#include "llvm/ADT/SetVector.h"
#include "llvm/ADT/StringExtras.h"
#include "llvm/Analysis/AliasAnalysis.h"
-#include "llvm/Analysis/AssumptionTracker.h"
+#include "llvm/Analysis/AssumptionCache.h"
#include "llvm/Analysis/CallGraph.h"
#include "llvm/Analysis/CaptureTracking.h"
#include "llvm/Analysis/InstructionSimplify.h"
#include "llvm/IR/DataLayout.h"
#include "llvm/IR/DebugInfo.h"
#include "llvm/IR/DerivedTypes.h"
+#include "llvm/IR/DIBuilder.h"
#include "llvm/IR/Dominators.h"
#include "llvm/IR/IRBuilder.h"
#include "llvm/IR/Instructions.h"
// Walk the existing metadata, adding the complete (perhaps cyclic) chain to
// the set.
- SmallVector<const Value *, 16> Queue(MD.begin(), MD.end());
+ SmallVector<const Metadata *, 16> Queue(MD.begin(), MD.end());
while (!Queue.empty()) {
const MDNode *M = cast<MDNode>(Queue.pop_back_val());
for (unsigned i = 0, ie = M->getNumOperands(); i != ie; ++i)
// Now we have a complete set of all metadata in the chains used to specify
// the noalias scopes and the lists of those scopes.
- SmallVector<MDNode *, 16> DummyNodes;
- DenseMap<const MDNode *, TrackingVH<MDNode> > MDMap;
+ SmallVector<TempMDTuple, 16> DummyNodes;
+ DenseMap<const MDNode *, TrackingMDNodeRef> MDMap;
for (SetVector<const MDNode *>::iterator I = MD.begin(), IE = MD.end();
I != IE; ++I) {
- MDNode *Dummy = MDNode::getTemporary(CalledFunc->getContext(), None);
- DummyNodes.push_back(Dummy);
- MDMap[*I] = Dummy;
+ DummyNodes.push_back(MDTuple::getTemporary(CalledFunc->getContext(), None));
+ MDMap[*I].reset(DummyNodes.back().get());
}
// Create new metadata nodes to replace the dummy nodes, replacing old
// node.
for (SetVector<const MDNode *>::iterator I = MD.begin(), IE = MD.end();
I != IE; ++I) {
- SmallVector<Value *, 4> NewOps;
+ SmallVector<Metadata *, 4> NewOps;
for (unsigned i = 0, ie = (*I)->getNumOperands(); i != ie; ++i) {
- const Value *V = (*I)->getOperand(i);
+ const Metadata *V = (*I)->getOperand(i);
if (const MDNode *M = dyn_cast<MDNode>(V))
NewOps.push_back(MDMap[M]);
else
- NewOps.push_back(const_cast<Value *>(V));
+ NewOps.push_back(const_cast<Metadata *>(V));
}
- MDNode *NewM = MDNode::get(CalledFunc->getContext(), NewOps),
- *TempM = MDMap[*I];
+ MDNode *NewM = MDNode::get(CalledFunc->getContext(), NewOps);
+ MDTuple *TempM = cast<MDTuple>(MDMap[*I]);
+ assert(TempM->isTemporary() && "Expected temporary node");
TempM->replaceAllUsesWith(NewM);
}
// which instructions inside it might belong), propagate those scopes to
// the inlined instructions.
if (MDNode *CSM =
- CS.getInstruction()->getMetadata(LLVMContext::MD_alias_scope))
+ CS.getInstruction()->getMetadata(LLVMContext::MD_alias_scope))
NewMD = MDNode::concatenate(NewMD, CSM);
NI->setMetadata(LLVMContext::MD_alias_scope, NewMD);
} else if (NI->mayReadOrWriteMemory()) {
if (MDNode *M =
- CS.getInstruction()->getMetadata(LLVMContext::MD_alias_scope))
+ CS.getInstruction()->getMetadata(LLVMContext::MD_alias_scope))
NI->setMetadata(LLVMContext::MD_alias_scope, M);
}
// which instructions inside it don't alias), propagate those scopes to
// the inlined instructions.
if (MDNode *CSM =
- CS.getInstruction()->getMetadata(LLVMContext::MD_noalias))
+ CS.getInstruction()->getMetadata(LLVMContext::MD_noalias))
NewMD = MDNode::concatenate(NewMD, CSM);
NI->setMetadata(LLVMContext::MD_noalias, NewMD);
} else if (NI->mayReadOrWriteMemory()) {
- if (MDNode *M =
- CS.getInstruction()->getMetadata(LLVMContext::MD_noalias))
+ if (MDNode *M = CS.getInstruction()->getMetadata(LLVMContext::MD_noalias))
NI->setMetadata(LLVMContext::MD_noalias, M);
}
}
-
- // Now that everything has been replaced, delete the dummy nodes.
