMap.erase(*AI);
}
CopyMap[Def] = MI;
AvailCopyMap[Def] = MI;
for (MCSubRegIterator SR(Def, TRI); SR.isValid(); ++SR) {
CopyMap[*SR] = MI;
AvailCopyMap[*SR] = MI;
}
// Remember source that's copied to Def. Once it's clobbered, then
// it's no longer available for copy propagation.
if (std::find(SrcMap[Src].begin(), SrcMap[Src].end(), Def) ==
SrcMap[Src].end()) {
SrcMap[Src].push_back(Def);
}
continue;
}
// Not a copy.
SmallVector<unsigned, 2> Defs;
int RegMaskOpNum = -1;
for (unsigned i = 0, e = MI->getNumOperands(); i != e; ++i) {
MachineOperand &MO = MI->getOperand(i);
if (MO.isRegMask())
RegMaskOpNum = i;
if (!MO.isReg())
continue;
unsigned Reg = MO.getReg();
if (!Reg)
continue;
if (TargetRegisterInfo::isVirtualRegister(Reg))
report_fatal_error("MachineCopyPropagation should be run after"
" register allocation!");
if (MO.isDef()) {
Defs.push_back(Reg);
continue;
}
// If 'Reg' is defined by a copy, the copy is no longer a candidate
// for elimination.
for (MCRegAliasIterator AI(Reg, TRI, true); AI.isValid(); ++AI) {
DenseMap<unsigned, MachineInstr*>::iterator CI = CopyMap.find(*AI);
if (CI != CopyMap.end())
MaybeDeadCopies.remove(CI->second);
}
}
// The instruction has a register mask operand which means that it clobbers
// a large set of registers. It is possible to use the register mask to
// prune the available copies, but treat it like a basic block boundary for
// now.
if (RegMaskOpNum >= 0) {
// Erase any MaybeDeadCopies whose destination register is clobbered.
const MachineOperand &MaskMO = MI->getOperand(RegMaskOpNum);
for (SmallSetVector<MachineInstr*, 8>::iterator
DI = MaybeDeadCopies.begin(), DE = MaybeDeadCopies.end();
DI != DE; ++DI) {
unsigned Reg = (*DI)->getOperand(0).getReg();
if (MRI->isReserved(Reg) || !MaskMO.clobbersPhysReg(Reg))
continue;
removeCopy(*DI);
Changed = true;
++NumDeletes;
}
// Clear all data structures as if we were beginning a new basic block.
MaybeDeadCopies.clear();
AvailCopyMap.clear();
CopyMap.clear();
SrcMap.clear();
continue;
}
for (unsigned i = 0, e = Defs.size(); i != e; ++i) {
unsigned Reg = Defs[i];
// No longer defined by a copy.
for (MCRegAliasIterator AI(Reg, TRI, true); AI.isValid(); ++AI) {
CopyMap.erase(*AI);
AvailCopyMap.erase(*AI);
}
// If 'Reg' is previously source of a copy, it is no longer available for
// copy propagation.
SourceNoLongerAvailable(Reg, SrcMap, AvailCopyMap);
}
}
// If MBB doesn't have successors, delete the copies whose defs are not used.
// If MBB does have successors, then conservative assume the defs are live-out
// since we don't want to trust live-in lists.
if (MBB.succ_empty()) {
for (SmallSetVector<MachineInstr*, 8>::iterator
DI = MaybeDeadCopies.begin(), DE = MaybeDeadCopies.end();
DI != DE; ++DI) {
if (!MRI->isReserved((*DI)->getOperand(0).getReg())) {
removeCopy(*DI);
Changed = true;
++NumDeletes;
}
}
}
return Changed;
}
/// For every instruction from the worklist, check to see if it has any uses
/// that are outside the current loop. If so, insert LCSSA PHI nodes and
/// rewrite the uses.
bool llvm::formLCSSAForInstructions(SmallVectorImpl<Instruction *> &Worklist,
DominatorTree &DT, LoopInfo &LI) {
SmallVector<Use *, 16> UsesToRewrite;
SmallSetVector<PHINode *, 16> PHIsToRemove;
PredIteratorCache PredCache;
bool Changed = false;
// Cache the Loop ExitBlocks across this loop. We expect to get a lot of
// instructions within the same loops, computing the exit blocks is
// expensive, and we're not mutating the loop structure.
SmallDenseMap<Loop*, SmallVector<BasicBlock *,1>> LoopExitBlocks;
while (!Worklist.empty()) {
UsesToRewrite.clear();
Instruction *I = Worklist.pop_back_val();
BasicBlock *InstBB = I->getParent();
Loop *L = LI.getLoopFor(InstBB);
if (!LoopExitBlocks.count(L))
L->getExitBlocks(LoopExitBlocks[L]);
assert(LoopExitBlocks.count(L));
const SmallVectorImpl<BasicBlock *> &ExitBlocks = LoopExitBlocks[L];
if (ExitBlocks.empty())
continue;
// Tokens cannot be used in PHI nodes, so we skip over them.
// We can run into tokens which are live out of a loop with catchswitch
// instructions in Windows EH if the catchswitch has one catchpad which
// is inside the loop and another which is not.
if (I->getType()->isTokenTy())
continue;
for (Use &U : I->uses()) {
Instruction *User = cast<Instruction>(U.getUser());
BasicBlock *UserBB = User->getParent();
if (PHINode *PN = dyn_cast<PHINode>(User))
UserBB = PN->getIncomingBlock(U);
if (InstBB != UserBB && !L->contains(UserBB))
UsesToRewrite.push_back(&U);
}
// If there are no uses outside the loop, exit with no change.
if (UsesToRewrite.empty())
continue;
++NumLCSSA; // We are applying the transformation
// Invoke instructions are special in that their result value is not
// available along their unwind edge. The code below tests to see whether
// DomBB dominates the value, so adjust DomBB to the normal destination
// block, which is effectively where the value is first usable.
BasicBlock *DomBB = InstBB;
if (InvokeInst *Inv = dyn_cast<InvokeInst>(I))
DomBB = Inv->getNormalDest();
DomTreeNode *DomNode = DT.getNode(DomBB);
SmallVector<PHINode *, 16> AddedPHIs;
SmallVector<PHINode *, 8> PostProcessPHIs;
SmallVector<PHINode *, 4> InsertedPHIs;
SSAUpdater SSAUpdate(&InsertedPHIs);
SSAUpdate.Initialize(I->getType(), I->getName());
// Insert the LCSSA phi's into all of the exit blocks dominated by the
// value, and add them to the Phi's map.
for (BasicBlock *ExitBB : ExitBlocks) {
if (!DT.dominates(DomNode, DT.getNode(ExitBB)))
continue;
// If we already inserted something for this BB, don't reprocess it.
if (SSAUpdate.HasValueForBlock(ExitBB))
continue;
PHINode *PN = PHINode::Create(I->getType(), PredCache.size(ExitBB),
I->getName() + ".lcssa", &ExitBB->front());
// Add inputs from inside the loop for this PHI.
for (BasicBlock *Pred : PredCache.get(ExitBB)) {
PN->addIncoming(I, Pred);
// If the exit block has a predecessor not within the loop, arrange for
// the incoming value use corresponding to that predecessor to be
// rewritten in terms of a different LCSSA PHI.
if (!L->contains(Pred))
UsesToRewrite.push_back(
&PN->getOperandUse(PN->getOperandNumForIncomingValue(
PN->getNumIncomingValues() - 1)));
}
AddedPHIs.push_back(PN);
// Remember that this phi makes the value alive in this block.
SSAUpdate.AddAvailableValue(ExitBB, PN);
// LoopSimplify might fail to simplify some loops (e.g. when indirect
// branches are involved). In such situations, it might happen that an
// exit for Loop L1 is the header of a disjoint Loop L2. Thus, when we
// create PHIs in such an exit block, we are also inserting PHIs into L2's
// header. This could break LCSSA form for L2 because these inserted PHIs
// can also have uses outside of L2. Remember all PHIs in such situation
// as to revisit than later on. FIXME: Remove this if indirectbr support
// into LoopSimplify gets improved.
if (auto *OtherLoop = LI.getLoopFor(ExitBB))
if (!L->contains(OtherLoop))
PostProcessPHIs.push_back(PN);
}
// Rewrite all uses outside the loop in terms of the new PHIs we just
// inserted.
for (Use *UseToRewrite : UsesToRewrite) {
// If this use is in an exit block, rewrite to use the newly inserted PHI.
// This is required for correctness because SSAUpdate doesn't handle uses
// in the same block. It assumes the PHI we inserted is at the end of the
// block.
Instruction *User = cast<Instruction>(UseToRewrite->getUser());
BasicBlock *UserBB = User->getParent();
if (PHINode *PN = dyn_cast<PHINode>(User))
UserBB = PN->getIncomingBlock(*UseToRewrite);
if (isa<PHINode>(UserBB->begin()) && isExitBlock(UserBB, ExitBlocks)) {
// Tell the VHs that the uses changed. This updates SCEV's caches.
if (UseToRewrite->get()->hasValueHandle())
ValueHandleBase::ValueIsRAUWd(*UseToRewrite, &UserBB->front());
UseToRewrite->set(&UserBB->front());
continue;
}
// Otherwise, do full PHI insertion.
SSAUpdate.RewriteUse(*UseToRewrite);
}
// SSAUpdater might have inserted phi-nodes inside other loops. We'll need
// to post-process them to keep LCSSA form.
for (PHINode *InsertedPN : InsertedPHIs) {
if (auto *OtherLoop = LI.getLoopFor(InsertedPN->getParent()))
if (!L->contains(OtherLoop))
PostProcessPHIs.push_back(InsertedPN);
}
// Post process PHI instructions that were inserted into another disjoint
// loop and update their exits properly.
for (auto *PostProcessPN : PostProcessPHIs) {
if (PostProcessPN->use_empty())
continue;
// Reprocess each PHI instruction.
Worklist.push_back(PostProcessPN);
}
// Keep track of PHI nodes that we want to remove because they did not have
// any uses rewritten.
for (PHINode *PN : AddedPHIs)
if (PN->use_empty())
PHIsToRemove.insert(PN);
Changed = true;
}
// Remove PHI nodes that did not have any uses rewritten.
for (PHINode *PN : PHIsToRemove) {
assert (PN->use_empty() && "Trying to remove a phi with uses.");
PN->eraseFromParent();
}
return Changed;
}
void AAEvaluator::runInternal(Function &F, AAResults &AA) {
const DataLayout &DL = F.getParent()->getDataLayout();
++FunctionCount;
SetVector<Value *> Pointers;
SmallSetVector<CallBase *, 16> Calls;
SetVector<Value *> Loads;
SetVector<Value *> Stores;
for (auto &I : F.args())
if (I.getType()->isPointerTy()) // Add all pointer arguments.
Pointers.insert(&I);
for (inst_iterator I = inst_begin(F), E = inst_end(F); I != E; ++I) {
if (I->getType()->isPointerTy()) // Add all pointer instructions.
Pointers.insert(&*I);
if (EvalAAMD && isa<LoadInst>(&*I))
Loads.insert(&*I);
if (EvalAAMD && isa<StoreInst>(&*I))
Stores.insert(&*I);
Instruction &Inst = *I;
if (auto *Call = dyn_cast<CallBase>(&Inst)) {
Value *Callee = Call->getCalledValue();
// Skip actual functions for direct function calls.
if (!isa<Function>(Callee) && isInterestingPointer(Callee))
Pointers.insert(Callee);
// Consider formals.
for (Use &DataOp : Call->data_ops())
if (isInterestingPointer(DataOp))
Pointers.insert(DataOp);
Calls.insert(Call);
} else {
// Consider all operands.
for (Instruction::op_iterator OI = Inst.op_begin(), OE = Inst.op_end();
OI != OE; ++OI)
if (isInterestingPointer(*OI))
Pointers.insert(*OI);
}
}
if (PrintAll || PrintNoAlias || PrintMayAlias || PrintPartialAlias ||
PrintMustAlias || PrintNoModRef || PrintMod || PrintRef || PrintModRef)
errs() << "Function: " << F.getName() << ": " << Pointers.size()
<< " pointers, " << Calls.size() << " call sites\n";
// iterate over the worklist, and run the full (n^2)/2 disambiguations
for (SetVector<Value *>::iterator I1 = Pointers.begin(), E = Pointers.end();
I1 != E; ++I1) {
auto I1Size = LocationSize::unknown();
Type *I1ElTy = cast<PointerType>((*I1)->getType())->getElementType();
if (I1ElTy->isSized())
I1Size = LocationSize::precise(DL.getTypeStoreSize(I1ElTy));
for (SetVector<Value *>::iterator I2 = Pointers.begin(); I2 != I1; ++I2) {
auto I2Size = LocationSize::unknown();
Type *I2ElTy = cast<PointerType>((*I2)->getType())->getElementType();
if (I2ElTy->isSized())
I2Size = LocationSize::precise(DL.getTypeStoreSize(I2ElTy));
AliasResult AR = AA.alias(*I1, I1Size, *I2, I2Size);
switch (AR) {
case NoAlias:
PrintResults(AR, PrintNoAlias, *I1, *I2, F.getParent());
++NoAliasCount;
break;
case MayAlias:
PrintResults(AR, PrintMayAlias, *I1, *I2, F.getParent());
++MayAliasCount;
break;
case PartialAlias:
PrintResults(AR, PrintPartialAlias, *I1, *I2, F.getParent());
++PartialAliasCount;
break;
case MustAlias:
PrintResults(AR, PrintMustAlias, *I1, *I2, F.getParent());
++MustAliasCount;
break;
}
}
}
if (EvalAAMD) {
// iterate over all pairs of load, store
for (Value *Load : Loads) {
for (Value *Store : Stores) {
AliasResult AR = AA.alias(MemoryLocation::get(cast<LoadInst>(Load)),
MemoryLocation::get(cast<StoreInst>(Store)));
switch (AR) {
case NoAlias:
PrintLoadStoreResults(AR, PrintNoAlias, Load, Store, F.getParent());
++NoAliasCount;
break;
case MayAlias:
PrintLoadStoreResults(AR, PrintMayAlias, Load, Store, F.getParent());
++MayAliasCount;
break;
case PartialAlias:
PrintLoadStoreResults(AR, PrintPartialAlias, Load, Store, F.getParent());
++PartialAliasCount;
break;
case MustAlias:
PrintLoadStoreResults(AR, PrintMustAlias, Load, Store, F.getParent());
++MustAliasCount;
break;
}
}
}
// iterate over all pairs of store, store
for (SetVector<Value *>::iterator I1 = Stores.begin(), E = Stores.end();
I1 != E; ++I1) {
for (SetVector<Value *>::iterator I2 = Stores.begin(); I2 != I1; ++I2) {
AliasResult AR = AA.alias(MemoryLocation::get(cast<StoreInst>(*I1)),
MemoryLocation::get(cast<StoreInst>(*I2)));
switch (AR) {
case NoAlias:
PrintLoadStoreResults(AR, PrintNoAlias, *I1, *I2, F.getParent());
++NoAliasCount;
break;
case MayAlias:
PrintLoadStoreResults(AR, PrintMayAlias, *I1, *I2, F.getParent());
++MayAliasCount;
break;
case PartialAlias:
PrintLoadStoreResults(AR, PrintPartialAlias, *I1, *I2, F.getParent());
++PartialAliasCount;
break;
case MustAlias:
PrintLoadStoreResults(AR, PrintMustAlias, *I1, *I2, F.getParent());
++MustAliasCount;
break;
}
}
}
}
// Mod/ref alias analysis: compare all pairs of calls and values
for (CallBase *Call : Calls) {
for (auto Pointer : Pointers) {
auto Size = LocationSize::unknown();
Type *ElTy = cast<PointerType>(Pointer->getType())->getElementType();
if (ElTy->isSized())
Size = LocationSize::precise(DL.getTypeStoreSize(ElTy));
switch (AA.getModRefInfo(Call, Pointer, Size)) {
case ModRefInfo::NoModRef:
PrintModRefResults("NoModRef", PrintNoModRef, Call, Pointer,
F.getParent());
++NoModRefCount;
break;
case ModRefInfo::Mod:
PrintModRefResults("Just Mod", PrintMod, Call, Pointer, F.getParent());
++ModCount;
break;
case ModRefInfo::Ref:
PrintModRefResults("Just Ref", PrintRef, Call, Pointer, F.getParent());
++RefCount;
break;
case ModRefInfo::ModRef:
PrintModRefResults("Both ModRef", PrintModRef, Call, Pointer,
F.getParent());
++ModRefCount;
break;
case ModRefInfo::Must:
PrintModRefResults("Must", PrintMust, Call, Pointer, F.getParent());
++MustCount;
break;
case ModRefInfo::MustMod:
PrintModRefResults("Just Mod (MustAlias)", PrintMustMod, Call, Pointer,
F.getParent());
++MustModCount;
break;
case ModRefInfo::MustRef:
PrintModRefResults("Just Ref (MustAlias)", PrintMustRef, Call, Pointer,
F.getParent());
++MustRefCount;
break;
case ModRefInfo::MustModRef:
PrintModRefResults("Both ModRef (MustAlias)", PrintMustModRef, Call,
Pointer, F.getParent());
++MustModRefCount;
break;
}
}
}
// Mod/ref alias analysis: compare all pairs of calls
for (CallBase *CallA : Calls) {
for (CallBase *CallB : Calls) {
if (CallA == CallB)
continue;
switch (AA.getModRefInfo(CallA, CallB)) {
case ModRefInfo::NoModRef:
PrintModRefResults("NoModRef", PrintNoModRef, CallA, CallB,
F.getParent());
++NoModRefCount;
break;
case ModRefInfo::Mod:
PrintModRefResults("Just Mod", PrintMod, CallA, CallB, F.getParent());
++ModCount;
break;
case ModRefInfo::Ref:
PrintModRefResults("Just Ref", PrintRef, CallA, CallB, F.getParent());
++RefCount;
break;
case ModRefInfo::ModRef:
PrintModRefResults("Both ModRef", PrintModRef, CallA, CallB,
F.getParent());
++ModRefCount;
break;
case ModRefInfo::Must:
PrintModRefResults("Must", PrintMust, CallA, CallB, F.getParent());
++MustCount;
break;
case ModRefInfo::MustMod:
PrintModRefResults("Just Mod (MustAlias)", PrintMustMod, CallA, CallB,
F.getParent());
++MustModCount;
break;
case ModRefInfo::MustRef:
PrintModRefResults("Just Ref (MustAlias)", PrintMustRef, CallA, CallB,
F.getParent());
++MustRefCount;
break;
case ModRefInfo::MustModRef:
PrintModRefResults("Both ModRef (MustAlias)", PrintMustModRef, CallA,
CallB, F.getParent());
++MustModRefCount;
break;
}
}
}
}
LazyCallGraph::SCC &llvm::updateCGAndAnalysisManagerForFunctionPass(
LazyCallGraph &G, LazyCallGraph::SCC &InitialC, LazyCallGraph::Node &N,
CGSCCAnalysisManager &AM, CGSCCUpdateResult &UR) {
using Node = LazyCallGraph::Node;
using Edge = LazyCallGraph::Edge;
using SCC = LazyCallGraph::SCC;
using RefSCC = LazyCallGraph::RefSCC;
RefSCC &InitialRC = InitialC.getOuterRefSCC();
SCC *C = &InitialC;
RefSCC *RC = &InitialRC;
Function &F = N.getFunction();
// Walk the function body and build up the set of retained, promoted, and
// demoted edges.
