//===----------------------------------------------------------------------===//
#include "llvm/Transforms/Utils/Cloning.h"
+#include "llvm/ADT/SetVector.h"
#include "llvm/ADT/SmallSet.h"
#include "llvm/ADT/SmallVector.h"
-#include "llvm/ADT/SetVector.h"
#include "llvm/ADT/StringExtras.h"
#include "llvm/Analysis/AliasAnalysis.h"
#include "llvm/Analysis/AssumptionCache.h"
#include "llvm/Transforms/Utils/Local.h"
#include "llvm/Support/CommandLine.h"
#include <algorithm>
+
using namespace llvm;
static cl::opt<bool>
cl::desc("Convert align attributes to assumptions during inlining."));
bool llvm::InlineFunction(CallInst *CI, InlineFunctionInfo &IFI,
- bool InsertLifetime) {
- return InlineFunction(CallSite(CI), IFI, InsertLifetime);
+ AAResults *CalleeAAR, bool InsertLifetime) {
+ return InlineFunction(CallSite(CI), IFI, CalleeAAR, InsertLifetime);
}
bool llvm::InlineFunction(InvokeInst *II, InlineFunctionInfo &IFI,
- bool InsertLifetime) {
- return InlineFunction(CallSite(II), IFI, InsertLifetime);
+ AAResults *CalleeAAR, bool InsertLifetime) {
+ return InlineFunction(CallSite(II), IFI, CalleeAAR, InsertLifetime);
}
namespace {
- /// A class for recording information about inlining through an invoke.
- class InvokeInliningInfo {
+ /// A class for recording information about inlining a landing pad.
+ class LandingPadInliningInfo {
BasicBlock *OuterResumeDest; ///< Destination of the invoke's unwind.
BasicBlock *InnerResumeDest; ///< Destination for the callee's resume.
LandingPadInst *CallerLPad; ///< LandingPadInst associated with the invoke.
SmallVector<Value*, 8> UnwindDestPHIValues;
public:
- InvokeInliningInfo(InvokeInst *II)
+ LandingPadInliningInfo(InvokeInst *II)
: OuterResumeDest(II->getUnwindDest()), InnerResumeDest(nullptr),
CallerLPad(nullptr), InnerEHValuesPHI(nullptr) {
// If there are PHI nodes in the unwind destination block, we need to keep
CallerLPad = cast<LandingPadInst>(I);
}
- /// getOuterResumeDest - The outer unwind destination is the target of
+ /// The outer unwind destination is the target of
/// unwind edges introduced for calls within the inlined function.
BasicBlock *getOuterResumeDest() const {
return OuterResumeDest;
LandingPadInst *getLandingPadInst() const { return CallerLPad; }
- /// forwardResume - Forward the 'resume' instruction to the caller's landing
- /// pad block. When the landing pad block has only one predecessor, this is
+ /// Forward the 'resume' instruction to the caller's landing pad block.
+ /// When the landing pad block has only one predecessor, this is
/// a simple branch. When there is more than one predecessor, we need to
/// split the landing pad block after the landingpad instruction and jump
/// to there.
void forwardResume(ResumeInst *RI,
SmallPtrSetImpl<LandingPadInst*> &InlinedLPads);
- /// addIncomingPHIValuesFor - Add incoming-PHI values to the unwind
- /// destination block for the given basic block, using the values for the
- /// original invoke's source block.
+ /// Add incoming-PHI values to the unwind destination block for the given
+ /// basic block, using the values for the original invoke's source block.
void addIncomingPHIValuesFor(BasicBlock *BB) const {
addIncomingPHIValuesForInto(BB, OuterResumeDest);
}
}
}
};
-}
+} // anonymous namespace
-/// getInnerResumeDest - Get or create a target for the branch from ResumeInsts.
-BasicBlock *InvokeInliningInfo::getInnerResumeDest() {
+/// Get or create a target for the branch from ResumeInsts.
+BasicBlock *LandingPadInliningInfo::getInnerResumeDest() {
if (InnerResumeDest) return InnerResumeDest;
// Split the landing pad.
- BasicBlock::iterator SplitPoint = CallerLPad; ++SplitPoint;
+ BasicBlock::iterator SplitPoint = ++CallerLPad->getIterator();
InnerResumeDest =
OuterResumeDest->splitBasicBlock(SplitPoint,
OuterResumeDest->getName() + ".body");
const unsigned PHICapacity = 2;
// Create corresponding new PHIs for all the PHIs in the outer landing pad.
- BasicBlock::iterator InsertPoint = InnerResumeDest->begin();
+ Instruction *InsertPoint = &InnerResumeDest->front();
BasicBlock::iterator I = OuterResumeDest->begin();
for (unsigned i = 0, e = UnwindDestPHIValues.size(); i != e; ++i, ++I) {
PHINode *OuterPHI = cast<PHINode>(I);
return InnerResumeDest;
}
-/// forwardResume - Forward the 'resume' instruction to the caller's landing pad
-/// block. When the landing pad block has only one predecessor, this is a simple
+/// Forward the 'resume' instruction to the caller's landing pad block.
+/// When the landing pad block has only one predecessor, this is a simple
/// branch. When there is more than one predecessor, we need to split the
/// landing pad block after the landingpad instruction and jump to there.
-void InvokeInliningInfo::forwardResume(ResumeInst *RI,
- SmallPtrSetImpl<LandingPadInst*> &InlinedLPads) {
+void LandingPadInliningInfo::forwardResume(
+ ResumeInst *RI, SmallPtrSetImpl<LandingPadInst *> &InlinedLPads) {
BasicBlock *Dest = getInnerResumeDest();
BasicBlock *Src = RI->getParent();
RI->eraseFromParent();
}
-/// HandleCallsInBlockInlinedThroughInvoke - When we inline a basic block into
-/// an invoke, we have to turn all of the calls that can throw into
-/// invokes. This function analyze BB to see if there are any calls, and if so,
+/// When we inline a basic block into an invoke,
+/// we have to turn all of the calls that can throw into invokes.