- for (unsigned i = 0, ie = DummyNodes.size(); i != ie; ++i)
- MDNode::deleteTemporary(DummyNodes[i]);
}
/// AddAliasScopeMetadata - If the inlined function has noalias arguments, then
/// 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, AliasAnalysis *AA) {
if (!EnableNoAliasConversion)
return;
// need to go through several PHIs to see it, and thus could be
// repeated in the Objects list.
SmallPtrSet<const Value *, 4> ObjSet;
- SmallVector<Value *, 4> Scopes, NoAliases;
+ SmallVector<Metadata *, 4> Scopes, NoAliases;
SmallSetVector<const Argument *, 4> NAPtrArgs;
for (unsigned i = 0, ie = PtrArgs.size(); i != ie; ++i) {
}
if (!NoAliases.empty())
- NI->setMetadata(LLVMContext::MD_noalias, MDNode::concatenate(
- NI->getMetadata(LLVMContext::MD_noalias),
- MDNode::get(CalledFunc->getContext(), NoAliases)));
+ NI->setMetadata(LLVMContext::MD_noalias,
+ MDNode::concatenate(
+ NI->getMetadata(LLVMContext::MD_noalias),
+ MDNode::get(CalledFunc->getContext(), NoAliases)));
// Next, we want to figure out all of the sets to which we might belong.
// We might belong to a set if the noalias argument is in the set of
}
if (!Scopes.empty())
- NI->setMetadata(LLVMContext::MD_alias_scope, MDNode::concatenate(
- NI->getMetadata(LLVMContext::MD_alias_scope),
- MDNode::get(CalledFunc->getContext(), Scopes)));
+ NI->setMetadata(
+ LLVMContext::MD_alias_scope,
+ MDNode::concatenate(NI->getMetadata(LLVMContext::MD_alias_scope),
+ MDNode::get(CalledFunc->getContext(), Scopes)));
}
}
}
/// 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.
DominatorTree DT;
bool DTCalculated = false;
- const Function *CalledFunc = CS.getCalledFunction();
- for (Function::const_arg_iterator I = CalledFunc->arg_begin(),
- E = CalledFunc->arg_end(); I != E; ++I) {
+ Function *CalledFunc = CS.getCalledFunction();
+ for (Function::arg_iterator I = CalledFunc->arg_begin(),
+ E = CalledFunc->arg_end();
+ I != E; ++I) {
unsigned Align = I->getType()->isPointerTy() ? I->getParamAlignment() : 0;
if (Align && !I->hasByValOrInAllocaAttr() && !I->hasNUses(0)) {
if (!DTCalculated) {
// 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.AT, CS.getInstruction(),
+ if (getKnownAlignment(Arg, DL, CS.getInstruction(),
+ &IFI.ACT->getAssumptionCache(*CalledFunc),
&DT) >= Align)
continue;
- IRBuilder<>(CS.getInstruction()).CreateAlignmentAssumption(*IFI.DL, Arg,
- Align);
+ IRBuilder<>(CS.getInstruction())
+ .CreateAlignmentAssumption(DL, Arg, Align);
}
}
}
Type *AggTy = cast<PointerType>(Src->getType())->getElementType();
IRBuilder<> Builder(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
static Value *HandleByValArgument(Value *Arg, Instruction *TheCall,
const Function *CalledFunc,
InlineFunctionInfo &IFI,
- unsigned ByValAlignment,
- bool &AddedNewAllocas) {
+ unsigned ByValAlignment) {
PointerType *ArgTy = cast<PointerType>(Arg->getType());
Type *AggTy = ArgTy->getElementType();
+ Function *Caller = TheCall->getParent()->getParent();
+
// If the called function is readonly, then it could not mutate the caller's
// copy of the byval'd memory. In this case, it is safe to elide the copy and
// temporary.
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.AT, 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).
Align = std::max(Align, ByValAlignment);
- Function *Caller = TheCall->getParent()->getParent();
-
Value *NewAlloca = new AllocaInst(AggTy, nullptr, Align, Arg->getName(),
&*Caller->begin()->begin());
IFI.StaticAllocas.push_back(cast<AllocaInst>(NewAlloca));
// Uses of the argument in the function should use our new alloca
// instead.
- AddedNewAllocas = true;
return NewAlloca;
}
return false;
}
-/// updateInlinedAtInfo - Helper function used by fixupLineNumbers to
-/// recursively update InlinedAtEntry of a DebugLoc.