SmallVector<Constant *, 16> Worklist;
SmallPtrSet<Constant *, 16> Visited;
SmallPtrSet<Node *, 16> RetainedEdges;
SmallSetVector<Node *, 4> PromotedRefTargets;
SmallSetVector<Node *, 4> DemotedCallTargets;
// First walk the function and handle all called functions. We do this first
// because if there is a single call edge, whether there are ref edges is
// irrelevant.
for (Instruction &I : instructions(F))
if (auto CS = CallSite(&I))
if (Function *Callee = CS.getCalledFunction())
if (Visited.insert(Callee).second && !Callee->isDeclaration()) {
Node &CalleeN = *G.lookup(*Callee);
Edge *E = N->lookup(CalleeN);
// FIXME: We should really handle adding new calls. While it will
// make downstream usage more complex, there is no fundamental
// limitation and it will allow passes within the CGSCC to be a bit
// more flexible in what transforms they can do. Until then, we
// verify that new calls haven't been introduced.
assert(E && "No function transformations should introduce *new* "
"call edges! Any new calls should be modeled as "
"promoted existing ref edges!");
bool Inserted = RetainedEdges.insert(&CalleeN).second;
(void)Inserted;
assert(Inserted && "We should never visit a function twice.");
if (!E->isCall())
PromotedRefTargets.insert(&CalleeN);
}
// Now walk all references.
for (Instruction &I : instructions(F))
for (Value *Op : I.operand_values())
if (auto *C = dyn_cast<Constant>(Op))
if (Visited.insert(C).second)
Worklist.push_back(C);
auto VisitRef = [&](Function &Referee) {
Node &RefereeN = *G.lookup(Referee);
Edge *E = N->lookup(RefereeN);
// FIXME: Similarly to new calls, we also currently preclude
// introducing new references. See above for details.
assert(E && "No function transformations should introduce *new* ref "
"edges! Any new ref edges would require IPO which "
"function passes aren't allowed to do!");
bool Inserted = RetainedEdges.insert(&RefereeN).second;
(void)Inserted;
assert(Inserted && "We should never visit a function twice.");
if (E->isCall())
DemotedCallTargets.insert(&RefereeN);
};
LazyCallGraph::visitReferences(Worklist, Visited, VisitRef);
// Include synthetic reference edges to known, defined lib functions.
for (auto *F : G.getLibFunctions())
// While the list of lib functions doesn't have repeats, don't re-visit
// anything handled above.
if (!Visited.count(F))
VisitRef(*F);
// First remove all of the edges that are no longer present in this function.
// The first step makes these edges uniformly ref edges and accumulates them
// into a separate data structure so removal doesn't invalidate anything.
SmallVector<Node *, 4> DeadTargets;
for (Edge &E : *N) {
if (RetainedEdges.count(&E.getNode()))
continue;
SCC &TargetC = *G.lookupSCC(E.getNode());
RefSCC &TargetRC = TargetC.getOuterRefSCC();
if (&TargetRC == RC && E.isCall()) {
if (C != &TargetC) {
// For separate SCCs this is trivial.
RC->switchTrivialInternalEdgeToRef(N, E.getNode());
} else {
// Now update the call graph.
C = incorporateNewSCCRange(RC->switchInternalEdgeToRef(N, E.getNode()),
G, N, C, AM, UR);
}
}
// Now that this is ready for actual removal, put it into our list.
DeadTargets.push_back(&E.getNode());
}
// Remove the easy cases quickly and actually pull them out of our list.
DeadTargets.erase(
llvm::remove_if(DeadTargets,
[&](Node *TargetN) {
SCC &TargetC = *G.lookupSCC(*TargetN);
RefSCC &TargetRC = TargetC.getOuterRefSCC();
// We can't trivially remove internal targets, so skip
// those.
if (&TargetRC == RC)
return false;
RC->removeOutgoingEdge(N, *TargetN);
LLVM_DEBUG(dbgs() << "Deleting outgoing edge from '"
<< N << "' to '" << TargetN << "'\n");
return true;
}),
DeadTargets.end());
// Now do a batch removal of the internal ref edges left.
auto NewRefSCCs = RC->removeInternalRefEdge(N, DeadTargets);
if (!NewRefSCCs.empty()) {
// The old RefSCC is dead, mark it as such.
UR.InvalidatedRefSCCs.insert(RC);
// Note that we don't bother to invalidate analyses as ref-edge
// connectivity is not really observable in any way and is intended
// exclusively to be used for ordering of transforms rather than for
// analysis conclusions.
// Update RC to the "bottom".
assert(G.lookupSCC(N) == C && "Changed the SCC when splitting RefSCCs!");
RC = &C->getOuterRefSCC();
assert(G.lookupRefSCC(N) == RC && "Failed to update current RefSCC!");
// The RC worklist is in reverse postorder, so we enqueue the new ones in
// RPO except for the one which contains the source node as that is the
// "bottom" we will continue processing in the bottom-up walk.
assert(NewRefSCCs.front() == RC &&
"New current RefSCC not first in the returned list!");
for (RefSCC *NewRC : llvm::reverse(make_range(std::next(NewRefSCCs.begin()),
NewRefSCCs.end()))) {
assert(NewRC != RC && "Should not encounter the current RefSCC further "
"in the postorder list of new RefSCCs.");
UR.RCWorklist.insert(NewRC);
LLVM_DEBUG(dbgs() << "Enqueuing a new RefSCC in the update worklist: "
<< *NewRC << "\n");
}
}
// Next demote all the call edges that are now ref edges. This helps make
// the SCCs small which should minimize the work below as we don't want to
// form cycles that this would break.
for (Node *RefTarget : DemotedCallTargets) {
SCC &TargetC = *G.lookupSCC(*RefTarget);
RefSCC &TargetRC = TargetC.getOuterRefSCC();
// The easy case is when the target RefSCC is not this RefSCC. This is
// only supported when the target RefSCC is a child of this RefSCC.
if (&TargetRC != RC) {
assert(RC->isAncestorOf(TargetRC) &&
"Cannot potentially form RefSCC cycles here!");
RC->switchOutgoingEdgeToRef(N, *RefTarget);
LLVM_DEBUG(dbgs() << "Switch outgoing call edge to a ref edge from '" << N
<< "' to '" << *RefTarget << "'\n");
continue;
}
// We are switching an internal call edge to a ref edge. This may split up
// some SCCs.
if (C != &TargetC) {
// For separate SCCs this is trivial.
RC->switchTrivialInternalEdgeToRef(N, *RefTarget);
continue;
}
// Now update the call graph.
C = incorporateNewSCCRange(RC->switchInternalEdgeToRef(N, *RefTarget), G, N,
C, AM, UR);
}
// Now promote ref edges into call edges.
for (Node *CallTarget : PromotedRefTargets) {
SCC &TargetC = *G.lookupSCC(*CallTarget);
RefSCC &TargetRC = TargetC.getOuterRefSCC();
// The easy case is when the target RefSCC is not this RefSCC. This is
// only supported when the target RefSCC is a child of this RefSCC.
if (&TargetRC != RC) {
assert(RC->isAncestorOf(TargetRC) &&
"Cannot potentially form RefSCC cycles here!");
RC->switchOutgoingEdgeToCall(N, *CallTarget);
LLVM_DEBUG(dbgs() << "Switch outgoing ref edge to a call edge from '" << N
<< "' to '" << *CallTarget << "'\n");
continue;
}
LLVM_DEBUG(dbgs() << "Switch an internal ref edge to a call edge from '"
<< N << "' to '" << *CallTarget << "'\n");
// Otherwise we are switching an internal ref edge to a call edge. This
// may merge away some SCCs, and we add those to the UpdateResult. We also
// need to make sure to update the worklist in the event SCCs have moved
// before the current one in the post-order sequence
bool HasFunctionAnalysisProxy = false;
auto InitialSCCIndex = RC->find(*C) - RC->begin();
bool FormedCycle = RC->switchInternalEdgeToCall(
N, *CallTarget, [&](ArrayRef<SCC *> MergedSCCs) {
for (SCC *MergedC : MergedSCCs) {
assert(MergedC != &TargetC && "Cannot merge away the target SCC!");
HasFunctionAnalysisProxy |=
AM.getCachedResult<FunctionAnalysisManagerCGSCCProxy>(
*MergedC) != nullptr;
// Mark that this SCC will no longer be valid.
UR.InvalidatedSCCs.insert(MergedC);
// FIXME: We should really do a 'clear' here to forcibly release
// memory, but we don't have a good way of doing that and
// preserving the function analyses.
auto PA = PreservedAnalyses::allInSet<AllAnalysesOn<Function>>();
PA.preserve<FunctionAnalysisManagerCGSCCProxy>();
AM.invalidate(*MergedC, PA);
}
});
// If we formed a cycle by creating this call, we need to update more data
// structures.
if (FormedCycle) {
C = &TargetC;
assert(G.lookupSCC(N) == C && "Failed to update current SCC!");
// If one of the invalidated SCCs had a cached proxy to a function
// analysis manager, we need to create a proxy in the new current SCC as
// the invalidated SCCs had their functions moved.
if (HasFunctionAnalysisProxy)
AM.getResult<FunctionAnalysisManagerCGSCCProxy>(*C, G);
// Any analyses cached for this SCC are no longer precise as the shape
// has changed by introducing this cycle. However, we have taken care to
// update the proxies so it remains valide.
auto PA = PreservedAnalyses::allInSet<AllAnalysesOn<Function>>();
PA.preserve<FunctionAnalysisManagerCGSCCProxy>();
AM.invalidate(*C, PA);
}
auto NewSCCIndex = RC->find(*C) - RC->begin();
// If we have actually moved an SCC to be topologically "below" the current
// one due to merging, we will need to revisit the current SCC after
// visiting those moved SCCs.
//
// It is critical that we *do not* revisit the current SCC unless we
// actually move SCCs in the process of merging because otherwise we may
// form a cycle where an SCC is split apart, merged, split, merged and so
// on infinitely.
if (InitialSCCIndex < NewSCCIndex) {
// Put our current SCC back onto the worklist as we'll visit other SCCs
// that are now definitively ordered prior to the current one in the
// post-order sequence, and may end up observing more precise context to
// optimize the current SCC.
UR.CWorklist.insert(C);
LLVM_DEBUG(dbgs() << "Enqueuing the existing SCC in the worklist: " << *C
<< "\n");
// Enqueue in reverse order as we pop off the back of the worklist.
for (SCC &MovedC : llvm::reverse(make_range(RC->begin() + InitialSCCIndex,
RC->begin() + NewSCCIndex))) {
UR.CWorklist.insert(&MovedC);
LLVM_DEBUG(dbgs() << "Enqueuing a newly earlier in post-order SCC: "
<< MovedC << "\n");
}
}
}
assert(!UR.InvalidatedSCCs.count(C) && "Invalidated the current SCC!");
assert(!UR.InvalidatedRefSCCs.count(RC) && "Invalidated the current RefSCC!");
assert(&C->getOuterRefSCC() == RC && "Current SCC not in current RefSCC!");
// Record the current RefSCC and SCC for higher layers of the CGSCC pass
// manager now that all the updates have been applied.
if (RC != &InitialRC)
UR.UpdatedRC = RC;
if (C != &InitialC)
UR.UpdatedC = C;
return *C;
}
bool AArch64RedundantCopyElimination::optimizeCopy(MachineBasicBlock *MBB) {
// Check if the current basic block has a single predecessor.
if (MBB->pred_size() != 1)
return false;
MachineBasicBlock *PredMBB = *MBB->pred_begin();
MachineBasicBlock::iterator CompBr = PredMBB->getLastNonDebugInstr();
if (CompBr == PredMBB->end() || PredMBB->succ_size() != 2)
return false;
++CompBr;
do {
--CompBr;
if (guaranteesZeroRegInBlock(*CompBr, MBB))
break;
} while (CompBr != PredMBB->begin() && CompBr->isTerminator());
// We've not found a CBZ/CBNZ, time to bail out.
if (!guaranteesZeroRegInBlock(*CompBr, MBB))
return false;
unsigned TargetReg = CompBr->getOperand(0).getReg();
if (!TargetReg)
return false;
assert(TargetRegisterInfo::isPhysicalRegister(TargetReg) &&
"Expect physical register");
// Remember all registers aliasing with TargetReg.
SmallSetVector<unsigned, 8> TargetRegs;
for (MCRegAliasIterator AI(TargetReg, TRI, true); AI.isValid(); ++AI)
TargetRegs.insert(*AI);
bool Changed = false;
MachineBasicBlock::iterator LastChange = MBB->begin();
unsigned SmallestDef = TargetReg;
// Remove redundant Copy instructions unless TargetReg is modified.
for (MachineBasicBlock::iterator I = MBB->begin(), E = MBB->end(); I != E;) {
MachineInstr *MI = &*I;
++I;
if (MI->isCopy() && MI->getOperand(0).isReg() &&
MI->getOperand(1).isReg()) {
unsigned DefReg = MI->getOperand(0).getReg();
unsigned SrcReg = MI->getOperand(1).getReg();
if ((SrcReg == AArch64::XZR || SrcReg == AArch64::WZR) &&
!MRI->isReserved(DefReg) &&
(TargetReg == DefReg || TRI->isSuperRegister(DefReg, TargetReg))) {
DEBUG(dbgs() << "Remove redundant Copy : ");
DEBUG((MI)->print(dbgs()));
MI->eraseFromParent();
Changed = true;
LastChange = I;
NumCopiesRemoved++;
SmallestDef =
TRI->isSubRegister(SmallestDef, DefReg) ? DefReg : SmallestDef;
continue;
}
}
if (MI->modifiesRegister(TargetReg, TRI))
break;
}
if (!Changed)
return false;
// Otherwise, we have to fixup the use-def chain, starting with the
// CBZ/CBNZ. Conservatively mark as much as we can live.
CompBr->clearRegisterKills(SmallestDef, TRI);
if (std::none_of(TargetRegs.begin(), TargetRegs.end(),
[&](unsigned Reg) { return MBB->isLiveIn(Reg); }))
MBB->addLiveIn(TargetReg);
// Clear any kills of TargetReg between CompBr and the last removed COPY.
for (MachineInstr &MMI :
make_range(MBB->begin()->getIterator(), LastChange->getIterator()))
MMI.clearRegisterKills(SmallestDef, TRI);
return true;
}
/// After FromBB is tail duplicated into its predecessor blocks, the successors
/// have gained new predecessors. Update the PHI instructions in them
/// accordingly.
void
TailDuplicatePass::UpdateSuccessorsPHIs(MachineBasicBlock *FromBB, bool isDead,
SmallVectorImpl<MachineBasicBlock *> &TDBBs,
SmallSetVector<MachineBasicBlock*,8> &Succs) {
for (SmallSetVector<MachineBasicBlock*, 8>::iterator SI = Succs.begin(),
SE = Succs.end(); SI != SE; ++SI) {
MachineBasicBlock *SuccBB = *SI;
for (MachineBasicBlock::iterator II = SuccBB->begin(), EE = SuccBB->end();
II != EE; ++II) {
if (!II->isPHI())
break;
MachineInstrBuilder MIB(*FromBB->getParent(), II);
unsigned Idx = 0;
for (unsigned i = 1, e = II->getNumOperands(); i != e; i += 2) {
MachineOperand &MO = II->getOperand(i+1);
if (MO.getMBB() == FromBB) {
Idx = i;
break;
}
}
assert(Idx != 0);
MachineOperand &MO0 = II->getOperand(Idx);
unsigned Reg = MO0.getReg();
if (isDead) {
// Folded into the previous BB.