+/// This function analyze BB to see if there are any calls, and if so,
/// it rewrites them to be invokes that jump to InvokeDest and fills in the PHI
/// nodes in that block with the values specified in InvokeDestPHIValues.
-static void HandleCallsInBlockInlinedThroughInvoke(BasicBlock *BB,
- InvokeInliningInfo &Invoke) {
+static BasicBlock *
+HandleCallsInBlockInlinedThroughInvoke(BasicBlock *BB, BasicBlock *UnwindEdge) {
for (BasicBlock::iterator BBI = BB->begin(), E = BB->end(); BBI != E; ) {
- Instruction *I = BBI++;
+ Instruction *I = &*BBI++;
// We only need to check for function calls: inlined invoke
// instructions require no special handling.
// Convert this function call into an invoke instruction. First, split the
// basic block.
- BasicBlock *Split = BB->splitBasicBlock(CI, CI->getName()+".noexc");
+ BasicBlock *Split =
+ BB->splitBasicBlock(CI->getIterator(), CI->getName() + ".noexc");
// Delete the unconditional branch inserted by splitBasicBlock
BB->getInstList().pop_back();
// Create the new invoke instruction.
ImmutableCallSite CS(CI);
SmallVector<Value*, 8> InvokeArgs(CS.arg_begin(), CS.arg_end());
- InvokeInst *II = InvokeInst::Create(CI->getCalledValue(), Split,
- Invoke.getOuterResumeDest(),
- InvokeArgs, CI->getName(), BB);
+ SmallVector<OperandBundleDef, 1> OpBundles;
+
+ // Copy the OperandBundeUse instances to OperandBundleDefs. These two are
+ // *different* representations of operand bundles: see the documentation in
+ // InstrTypes.h for more details.
+ for (unsigned i = 0, e = CS.getNumOperandBundles(); i != e; ++i)
+ OpBundles.emplace_back(CS.getOperandBundleAt(i));
+
+ // Note: we're round tripping operand bundles through memory here, and that
+ // can potentially be avoided with a cleverer API design that we do not have
+ // as of this time.
+
+ InvokeInst *II =
+ InvokeInst::Create(CI->getCalledValue(), Split, UnwindEdge, InvokeArgs,
+ OpBundles, CI->getName(), BB);
II->setDebugLoc(CI->getDebugLoc());
II->setCallingConv(CI->getCallingConv());
II->setAttributes(CI->getAttributes());
// Delete the original call
Split->getInstList().pop_front();
-
- // Update any PHI nodes in the exceptional block to indicate that there is
- // now a new entry in them.
- Invoke.addIncomingPHIValuesFor(BB);
- return;
+ return BB;
}
+ return nullptr;
}
-/// HandleInlinedInvoke - If we inlined an invoke site, we need to convert calls
+/// If we inlined an invoke site, we need to convert calls
/// in the body of the inlined function into invokes.
///
/// II is the invoke instruction being inlined. FirstNewBlock is the first
/// block of the inlined code (the last block is the end of the function),
/// and InlineCodeInfo is information about the code that got inlined.
-static void HandleInlinedInvoke(InvokeInst *II, BasicBlock *FirstNewBlock,
- ClonedCodeInfo &InlinedCodeInfo) {
+static void HandleInlinedLandingPad(InvokeInst *II, BasicBlock *FirstNewBlock,
+ ClonedCodeInfo &InlinedCodeInfo) {
BasicBlock *InvokeDest = II->getUnwindDest();
Function *Caller = FirstNewBlock->getParent();
// The inlined code is currently at the end of the function, scan from the
// start of the inlined code to its end, checking for stuff we need to
// rewrite.
- InvokeInliningInfo Invoke(II);
+ LandingPadInliningInfo Invoke(II);
// Get all of the inlined landing pad instructions.
SmallPtrSet<LandingPadInst*, 16> InlinedLPads;
- for (Function::iterator I = FirstNewBlock, E = Caller->end(); I != E; ++I)
+ for (Function::iterator I = FirstNewBlock->getIterator(), E = Caller->end();
+ I != E; ++I)
if (InvokeInst *II = dyn_cast<InvokeInst>(I->getTerminator()))
InlinedLPads.insert(II->getLandingPadInst());
InlinedLPad->setCleanup(true);
}
- for (Function::iterator BB = FirstNewBlock, E = Caller->end(); BB != E; ++BB){
+ for (Function::iterator BB = FirstNewBlock->getIterator(), E = Caller->end();
+ BB != E; ++BB) {
if (InlinedCodeInfo.ContainsCalls)
- HandleCallsInBlockInlinedThroughInvoke(BB, Invoke);
+ if (BasicBlock *NewBB = HandleCallsInBlockInlinedThroughInvoke(
+ &*BB, Invoke.getOuterResumeDest()))
+ // Update any PHI nodes in the exceptional block to indicate that there
+ // is now a new entry in them.
+ Invoke.addIncomingPHIValuesFor(NewBB);
// Forward any resumes that are remaining here.
if (ResumeInst *RI = dyn_cast<ResumeInst>(BB->getTerminator()))
InvokeDest->removePredecessor(II->getParent());
}
-/// CloneAliasScopeMetadata - When inlining a function that contains noalias
-/// scope metadata, this metadata needs to be cloned so that the inlined blocks
+/// If we inlined an invoke site, we need to convert calls
+/// in the body of the inlined function into invokes.