-static DebugLoc updateInlinedAtInfo(const DebugLoc &DL,
- const DebugLoc &InlinedAtDL,
- LLVMContext &Ctx) {
- if (MDNode *IA = DL.getInlinedAt(Ctx)) {
- DebugLoc NewInlinedAtDL
- = updateInlinedAtInfo(DebugLoc::getFromDILocation(IA), InlinedAtDL, Ctx);
- return DebugLoc::get(DL.getLine(), DL.getCol(), DL.getScope(Ctx),
- NewInlinedAtDL.getAsMDNode(Ctx));
+/// 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;
+
+ // Gather all the inlined-at nodes
+ while (MDLocation *IA =
+ cast_or_null<MDLocation>(CurInlinedAt.getInlinedAt(Ctx))) {
+ // Skip any we've already built nodes for
+ if (MDLocation *Found = IANodes[IA]) {
+ Last = Found;
+ break;
+ }
+
+ InlinedAtLocations.push_back(IA);
+ CurInlinedAt = DebugLoc::getFromDILocation(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(
+ Ctx, MD->getLine(), MD->getColumn(), MD->getScope(), Last);
}
- return DebugLoc::get(DL.getLine(), DL.getCol(), DL.getScope(Ctx),
- InlinedAtDL.getAsMDNode(Ctx));
+ // 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);
}
/// fixupLineNumbers - Update inlined instructions' line numbers to
if (TheCallDL.isUnknown())
return;
+ auto &Ctx = Fn->getContext();
+ auto *InlinedAtNode = cast<MDLocation>(TheCallDL.getAsMDNode(Ctx));
+
+ // Create a unique call site, not to be confused with any other call from the
+ // same location.
+ InlinedAtNode = MDLocation::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;
+
for (; FI != Fn->end(); ++FI) {
for (BasicBlock::iterator BI = FI->begin(), BE = FI->end();
BI != BE; ++BI) {
BI->setDebugLoc(TheCallDL);
} else {
- BI->setDebugLoc(updateInlinedAtInfo(DL, TheCallDL, BI->getContext()));
+ 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, createInlinedVariable(DVI->getVariable(),
- InlinedAt, 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)));
}
}
}
SmallVector<ReturnInst*, 8> Returns;
ClonedCodeInfo InlinedFunctionInfo;
Function::iterator FirstNewBlock;
- bool AddedNewAllocas = false;
{ // Scope to destroy VMap after cloning.
ValueToValueMapTy VMap;
// 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!");
// modify the struct.
if (CS.isByValArgument(ArgNo)) {
ActualArg = HandleByValArgument(ActualArg, TheCall, CalledFunc, IFI,
- CalledFunc->getParamAlignment(ArgNo+1),
- AddedNewAllocas);
+ CalledFunc->getParamAlignment(ArgNo+1));
if (ActualArg != *AI)
ByValInit.push_back(std::make_pair(ActualArg, (Value*) *AI));
}
// 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;
CloneAliasScopeMetadata(CS, VMap);
// Add noalias metadata if necessary.
- AddAliasScopeMetadata(CS, VMap, IFI.DL, IFI.AA);
+ AddAliasScopeMetadata(CS, VMap, DL, IFI.AA);
// FIXME: We could register any cloned assumptions instead of clearing the
// whole function's cache.
- if (IFI.AT)
- IFI.AT->forgetCachedAssumptions(Caller);
+ if (IFI.ACT)
+ IFI.ACT->getAssumptionCache(*Caller).clear();
}
// If there are any alloca instructions in the block that used to be the entry
FirstNewBlock->getInstList(),
AI, I);
}
+ // Move any dbg.declares describing the allocas into the entry basic block.
+ DIBuilder DIB(*Caller->getParent());
+ for (auto &AI : IFI.StaticAllocas)
+ replaceDbgDeclareForAlloca(AI, AI, DIB, /*Deref=*/false);
}
bool InlinedMustTailCalls = false;
// f -> musttail g -> tail f ==> f -> tail f
// f -> g -> musttail f ==> f -> f
// f -> g -> tail f ==> f -> f
- //
- // If an alloca was introduced in the frame due to a byval parameter
- // being passed to a subsequent call, tail calls must have the tail
- // stripped as they may not access variables in the caller's stack.
- // A single alloca ripples through out as the alloca may be aliased by
- // bitcasts or may escape and be mutated outside of the function.
CallInst::TailCallKind ChildTCK = CI->getTailCallKind();
ChildTCK = std::min(CallSiteTailKind, ChildTCK);
- if (AddedNewAllocas)
- CI->setTailCallKind(CallInst::TCK_None);
- else
- CI->setTailCallKind(ChildTCK);
+ CI->setTailCallKind(ChildTCK);
InlinedMustTailCalls |= CI->isMustTailCall();
// Calls inlined through a 'nounwind' call site should be marked
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();
+ 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);
}
}
// 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, IFI.AT)) {
+ auto &DL = Caller->getParent()->getDataLayout();
+ if (Value *V = SimplifyInstruction(PHI, DL, nullptr, nullptr,
+ &IFI.ACT->getAssumptionCache(*Caller))) {
PHI->replaceAllUsesWith(V);
PHI->eraseFromParent();
}