// There could be duplicate phi source entries. FIXME: Should sdisel
// or earlier pass fixed this?
for (unsigned i = II->getNumOperands()-2; i != Idx; i -= 2) {
MachineOperand &MO = II->getOperand(i+1);
if (MO.getMBB() == FromBB) {
II->RemoveOperand(i+1);
II->RemoveOperand(i);
}
}
} else
Idx = 0;
// If Idx is set, the operands at Idx and Idx+1 must be removed.
// We reuse the location to avoid expensive RemoveOperand calls.
DenseMap<unsigned,AvailableValsTy>::iterator LI=SSAUpdateVals.find(Reg);
if (LI != SSAUpdateVals.end()) {
// This register is defined in the tail block.
for (unsigned j = 0, ee = LI->second.size(); j != ee; ++j) {
MachineBasicBlock *SrcBB = LI->second[j].first;
// If we didn't duplicate a bb into a particular predecessor, we
// might still have added an entry to SSAUpdateVals to correcly
// recompute SSA. If that case, avoid adding a dummy extra argument
// this PHI.
if (!SrcBB->isSuccessor(SuccBB))
continue;
unsigned SrcReg = LI->second[j].second;
if (Idx != 0) {
II->getOperand(Idx).setReg(SrcReg);
II->getOperand(Idx+1).setMBB(SrcBB);
Idx = 0;
} else {
MIB.addReg(SrcReg).addMBB(SrcBB);
}
}
} else {
// Live in tail block, must also be live in predecessors.
for (unsigned j = 0, ee = TDBBs.size(); j != ee; ++j) {
MachineBasicBlock *SrcBB = TDBBs[j];
if (Idx != 0) {
II->getOperand(Idx).setReg(Reg);
II->getOperand(Idx+1).setMBB(SrcBB);
Idx = 0;
} else {
MIB.addReg(Reg).addMBB(SrcBB);
}
}
}
if (Idx != 0) {
II->RemoveOperand(Idx+1);
II->RemoveOperand(Idx);
}
}
}
}
/// linkModuleFlagsMetadata - Merge the linker flags in Src into the Dest
/// module.
bool ModuleLinker::linkModuleFlagsMetadata() {
const NamedMDNode *SrcModFlags = SrcM->getModuleFlagsMetadata();
if (!SrcModFlags) return false;
NamedMDNode *DstModFlags = DstM->getOrInsertModuleFlagsMetadata();
// If the destination module doesn't have module flags yet, then just copy
// over the source module's flags.
if (DstModFlags->getNumOperands() == 0) {
for (unsigned I = 0, E = SrcModFlags->getNumOperands(); I != E; ++I)
DstModFlags->addOperand(SrcModFlags->getOperand(I));
return false;
}
bool HasErr = false;
// Otherwise, we have to merge them based on their behaviors. First,
// categorize all of the nodes in the modules' module flags. If an error or
// warning occurs, then emit the appropriate message(s).
DenseMap<MDString*, MDNode*> ErrorNode;
DenseMap<MDString*, MDNode*> WarningNode;
DenseMap<MDString*, MDNode*> OverrideNode;
DenseMap<MDString*, SmallSetVector<MDNode*, 8> > RequireNodes;
SmallSetVector<MDString*, 16> SeenIDs;
HasErr |= categorizeModuleFlagNodes(SrcModFlags, ErrorNode, WarningNode,
OverrideNode, RequireNodes, SeenIDs);
HasErr |= categorizeModuleFlagNodes(DstModFlags, ErrorNode, WarningNode,
OverrideNode, RequireNodes, SeenIDs);
// Check that there isn't both an error and warning node for a flag.
for (SmallSetVector<MDString*, 16>::iterator
I = SeenIDs.begin(), E = SeenIDs.end(); I != E; ++I) {
MDString *ID = *I;
if (ErrorNode[ID] && WarningNode[ID])
HasErr = emitError("linking module flags '" + ID->getString() +
"': IDs have conflicting behaviors");
}
// Early exit if we had an error.
if (HasErr) return true;
// Get the destination's module flags ready for new operands.
DstModFlags->dropAllReferences();
// Add all of the module flags to the destination module.
DenseMap<MDString*, SmallVector<MDNode*, 4> > AddedNodes;
for (SmallSetVector<MDString*, 16>::iterator
I = SeenIDs.begin(), E = SeenIDs.end(); I != E; ++I) {
MDString *ID = *I;
if (OverrideNode[ID]) {
DstModFlags->addOperand(OverrideNode[ID]);
AddedNodes[ID].push_back(OverrideNode[ID]);
} else if (ErrorNode[ID]) {
DstModFlags->addOperand(ErrorNode[ID]);
AddedNodes[ID].push_back(ErrorNode[ID]);
} else if (WarningNode[ID]) {
DstModFlags->addOperand(WarningNode[ID]);
AddedNodes[ID].push_back(WarningNode[ID]);
}
for (SmallSetVector<MDNode*, 8>::iterator
II = RequireNodes[ID].begin(), IE = RequireNodes[ID].end();
II != IE; ++II)
DstModFlags->addOperand(*II);
}
// Now check that all of the requirements have been satisfied.
for (SmallSetVector<MDString*, 16>::iterator
I = SeenIDs.begin(), E = SeenIDs.end(); I != E; ++I) {
MDString *ID = *I;
SmallSetVector<MDNode*, 8> &Set = RequireNodes[ID];
for (SmallSetVector<MDNode*, 8>::iterator
II = Set.begin(), IE = Set.end(); II != IE; ++II) {
MDNode *Node = *II;
assert(isa<MDNode>(Node->getOperand(2)) &&
"Module flag's third operand must be an MDNode!");
MDNode *Val = cast<MDNode>(Node->getOperand(2));
MDString *ReqID = cast<MDString>(Val->getOperand(0));
Value *ReqVal = Val->getOperand(1);
bool HasValue = false;
for (SmallVectorImpl<MDNode*>::iterator
RI = AddedNodes[ReqID].begin(), RE = AddedNodes[ReqID].end();
RI != RE; ++RI) {
MDNode *ReqNode = *RI;
if (ReqNode->getOperand(2) == ReqVal) {
HasValue = true;
break;
}
}
if (!HasValue)
HasErr = emitError("linking module flags '" + ReqID->getString() +
"': does not have the required value");
}
}
return HasErr;
}
/// setupEntryBlockAndCallSites - Setup the entry block by creating and filling
/// the function context and marking the call sites with the appropriate
/// values. These values are used by the DWARF EH emitter.
bool SjLjEHPrepare::setupEntryBlockAndCallSites(Function &F) {
SmallVector<ReturnInst *, 16> Returns;
SmallVector<InvokeInst *, 16> Invokes;
SmallSetVector<LandingPadInst *, 16> LPads;
// Look through the terminators of the basic blocks to find invokes.
for (BasicBlock &BB : F)
if (auto *II = dyn_cast<InvokeInst>(BB.getTerminator())) {
if (Function *Callee = II->getCalledFunction())
if (Callee->getIntrinsicID() == Intrinsic::donothing) {
// Remove the NOP invoke.
BranchInst::Create(II->getNormalDest(), II);
II->eraseFromParent();
continue;
}
Invokes.push_back(II);
LPads.insert(II->getUnwindDest()->getLandingPadInst());
} else if (auto *RI = dyn_cast<ReturnInst>(BB.getTerminator())) {
Returns.push_back(RI);
}
if (Invokes.empty())
return false;
NumInvokes += Invokes.size();
lowerIncomingArguments(F);
lowerAcrossUnwindEdges(F, Invokes);
Value *FuncCtx =
setupFunctionContext(F, makeArrayRef(LPads.begin(), LPads.end()));
BasicBlock *EntryBB = &F.front();
IRBuilder<> Builder(EntryBB->getTerminator());
// Get a reference to the jump buffer.
Value *JBufPtr =
Builder.CreateConstGEP2_32(FunctionContextTy, FuncCtx, 0, 5, "jbuf_gep");
// Save the frame pointer.
Value *FramePtr = Builder.CreateConstGEP2_32(doubleUnderJBufTy, JBufPtr, 0, 0,
"jbuf_fp_gep");
Value *Val = Builder.CreateCall(FrameAddrFn, Builder.getInt32(0), "fp");
Builder.CreateStore(Val, FramePtr, /*isVolatile=*/true);
// Save the stack pointer.
Value *StackPtr = Builder.CreateConstGEP2_32(doubleUnderJBufTy, JBufPtr, 0, 2,
"jbuf_sp_gep");
Val = Builder.CreateCall(StackAddrFn, {}, "sp");
Builder.CreateStore(Val, StackPtr, /*isVolatile=*/true);
// Call the setup_dispatch instrinsic. It fills in the rest of the jmpbuf.
Builder.CreateCall(BuiltinSetupDispatchFn, {});
// Store a pointer to the function context so that the back-end will know
// where to look for it.
Value *FuncCtxArg = Builder.CreateBitCast(FuncCtx, Builder.getInt8PtrTy());
Builder.CreateCall(FuncCtxFn, FuncCtxArg);
// At this point, we are all set up, update the invoke instructions to mark
// their call_site values.
for (unsigned I = 0, E = Invokes.size(); I != E; ++I) {
insertCallSiteStore(Invokes[I], I + 1);
ConstantInt *CallSiteNum =
ConstantInt::get(Type::getInt32Ty(F.getContext()), I + 1);
// Record the call site value for the back end so it stays associated with
// the invoke.
CallInst::Create(CallSiteFn, CallSiteNum, "", Invokes[I]);
}
// Mark call instructions that aren't nounwind as no-action (call_site ==
// -1). Skip the entry block, as prior to then, no function context has been
// created for this function and any unexpected exceptions thrown will go
// directly to the caller's context, which is what we want anyway, so no need
// to do anything here.
for (BasicBlock &BB : F) {
if (&BB == &F.front())
continue;
for (Instruction &I : BB)
if (I.mayThrow())
insertCallSiteStore(&I, -1);
}
// Register the function context and make sure it's known to not throw
CallInst *Register =
CallInst::Create(RegisterFn, FuncCtx, "", EntryBB->getTerminator());
Register->setDoesNotThrow();
// Following any allocas not in the entry block, update the saved SP in the
// jmpbuf to the new value.
for (BasicBlock &BB : F) {
if (&BB == &F.front())
continue;
for (Instruction &I : BB) {
if (auto *CI = dyn_cast<CallInst>(&I)) {
if (CI->getCalledFunction() != StackRestoreFn)
continue;
} else if (!isa<AllocaInst>(&I)) {
continue;
}
Instruction *StackAddr = CallInst::Create(StackAddrFn, "sp");
StackAddr->insertAfter(&I);
Instruction *StoreStackAddr = new StoreInst(StackAddr, StackPtr, true);
StoreStackAddr->insertAfter(StackAddr);
}
}
// Finally, for any returns from this function, if this function contains an
// invoke, add a call to unregister the function context.
for (ReturnInst *Return : Returns)
CallInst::Create(UnregisterFn, FuncCtx, "", Return);
return true;
}
bool MachineCopyPropagation::CopyPropagateBlock(MachineBasicBlock &MBB) {
SmallSetVector<MachineInstr*, 8> MaybeDeadCopies; // Candidates for deletion
DenseMap<unsigned, MachineInstr*> AvailCopyMap; // Def -> available copies map
DenseMap<unsigned, MachineInstr*> CopyMap; // Def -> copies map
DenseMap<unsigned, unsigned> SrcMap; // Src -> Def map
bool Changed = false;
for (MachineBasicBlock::iterator I = MBB.begin(), E = MBB.end(); I != E; ) {
MachineInstr *MI = &*I;
++I;
if (MI->isCopy()) {
unsigned Def = MI->getOperand(0).getReg();
unsigned Src = MI->getOperand(1).getReg();
if (TargetRegisterInfo::isVirtualRegister(Def) ||
TargetRegisterInfo::isVirtualRegister(Src))
report_fatal_error("MachineCopyPropagation should be run after"
" register allocation!");
DenseMap<unsigned, MachineInstr*>::iterator CI = AvailCopyMap.find(Src);
if (CI != AvailCopyMap.end()) {
MachineInstr *CopyMI = CI->second;
unsigned SrcSrc = CopyMI->getOperand(1).getReg();
if (!ReservedRegs.test(Def) &&
(!ReservedRegs.test(Src) || NoInterveningSideEffect(CopyMI, MI)) &&
(SrcSrc == Def || TRI->isSubRegister(SrcSrc, Def))) {
// The two copies cancel out and the source of the first copy
// hasn't been overridden, eliminate the second one. e.g.
// %ECX<def> = COPY %EAX<kill>
// ... nothing clobbered EAX.
// %EAX<def> = COPY %ECX
// =>
// %ECX<def> = COPY %EAX
//
// Also avoid eliminating a copy from reserved registers unless the
// definition is proven not clobbered. e.g.
// %RSP<def> = COPY %RAX
// CALL
// %RAX<def> = COPY %RSP
CopyMI->getOperand(1).setIsKill(false);
MI->eraseFromParent();
Changed = true;
++NumDeletes;
continue;
}
}
// If Src is defined by a previous copy, it cannot be eliminated.
CI = CopyMap.find(Src);
if (CI != CopyMap.end())
MaybeDeadCopies.remove(CI->second);
for (const unsigned *AS = TRI->getAliasSet(Src); *AS; ++AS) {
CI = CopyMap.find(*AS);
if (CI != CopyMap.end())
MaybeDeadCopies.remove(CI->second);
}
// Copy is now a candidate for deletion.
MaybeDeadCopies.insert(MI);
// If 'Src' is previously source of another copy, then this earlier copy's
// source is no longer available. e.g.
// %xmm9<def> = copy %xmm2
// ...
// %xmm2<def> = copy %xmm0
// ...
// %xmm2<def> = copy %xmm9
SourceNoLongerAvailable(Def, SrcMap, AvailCopyMap);
// Remember Def is defined by the copy.
CopyMap[Def] = MI;
AvailCopyMap[Def] = MI;
for (const unsigned *SR = TRI->getSubRegisters(Def); *SR; ++SR) {
CopyMap[*SR] = MI;
AvailCopyMap[*SR] = MI;
}
// Remember source that's copied to Def. Once it's clobbered, then
// it's no longer available for copy propagation.
SrcMap[Src] = Def;
continue;
}
// Not a copy.
SmallVector<unsigned, 2> Defs;
for (unsigned i = 0, e = MI->getNumOperands(); i != e; ++i) {
MachineOperand &MO = MI->getOperand(i);
if (!MO.isReg())
continue;
unsigned Reg = MO.getReg();
if (!Reg)
continue;
if (TargetRegisterInfo::isVirtualRegister(Reg))
report_fatal_error("MachineCopyPropagation should be run after"
" register allocation!");
if (MO.isDef()) {
Defs.push_back(Reg);
continue;
}
// If 'Reg' is defined by a copy, the copy is no longer a candidate
// for elimination.
DenseMap<unsigned, MachineInstr*>::iterator CI = CopyMap.find(Reg);
if (CI != CopyMap.end())
MaybeDeadCopies.remove(CI->second);
for (const unsigned *AS = TRI->getAliasSet(Reg); *AS; ++AS) {
CI = CopyMap.find(*AS);
if (CI != CopyMap.end())
MaybeDeadCopies.remove(CI->second);
}
}
for (unsigned i = 0, e = Defs.size(); i != e; ++i) {
unsigned Reg = Defs[i];
// No longer defined by a copy.
CopyMap.erase(Reg);
AvailCopyMap.erase(Reg);
for (const unsigned *AS = TRI->getAliasSet(Reg); *AS; ++AS) {
CopyMap.erase(*AS);
AvailCopyMap.erase(*AS);
}
// If 'Reg' is previously source of a copy, it is no longer available for
// copy propagation.
SourceNoLongerAvailable(Reg, SrcMap, AvailCopyMap);
}
}
// If MBB doesn't have successors, delete the copies whose defs are not used.
// If MBB does have successors, then conservative assume the defs are live-out
// since we don't want to trust live-in lists.
if (MBB.succ_empty()) {
for (SmallSetVector<MachineInstr*, 8>::iterator
DI = MaybeDeadCopies.begin(), DE = MaybeDeadCopies.end();
DI != DE; ++DI) {
if (!ReservedRegs.test((*DI)->getOperand(0).getReg())) {
(*DI)->eraseFromParent();
Changed = true;
++NumDeletes;
}
}
}
return Changed;
}
/// Remove dead stores to stack-allocated locations in the function end block.
/// Ex:
/// %A = alloca i32
/// ...
/// store i32 1, i32* %A
/// ret void
static bool handleEndBlock(BasicBlock &BB, AliasAnalysis *AA,
MemoryDependenceResults *MD,
const TargetLibraryInfo *TLI,
InstOverlapIntervalsTy &IOL,
DenseMap<Instruction*, size_t> *InstrOrdering) {
bool MadeChange = false;
// Keep track of all of the stack objects that are dead at the end of the
// function.
SmallSetVector<Value*, 16> DeadStackObjects;
// Find all of the alloca'd pointers in the entry block.