+///
+/// II is the invoke instruction being inlined. FirstNewBlock is the first
+/// block of the inlined code (the last block is the end of the function),
+/// and InlineCodeInfo is information about the code that got inlined.
+static void HandleInlinedEHPad(InvokeInst *II, BasicBlock *FirstNewBlock,
+ ClonedCodeInfo &InlinedCodeInfo) {
+ BasicBlock *UnwindDest = II->getUnwindDest();
+ Function *Caller = FirstNewBlock->getParent();
+
+ assert(UnwindDest->getFirstNonPHI()->isEHPad() && "unexpected BasicBlock!");
+
+ // If there are PHI nodes in the unwind destination block, we need to keep
+ // track of which values came into them from the invoke before removing the
+ // edge from this block.
+ SmallVector<Value *, 8> UnwindDestPHIValues;
+ llvm::BasicBlock *InvokeBB = II->getParent();
+ for (Instruction &I : *UnwindDest) {
+ // Save the value to use for this edge.
+ PHINode *PHI = dyn_cast<PHINode>(&I);
+ if (!PHI)
+ break;
+ UnwindDestPHIValues.push_back(PHI->getIncomingValueForBlock(InvokeBB));
+ }
+
+ // Add incoming-PHI values to the unwind destination block for the given basic
+ // block, using the values for the original invoke's source block.
+ auto UpdatePHINodes = [&](BasicBlock *Src) {
+ BasicBlock::iterator I = UnwindDest->begin();
+ for (Value *V : UnwindDestPHIValues) {
+ PHINode *PHI = cast<PHINode>(I);
+ PHI->addIncoming(V, Src);
+ ++I;
+ }
+ };
+
+ // Forward EH terminator instructions to the caller's invoke destination.
+ // This is as simple as connect all the instructions which 'unwind to caller'
+ // to the invoke destination.
+ for (Function::iterator BB = FirstNewBlock->getIterator(), E = Caller->end();
+ BB != E; ++BB) {
+ Instruction *I = BB->getFirstNonPHI();
+ if (I->isEHPad()) {
+ if (auto *CEPI = dyn_cast<CatchEndPadInst>(I)) {
+ if (CEPI->unwindsToCaller()) {
+ CatchEndPadInst::Create(CEPI->getContext(), UnwindDest, CEPI);
+ CEPI->eraseFromParent();
+ UpdatePHINodes(&*BB);
+ }
+ } else if (auto *CEPI = dyn_cast<CleanupEndPadInst>(I)) {
+ if (CEPI->unwindsToCaller()) {
+ CleanupEndPadInst::Create(CEPI->getCleanupPad(), UnwindDest, CEPI);
+ CEPI->eraseFromParent();
+ UpdatePHINodes(&*BB);
+ }
+ } else if (auto *TPI = dyn_cast<TerminatePadInst>(I)) {
+ if (TPI->unwindsToCaller()) {
+ SmallVector<Value *, 3> TerminatePadArgs;
+ for (Value *ArgOperand : TPI->arg_operands())
+ TerminatePadArgs.push_back(ArgOperand);
+ TerminatePadInst::Create(TPI->getContext(), UnwindDest,
+ TerminatePadArgs, TPI);
+ TPI->eraseFromParent();
+ UpdatePHINodes(&*BB);
+ }
+ } else {
+ assert(isa<CatchPadInst>(I) || isa<CleanupPadInst>(I));
+ }
+ }
+
+ if (auto *CRI = dyn_cast<CleanupReturnInst>(BB->getTerminator())) {
+ if (CRI->unwindsToCaller()) {
+ CleanupReturnInst::Create(CRI->getCleanupPad(), UnwindDest, CRI);
+ CRI->eraseFromParent();
+ UpdatePHINodes(&*BB);
+ }
+ }
+ }
+
+ if (InlinedCodeInfo.ContainsCalls)
+ for (Function::iterator BB = FirstNewBlock->getIterator(),
+ E = Caller->end();
+ BB != E; ++BB)
+ if (BasicBlock *NewBB =
+ HandleCallsInBlockInlinedThroughInvoke(&*BB, UnwindDest))
+ // Update any PHI nodes in the exceptional block to indicate that there
+ // is now a new entry in them.
+ UpdatePHINodes(NewBB);
+
+ // Now that everything is happy, we have one final detail. The PHI nodes in
+ // the exception destination block still have entries due to the original
+ // invoke instruction. Eliminate these entries (which might even delete the
+ // PHI node) now.
+ UnwindDest->removePredecessor(InvokeBB);
+}
+
+/// When inlining a function that contains noalias scope metadata,
+/// this metadata needs to be cloned so that the inlined blocks
/// have different "unqiue scopes" at every call site. Were this not done, then
/// aliasing scopes from a function inlined into a caller multiple times could
/// not be differentiated (and this would lead to miscompiles because the
}
}
-/// AddAliasScopeMetadata - If the inlined function has noalias arguments, then
-/// add new alias scopes for each noalias argument, tag the mapped noalias
+/// If the inlined function has noalias arguments,
+/// then add new alias scopes for each noalias argument, tag the mapped noalias
/// parameters with noalias metadata specifying the new scope, and tag all
/// non-derived loads, stores and memory intrinsics with the new alias scopes.