BasicBlock &Entry = BB.getParent()->front();
for (Instruction &I : Entry) {
if (isa<AllocaInst>(&I))
DeadStackObjects.insert(&I);
// Okay, so these are dead heap objects, but if the pointer never escapes
// then it's leaked by this function anyways.
else if (isAllocLikeFn(&I, TLI) && !PointerMayBeCaptured(&I, true, true))
DeadStackObjects.insert(&I);
}
// Treat byval or inalloca arguments the same, stores to them are dead at the
// end of the function.
for (Argument &AI : BB.getParent()->args())
if (AI.hasByValOrInAllocaAttr())
DeadStackObjects.insert(&AI);
const DataLayout &DL = BB.getModule()->getDataLayout();
// Scan the basic block backwards
for (BasicBlock::iterator BBI = BB.end(); BBI != BB.begin(); ){
--BBI;
// If we find a store, check to see if it points into a dead stack value.
if (hasMemoryWrite(&*BBI, *TLI) && isRemovable(&*BBI)) {
// See through pointer-to-pointer bitcasts
SmallVector<Value *, 4> Pointers;
GetUnderlyingObjects(getStoredPointerOperand(&*BBI), Pointers, DL);
// Stores to stack values are valid candidates for removal.
bool AllDead = true;
for (Value *Pointer : Pointers)
if (!DeadStackObjects.count(Pointer)) {
AllDead = false;
break;
}
if (AllDead) {
Instruction *Dead = &*BBI;
DEBUG(dbgs() << "DSE: Dead Store at End of Block:\n DEAD: "
<< *Dead << "\n Objects: ";
for (SmallVectorImpl<Value *>::iterator I = Pointers.begin(),
E = Pointers.end(); I != E; ++I) {
dbgs() << **I;
if (std::next(I) != E)
dbgs() << ", ";
}
dbgs() << '\n');
// DCE instructions only used to calculate that store.
deleteDeadInstruction(Dead, &BBI, *MD, *TLI, IOL, InstrOrdering, &DeadStackObjects);
++NumFastStores;
MadeChange = true;
continue;
}
}
// Remove any dead non-memory-mutating instructions.
if (isInstructionTriviallyDead(&*BBI, TLI)) {
DEBUG(dbgs() << "DSE: Removing trivially dead instruction:\n DEAD: "
<< *&*BBI << '\n');
deleteDeadInstruction(&*BBI, &BBI, *MD, *TLI, IOL, InstrOrdering, &DeadStackObjects);
++NumFastOther;
MadeChange = true;
continue;
}
if (isa<AllocaInst>(BBI)) {
// Remove allocas from the list of dead stack objects; there can't be
// any references before the definition.
DeadStackObjects.remove(&*BBI);
continue;
}
if (auto CS = CallSite(&*BBI)) {
// Remove allocation function calls from the list of dead stack objects;
// there can't be any references before the definition.
if (isAllocLikeFn(&*BBI, TLI))
DeadStackObjects.remove(&*BBI);
// If this call does not access memory, it can't be loading any of our
// pointers.
if (AA->doesNotAccessMemory(CS))
continue;
// If the call might load from any of our allocas, then any store above
// the call is live.
DeadStackObjects.remove_if([&](Value *I) {
// See if the call site touches the value.
ModRefInfo A = AA->getModRefInfo(CS, I, getPointerSize(I, DL, *TLI));
return A == MRI_ModRef || A == MRI_Ref;
});
// If all of the allocas were clobbered by the call then we're not going
// to find anything else to process.
if (DeadStackObjects.empty())
break;
continue;
}
// We can remove the dead stores, irrespective of the fence and its ordering
// (release/acquire/seq_cst). Fences only constraints the ordering of
// already visible stores, it does not make a store visible to other
// threads. So, skipping over a fence does not change a store from being
// dead.
if (isa<FenceInst>(*BBI))
continue;
MemoryLocation LoadedLoc;
// If we encounter a use of the pointer, it is no longer considered dead
if (LoadInst *L = dyn_cast<LoadInst>(BBI)) {
if (!L->isUnordered()) // Be conservative with atomic/volatile load
break;
LoadedLoc = MemoryLocation::get(L);
} else if (VAArgInst *V = dyn_cast<VAArgInst>(BBI)) {
LoadedLoc = MemoryLocation::get(V);
} else if (MemTransferInst *MTI = dyn_cast<MemTransferInst>(BBI)) {
LoadedLoc = MemoryLocation::getForSource(MTI);
} else if (!BBI->mayReadFromMemory()) {
// Instruction doesn't read memory. Note that stores that weren't removed
// above will hit this case.
continue;
} else {
// Unknown inst; assume it clobbers everything.
break;
}
// Remove any allocas from the DeadPointer set that are loaded, as this
// makes any stores above the access live.
removeAccessedObjects(LoadedLoc, DeadStackObjects, DL, AA, TLI);
// If all of the allocas were clobbered by the access then we're not going
// to find anything else to process.
if (DeadStackObjects.empty())
break;
}
/// Sort the blocks in RPO, taking special care to make sure that loops are
/// contiguous even in the case of split backedges.
///
/// TODO: Determine whether RPO is actually worthwhile, or whether we should
/// move to just a stable-topological-sort-based approach that would preserve
/// more of the original order.
static void SortBlocks(MachineFunction &MF, const MachineLoopInfo &MLI) {
// Note that we do our own RPO rather than using
// "llvm/ADT/PostOrderIterator.h" because we want control over the order that
// successors are visited in (see above). Also, we can sort the blocks in the
// MachineFunction as we go.
SmallPtrSet<MachineBasicBlock *, 16> Visited;
SmallVector<POStackEntry, 16> Stack;
MachineBasicBlock *EntryBlock = &*MF.begin();
Visited.insert(EntryBlock);
Stack.push_back(POStackEntry(EntryBlock, MF, MLI));
for (;;) {
POStackEntry &Entry = Stack.back();
SmallVectorImpl<MachineBasicBlock *> &Succs = Entry.Succs;
if (!Succs.empty()) {
MachineBasicBlock *Succ = Succs.pop_back_val();
if (Visited.insert(Succ).second)
Stack.push_back(POStackEntry(Succ, MF, MLI));
continue;
}
// Put the block in its position in the MachineFunction.
MachineBasicBlock &MBB = *Entry.MBB;
MBB.moveBefore(&*MF.begin());
// Branch instructions may utilize a fallthrough, so update them if a
// fallthrough has been added or removed.
if (!MBB.empty() && MBB.back().isTerminator() && !MBB.back().isBranch() &&
!MBB.back().isBarrier())
report_fatal_error(
"Non-branch terminator with fallthrough cannot yet be rewritten");
if (MBB.empty() || !MBB.back().isTerminator() || MBB.back().isBranch())
MBB.updateTerminator();
Stack.pop_back();
if (Stack.empty())
break;
}
// Now that we've sorted the blocks in RPO, renumber them.
MF.RenumberBlocks();
#ifndef NDEBUG
SmallSetVector<MachineLoop *, 8> OnStack;
// Insert a sentinel representing the degenerate loop that starts at the
// function entry block and includes the entire function as a "loop" that
// executes once.
OnStack.insert(nullptr);
for (auto &MBB : MF) {
assert(MBB.getNumber() >= 0 && "Renumbered blocks should be non-negative.");
MachineLoop *Loop = MLI.getLoopFor(&MBB);
if (Loop && &MBB == Loop->getHeader()) {
// Loop header. The loop predecessor should be sorted above, and the other
// predecessors should be backedges below.
for (auto Pred : MBB.predecessors())
assert(
(Pred->getNumber() < MBB.getNumber() || Loop->contains(Pred)) &&
"Loop header predecessors must be loop predecessors or backedges");
assert(OnStack.insert(Loop) && "Loops should be declared at most once.");
} else {
// Not a loop header. All predecessors should be sorted above.
for (auto Pred : MBB.predecessors())
assert(Pred->getNumber() < MBB.getNumber() &&
"Non-loop-header predecessors should be topologically sorted");
assert(OnStack.count(MLI.getLoopFor(&MBB)) &&
"Blocks must be nested in their loops");
}
while (OnStack.size() > 1 && &MBB == LoopBottom(OnStack.back()))
OnStack.pop_back();
}
assert(OnStack.pop_back_val() == nullptr &&
"The function entry block shouldn't actually be a loop header");
assert(OnStack.empty() &&
"Control flow stack pushes and pops should be balanced.");
#endif
}
/// Tests whether this function is known to not return null.
///
/// Requires that the function returns a pointer.
///
/// Returns true if it believes the function will not return a null, and sets
/// \p Speculative based on whether the returned conclusion is a speculative
/// conclusion due to SCC calls.
static bool isReturnNonNull(Function *F, const SCCNodeSet &SCCNodes,
const TargetLibraryInfo &TLI, bool &Speculative) {
assert(F->getReturnType()->isPointerTy() &&
"nonnull only meaningful on pointer types");
Speculative = false;
SmallSetVector<Value *, 8> FlowsToReturn;
for (BasicBlock &BB : *F)
if (auto *Ret = dyn_cast<ReturnInst>(BB.getTerminator()))
FlowsToReturn.insert(Ret->getReturnValue());
for (unsigned i = 0; i != FlowsToReturn.size(); ++i) {
Value *RetVal = FlowsToReturn[i];
// If this value is locally known to be non-null, we're good
if (isKnownNonNull(RetVal, &TLI))
continue;
// Otherwise, we need to look upwards since we can't make any local
// conclusions.
Instruction *RVI = dyn_cast<Instruction>(RetVal);
if (!RVI)
return false;
switch (RVI->getOpcode()) {
// Extend the analysis by looking upwards.
case Instruction::BitCast:
case Instruction::GetElementPtr:
case Instruction::AddrSpaceCast:
FlowsToReturn.insert(RVI->getOperand(0));
continue;
case Instruction::Select: {
SelectInst *SI = cast<SelectInst>(RVI);
FlowsToReturn.insert(SI->getTrueValue());
FlowsToReturn.insert(SI->getFalseValue());
continue;
}
case Instruction::PHI: {
PHINode *PN = cast<PHINode>(RVI);
for (int i = 0, e = PN->getNumIncomingValues(); i != e; ++i)
FlowsToReturn.insert(PN->getIncomingValue(i));
continue;
}
case Instruction::Call:
case Instruction::Invoke: {
CallSite CS(RVI);
Function *Callee = CS.getCalledFunction();
// A call to a node within the SCC is assumed to return null until
// proven otherwise
if (Callee && SCCNodes.count(Callee)) {
Speculative = true;
continue;
}
return false;
}
default:
return false; // Unknown source, may be null
};
llvm_unreachable("should have either continued or returned");
}
return true;
}
/// Merge the linker flags in Src into the Dest module.
Error IRLinker::linkModuleFlagsMetadata() {
// If the source module has no module flags, we are done.
const NamedMDNode *SrcModFlags = SrcM->getModuleFlagsMetadata();
if (!SrcModFlags)
return Error::success();
// If the destination module doesn't have module flags yet, then just copy
// over the source module's flags.
NamedMDNode *DstModFlags = DstM.getOrInsertModuleFlagsMetadata();
if (DstModFlags->getNumOperands() == 0) {
for (unsigned I = 0, E = SrcModFlags->getNumOperands(); I != E; ++I)
DstModFlags->addOperand(SrcModFlags->getOperand(I));
return Error::success();
}
// First build a map of the existing module flags and requirements.
DenseMap<MDString *, std::pair<MDNode *, unsigned>> Flags;
SmallSetVector<MDNode *, 16> Requirements;
for (unsigned I = 0, E = DstModFlags->getNumOperands(); I != E; ++I) {
MDNode *Op = DstModFlags->getOperand(I);
ConstantInt *Behavior = mdconst::extract<ConstantInt>(Op->getOperand(0));
MDString *ID = cast<MDString>(Op->getOperand(1));
if (Behavior->getZExtValue() == Module::Require) {
Requirements.insert(cast<MDNode>(Op->getOperand(2)));
} else {
Flags[ID] = std::make_pair(Op, I);
}
}
// Merge in the flags from the source module, and also collect its set of
// requirements.
for (unsigned I = 0, E = SrcModFlags->getNumOperands(); I != E; ++I) {
MDNode *SrcOp = SrcModFlags->getOperand(I);
ConstantInt *SrcBehavior =
mdconst::extract<ConstantInt>(SrcOp->getOperand(0));
MDString *ID = cast<MDString>(SrcOp->getOperand(1));
MDNode *DstOp;
unsigned DstIndex;
std::tie(DstOp, DstIndex) = Flags.lookup(ID);
unsigned SrcBehaviorValue = SrcBehavior->getZExtValue();
// If this is a requirement, add it and continue.
if (SrcBehaviorValue == Module::Require) {
// If the destination module does not already have this requirement, add
// it.
if (Requirements.insert(cast<MDNode>(SrcOp->getOperand(2)))) {
DstModFlags->addOperand(SrcOp);
}
continue;
}
// If there is no existing flag with this ID, just add it.
if (!DstOp) {
Flags[ID] = std::make_pair(SrcOp, DstModFlags->getNumOperands());
DstModFlags->addOperand(SrcOp);
continue;
}
// Otherwise, perform a merge.
ConstantInt *DstBehavior =
mdconst::extract<ConstantInt>(DstOp->getOperand(0));
unsigned DstBehaviorValue = DstBehavior->getZExtValue();
// If either flag has override behavior, handle it first.
if (DstBehaviorValue == Module::Override) {
// Diagnose inconsistent flags which both have override behavior.
if (SrcBehaviorValue == Module::Override &&
SrcOp->getOperand(2) != DstOp->getOperand(2))
return stringErr("linking module flags '" + ID->getString() +
"': IDs have conflicting override values");
continue;
} else if (SrcBehaviorValue == Module::Override) {
// Update the destination flag to that of the source.
DstModFlags->setOperand(DstIndex, SrcOp);
Flags[ID].first = SrcOp;
continue;
}
// Diagnose inconsistent merge behavior types.
if (SrcBehaviorValue != DstBehaviorValue)
return stringErr("linking module flags '" + ID->getString() +
"': IDs have conflicting behaviors");
auto replaceDstValue = [&](MDNode *New) {
Metadata *FlagOps[] = {DstOp->getOperand(0), ID, New};
MDNode *Flag = MDNode::get(DstM.getContext(), FlagOps);
DstModFlags->setOperand(DstIndex, Flag);
Flags[ID].first = Flag;
};
// Perform the merge for standard behavior types.
switch (SrcBehaviorValue) {
case Module::Require:
case Module::Override:
llvm_unreachable("not possible");
case Module::Error: {
// Emit an error if the values differ.
if (SrcOp->getOperand(2) != DstOp->getOperand(2))
return stringErr("linking module flags '" + ID->getString() +
"': IDs have conflicting values");
continue;
}
case Module::Warning: {
// Emit a warning if the values differ.
if (SrcOp->getOperand(2) != DstOp->getOperand(2)) {
emitWarning("linking module flags '" + ID->getString() +
"': IDs have conflicting values");
}
continue;
}
case Module::Append: {
MDNode *DstValue = cast<MDNode>(DstOp->getOperand(2));
MDNode *SrcValue = cast<MDNode>(SrcOp->getOperand(2));
SmallVector<Metadata *, 8> MDs;
MDs.reserve(DstValue->getNumOperands() + SrcValue->getNumOperands());
MDs.append(DstValue->op_begin(), DstValue->op_end());
MDs.append(SrcValue->op_begin(), SrcValue->op_end());
replaceDstValue(MDNode::get(DstM.getContext(), MDs));
break;
}
case Module::AppendUnique: {
SmallSetVector<Metadata *, 16> Elts;
MDNode *DstValue = cast<MDNode>(DstOp->getOperand(2));
MDNode *SrcValue = cast<MDNode>(SrcOp->getOperand(2));
Elts.insert(DstValue->op_begin(), DstValue->op_end());
Elts.insert(SrcValue->op_begin(), SrcValue->op_end());
replaceDstValue(MDNode::get(DstM.getContext(),
makeArrayRef(Elts.begin(), Elts.end())));
break;
}
}
}
// Check all of the requirements.
for (unsigned I = 0, E = Requirements.size(); I != E; ++I) {
MDNode *Requirement = Requirements[I];
MDString *Flag = cast<MDString>(Requirement->getOperand(0));
Metadata *ReqValue = Requirement->getOperand(1);
MDNode *Op = Flags[Flag].first;
if (!Op || Op->getOperand(2) != ReqValue)
return stringErr("linking module flags '" + Flag->getString() +
"': does not have the required value");
}
return Error::success();
}
void MCObjectDisassembler::buildCFG(MCModule *Module) {
typedef std::map<uint64_t, BBInfo> BBInfoByAddrTy;
BBInfoByAddrTy BBInfos;
AddressSetTy Splits;
AddressSetTy Calls;
error_code ec;
for (symbol_iterator SI = Obj.begin_symbols(), SE = Obj.end_symbols();
SI != SE; SI.increment(ec)) {
if (ec)
break;
SymbolRef::Type SymType;
SI->getType(SymType);
if (SymType == SymbolRef::ST_Function) {
uint64_t SymAddr;
SI->getAddress(SymAddr);
SymAddr = getEffectiveLoadAddr(SymAddr);
Calls.push_back(SymAddr);
Splits.push_back(SymAddr);
}
}
assert(Module->func_begin() == Module->func_end()
&& "Module already has a CFG!");
// First, determine the basic block boundaries and call targets.
for (MCModule::atom_iterator AI = Module->atom_begin(),
AE = Module->atom_end();
AI != AE; ++AI) {
MCTextAtom *TA = dyn_cast<MCTextAtom>(*AI);
if (!TA) continue;
Calls.push_back(TA->getBeginAddr());
BBInfos[TA->getBeginAddr()].Atom = TA;
for (MCTextAtom::const_iterator II = TA->begin(), IE = TA->end();
II != IE; ++II) {
if (MIA.isTerminator(II->Inst))
Splits.push_back(II->Address + II->Size);
uint64_t Target;
if (MIA.evaluateBranch(II->Inst, II->Address, II->Size, Target)) {
if (MIA.isCall(II->Inst))
Calls.push_back(Target);
Splits.push_back(Target);
}
}
}
RemoveDupsFromAddressVector(Splits);
RemoveDupsFromAddressVector(Calls);
// Split text atoms into basic block atoms.
for (AddressSetTy::const_iterator SI = Splits.begin(), SE = Splits.end();
SI != SE; ++SI) {
MCAtom *A = Module->findAtomContaining(*SI);
if (!A) continue;
MCTextAtom *TA = cast<MCTextAtom>(A);
if (TA->getBeginAddr() == *SI)
continue;
MCTextAtom *NewAtom = TA->split(*SI);
BBInfos[NewAtom->getBeginAddr()].Atom = NewAtom;
StringRef BBName = TA->getName();
BBName = BBName.substr(0, BBName.find_last_of(':'));
NewAtom->setName((BBName + ":" + utohexstr(*SI)).str());
}
// Compute succs/preds.
for (MCModule::atom_iterator AI = Module->atom_begin(),
AE = Module->atom_end();
AI != AE; ++AI) {
MCTextAtom *TA = dyn_cast<MCTextAtom>(*AI);
if (!TA) continue;
BBInfo &CurBB = BBInfos[TA->getBeginAddr()];
const MCDecodedInst &LI = TA->back();
if (MIA.isBranch(LI.Inst)) {
uint64_t Target;
if (MIA.evaluateBranch(LI.Inst, LI.Address, LI.Size, Target))
CurBB.addSucc(BBInfos[Target]);
if (MIA.isConditionalBranch(LI.Inst))
CurBB.addSucc(BBInfos[LI.Address + LI.Size]);
} else if (!MIA.isTerminator(LI.Inst))
CurBB.addSucc(BBInfos[LI.Address + LI.Size]);
}
// Create functions and basic blocks.
for (AddressSetTy::const_iterator CI = Calls.begin(), CE = Calls.end();
CI != CE; ++CI) {
BBInfo &BBI = BBInfos[*CI];
if (!BBI.Atom) continue;
MCFunction &MCFN = *Module->createFunction(BBI.Atom->getName());
// Create MCBBs.