static void AddAliasScopeMetadata(CallSite CS, ValueToValueMapTy &VMap,
- const DataLayout *DL, AliasAnalysis *AA) {
+ const DataLayout &DL, AAResults *CalleeAAR) {
if (!EnableNoAliasConversion)
return;
const Function *CalledFunc = CS.getCalledFunction();
SmallVector<const Argument *, 4> NoAliasArgs;
- for (Function::const_arg_iterator I = CalledFunc->arg_begin(),
- E = CalledFunc->arg_end(); I != E; ++I) {
- if (I->hasNoAliasAttr() && !I->hasNUses(0))
- NoAliasArgs.push_back(I);
+ for (const Argument &I : CalledFunc->args()) {
+ if (I.hasNoAliasAttr() && !I.hasNUses(0))
+ NoAliasArgs.push_back(&I);
}
if (NoAliasArgs.empty())
continue;
IsFuncCall = true;
- if (AA) {
- AliasAnalysis::ModRefBehavior MRB = AA->getModRefBehavior(ICS);
- if (MRB == AliasAnalysis::OnlyAccessesArgumentPointees ||
- MRB == AliasAnalysis::OnlyReadsArgumentPointees)
+ if (CalleeAAR) {
+ FunctionModRefBehavior MRB = CalleeAAR->getModRefBehavior(ICS);
+ if (MRB == FMRB_OnlyAccessesArgumentPointees ||
+ MRB == FMRB_OnlyReadsArgumentPointees)
IsArgMemOnlyCall = true;
}
for (unsigned i = 0, ie = PtrArgs.size(); i != ie; ++i) {
SmallVector<Value *, 4> Objects;
GetUnderlyingObjects(const_cast<Value*>(PtrArgs[i]),
- Objects, DL, /* MaxLookup = */ 0);
+ Objects, DL, /* LI = */ nullptr);
for (Value *O : Objects)
ObjSet.insert(O);
/// If the inlined function has non-byval align arguments, then
/// add @llvm.assume-based alignment assumptions to preserve this information.
static void AddAlignmentAssumptions(CallSite CS, InlineFunctionInfo &IFI) {
- if (!PreserveAlignmentAssumptions || !IFI.DL)
+ if (!PreserveAlignmentAssumptions)
return;
+ auto &DL = CS.getCaller()->getParent()->getDataLayout();
// To avoid inserting redundant assumptions, we should check for assumptions
// already in the caller. To do this, we might need a DT of the caller.
// If we can already prove the asserted alignment in the context of the
// caller, then don't bother inserting the assumption.
Value *Arg = CS.getArgument(I->getArgNo());
- if (getKnownAlignment(Arg, IFI.DL,
- &IFI.ACT->getAssumptionCache(*CalledFunc),
- CS.getInstruction(), &DT) >= Align)
+ if (getKnownAlignment(Arg, DL, CS.getInstruction(),
+ &IFI.ACT->getAssumptionCache(*CS.getCaller()),
+ &DT) >= Align)
continue;
- IRBuilder<>(CS.getInstruction()).CreateAlignmentAssumption(*IFI.DL, Arg,
- Align);
+ IRBuilder<>(CS.getInstruction())
+ .CreateAlignmentAssumption(DL, Arg, Align);
}
}
}
-/// UpdateCallGraphAfterInlining - Once we have cloned code over from a callee
-/// into the caller, update the specified callgraph to reflect the changes we
-/// made. Note that it's possible that not all code was copied over, so only
+/// Once we have cloned code over from a callee into the caller,
+/// update the specified callgraph to reflect the changes we made.
+/// Note that it's possible that not all code was copied over, so only
/// some edges of the callgraph may remain.
static void UpdateCallGraphAfterInlining(CallSite CS,
Function::iterator FirstNewBlock,
// If the call was inlined, but then constant folded, there is no edge to
// add. Check for this case.
Instruction *NewCall = dyn_cast<Instruction>(VMI->second);
- if (!NewCall) continue;
+ if (!NewCall)
+ continue;
+ // We do not treat intrinsic calls like real function calls because we
+ // expect them to become inline code; do not add an edge for an intrinsic.
+ CallSite CS = CallSite(NewCall);
+ if (CS && CS.getCalledFunction() && CS.getCalledFunction()->isIntrinsic())
+ continue;
+
// Remember that this call site got inlined for the client of
// InlineFunction.
IFI.InlinedCalls.push_back(NewCall);
BasicBlock *InsertBlock,
InlineFunctionInfo &IFI) {
Type *AggTy = cast<PointerType>(Src->getType())->getElementType();
- IRBuilder<> Builder(InsertBlock->begin());
+ IRBuilder<> Builder(InsertBlock, InsertBlock->begin());
- Value *Size;
- if (IFI.DL == nullptr)
- Size = ConstantExpr::getSizeOf(AggTy);
- else
- Size = Builder.getInt64(IFI.DL->getTypeStoreSize(AggTy));
+ Value *Size = Builder.getInt64(M->getDataLayout().getTypeStoreSize(AggTy));
// Always generate a memcpy of alignment 1 here because we don't know
// the alignment of the src pointer. Other optimizations can infer
Builder.CreateMemCpy(Dst, Src, Size, /*Align=*/1);
}
-/// HandleByValArgument - When inlining a call site that has a byval argument,
+/// When inlining a call site that has a byval argument,
/// we have to make the implicit memcpy explicit by adding it.
static Value *HandleByValArgument(Value *Arg, Instruction *TheCall,
const Function *CalledFunc,
if (ByValAlignment <= 1) // 0 = unspecified, 1 = no particular alignment.
return Arg;
+ const DataLayout &DL = Caller->getParent()->getDataLayout();
+
// If the pointer is already known to be sufficiently aligned, or if we can
// round it up to a larger alignment, then we don't need a temporary.
- if (getOrEnforceKnownAlignment(Arg, ByValAlignment, IFI.DL,
- &IFI.ACT->getAssumptionCache(*Caller),
- TheCall) >= ByValAlignment)
+ if (getOrEnforceKnownAlignment(Arg, ByValAlignment, DL, TheCall,
+ &IFI.ACT->getAssumptionCache(*Caller)) >=
+ ByValAlignment)
return Arg;
// Otherwise, we have to make a memcpy to get a safe alignment. This is bad
}
// Create the alloca. If we have DataLayout, use nice alignment.
- unsigned Align = 1;
- if (IFI.DL)
- Align = IFI.DL->getPrefTypeAlignment(AggTy);
-
+ unsigned Align =
+ Caller->getParent()->getDataLayout().getPrefTypeAlignment(AggTy);
+
// If the byval had an alignment specified, we *must* use at least that
// alignment, as it is required by the byval argument (and uses of the
// pointer inside the callee).
return NewAlloca;
}
-// isUsedByLifetimeMarker - Check whether this Value is used by a lifetime
-// intrinsic.
+// Check whether this Value is used by a lifetime intrinsic.
static bool isUsedByLifetimeMarker(Value *V) {
for (User *U : V->users()) {
if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(U)) {
return false;
}
-// hasLifetimeMarkers - Check whether the given alloca already has
+// Check whether the given alloca already has
// lifetime.start or lifetime.end intrinsics.
static bool hasLifetimeMarkers(AllocaInst *AI) {
Type *Ty = AI->getType();
/// Rebuild the entire inlined-at chain for this instruction so that the top of
/// the chain now is inlined-at the new call site.
static DebugLoc
-updateInlinedAtInfo(DebugLoc DL, MDLocation *InlinedAtNode,
- LLVMContext &Ctx,
- DenseMap<const MDLocation *, MDLocation *> &IANodes) {
- SmallVector<MDLocation*, 3> InlinedAtLocations;
- MDLocation *Last = InlinedAtNode;
- DebugLoc CurInlinedAt = DL;
+updateInlinedAtInfo(DebugLoc DL, DILocation *InlinedAtNode, LLVMContext &Ctx,
+ DenseMap<const DILocation *, DILocation *> &IANodes) {
+ SmallVector<DILocation *, 3> InlinedAtLocations;
+ DILocation *Last = InlinedAtNode;
+ DILocation *CurInlinedAt = DL;
// Gather all the inlined-at nodes
- while (MDLocation *IA =
- cast_or_null<MDLocation>(CurInlinedAt.getInlinedAt(Ctx))) {
+ while (DILocation *IA = CurInlinedAt->getInlinedAt()) {
// Skip any we've already built nodes for
- if (MDLocation *Found = IANodes[IA]) {
+ if (DILocation *Found = IANodes[IA]) {
Last = Found;
break;
}
InlinedAtLocations.push_back(IA);
- CurInlinedAt = DebugLoc::getFromDILocation(IA);
+ CurInlinedAt = IA;
}
// Starting from the top, rebuild the nodes to point to the new inlined-at
// location (then rebuilding the rest of the chain behind it) and update the
// map of already-constructed inlined-at nodes.
- for (auto I = InlinedAtLocations.rbegin(), E = InlinedAtLocations.rend();
- I != E; ++I) {
- const MDLocation *MD = *I;
- Last = IANodes[MD] = MDLocation::getDistinct(
+ for (const DILocation *MD : make_range(InlinedAtLocations.rbegin(),
+ InlinedAtLocations.rend())) {
+ Last = IANodes[MD] = DILocation::getDistinct(
Ctx, MD->getLine(), MD->getColumn(), MD->getScope(), Last);
}
// And finally create the normal location for this instruction, referring to
// the new inlined-at chain.
- return DebugLoc::get(DL.getLine(), DL.getCol(), DL.getScope(Ctx), Last);
+ return DebugLoc::get(DL.getLine(), DL.getCol(), DL.getScope(), Last);
}
-/// fixupLineNumbers - Update inlined instructions' line numbers to
+/// Update inlined instructions' line numbers to
/// to encode location where these instructions are inlined.
static void fixupLineNumbers(Function *Fn, Function::iterator FI,
Instruction *TheCall) {
DebugLoc TheCallDL = TheCall->getDebugLoc();
- if (TheCallDL.isUnknown())
+ if (!TheCallDL)
return;
auto &Ctx = Fn->getContext();
- auto *InlinedAtNode = cast<MDLocation>(TheCallDL.getAsMDNode(Ctx));
+ DILocation *InlinedAtNode = TheCallDL;
// Create a unique call site, not to be confused with any other call from the
// same location.
- InlinedAtNode = MDLocation::getDistinct(
+ InlinedAtNode = DILocation::getDistinct(
Ctx, InlinedAtNode->getLine(), InlinedAtNode->getColumn(),
InlinedAtNode->getScope(), InlinedAtNode->getInlinedAt());
// Cache the inlined-at nodes as they're built so they are reused, without
// this every instruction's inlined-at chain would become distinct from each
// other.