SmallSetVector<BBInfo*, 16> Worklist;
Worklist.insert(&BBI);
for (size_t wi = 0; wi < Worklist.size(); ++wi) {
BBInfo *BBI = Worklist[wi];
if (!BBI->Atom)
continue;
BBI->BB = &MCFN.createBlock(*BBI->Atom);
// Add all predecessors and successors to the worklist.
for (BBInfoSetTy::iterator SI = BBI->Succs.begin(), SE = BBI->Succs.end();
SI != SE; ++SI)
Worklist.insert(*SI);
for (BBInfoSetTy::iterator PI = BBI->Preds.begin(), PE = BBI->Preds.end();
PI != PE; ++PI)
Worklist.insert(*PI);
}
// Set preds/succs.
for (size_t wi = 0; wi < Worklist.size(); ++wi) {
BBInfo *BBI = Worklist[wi];
MCBasicBlock *MCBB = BBI->BB;
if (!MCBB)
continue;
for (BBInfoSetTy::iterator SI = BBI->Succs.begin(), SE = BBI->Succs.end();
SI != SE; ++SI)
if ((*SI)->BB)
MCBB->addSuccessor((*SI)->BB);
for (BBInfoSetTy::iterator PI = BBI->Preds.begin(), PE = BBI->Preds.end();
PI != PE; ++PI)
if ((*PI)->BB)
MCBB->addPredecessor((*PI)->BB);
}
}
}
PreservedAnalyses InlinerPass::run(LazyCallGraph::SCC &InitialC,
CGSCCAnalysisManager &AM, LazyCallGraph &CG,
CGSCCUpdateResult &UR) {
const ModuleAnalysisManager &MAM =
AM.getResult<ModuleAnalysisManagerCGSCCProxy>(InitialC, CG).getManager();
bool Changed = false;
assert(InitialC.size() > 0 && "Cannot handle an empty SCC!");
Module &M = *InitialC.begin()->getFunction().getParent();
ProfileSummaryInfo *PSI = MAM.getCachedResult<ProfileSummaryAnalysis>(M);
if (!ImportedFunctionsStats &&
InlinerFunctionImportStats != InlinerFunctionImportStatsOpts::No) {
ImportedFunctionsStats =
llvm::make_unique<ImportedFunctionsInliningStatistics>();
ImportedFunctionsStats->setModuleInfo(M);
}
// We use a single common worklist for calls across the entire SCC. We
// process these in-order and append new calls introduced during inlining to
// the end.
//
// Note that this particular order of processing is actually critical to
// avoid very bad behaviors. Consider *highly connected* call graphs where
// each function contains a small amonut of code and a couple of calls to
// other functions. Because the LLVM inliner is fundamentally a bottom-up
// inliner, it can handle gracefully the fact that these all appear to be
// reasonable inlining candidates as it will flatten things until they become
// too big to inline, and then move on and flatten another batch.
//
// However, when processing call edges *within* an SCC we cannot rely on this
// bottom-up behavior. As a consequence, with heavily connected *SCCs* of
// functions we can end up incrementally inlining N calls into each of
// N functions because each incremental inlining decision looks good and we
// don't have a topological ordering to prevent explosions.
//
// To compensate for this, we don't process transitive edges made immediate
// by inlining until we've done one pass of inlining across the entire SCC.
// Large, highly connected SCCs still lead to some amount of code bloat in
// this model, but it is uniformly spread across all the functions in the SCC
// and eventually they all become too large to inline, rather than
// incrementally maknig a single function grow in a super linear fashion.
SmallVector<std::pair<CallSite, int>, 16> Calls;
FunctionAnalysisManager &FAM =
AM.getResult<FunctionAnalysisManagerCGSCCProxy>(InitialC, CG)
.getManager();
// Populate the initial list of calls in this SCC.
for (auto &N : InitialC) {
auto &ORE =
FAM.getResult<OptimizationRemarkEmitterAnalysis>(N.getFunction());
// We want to generally process call sites top-down in order for
// simplifications stemming from replacing the call with the returned value
// after inlining to be visible to subsequent inlining decisions.
// FIXME: Using instructions sequence is a really bad way to do this.
// Instead we should do an actual RPO walk of the function body.
for (Instruction &I : instructions(N.getFunction()))
if (auto CS = CallSite(&I))
if (Function *Callee = CS.getCalledFunction()) {
if (!Callee->isDeclaration())
Calls.push_back({CS, -1});
else if (!isa<IntrinsicInst>(I)) {
using namespace ore;
ORE.emit([&]() {
return OptimizationRemarkMissed(DEBUG_TYPE, "NoDefinition", &I)
<< NV("Callee", Callee) << " will not be inlined into "
<< NV("Caller", CS.getCaller())
<< " because its definition is unavailable"
<< setIsVerbose();
});
}
}
}
if (Calls.empty())
return PreservedAnalyses::all();
// Capture updatable variables for the current SCC and RefSCC.
auto *C = &InitialC;
auto *RC = &C->getOuterRefSCC();
// When inlining a callee produces new call sites, we want to keep track of
// the fact that they were inlined from the callee. This allows us to avoid
// infinite inlining in some obscure cases. To represent this, we use an
// index into the InlineHistory vector.
SmallVector<std::pair<Function *, int>, 16> InlineHistory;
// Track a set vector of inlined callees so that we can augment the caller
// with all of their edges in the call graph before pruning out the ones that
// got simplified away.
SmallSetVector<Function *, 4> InlinedCallees;
// Track the dead functions to delete once finished with inlining calls. We
// defer deleting these to make it easier to handle the call graph updates.
SmallVector<Function *, 4> DeadFunctions;
// Loop forward over all of the calls. Note that we cannot cache the size as
// inlining can introduce new calls that need to be processed.
for (int i = 0; i < (int)Calls.size(); ++i) {
// We expect the calls to typically be batched with sequences of calls that
// have the same caller, so we first set up some shared infrastructure for
// this caller. We also do any pruning we can at this layer on the caller
// alone.
Function &F = *Calls[i].first.getCaller();
LazyCallGraph::Node &N = *CG.lookup(F);
if (CG.lookupSCC(N) != C)
continue;
if (F.hasFnAttribute(Attribute::OptimizeNone))
continue;
LLVM_DEBUG(dbgs() << "Inlining calls in: " << F.getName() << "\n");
// Get a FunctionAnalysisManager via a proxy for this particular node. We
// do this each time we visit a node as the SCC may have changed and as
// we're going to mutate this particular function we want to make sure the
// proxy is in place to forward any invalidation events. We can use the
// manager we get here for looking up results for functions other than this
// node however because those functions aren't going to be mutated by this
// pass.
FunctionAnalysisManager &FAM =
AM.getResult<FunctionAnalysisManagerCGSCCProxy>(*C, CG)
.getManager();
// Get the remarks emission analysis for the caller.
auto &ORE = FAM.getResult<OptimizationRemarkEmitterAnalysis>(F);
std::function<AssumptionCache &(Function &)> GetAssumptionCache =
[&](Function &F) -> AssumptionCache & {
return FAM.getResult<AssumptionAnalysis>(F);
};
auto GetBFI = [&](Function &F) -> BlockFrequencyInfo & {
return FAM.getResult<BlockFrequencyAnalysis>(F);
};
auto GetInlineCost = [&](CallSite CS) {
Function &Callee = *CS.getCalledFunction();
auto &CalleeTTI = FAM.getResult<TargetIRAnalysis>(Callee);
return getInlineCost(CS, Params, CalleeTTI, GetAssumptionCache, {GetBFI},
PSI, &ORE);
};
// Now process as many calls as we have within this caller in the sequnece.
// We bail out as soon as the caller has to change so we can update the
// call graph and prepare the context of that new caller.
bool DidInline = false;
for (; i < (int)Calls.size() && Calls[i].first.getCaller() == &F; ++i) {
int InlineHistoryID;
CallSite CS;
std::tie(CS, InlineHistoryID) = Calls[i];
Function &Callee = *CS.getCalledFunction();
if (InlineHistoryID != -1 &&
InlineHistoryIncludes(&Callee, InlineHistoryID, InlineHistory))
continue;
// Check if this inlining may repeat breaking an SCC apart that has
// already been split once before. In that case, inlining here may
// trigger infinite inlining, much like is prevented within the inliner
// itself by the InlineHistory above, but spread across CGSCC iterations
// and thus hidden from the full inline history.
if (CG.lookupSCC(*CG.lookup(Callee)) == C &&
UR.InlinedInternalEdges.count({&N, C})) {
LLVM_DEBUG(dbgs() << "Skipping inlining internal SCC edge from a node "
"previously split out of this SCC by inlining: "
<< F.getName() << " -> " << Callee.getName() << "\n");
continue;
}
Optional<InlineCost> OIC = shouldInline(CS, GetInlineCost, ORE);
// Check whether we want to inline this callsite.
if (!OIC)
continue;
// Setup the data structure used to plumb customization into the
// `InlineFunction` routine.
InlineFunctionInfo IFI(
/*cg=*/nullptr, &GetAssumptionCache, PSI,
&FAM.getResult<BlockFrequencyAnalysis>(*(CS.getCaller())),
&FAM.getResult<BlockFrequencyAnalysis>(Callee));
// Get DebugLoc to report. CS will be invalid after Inliner.
DebugLoc DLoc = CS->getDebugLoc();
BasicBlock *Block = CS.getParent();
using namespace ore;
if (!InlineFunction(CS, IFI)) {
ORE.emit([&]() {
return OptimizationRemarkMissed(DEBUG_TYPE, "NotInlined", DLoc, Block)
<< NV("Callee", &Callee) << " will not be inlined into "
<< NV("Caller", &F);
});
continue;
}
DidInline = true;
InlinedCallees.insert(&Callee);
ORE.emit([&]() {
bool AlwaysInline = OIC->isAlways();
StringRef RemarkName = AlwaysInline ? "AlwaysInline" : "Inlined";
OptimizationRemark R(DEBUG_TYPE, RemarkName, DLoc, Block);
R << NV("Callee", &Callee) << " inlined into ";
R << NV("Caller", &F);
if (AlwaysInline)
R << " with cost=always";
else {
R << " with cost=" << NV("Cost", OIC->getCost());
R << " (threshold=" << NV("Threshold", OIC->getThreshold());
R << ")";
}
return R;
});
// Add any new callsites to defined functions to the worklist.
if (!IFI.InlinedCallSites.empty()) {
int NewHistoryID = InlineHistory.size();
InlineHistory.push_back({&Callee, InlineHistoryID});
for (CallSite &CS : reverse(IFI.InlinedCallSites))
if (Function *NewCallee = CS.getCalledFunction())
if (!NewCallee->isDeclaration())
Calls.push_back({CS, NewHistoryID});
}
if (InlinerFunctionImportStats != InlinerFunctionImportStatsOpts::No)
ImportedFunctionsStats->recordInline(F, Callee);
// Merge the attributes based on the inlining.
AttributeFuncs::mergeAttributesForInlining(F, Callee);
// For local functions, check whether this makes the callee trivially
// dead. In that case, we can drop the body of the function eagerly
// which may reduce the number of callers of other functions to one,
// changing inline cost thresholds.
if (Callee.hasLocalLinkage()) {
// To check this we also need to nuke any dead constant uses (perhaps
// made dead by this operation on other functions).
Callee.removeDeadConstantUsers();
if (Callee.use_empty() && !CG.isLibFunction(Callee)) {
Calls.erase(
std::remove_if(Calls.begin() + i + 1, Calls.end(),
[&Callee](const std::pair<CallSite, int> &Call) {
return Call.first.getCaller() == &Callee;
}),
Calls.end());
// Clear the body and queue the function itself for deletion when we
// finish inlining and call graph updates.
// Note that after this point, it is an error to do anything other
// than use the callee's address or delete it.
Callee.dropAllReferences();
assert(find(DeadFunctions, &Callee) == DeadFunctions.end() &&
"Cannot put cause a function to become dead twice!");
DeadFunctions.push_back(&Callee);
}
}
}
// Back the call index up by one to put us in a good position to go around
// the outer loop.
--i;
if (!DidInline)
continue;
Changed = true;
// Add all the inlined callees' edges as ref edges to the caller. These are
// by definition trivial edges as we always have *some* transitive ref edge
// chain. While in some cases these edges are direct calls inside the
// callee, they have to be modeled in the inliner as reference edges as
// there may be a reference edge anywhere along the chain from the current
// caller to the callee that causes the whole thing to appear like
// a (transitive) reference edge that will require promotion to a call edge
// below.
for (Function *InlinedCallee : InlinedCallees) {
LazyCallGraph::Node &CalleeN = *CG.lookup(*InlinedCallee);
for (LazyCallGraph::Edge &E : *CalleeN)
RC->insertTrivialRefEdge(N, E.getNode());
}
// At this point, since we have made changes we have at least removed
// a call instruction. However, in the process we do some incremental
// simplification of the surrounding code. This simplification can
// essentially do all of the same things as a function pass and we can
// re-use the exact same logic for updating the call graph to reflect the
// change.
LazyCallGraph::SCC *OldC = C;
C = &updateCGAndAnalysisManagerForFunctionPass(CG, *C, N, AM, UR);
LLVM_DEBUG(dbgs() << "Updated inlining SCC: " << *C << "\n");
RC = &C->getOuterRefSCC();
// If this causes an SCC to split apart into multiple smaller SCCs, there
// is a subtle risk we need to prepare for. Other transformations may
// expose an "infinite inlining" opportunity later, and because of the SCC
// mutation, we will revisit this function and potentially re-inline. If we
// do, and that re-inlining also has the potentially to mutate the SCC
// structure, the infinite inlining problem can manifest through infinite
// SCC splits and merges. To avoid this, we capture the originating caller
// node and the SCC containing the call edge. This is a slight over
// approximation of the possible inlining decisions that must be avoided,
// but is relatively efficient to store.
// FIXME: This seems like a very heavyweight way of retaining the inline
// history, we should look for a more efficient way of tracking it.
if (C != OldC && llvm::any_of(InlinedCallees, [&](Function *Callee) {
return CG.lookupSCC(*CG.lookup(*Callee)) == OldC;
})) {
LLVM_DEBUG(dbgs() << "Inlined an internal call edge and split an SCC, "
"retaining this to avoid infinite inlining.\n");
UR.InlinedInternalEdges.insert({&N, OldC});
}
InlinedCallees.clear();
}
// Now that we've finished inlining all of the calls across this SCC, delete
// all of the trivially dead functions, updating the call graph and the CGSCC
// pass manager in the process.
//
// Note that this walks a pointer set which has non-deterministic order but
// that is OK as all we do is delete things and add pointers to unordered
// sets.
for (Function *DeadF : DeadFunctions) {
// Get the necessary information out of the call graph and nuke the
// function there. Also, cclear out any cached analyses.
auto &DeadC = *CG.lookupSCC(*CG.lookup(*DeadF));
FunctionAnalysisManager &FAM =
AM.getResult<FunctionAnalysisManagerCGSCCProxy>(DeadC, CG)
.getManager();
FAM.clear(*DeadF, DeadF->getName());
AM.clear(DeadC, DeadC.getName());
auto &DeadRC = DeadC.getOuterRefSCC();
CG.removeDeadFunction(*DeadF);
// Mark the relevant parts of the call graph as invalid so we don't visit
// them.