- DenseMap<const MDLocation *, MDLocation *> IANodes;
+ DenseMap<const DILocation *, DILocation *> IANodes;
for (; FI != Fn->end(); ++FI) {
for (BasicBlock::iterator BI = FI->begin(), BE = FI->end();
BI != BE; ++BI) {
DebugLoc DL = BI->getDebugLoc();
- if (DL.isUnknown()) {
+ if (!DL) {
// If the inlined instruction has no line number, make it look as if it
// originates from the call location. This is important for
// ((__always_inline__, __nodebug__)) functions which must use caller
BI->setDebugLoc(TheCallDL);
} else {
BI->setDebugLoc(updateInlinedAtInfo(DL, InlinedAtNode, BI->getContext(), IANodes));
- if (DbgValueInst *DVI = dyn_cast<DbgValueInst>(BI)) {
- LLVMContext &Ctx = BI->getContext();
- MDNode *InlinedAt = BI->getDebugLoc().getInlinedAt(Ctx);
- DVI->setOperand(2, MetadataAsValue::get(
- Ctx, createInlinedVariable(DVI->getVariable(),
- InlinedAt, Ctx)));
- } else if (DbgDeclareInst *DDI = dyn_cast<DbgDeclareInst>(BI)) {
- LLVMContext &Ctx = BI->getContext();
- MDNode *InlinedAt = BI->getDebugLoc().getInlinedAt(Ctx);
- DDI->setOperand(1, MetadataAsValue::get(
- Ctx, createInlinedVariable(DDI->getVariable(),
- InlinedAt, Ctx)));
- }
}
}
}
}
-/// InlineFunction - This function inlines the called function into the basic
-/// block of the caller. This returns false if it is not possible to inline
-/// this call. The program is still in a well defined state if this occurs
-/// though.
+/// This function inlines the called function into the basic block of the
+/// caller. This returns false if it is not possible to inline this call.
+/// The program is still in a well defined state if this occurs though.
///
/// Note that this only does one level of inlining. For example, if the
/// instruction 'call B' is inlined, and 'B' calls 'C', then the call to 'C' now
/// exists in the instruction stream. Similarly this will inline a recursive
/// function by one level.
bool llvm::InlineFunction(CallSite CS, InlineFunctionInfo &IFI,
- bool InsertLifetime) {
+ AAResults *CalleeAAR, bool InsertLifetime) {
Instruction *TheCall = CS.getInstruction();
assert(TheCall->getParent() && TheCall->getParent()->getParent() &&
"Instruction not in function!");
CalledFunc->isDeclaration() || // call, or call to a vararg function!
CalledFunc->getFunctionType()->isVarArg()) return false;
+ // The inliner does not know how to inline through calls with operand bundles
+ // in general ...
+ if (CS.hasOperandBundles()) {
+ // ... but it knows how to inline through "deopt" operand bundles.
+ bool CanInline =
+ CS.getNumOperandBundles() == 1 &&
+ CS.getOperandBundleAt(0).getTagID() == LLVMContext::OB_deopt;
+ if (!CanInline)
+ return false;
+ }
+
// If the call to the callee cannot throw, set the 'nounwind' flag on any
// calls that we inline.
bool MarkNoUnwind = CS.doesNotThrow();
}
// Get the personality function from the callee if it contains a landing pad.
- Value *CalleePersonality = nullptr;
- for (Function::const_iterator I = CalledFunc->begin(), E = CalledFunc->end();
- I != E; ++I)
- if (const InvokeInst *II = dyn_cast<InvokeInst>(I->getTerminator())) {
- const BasicBlock *BB = II->getUnwindDest();
- const LandingPadInst *LP = BB->getLandingPadInst();
- CalleePersonality = LP->getPersonalityFn();
- break;
- }
+ Constant *CalledPersonality =
+ CalledFunc->hasPersonalityFn()
+ ? CalledFunc->getPersonalityFn()->stripPointerCasts()
+ : nullptr;
// Find the personality function used by the landing pads of the caller. If it
// exists, then check to see that it matches the personality function used in
// the callee.
- if (CalleePersonality) {
- for (Function::const_iterator I = Caller->begin(), E = Caller->end();
- I != E; ++I)
- if (const InvokeInst *II = dyn_cast<InvokeInst>(I->getTerminator())) {
- const BasicBlock *BB = II->getUnwindDest();
- const LandingPadInst *LP = BB->getLandingPadInst();
-
- // If the personality functions match, then we can perform the
- // inlining. Otherwise, we can't inline.
- // TODO: This isn't 100% true. Some personality functions are proper
- // supersets of others and can be used in place of the other.
- if (LP->getPersonalityFn() != CalleePersonality)
- return false;
-
- break;
- }
+ Constant *CallerPersonality =
+ Caller->hasPersonalityFn()
+ ? Caller->getPersonalityFn()->stripPointerCasts()
+ : nullptr;
+ if (CalledPersonality) {
+ if (!CallerPersonality)
+ Caller->setPersonalityFn(CalledPersonality);
+ // If the personality functions match, then we can perform the
+ // inlining. Otherwise, we can't inline.
+ // TODO: This isn't 100% true. Some personality functions are proper
+ // supersets of others and can be used in place of the other.
+ else if (CalledPersonality != CallerPersonality)
+ return false;
}
// Get an iterator to the last basic block in the function, which will have
// the new function inlined after it.
- Function::iterator LastBlock = &Caller->back();
+ Function::iterator LastBlock = --Caller->end();
// Make sure to capture all of the return instructions from the cloned
// function.
// Keep a list of pair (dst, src) to emit byval initializations.
SmallVector<std::pair<Value*, Value*>, 4> ByValInit;
+ auto &DL = Caller->getParent()->getDataLayout();
+
assert(CalledFunc->arg_size() == CS.arg_size() &&
"No varargs calls can be inlined!");
ByValInit.push_back(std::make_pair(ActualArg, (Value*) *AI));
}
- VMap[I] = ActualArg;
+ VMap[&*I] = ActualArg;
}
// Add alignment assumptions if necessary. We do this before the inlined
// have no dead or constant instructions leftover after inlining occurs
// (which can happen, e.g., because an argument was constant), but we'll be
// happy with whatever the cloner can do.
- CloneAndPruneFunctionInto(Caller, CalledFunc, VMap,
+ CloneAndPruneFunctionInto(Caller, CalledFunc, VMap,
/*ModuleLevelChanges=*/false, Returns, ".i",
- &InlinedFunctionInfo, IFI.DL, TheCall);
+ &InlinedFunctionInfo, TheCall);
// Remember the first block that is newly cloned over.
FirstNewBlock = LastBlock; ++FirstNewBlock;
// Inject byval arguments initialization.
for (std::pair<Value*, Value*> &Init : ByValInit)
HandleByValArgumentInit(Init.first, Init.second, Caller->getParent(),
- FirstNewBlock, IFI);
+ &*FirstNewBlock, IFI);
+
+ if (CS.hasOperandBundles()) {
+ auto ParentDeopt = CS.getOperandBundleAt(0);
+ assert(ParentDeopt.getTagID() == LLVMContext::OB_deopt &&
+ "Checked on entry!");
+
+ SmallVector<OperandBundleDef, 2> OpDefs;
+
+ for (auto &VH : InlinedFunctionInfo.OperandBundleCallSites) {
+ Instruction *I = VH;
+
+ OpDefs.clear();
+
+ CallSite ICS(I);
+ OpDefs.reserve(ICS.getNumOperandBundles());
+
+ for (unsigned i = 0, e = ICS.getNumOperandBundles(); i < e; ++i) {
+ auto ChildOB = ICS.getOperandBundleAt(i);
+ if (ChildOB.getTagID() != LLVMContext::OB_deopt) {
+ // If the inlined call has other operand bundles, let them be
+ OpDefs.emplace_back(ChildOB);
+ continue;
+ }
+
+ // It may be useful to separate this logic (of handling operand
+ // bundles) out to a separate "policy" component if this gets crowded.
+ // Prepend the parent's deoptimization continuation to the newly
+ // inlined call's deoptimization continuation.
+ std::vector<Value *> MergedDeoptArgs;
+ MergedDeoptArgs.reserve(ParentDeopt.Inputs.size() +
+ ChildOB.Inputs.size());
+
+ MergedDeoptArgs.insert(MergedDeoptArgs.end(),
+ ParentDeopt.Inputs.begin(),
+ ParentDeopt.Inputs.end());
+ MergedDeoptArgs.insert(MergedDeoptArgs.end(), ChildOB.Inputs.begin(),
+ ChildOB.Inputs.end());
+
+ OpDefs.emplace_back(StringRef("deopt"), std::move(MergedDeoptArgs));
+ }
+
+ Instruction *NewI = nullptr;
+ if (isa<CallInst>(I))
+ NewI = CallInst::Create(cast<CallInst>(I), OpDefs, I);
+ else
+ NewI = InvokeInst::Create(cast<InvokeInst>(I), OpDefs, I);
+
+ // Note: the RAUW does the appropriate fixup in VMap, so we need to do
+ // this even if the call returns void.
+ I->replaceAllUsesWith(NewI);
+
+ VH = nullptr;
+ I->eraseFromParent();
+ }
+ }
// Update the callgraph if requested.
if (IFI.CG)
CloneAliasScopeMetadata(CS, VMap);
// Add noalias metadata if necessary.
- AddAliasScopeMetadata(CS, VMap, IFI.DL, IFI.AA);
+ AddAliasScopeMetadata(CS, VMap, DL, CalleeAAR);
// FIXME: We could register any cloned assumptions instead of clearing the
// whole function's cache.
// Transfer all of the allocas over in a block. Using splice means
// that the instructions aren't removed from the symbol table, then
// reinserted.
- Caller->getEntryBlock().getInstList().splice(InsertPoint,
- FirstNewBlock->getInstList(),
- AI, I);
+ Caller->getEntryBlock().getInstList().splice(
+ InsertPoint, FirstNewBlock->getInstList(), AI->getIterator(), I);
}
// Move any dbg.declares describing the allocas into the entry basic block.
DIBuilder DIB(*Caller->getParent());
// Leave lifetime markers for the static alloca's, scoping them to the
// function we just inlined.
if (InsertLifetime && !IFI.StaticAllocas.empty()) {
- IRBuilder<> builder(FirstNewBlock->begin());
+ IRBuilder<> builder(&FirstNewBlock->front());
for (unsigned ai = 0, ae = IFI.StaticAllocas.size(); ai != ae; ++ai) {
AllocaInst *AI = IFI.StaticAllocas[ai];
ConstantInt *AllocaSize = nullptr;
if (ConstantInt *AIArraySize =
dyn_cast<ConstantInt>(AI->getArraySize())) {
- if (IFI.DL) {
- Type *AllocaType = AI->getAllocatedType();
- uint64_t AllocaTypeSize = IFI.DL->getTypeAllocSize(AllocaType);
- uint64_t AllocaArraySize = AIArraySize->getLimitedValue();
- assert(AllocaArraySize > 0 && "array size of AllocaInst is zero");
- // Check that array size doesn't saturate uint64_t and doesn't
- // overflow when it's multiplied by type size.
- if (AllocaArraySize != ~0ULL &&
- UINT64_MAX / AllocaArraySize >= AllocaTypeSize) {
- AllocaSize = ConstantInt::get(Type::getInt64Ty(AI->getContext()),
- AllocaArraySize * AllocaTypeSize);
- }
+ auto &DL = Caller->getParent()->getDataLayout();
+ Type *AllocaType = AI->getAllocatedType();
+ uint64_t AllocaTypeSize = DL.getTypeAllocSize(AllocaType);
+ uint64_t AllocaArraySize = AIArraySize->getLimitedValue();
+
+ // Don't add markers for zero-sized allocas.