UR.InvalidatedSCCs.insert(&DeadC);
UR.InvalidatedRefSCCs.insert(&DeadRC);
// And delete the actual function from the module.
M.getFunctionList().erase(DeadF);
}
if (!Changed)
return PreservedAnalyses::all();
// Even if we change the IR, we update the core CGSCC data structures and so
// can preserve the proxy to the function analysis manager.
PreservedAnalyses PA;
PA.preserve<FunctionAnalysisManagerCGSCCProxy>();
return PA;
}
/// propagateSiblingValue - Propagate the value in SVI to dependents if it is
/// known. Otherwise remember the dependency for later.
///
/// @param SVIIter SibValues entry to propagate.
/// @param VNI Dependent value, or NULL to propagate to all saved dependents.
void InlineSpiller::propagateSiblingValue(SibValueMap::iterator SVIIter,
VNInfo *VNI) {
SibValueMap::value_type *SVI = &*SVIIter;
// When VNI is non-NULL, add it to SVI's deps, and only propagate to that.
TinyPtrVector<VNInfo*> FirstDeps;
if (VNI) {
FirstDeps.push_back(VNI);
SVI->second.Deps.push_back(VNI);
}
// Has the value been completely determined yet? If not, defer propagation.
if (!SVI->second.hasDef())
return;
// Work list of values to propagate.
SmallSetVector<SibValueMap::value_type *, 8> WorkList;
WorkList.insert(SVI);
do {
SVI = WorkList.pop_back_val();
TinyPtrVector<VNInfo*> *Deps = VNI ? &FirstDeps : &SVI->second.Deps;
VNI = nullptr;
SibValueInfo &SV = SVI->second;
if (!SV.SpillMBB)
SV.SpillMBB = LIS.getMBBFromIndex(SV.SpillVNI->def);
DEBUG(dbgs() << " prop to " << Deps->size() << ": "
<< SVI->first->id << '@' << SVI->first->def << ":\t" << SV);
assert(SV.hasDef() && "Propagating undefined value");
// Should this value be propagated as a preferred spill candidate? We don't
// propagate values of registers that are about to spill.
bool PropSpill = !DisableHoisting && !isRegToSpill(SV.SpillReg);
unsigned SpillDepth = ~0u;
for (TinyPtrVector<VNInfo*>::iterator DepI = Deps->begin(),
DepE = Deps->end(); DepI != DepE; ++DepI) {
SibValueMap::iterator DepSVI = SibValues.find(*DepI);
assert(DepSVI != SibValues.end() && "Dependent value not in SibValues");
SibValueInfo &DepSV = DepSVI->second;
if (!DepSV.SpillMBB)
DepSV.SpillMBB = LIS.getMBBFromIndex(DepSV.SpillVNI->def);
bool Changed = false;
// Propagate defining instruction.
if (!DepSV.hasDef()) {
Changed = true;
DepSV.DefMI = SV.DefMI;
DepSV.DefByOrigPHI = SV.DefByOrigPHI;
}
// Propagate AllDefsAreReloads. For PHI values, this computes an AND of
// all predecessors.
if (!SV.AllDefsAreReloads && DepSV.AllDefsAreReloads) {
Changed = true;
DepSV.AllDefsAreReloads = false;
}
// Propagate best spill value.
if (PropSpill && SV.SpillVNI != DepSV.SpillVNI) {
if (SV.SpillMBB == DepSV.SpillMBB) {
// DepSV is in the same block. Hoist when dominated.
if (DepSV.KillsSource && SV.SpillVNI->def < DepSV.SpillVNI->def) {
// This is an alternative def earlier in the same MBB.
// Hoist the spill as far as possible in SpillMBB. This can ease
// register pressure:
//
// x = def
// y = use x
// s = copy x
//
// Hoisting the spill of s to immediately after the def removes the
// interference between x and y:
//
// x = def
// spill x
// y = use x<kill>
//
// This hoist only helps when the DepSV copy kills its source.
Changed = true;
DepSV.SpillReg = SV.SpillReg;
DepSV.SpillVNI = SV.SpillVNI;
DepSV.SpillMBB = SV.SpillMBB;
}
} else {
// DepSV is in a different block.
if (SpillDepth == ~0u)
SpillDepth = Loops.getLoopDepth(SV.SpillMBB);
// Also hoist spills to blocks with smaller loop depth, but make sure
// that the new value dominates. Non-phi dependents are always
// dominated, phis need checking.
const BranchProbability MarginProb(4, 5); // 80%
// Hoist a spill to outer loop if there are multiple dependents (it
// can be beneficial if more than one dependents are hoisted) or
// if DepSV (the hoisting source) is hotter than SV (the hoisting
// destination) (we add a 80% margin to bias a little towards
// loop depth).
bool HoistCondition =
(MBFI.getBlockFreq(DepSV.SpillMBB) >=
(MBFI.getBlockFreq(SV.SpillMBB) * MarginProb)) ||
Deps->size() > 1;
if ((Loops.getLoopDepth(DepSV.SpillMBB) > SpillDepth) &&
HoistCondition &&
(!DepSVI->first->isPHIDef() ||
MDT.dominates(SV.SpillMBB, DepSV.SpillMBB))) {
Changed = true;
DepSV.SpillReg = SV.SpillReg;
DepSV.SpillVNI = SV.SpillVNI;
DepSV.SpillMBB = SV.SpillMBB;
}
}
}
if (!Changed)
continue;
// Something changed in DepSVI. Propagate to dependents.
WorkList.insert(&*DepSVI);
DEBUG(dbgs() << " update " << DepSVI->first->id << '@'
<< DepSVI->first->def << " to:\t" << DepSV);
}
} while (!WorkList.empty());
}
MDNode *MDNode::getMostGenericTBAA(MDNode *A, MDNode *B) {
if (!A || !B)
return nullptr;
if (A == B)
return A;
// For struct-path aware TBAA, we use the access type of the tag.
bool StructPath = isStructPathTBAA(A) && isStructPathTBAA(B);
if (StructPath) {
A = cast_or_null<MDNode>(A->getOperand(1));
if (!A)
return nullptr;
B = cast_or_null<MDNode>(B->getOperand(1));
if (!B)
return nullptr;
}
SmallSetVector<MDNode *, 4> PathA;
MDNode *T = A;
while (T) {
if (PathA.count(T))
report_fatal_error("Cycle found in TBAA metadata.");
PathA.insert(T);
T = T->getNumOperands() >= 2 ? cast_or_null<MDNode>(T->getOperand(1))
: nullptr;
}
SmallSetVector<MDNode *, 4> PathB;
T = B;
while (T) {
if (PathB.count(T))
report_fatal_error("Cycle found in TBAA metadata.");
PathB.insert(T);
T = T->getNumOperands() >= 2 ? cast_or_null<MDNode>(T->getOperand(1))
: nullptr;
}
int IA = PathA.size() - 1;
int IB = PathB.size() - 1;
MDNode *Ret = nullptr;
while (IA >= 0 && IB >= 0) {
if (PathA[IA] == PathB[IB])
Ret = PathA[IA];
else
break;
--IA;
--IB;
}
if (!StructPath)
return Ret;
if (!Ret)
return nullptr;
// We need to convert from a type node to a tag node.
Type *Int64 = IntegerType::get(A->getContext(), 64);
Metadata *Ops[3] = {Ret, Ret,
ConstantAsMetadata::get(ConstantInt::get(Int64, 0))};
return MDNode::get(A->getContext(), Ops);
}
/// handleEndBlock - Remove dead stores to stack-allocated locations in the
/// function end block. Ex:
/// %A = alloca i32
/// ...
/// store i32 1, i32* %A
/// ret void
bool DSE::handleEndBlock(BasicBlock &BB) {
bool MadeChange = false;
// Keep track of all of the stack objects that are dead at the end of the
// function.
SmallSetVector<Value*, 16> DeadStackObjects;
// Find all of the alloca'd pointers in the entry block.
BasicBlock *Entry = BB.getParent()->begin();
for (BasicBlock::iterator I = Entry->begin(), E = Entry->end(); I != E; ++I) {
if (isa<AllocaInst>(I))
DeadStackObjects.insert(I);
// Okay, so these are dead heap objects, but if the pointer never escapes
// then it's leaked by this function anyways.
else if (isAllocLikeFn(I, TLI) && !PointerMayBeCaptured(I, true, true))
DeadStackObjects.insert(I);
}
// Treat byval or inalloca arguments the same, stores to them are dead at the
// end of the function.
for (Function::arg_iterator AI = BB.getParent()->arg_begin(),
AE = BB.getParent()->arg_end(); AI != AE; ++AI)
if (AI->hasByValOrInAllocaAttr())
DeadStackObjects.insert(AI);
// Scan the basic block backwards
for (BasicBlock::iterator BBI = BB.end(); BBI != BB.begin(); ){
--BBI;
// If we find a store, check to see if it points into a dead stack value.
if (hasMemoryWrite(BBI, TLI) && isRemovable(BBI)) {
// See through pointer-to-pointer bitcasts
SmallVector<Value *, 4> Pointers;
GetUnderlyingObjects(getStoredPointerOperand(BBI), Pointers);
// Stores to stack values are valid candidates for removal.
bool AllDead = true;
for (SmallVectorImpl<Value *>::iterator I = Pointers.begin(),
E = Pointers.end(); I != E; ++I)
if (!DeadStackObjects.count(*I)) {
AllDead = false;
break;
}
if (AllDead) {
Instruction *Dead = BBI++;
DEBUG(dbgs() << "DSE: Dead Store at End of Block:\n DEAD: "
<< *Dead << "\n Objects: ";
for (SmallVectorImpl<Value *>::iterator I = Pointers.begin(),
E = Pointers.end(); I != E; ++I) {
dbgs() << **I;
if (std::next(I) != E)
dbgs() << ", ";
}
dbgs() << '\n');
// DCE instructions only used to calculate that store.
DeleteDeadInstruction(Dead, *MD, TLI, &DeadStackObjects);
++NumFastStores;
MadeChange = true;
continue;
}
}
// Remove any dead non-memory-mutating instructions.
if (isInstructionTriviallyDead(BBI, TLI)) {
Instruction *Inst = BBI++;
DeleteDeadInstruction(Inst, *MD, TLI, &DeadStackObjects);
++NumFastOther;
MadeChange = true;
continue;
}
if (isa<AllocaInst>(BBI)) {
// Remove allocas from the list of dead stack objects; there can't be
// any references before the definition.
DeadStackObjects.remove(BBI);
continue;
}
if (CallSite CS = cast<Value>(BBI)) {
// Remove allocation function calls from the list of dead stack objects;
// there can't be any references before the definition.
if (isAllocLikeFn(BBI, TLI))
DeadStackObjects.remove(BBI);
// If this call does not access memory, it can't be loading any of our
// pointers.
if (AA->doesNotAccessMemory(CS))
continue;
// If the call might load from any of our allocas, then any store above
// the call is live.
DeadStackObjects.remove_if([&](Value *I) {
// See if the call site touches the value.
AliasAnalysis::ModRefResult A =
AA->getModRefInfo(CS, I, getPointerSize(I, *AA));
return A == AliasAnalysis::ModRef || A == AliasAnalysis::Ref;
});
// If all of the allocas were clobbered by the call then we're not going
// to find anything else to process.
if (DeadStackObjects.empty())
break;
continue;
}
AliasAnalysis::Location LoadedLoc;
// If we encounter a use of the pointer, it is no longer considered dead
if (LoadInst *L = dyn_cast<LoadInst>(BBI)) {
if (!L->isUnordered()) // Be conservative with atomic/volatile load
break;
LoadedLoc = AA->getLocation(L);
} else if (VAArgInst *V = dyn_cast<VAArgInst>(BBI)) {
LoadedLoc = AA->getLocation(V);
} else if (MemTransferInst *MTI = dyn_cast<MemTransferInst>(BBI)) {
LoadedLoc = AA->getLocationForSource(MTI);
} else if (!BBI->mayReadFromMemory()) {
// Instruction doesn't read memory. Note that stores that weren't removed
// above will hit this case.
continue;
} else {
// Unknown inst; assume it clobbers everything.
break;
}
// Remove any allocas from the DeadPointer set that are loaded, as this
// makes any stores above the access live.
RemoveAccessedObjects(LoadedLoc, DeadStackObjects);
// If all of the allocas were clobbered by the access then we're not going
// to find anything else to process.
if (DeadStackObjects.empty())
break;
}
/// setupEntryBlockAndCallSites - Setup the entry block by creating and filling
/// the function context and marking the call sites with the appropriate
/// values. These values are used by the DWARF EH emitter.
bool SjLjEHPass::setupEntryBlockAndCallSites(Function &F) {
SmallVector<ReturnInst*, 16> Returns;
SmallVector<InvokeInst*, 16> Invokes;
SmallSetVector<LandingPadInst*, 16> LPads;
// Look through the terminators of the basic blocks to find invokes.
for (Function::iterator BB = F.begin(), E = F.end(); BB != E; ++BB)
if (InvokeInst *II = dyn_cast<InvokeInst>(BB->getTerminator())) {
Invokes.push_back(II);
LPads.insert(II->getUnwindDest()->getLandingPadInst());
} else if (ReturnInst *RI = dyn_cast<ReturnInst>(BB->getTerminator())) {
Returns.push_back(RI);
}
if (Invokes.empty()) return false;
NumInvokes += Invokes.size();
lowerIncomingArguments(F);
lowerAcrossUnwindEdges(F, Invokes);
Value *FuncCtx =
setupFunctionContext(F, makeArrayRef(LPads.begin(), LPads.end()));
BasicBlock *EntryBB = F.begin();
Type *Int32Ty = Type::getInt32Ty(F.getContext());
Value *Idxs[2] = {
ConstantInt::get(Int32Ty, 0), 0
};
// Get a reference to the jump buffer.
Idxs[1] = ConstantInt::get(Int32Ty, 5);
Value *JBufPtr = GetElementPtrInst::Create(FuncCtx, Idxs, "jbuf_gep",
EntryBB->getTerminator());
// Save the frame pointer.
Idxs[1] = ConstantInt::get(Int32Ty, 0);
Value *FramePtr = GetElementPtrInst::Create(JBufPtr, Idxs, "jbuf_fp_gep",
EntryBB->getTerminator());
Value *Val = CallInst::Create(FrameAddrFn,
ConstantInt::get(Int32Ty, 0),
"fp",
EntryBB->getTerminator());
new StoreInst(Val, FramePtr, true, EntryBB->getTerminator());
// Save the stack pointer.
Idxs[1] = ConstantInt::get(Int32Ty, 2);
Value *StackPtr = GetElementPtrInst::Create(JBufPtr, Idxs, "jbuf_sp_gep",
EntryBB->getTerminator());
Val = CallInst::Create(StackAddrFn, "sp", EntryBB->getTerminator());
new StoreInst(Val, StackPtr, true, EntryBB->getTerminator());
// Call the setjmp instrinsic. It fills in the rest of the jmpbuf.
Value *SetjmpArg = CastInst::Create(Instruction::BitCast, JBufPtr,
Type::getInt8PtrTy(F.getContext()), "",
EntryBB->getTerminator());
CallInst::Create(BuiltinSetjmpFn, SetjmpArg, "", EntryBB->getTerminator());
// Store a pointer to the function context so that the back-end will know
// where to look for it.
Value *FuncCtxArg = CastInst::Create(Instruction::BitCast, FuncCtx,
Type::getInt8PtrTy(F.getContext()), "",
EntryBB->getTerminator());
CallInst::Create(FuncCtxFn, FuncCtxArg, "", EntryBB->getTerminator());
// At this point, we are all set up, update the invoke instructions to mark
// their call_site values.
for (unsigned I = 0, E = Invokes.size(); I != E; ++I) {
insertCallSiteStore(Invokes[I], I + 1);
ConstantInt *CallSiteNum =
ConstantInt::get(Type::getInt32Ty(F.getContext()), I + 1);
// Record the call site value for the back end so it stays associated with
// the invoke.
CallInst::Create(CallSiteFn, CallSiteNum, "", Invokes[I]);
}
// Mark call instructions that aren't nounwind as no-action (call_site ==
// -1). Skip the entry block, as prior to then, no function context has been
// created for this function and any unexpected exceptions thrown will go
// directly to the caller's context, which is what we want anyway, so no need
// to do anything here.
for (Function::iterator BB = F.begin(), E = F.end(); ++BB != E;)
for (BasicBlock::iterator I = BB->begin(), end = BB->end(); I != end; ++I)
if (CallInst *CI = dyn_cast<CallInst>(I)) {
if (!CI->doesNotThrow())
insertCallSiteStore(CI, -1);
} else if (ResumeInst *RI = dyn_cast<ResumeInst>(I)) {
insertCallSiteStore(RI, -1);
}
// Register the function context and make sure it's known to not throw
CallInst *Register = CallInst::Create(RegisterFn, FuncCtx, "",
EntryBB->getTerminator());
Register->setDoesNotThrow();
// Following any allocas not in the entry block, update the saved SP in the
// jmpbuf to the new value.
for (Function::iterator BB = F.begin(), E = F.end(); BB != E; ++BB) {
if (BB == F.begin())
continue;
for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I) {
if (CallInst *CI = dyn_cast<CallInst>(I)) {
if (CI->getCalledFunction() != StackRestoreFn)
continue;
} else if (!isa<AllocaInst>(I)) {
continue;
}
Instruction *StackAddr = CallInst::Create(StackAddrFn, "sp");
StackAddr->insertAfter(I);
Instruction *StoreStackAddr = new StoreInst(StackAddr, StackPtr, true);
StoreStackAddr->insertAfter(StackAddr);
}
}
// Finally, for any returns from this function, if this function contains an
// invoke, add a call to unregister the function context.
for (unsigned I = 0, E = Returns.size(); I != E; ++I)
CallInst::Create(UnregisterFn, FuncCtx, "", Returns[I]);
return true;
}
/// This works like CloneAndPruneFunctionInto, except that it does not clone the
/// entire function. Instead it starts at an instruction provided by the caller
/// and copies (and prunes) only the code reachable from that instruction.
void llvm::CloneAndPruneIntoFromInst(Function *NewFunc, const Function *OldFunc,
const Instruction *StartingInst,
ValueToValueMapTy &VMap,
bool ModuleLevelChanges,
SmallVectorImpl<ReturnInst *> &Returns,
const char *NameSuffix,
ClonedCodeInfo *CodeInfo) {
assert(NameSuffix && "NameSuffix cannot be null!");
ValueMapTypeRemapper *TypeMapper = nullptr;
ValueMaterializer *Materializer = nullptr;
#ifndef NDEBUG
// If the cloning starts at the beginning of the function, verify that
// the function arguments are mapped.
if (!StartingInst)
for (const Argument &II : OldFunc->args())
assert(VMap.count(&II) && "No mapping from source argument specified!");
#endif
PruningFunctionCloner PFC(NewFunc, OldFunc, VMap, ModuleLevelChanges,
NameSuffix, CodeInfo);
const BasicBlock *StartingBB;
if (StartingInst)
StartingBB = StartingInst->getParent();
else {
StartingBB = &OldFunc->getEntryBlock();
StartingInst = &StartingBB->front();
}
// Clone the entry block, and anything recursively reachable from it.
std::vector<const BasicBlock*> CloneWorklist;
PFC.CloneBlock(StartingBB, StartingInst->getIterator(), CloneWorklist);
while (!CloneWorklist.empty()) {
const BasicBlock *BB = CloneWorklist.back();
CloneWorklist.pop_back();
PFC.CloneBlock(BB, BB->begin(), CloneWorklist);
}
// Loop over all of the basic blocks in the old function. If the block was
// reachable, we have cloned it and the old block is now in the value map:
// insert it into the new function in the right order. If not, ignore it.
//
// Defer PHI resolution until rest of function is resolved.
SmallVector<const PHINode*, 16> PHIToResolve;
for (const BasicBlock &BI : *OldFunc) {
Value *V = VMap.lookup(&BI);
BasicBlock *NewBB = cast_or_null<BasicBlock>(V);
if (!NewBB) continue; // Dead block.
// Add the new block to the new function.
NewFunc->getBasicBlockList().push_back(NewBB);
// Handle PHI nodes specially, as we have to remove references to dead
// blocks.
for (BasicBlock::const_iterator I = BI.begin(), E = BI.end(); I != E; ++I) {
// PHI nodes may have been remapped to non-PHI nodes by the caller or
// during the cloning process.
if (const PHINode *PN = dyn_cast<PHINode>(I)) {
if (isa<PHINode>(VMap[PN]))
PHIToResolve.push_back(PN);
else
break;
} else {
break;
}
}
// Finally, remap the terminator instructions, as those can't be remapped
// until all BBs are mapped.
RemapInstruction(NewBB->getTerminator(), VMap,
ModuleLevelChanges ? RF_None : RF_NoModuleLevelChanges,
TypeMapper, Materializer);
}
// Defer PHI resolution until rest of function is resolved, PHI resolution
// requires the CFG to be up-to-date.
for (unsigned phino = 0, e = PHIToResolve.size(); phino != e; ) {
const PHINode *OPN = PHIToResolve[phino];
unsigned NumPreds = OPN->getNumIncomingValues();
const BasicBlock *OldBB = OPN->getParent();
BasicBlock *NewBB = cast<BasicBlock>(VMap[OldBB]);
// Map operands for blocks that are live and remove operands for blocks
// that are dead.
for (; phino != PHIToResolve.size() &&
PHIToResolve[phino]->getParent() == OldBB; ++phino) {
OPN = PHIToResolve[phino];
PHINode *PN = cast<PHINode>(VMap[OPN]);
for (unsigned pred = 0, e = NumPreds; pred != e; ++pred) {
Value *V = VMap.lookup(PN->getIncomingBlock(pred));
if (BasicBlock *MappedBlock = cast_or_null<BasicBlock>(V)) {
Value *InVal = MapValue(PN->getIncomingValue(pred),
VMap,
ModuleLevelChanges ? RF_None : RF_NoModuleLevelChanges);
assert(InVal && "Unknown input value?");
PN->setIncomingValue(pred, InVal);
PN->setIncomingBlock(pred, MappedBlock);
} else {
PN->removeIncomingValue(pred, false);
--pred; // Revisit the next entry.
--e;
}
}
}
// The loop above has removed PHI entries for those blocks that are dead
// and has updated others. However, if a block is live (i.e. copied over)
// but its terminator has been changed to not go to this block, then our
// phi nodes will have invalid entries. Update the PHI nodes in this
// case.
PHINode *PN = cast<PHINode>(NewBB->begin());
NumPreds = std::distance(pred_begin(NewBB), pred_end(NewBB));
if (NumPreds != PN->getNumIncomingValues()) {
assert(NumPreds < PN->getNumIncomingValues());
// Count how many times each predecessor comes to this block.
std::map<BasicBlock*, unsigned> PredCount;
for (pred_iterator PI = pred_begin(NewBB), E = pred_end(NewBB);
PI != E; ++PI)
--PredCount[*PI];
// Figure out how many entries to remove from each PHI.
for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i)
++PredCount[PN->getIncomingBlock(i)];
// At this point, the excess predecessor entries are positive in the
// map. Loop over all of the PHIs and remove excess predecessor
// entries.
BasicBlock::iterator I = NewBB->begin();
for (; (PN = dyn_cast<PHINode>(I)); ++I) {
for (const auto &PCI : PredCount) {
BasicBlock *Pred = PCI.first;
for (unsigned NumToRemove = PCI.second; NumToRemove; --NumToRemove)
PN->removeIncomingValue(Pred, false);
}
}
}
// If the loops above have made these phi nodes have 0 or 1 operand,
// replace them with undef or the input value. We must do this for
// correctness, because 0-operand phis are not valid.
PN = cast<PHINode>(NewBB->begin());
if (PN->getNumIncomingValues() == 0) {
BasicBlock::iterator I = NewBB->begin();
BasicBlock::const_iterator OldI = OldBB->begin();
while ((PN = dyn_cast<PHINode>(I++))) {
Value *NV = UndefValue::get(PN->getType());
PN->replaceAllUsesWith(NV);
assert(VMap[&*OldI] == PN && "VMap mismatch");
VMap[&*OldI] = NV;
PN->eraseFromParent();
++OldI;
}
}
}
// Make a second pass over the PHINodes now that all of them have been
// remapped into the new function, simplifying the PHINode and performing any
// recursive simplifications exposed. This will transparently update the
// WeakVH in the VMap. Notably, we rely on that so that if we coalesce
// two PHINodes, the iteration over the old PHIs remains valid, and the
// mapping will just map us to the new node (which may not even be a PHI
// node).
const DataLayout &DL = NewFunc->getParent()->getDataLayout();
SmallSetVector<const Value *, 8> Worklist;
for (unsigned Idx = 0, Size = PHIToResolve.size(); Idx != Size; ++Idx)
if (isa<PHINode>(VMap[PHIToResolve[Idx]]))
Worklist.insert(PHIToResolve[Idx]);
// Note that we must test the size on each iteration, the worklist can grow.
for (unsigned Idx = 0; Idx != Worklist.size(); ++Idx) {
const Value *OrigV = Worklist[Idx];
auto *I = dyn_cast_or_null<Instruction>(VMap.lookup(OrigV));
if (!I)
continue;
// See if this instruction simplifies.
Value *SimpleV = SimplifyInstruction(I, DL);
if (!SimpleV)
continue;
// Stash away all the uses of the old instruction so we can check them for
// recursive simplifications after a RAUW. This is cheaper than checking all
// uses of To on the recursive step in most cases.
for (const User *U : OrigV->users())
Worklist.insert(cast<Instruction>(U));
// Replace the instruction with its simplified value.
I->replaceAllUsesWith(SimpleV);
// If the original instruction had no side effects, remove it.
if (isInstructionTriviallyDead(I))
I->eraseFromParent();
else
VMap[OrigV] = I;
}
// Now that the inlined function body has been fully constructed, go through
// and zap unconditional fall-through branches. This happens all the time when
// specializing code: code specialization turns conditional branches into
// uncond branches, and this code folds them.
Function::iterator Begin = cast<BasicBlock>(VMap[StartingBB])->getIterator();
Function::iterator I = Begin;
while (I != NewFunc->end()) {
// Check if this block has become dead during inlining or other
// simplifications. Note that the first block will appear dead, as it has
// not yet been wired up properly.
if (I != Begin && (pred_begin(&*I) == pred_end(&*I) ||
I->getSinglePredecessor() == &*I)) {
BasicBlock *DeadBB = &*I++;
DeleteDeadBlock(DeadBB);
continue;
}
// We need to simplify conditional branches and switches with a constant
// operand. We try to prune these out when cloning, but if the
// simplification required looking through PHI nodes, those are only
// available after forming the full basic block. That may leave some here,
// and we still want to prune the dead code as early as possible.
ConstantFoldTerminator(&*I);
BranchInst *BI = dyn_cast<BranchInst>(I->getTerminator());
if (!BI || BI->isConditional()) { ++I; continue; }
BasicBlock *Dest = BI->getSuccessor(0);
if (!Dest->getSinglePredecessor()) {
++I; continue;
}
// We shouldn't be able to get single-entry PHI nodes here, as instsimplify
// above should have zapped all of them..
assert(!isa<PHINode>(Dest->begin()));
// We know all single-entry PHI nodes in the inlined function have been
// removed, so we just need to splice the blocks.
BI->eraseFromParent();
// Make all PHI nodes that referred to Dest now refer to I as their source.
Dest->replaceAllUsesWith(&*I);
// Move all the instructions in the succ to the pred.
I->getInstList().splice(I->end(), Dest->getInstList());
// Remove the dest block.
Dest->eraseFromParent();
// Do not increment I, iteratively merge all things this block branches to.
}
// Make a final pass over the basic blocks from the old function to gather
// any return instructions which survived folding. We have to do this here
// because we can iteratively remove and merge returns above.
for (Function::iterator I = cast<BasicBlock>(VMap[StartingBB])->getIterator(),
E = NewFunc->end();
I != E; ++I)
if (ReturnInst *RI = dyn_cast<ReturnInst>(I->getTerminator()))
Returns.push_back(RI);
}
/// performStoreOnlyObjectElimination - Scan the graph of uses of the specified
/// object allocation. If the object does not escape and is only stored to
/// (this happens because GVN and other optimizations hoists forward substitutes
/// all stores to the object to eliminate all loads from it), then zap the
/// object and all accesses related to it.
static bool performStoreOnlyObjectElimination(CallInst &Allocation,
BasicBlock::iterator &BBI) {
DtorKind DtorInfo = analyzeDestructor(Allocation.getArgOperand(0));
// We can't delete the object if its destructor has side effects.
if (DtorInfo != DtorKind::NoSideEffects)
return false;
// Do a depth first search exploring all of the uses of the object pointer,
// following through casts, pointer adjustments etc. If we find any loads or
// any escape sites of the object, we give up. If we succeed in walking the
// entire graph of uses, we can remove the resultant set.
SmallSetVector<Instruction*, 16> InvolvedInstructions;
SmallVector<Instruction*, 16> Worklist;
Worklist.push_back(&Allocation);
// Stores - Keep track of all of the store instructions we see.
SmallVector<StoreInst*, 16> Stores;
while (!Worklist.empty()) {
Instruction *I = Worklist.pop_back_val();
// Insert the instruction into our InvolvedInstructions set. If we have
// already seen it, then don't reprocess all of the uses.
if (!InvolvedInstructions.insert(I)) continue;
// Okay, this is the first time we've seen this instruction, proceed.
switch (classifyInstruction(*I)) {
// These instructions should not reach here based on the pass ordering.
// i.e. LLVMARCOpt -> LLVMContractOpt.
case RT_RetainN:
case RT_UnknownRetainN:
case RT_BridgeRetainN:
case RT_ReleaseN:
case RT_UnknownReleaseN:
case RT_BridgeReleaseN:
llvm_unreachable("These are only created by LLVMARCContract !");
case RT_AllocObject:
// If this is a different swift_allocObject than we started with, then
// there is some computation feeding into a size or alignment computation
// that we have to keep... unless we can delete *that* entire object as
// well.
break;
case RT_NoMemoryAccessed:
// If no memory is accessed, then something is being done with the
// pointer: maybe it is bitcast or GEP'd. Since there are no side effects,
// it is perfectly fine to delete this instruction if all uses of the
// instruction are also eliminable.
if (I->mayHaveSideEffects() || isa<TerminatorInst>(I))
return false;
break;
case RT_Release:
case RT_Retain:
case RT_FixLifetime:
case RT_CheckUnowned:
// It is perfectly fine to eliminate various retains and releases of this
// object: we are zapping all accesses or none.
break;
// If this is an unknown instruction, we have more interesting things to
// consider.
case RT_Unknown:
case RT_ObjCRelease:
case RT_ObjCRetain:
case RT_UnknownRetain:
case RT_UnknownRelease:
case RT_BridgeRetain:
case RT_BridgeRelease:
case RT_RetainUnowned:
// Otherwise, this really is some unhandled instruction. Bail out.
return false;
}
// Okay, if we got here, the instruction can be eaten so-long as all of its
// uses can be. Scan through the uses and add them to the worklist for
// recursive processing.
for (auto UI = I->user_begin(), E = I->user_end(); UI != E; ++UI) {
Instruction *User = cast<Instruction>(*UI);
// Handle stores as a special case here: we want to make sure that the
// object is being stored *to*, not itself being stored (which would be an
// escape point). Since stores themselves don't have any uses, we can
// short-cut the classification scheme above.
if (StoreInst *SI = dyn_cast<StoreInst>(User)) {
// If this is a store *to* the object, we can zap it.
if (UI.getUse().getOperandNo() == StoreInst::getPointerOperandIndex()) {
InvolvedInstructions.insert(SI);
continue;
}
// Otherwise, using the object as a source (or size) is an escape.
return false;
}
if (MemIntrinsic *MI = dyn_cast<MemIntrinsic>(User)) {
// If this is a memset/memcpy/memmove *to* the object, we can zap it.
if (UI.getUse().getOperandNo() == 0) {
InvolvedInstructions.insert(MI);
continue;
}
// Otherwise, using the object as a source (or size) is an escape.
return false;
}
// Otherwise, normal instructions just go on the worklist for processing.
Worklist.push_back(User);
}
}
// Ok, we succeeded! This means we can zap all of the instructions that use
// the object. One thing we have to be careful of is to make sure that we
// don't invalidate "BBI" (the iterator the outer walk of the optimization
// pass is using, and indicates the next instruction to process). This would
// happen if we delete the instruction it is pointing to. Advance the
// iterator if that would happen.
while (InvolvedInstructions.count(&*BBI))
++BBI;
// Zap all of the instructions.
for (auto I : InvolvedInstructions) {
if (!I->use_empty())
I->replaceAllUsesWith(UndefValue::get(I->getType()));
I->eraseFromParent();
}
++NumStoreOnlyObjectsEliminated;
return true;
}
/// \brief Figure out if the loop is worth full unrolling.
///
/// Complete loop unrolling can make some loads constant, and we need to know
/// if that would expose any further optimization opportunities. This routine
/// estimates this optimization. It computes cost of unrolled loop
/// (UnrolledCost) and dynamic cost of the original loop (RolledDynamicCost). By
/// dynamic cost we mean that we won't count costs of blocks that are known not
/// to be executed (i.e. if we have a branch in the loop and we know that at the
/// given iteration its condition would be resolved to true, we won't add up the
/// cost of the 'false'-block).