+ if (AllocaArraySize == 0)
+ continue;
+
+ // Check that array size doesn't saturate uint64_t and doesn't
+ // overflow when it's multiplied by type size.
+ if (AllocaArraySize != ~0ULL &&
+ UINT64_MAX / AllocaArraySize >= AllocaTypeSize) {
+ AllocaSize = ConstantInt::get(Type::getInt64Ty(AI->getContext()),
+ AllocaArraySize * AllocaTypeSize);
}
}
Function *StackRestore=Intrinsic::getDeclaration(M,Intrinsic::stackrestore);
// Insert the llvm.stacksave.
- CallInst *SavedPtr = IRBuilder<>(FirstNewBlock, FirstNewBlock->begin())
- .CreateCall(StackSave, "savedstack");
+ CallInst *SavedPtr = IRBuilder<>(&*FirstNewBlock, FirstNewBlock->begin())
+ .CreateCall(StackSave, {}, "savedstack");
// Insert a call to llvm.stackrestore before any return instructions in the
// inlined function.
// If we are inlining for an invoke instruction, we must make sure to rewrite
// any call instructions into invoke instructions.
- if (InvokeInst *II = dyn_cast<InvokeInst>(TheCall))
- HandleInlinedInvoke(II, FirstNewBlock, InlinedFunctionInfo);
+ if (auto *II = dyn_cast<InvokeInst>(TheCall)) {
+ BasicBlock *UnwindDest = II->getUnwindDest();
+ Instruction *FirstNonPHI = UnwindDest->getFirstNonPHI();
+ if (isa<LandingPadInst>(FirstNonPHI)) {
+ HandleInlinedLandingPad(II, &*FirstNewBlock, InlinedFunctionInfo);
+ } else {
+ HandleInlinedEHPad(II, &*FirstNewBlock, InlinedFunctionInfo);
+ }
+ }
// Handle any inlined musttail call sites. In order for a new call site to be
// musttail, the source of the clone and the inlined call site must have been
// the calling basic block.
if (Returns.size() == 1 && std::distance(FirstNewBlock, Caller->end()) == 1) {
// Move all of the instructions right before the call.
- OrigBB->getInstList().splice(TheCall, FirstNewBlock->getInstList(),
+ OrigBB->getInstList().splice(TheCall->getIterator(),
+ FirstNewBlock->getInstList(),
FirstNewBlock->begin(), FirstNewBlock->end());
// Remove the cloned basic block.
Caller->getBasicBlockList().pop_back();
// Split the basic block. This guarantees that no PHI nodes will have to be
// updated due to new incoming edges, and make the invoke case more
// symmetric to the call case.
- AfterCallBB = OrigBB->splitBasicBlock(CreatedBranchToNormalDest,
- CalledFunc->getName()+".exit");
+ AfterCallBB =
+ OrigBB->splitBasicBlock(CreatedBranchToNormalDest->getIterator(),
+ CalledFunc->getName() + ".exit");
} else { // It's a call
// If this is a call instruction, we need to split the basic block that
// the call lives in.
//
- AfterCallBB = OrigBB->splitBasicBlock(TheCall,
- CalledFunc->getName()+".exit");
+ AfterCallBB = OrigBB->splitBasicBlock(TheCall->getIterator(),
+ CalledFunc->getName() + ".exit");
}
// Change the branch that used to go to AfterCallBB to branch to the first
TerminatorInst *Br = OrigBB->getTerminator();
assert(Br && Br->getOpcode() == Instruction::Br &&
"splitBasicBlock broken!");
- Br->setOperand(0, FirstNewBlock);
-
+ Br->setOperand(0, &*FirstNewBlock);
// Now that the function is correct, make it a little bit nicer. In
// particular, move the basic blocks inserted from the end of the function
// into the space made by splitting the source basic block.
- Caller->getBasicBlockList().splice(AfterCallBB, Caller->getBasicBlockList(),
- FirstNewBlock, Caller->end());
+ Caller->getBasicBlockList().splice(AfterCallBB->getIterator(),
+ Caller->getBasicBlockList(), FirstNewBlock,
+ Caller->end());
// Handle all of the return instructions that we just cloned in, and eliminate
// any users of the original call/invoke instruction.
// possible incoming values.
if (!TheCall->use_empty()) {
PHI = PHINode::Create(RTy, Returns.size(), TheCall->getName(),
- AfterCallBB->begin());
+ &AfterCallBB->front());
// Anything that used the result of the function call should now use the
// PHI node as their operand.
TheCall->replaceAllUsesWith(PHI);
}
}
-
// Add a branch to the merge points and remove return instructions.
DebugLoc Loc;
for (unsigned i = 0, e = Returns.size(); i != e; ++i) {
// Splice the code entry block into calling block, right before the
// unconditional branch.
CalleeEntry->replaceAllUsesWith(OrigBB); // Update PHI nodes
- OrigBB->getInstList().splice(Br, CalleeEntry->getInstList());
+ OrigBB->getInstList().splice(Br->getIterator(), CalleeEntry->getInstList());
// Remove the unconditional branch.
OrigBB->getInstList().erase(Br);
// the entries are the same or undef). If so, remove the PHI so it doesn't
// block other optimizations.
if (PHI) {
- if (Value *V = SimplifyInstruction(PHI, IFI.DL, nullptr, nullptr,
+ auto &DL = Caller->getParent()->getDataLayout();
+ if (Value *V = SimplifyInstruction(PHI, DL, nullptr, nullptr,
&IFI.ACT->getAssumptionCache(*Caller))) {
PHI->replaceAllUsesWith(V);
PHI->eraseFromParent();
projects / oota-llvm.git / blobdiff