/// \returns Optional value, holding the RolledDynamicCost and UnrolledCost. If
/// the analysis failed (no benefits expected from the unrolling, or the loop is
/// too big to analyze), the returned value is None.
static Optional<EstimatedUnrollCost>
analyzeLoopUnrollCost(const Loop *L, unsigned TripCount, DominatorTree &DT,
ScalarEvolution &SE, const TargetTransformInfo &TTI,
int MaxUnrolledLoopSize) {
// We want to be able to scale offsets by the trip count and add more offsets
// to them without checking for overflows, and we already don't want to
// analyze *massive* trip counts, so we force the max to be reasonably small.
assert(UnrollMaxIterationsCountToAnalyze < (INT_MAX / 2) &&
"The unroll iterations max is too large!");
// Only analyze inner loops. We can't properly estimate cost of nested loops
// and we won't visit inner loops again anyway.
if (!L->empty())
return None;
// Don't simulate loops with a big or unknown tripcount
if (!UnrollMaxIterationsCountToAnalyze || !TripCount ||
TripCount > UnrollMaxIterationsCountToAnalyze)
return None;
SmallSetVector<BasicBlock *, 16> BBWorklist;
SmallSetVector<std::pair<BasicBlock *, BasicBlock *>, 4> ExitWorklist;
DenseMap<Value *, Constant *> SimplifiedValues;
SmallVector<std::pair<Value *, Constant *>, 4> SimplifiedInputValues;
// The estimated cost of the unrolled form of the loop. We try to estimate
// this by simplifying as much as we can while computing the estimate.
int UnrolledCost = 0;
// We also track the estimated dynamic (that is, actually executed) cost in
// the rolled form. This helps identify cases when the savings from unrolling
// aren't just exposing dead control flows, but actual reduced dynamic
// instructions due to the simplifications which we expect to occur after
// unrolling.
int RolledDynamicCost = 0;
// We track the simplification of each instruction in each iteration. We use
// this to recursively merge costs into the unrolled cost on-demand so that
// we don't count the cost of any dead code. This is essentially a map from
// <instruction, int> to <bool, bool>, but stored as a densely packed struct.
DenseSet<UnrolledInstState, UnrolledInstStateKeyInfo> InstCostMap;
// A small worklist used to accumulate cost of instructions from each
// observable and reached root in the loop.
SmallVector<Instruction *, 16> CostWorklist;
// PHI-used worklist used between iterations while accumulating cost.
SmallVector<Instruction *, 4> PHIUsedList;
// Helper function to accumulate cost for instructions in the loop.
auto AddCostRecursively = [&](Instruction &RootI, int Iteration) {
assert(Iteration >= 0 && "Cannot have a negative iteration!");
assert(CostWorklist.empty() && "Must start with an empty cost list");
assert(PHIUsedList.empty() && "Must start with an empty phi used list");
CostWorklist.push_back(&RootI);
for (;; --Iteration) {
do {
Instruction *I = CostWorklist.pop_back_val();
// InstCostMap only uses I and Iteration as a key, the other two values
// don't matter here.
auto CostIter = InstCostMap.find({I, Iteration, 0, 0});
if (CostIter == InstCostMap.end())
// If an input to a PHI node comes from a dead path through the loop
// we may have no cost data for it here. What that actually means is
// that it is free.
continue;
auto &Cost = *CostIter;
if (Cost.IsCounted)
// Already counted this instruction.
continue;
// Mark that we are counting the cost of this instruction now.
Cost.IsCounted = true;
// If this is a PHI node in the loop header, just add it to the PHI set.
if (auto *PhiI = dyn_cast<PHINode>(I))
if (PhiI->getParent() == L->getHeader()) {
assert(Cost.IsFree && "Loop PHIs shouldn't be evaluated as they "
"inherently simplify during unrolling.");
if (Iteration == 0)
continue;
// Push the incoming value from the backedge into the PHI used list
// if it is an in-loop instruction. We'll use this to populate the
// cost worklist for the next iteration (as we count backwards).
if (auto *OpI = dyn_cast<Instruction>(
PhiI->getIncomingValueForBlock(L->getLoopLatch())))
if (L->contains(OpI))
PHIUsedList.push_back(OpI);
continue;
}
// First accumulate the cost of this instruction.
if (!Cost.IsFree) {
UnrolledCost += TTI.getUserCost(I);
DEBUG(dbgs() << "Adding cost of instruction (iteration " << Iteration
<< "): ");
DEBUG(I->dump());
}
// We must count the cost of every operand which is not free,
// recursively. If we reach a loop PHI node, simply add it to the set
// to be considered on the next iteration (backwards!).
for (Value *Op : I->operands()) {
// Check whether this operand is free due to being a constant or
// outside the loop.
auto *OpI = dyn_cast<Instruction>(Op);
if (!OpI || !L->contains(OpI))
continue;
// Otherwise accumulate its cost.
CostWorklist.push_back(OpI);
}
} while (!CostWorklist.empty());
if (PHIUsedList.empty())
// We've exhausted the search.
break;
assert(Iteration > 0 &&
"Cannot track PHI-used values past the first iteration!");
CostWorklist.append(PHIUsedList.begin(), PHIUsedList.end());
PHIUsedList.clear();
}
};
// Ensure that we don't violate the loop structure invariants relied on by
// this analysis.
assert(L->isLoopSimplifyForm() && "Must put loop into normal form first.");
assert(L->isLCSSAForm(DT) &&
"Must have loops in LCSSA form to track live-out values.");
DEBUG(dbgs() << "Starting LoopUnroll profitability analysis...\n");
// Simulate execution of each iteration of the loop counting instructions,
// which would be simplified.
// Since the same load will take different values on different iterations,
// we literally have to go through all loop's iterations.
for (unsigned Iteration = 0; Iteration < TripCount; ++Iteration) {
DEBUG(dbgs() << " Analyzing iteration " << Iteration << "\n");
// Prepare for the iteration by collecting any simplified entry or backedge
// inputs.
for (Instruction &I : *L->getHeader()) {
auto *PHI = dyn_cast<PHINode>(&I);
if (!PHI)
break;
// The loop header PHI nodes must have exactly two input: one from the
// loop preheader and one from the loop latch.
assert(
PHI->getNumIncomingValues() == 2 &&
"Must have an incoming value only for the preheader and the latch.");
Value *V = PHI->getIncomingValueForBlock(
Iteration == 0 ? L->getLoopPreheader() : L->getLoopLatch());
Constant *C = dyn_cast<Constant>(V);
if (Iteration != 0 && !C)
C = SimplifiedValues.lookup(V);
if (C)
SimplifiedInputValues.push_back({PHI, C});
}
// Now clear and re-populate the map for the next iteration.
SimplifiedValues.clear();
while (!SimplifiedInputValues.empty())
SimplifiedValues.insert(SimplifiedInputValues.pop_back_val());
UnrolledInstAnalyzer Analyzer(Iteration, SimplifiedValues, SE, L);
BBWorklist.clear();
BBWorklist.insert(L->getHeader());
// Note that we *must not* cache the size, this loop grows the worklist.
for (unsigned Idx = 0; Idx != BBWorklist.size(); ++Idx) {
BasicBlock *BB = BBWorklist[Idx];
// Visit all instructions in the given basic block and try to simplify
// it. We don't change the actual IR, just count optimization
// opportunities.
for (Instruction &I : *BB) {
// Track this instruction's expected baseline cost when executing the
// rolled loop form.
RolledDynamicCost += TTI.getUserCost(&I);
// Visit the instruction to analyze its loop cost after unrolling,
// and if the visitor returns true, mark the instruction as free after
// unrolling and continue.
bool IsFree = Analyzer.visit(I);
bool Inserted = InstCostMap.insert({&I, (int)Iteration,
(unsigned)IsFree,
/*IsCounted*/ false}).second;
(void)Inserted;
assert(Inserted && "Cannot have a state for an unvisited instruction!");
if (IsFree)
continue;
// If the instruction might have a side-effect recursively account for
// the cost of it and all the instructions leading up to it.
if (I.mayHaveSideEffects())
AddCostRecursively(I, Iteration);
// Can't properly model a cost of a call.
// FIXME: With a proper cost model we should be able to do it.
if(isa<CallInst>(&I))
return None;
// If unrolled body turns out to be too big, bail out.
if (UnrolledCost > MaxUnrolledLoopSize) {
DEBUG(dbgs() << " Exceeded threshold.. exiting.\n"
<< " UnrolledCost: " << UnrolledCost
<< ", MaxUnrolledLoopSize: " << MaxUnrolledLoopSize
<< "\n");
return None;
}
}
TerminatorInst *TI = BB->getTerminator();
// Add in the live successors by first checking whether we have terminator
// that may be simplified based on the values simplified by this call.
BasicBlock *KnownSucc = nullptr;
if (BranchInst *BI = dyn_cast<BranchInst>(TI)) {
if (BI->isConditional()) {
if (Constant *SimpleCond =
SimplifiedValues.lookup(BI->getCondition())) {
// Just take the first successor if condition is undef
if (isa<UndefValue>(SimpleCond))
KnownSucc = BI->getSuccessor(0);
else if (ConstantInt *SimpleCondVal =
dyn_cast<ConstantInt>(SimpleCond))
KnownSucc = BI->getSuccessor(SimpleCondVal->isZero() ? 1 : 0);
}
}
} else if (SwitchInst *SI = dyn_cast<SwitchInst>(TI)) {
if (Constant *SimpleCond =
SimplifiedValues.lookup(SI->getCondition())) {
// Just take the first successor if condition is undef
if (isa<UndefValue>(SimpleCond))
KnownSucc = SI->getSuccessor(0);
else if (ConstantInt *SimpleCondVal =
dyn_cast<ConstantInt>(SimpleCond))
KnownSucc = SI->findCaseValue(SimpleCondVal).getCaseSuccessor();
}
}
if (KnownSucc) {
if (L->contains(KnownSucc))
BBWorklist.insert(KnownSucc);
else
ExitWorklist.insert({BB, KnownSucc});
continue;
}
// Add BB's successors to the worklist.
for (BasicBlock *Succ : successors(BB))
if (L->contains(Succ))
BBWorklist.insert(Succ);
else
ExitWorklist.insert({BB, Succ});
AddCostRecursively(*TI, Iteration);
}
// If we found no optimization opportunities on the first iteration, we
// won't find them on later ones too.
if (UnrolledCost == RolledDynamicCost) {
DEBUG(dbgs() << " No opportunities found.. exiting.\n"
<< " UnrolledCost: " << UnrolledCost << "\n");
return None;
}
}
while (!ExitWorklist.empty()) {
BasicBlock *ExitingBB, *ExitBB;
std::tie(ExitingBB, ExitBB) = ExitWorklist.pop_back_val();
for (Instruction &I : *ExitBB) {
auto *PN = dyn_cast<PHINode>(&I);
if (!PN)
break;
Value *Op = PN->getIncomingValueForBlock(ExitingBB);
if (auto *OpI = dyn_cast<Instruction>(Op))
if (L->contains(OpI))
AddCostRecursively(*OpI, TripCount - 1);
}
}
DEBUG(dbgs() << "Analysis finished:\n"
<< "UnrolledCost: " << UnrolledCost << ", "
<< "RolledDynamicCost: " << RolledDynamicCost << "\n");
return {{UnrolledCost, RolledDynamicCost}};
}
/// \brief Figure out if the loop is worth full unrolling.
///
/// Complete loop unrolling can make some loads constant, and we need to know
/// if that would expose any further optimization opportunities. This routine
/// estimates this optimization. It computes cost of unrolled loop
/// (UnrolledCost) and dynamic cost of the original loop (RolledDynamicCost). By
/// dynamic cost we mean that we won't count costs of blocks that are known not
/// to be executed (i.e. if we have a branch in the loop and we know that at the
/// given iteration its condition would be resolved to true, we won't add up the
/// cost of the 'false'-block).
/// \returns Optional value, holding the RolledDynamicCost and UnrolledCost. If
/// the analysis failed (no benefits expected from the unrolling, or the loop is
/// too big to analyze), the returned value is None.
static Optional<EstimatedUnrollCost>
analyzeLoopUnrollCost(const Loop *L, unsigned TripCount, DominatorTree &DT,
ScalarEvolution &SE, const TargetTransformInfo &TTI,
int MaxUnrolledLoopSize) {
// We want to be able to scale offsets by the trip count and add more offsets
// to them without checking for overflows, and we already don't want to
// analyze *massive* trip counts, so we force the max to be reasonably small.
assert(UnrollMaxIterationsCountToAnalyze < (INT_MAX / 2) &&
"The unroll iterations max is too large!");
// Don't simulate loops with a big or unknown tripcount
if (!UnrollMaxIterationsCountToAnalyze || !TripCount ||
TripCount > UnrollMaxIterationsCountToAnalyze)
return None;
SmallSetVector<BasicBlock *, 16> BBWorklist;
DenseMap<Value *, Constant *> SimplifiedValues;
SmallVector<std::pair<Value *, Constant *>, 4> SimplifiedInputValues;
// The estimated cost of the unrolled form of the loop. We try to estimate
// this by simplifying as much as we can while computing the estimate.
int UnrolledCost = 0;
// We also track the estimated dynamic (that is, actually executed) cost in
// the rolled form. This helps identify cases when the savings from unrolling
// aren't just exposing dead control flows, but actual reduced dynamic
// instructions due to the simplifications which we expect to occur after
// unrolling.
int RolledDynamicCost = 0;
// Ensure that we don't violate the loop structure invariants relied on by
// this analysis.
assert(L->isLoopSimplifyForm() && "Must put loop into normal form first.");
assert(L->isLCSSAForm(DT) &&
"Must have loops in LCSSA form to track live-out values.");
DEBUG(dbgs() << "Starting LoopUnroll profitability analysis...\n");
// Simulate execution of each iteration of the loop counting instructions,
// which would be simplified.
// Since the same load will take different values on different iterations,
// we literally have to go through all loop's iterations.
for (unsigned Iteration = 0; Iteration < TripCount; ++Iteration) {
DEBUG(dbgs() << " Analyzing iteration " << Iteration << "\n");
// Prepare for the iteration by collecting any simplified entry or backedge
// inputs.
for (Instruction &I : *L->getHeader()) {
auto *PHI = dyn_cast<PHINode>(&I);
if (!PHI)
break;
// The loop header PHI nodes must have exactly two input: one from the
// loop preheader and one from the loop latch.
assert(
PHI->getNumIncomingValues() == 2 &&
"Must have an incoming value only for the preheader and the latch.");
Value *V = PHI->getIncomingValueForBlock(
Iteration == 0 ? L->getLoopPreheader() : L->getLoopLatch());
Constant *C = dyn_cast<Constant>(V);
if (Iteration != 0 && !C)
C = SimplifiedValues.lookup(V);
if (C)
SimplifiedInputValues.push_back({PHI, C});
}
// Now clear and re-populate the map for the next iteration.
SimplifiedValues.clear();
while (!SimplifiedInputValues.empty())
SimplifiedValues.insert(SimplifiedInputValues.pop_back_val());
UnrolledInstAnalyzer Analyzer(Iteration, SimplifiedValues, SE);
BBWorklist.clear();
BBWorklist.insert(L->getHeader());
// Note that we *must not* cache the size, this loop grows the worklist.
for (unsigned Idx = 0; Idx != BBWorklist.size(); ++Idx) {
BasicBlock *BB = BBWorklist[Idx];
// Visit all instructions in the given basic block and try to simplify
// it. We don't change the actual IR, just count optimization
// opportunities.
for (Instruction &I : *BB) {
int InstCost = TTI.getUserCost(&I);
// Visit the instruction to analyze its loop cost after unrolling,
// and if the visitor returns false, include this instruction in the
// unrolled cost.
if (!Analyzer.visit(I))
UnrolledCost += InstCost;
else {
DEBUG(dbgs() << " " << I
<< " would be simplified if loop is unrolled.\n");
(void)0;
}
// Also track this instructions expected cost when executing the rolled
// loop form.
RolledDynamicCost += InstCost;
// If unrolled body turns out to be too big, bail out.
if (UnrolledCost > MaxUnrolledLoopSize) {
DEBUG(dbgs() << " Exceeded threshold.. exiting.\n"
<< " UnrolledCost: " << UnrolledCost
<< ", MaxUnrolledLoopSize: " << MaxUnrolledLoopSize
<< "\n");
return None;
}
}
TerminatorInst *TI = BB->getTerminator();
// Add in the live successors by first checking whether we have terminator
// that may be simplified based on the values simplified by this call.
if (BranchInst *BI = dyn_cast<BranchInst>(TI)) {
if (BI->isConditional()) {
if (Constant *SimpleCond =
SimplifiedValues.lookup(BI->getCondition())) {
BasicBlock *Succ = nullptr;
// Just take the first successor if condition is undef
if (isa<UndefValue>(SimpleCond))
Succ = BI->getSuccessor(0);
else
Succ = BI->getSuccessor(
cast<ConstantInt>(SimpleCond)->isZero() ? 1 : 0);
if (L->contains(Succ))
BBWorklist.insert(Succ);
continue;
}
}
} else if (SwitchInst *SI = dyn_cast<SwitchInst>(TI)) {
if (Constant *SimpleCond =
SimplifiedValues.lookup(SI->getCondition())) {
BasicBlock *Succ = nullptr;
// Just take the first successor if condition is undef
if (isa<UndefValue>(SimpleCond))
Succ = SI->getSuccessor(0);
else
Succ = SI->findCaseValue(cast<ConstantInt>(SimpleCond))
.getCaseSuccessor();
if (L->contains(Succ))
BBWorklist.insert(Succ);
continue;
}
}
// Add BB's successors to the worklist.
for (BasicBlock *Succ : successors(BB))
if (L->contains(Succ))
BBWorklist.insert(Succ);
}
// If we found no optimization opportunities on the first iteration, we
// won't find them on later ones too.
if (UnrolledCost == RolledDynamicCost) {
DEBUG(dbgs() << " No opportunities found.. exiting.\n"
<< " UnrolledCost: " << UnrolledCost << "\n");
return None;
}
}
DEBUG(dbgs() << "Analysis finished:\n"
<< "UnrolledCost: " << UnrolledCost << ", "
<< "RolledDynamicCost: " << RolledDynamicCost << "\n");
return {{UnrolledCost, RolledDynamicCost}};
}
