aboutsummaryrefslogtreecommitdiffstats
path: root/lib/Transforms/Utils/InlineFunction.cpp
diff options
context:
space:
mode:
Diffstat (limited to 'lib/Transforms/Utils/InlineFunction.cpp')
-rw-r--r--lib/Transforms/Utils/InlineFunction.cpp132
1 files changed, 68 insertions, 64 deletions
diff --git a/lib/Transforms/Utils/InlineFunction.cpp b/lib/Transforms/Utils/InlineFunction.cpp
index c2ef1ac..df3e1d4 100644
--- a/lib/Transforms/Utils/InlineFunction.cpp
+++ b/lib/Transforms/Utils/InlineFunction.cpp
@@ -89,7 +89,7 @@ namespace {
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;
@@ -99,17 +99,16 @@ namespace {
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);
}
@@ -124,7 +123,7 @@ namespace {
};
}
-/// getInnerResumeDest - Get or create a target for the branch from ResumeInsts.
+/// Get or create a target for the branch from ResumeInsts.
BasicBlock *InvokeInliningInfo::getInnerResumeDest() {
if (InnerResumeDest) return InnerResumeDest;
@@ -159,8 +158,8 @@ BasicBlock *InvokeInliningInfo::getInnerResumeDest() {
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,
@@ -178,9 +177,9 @@ void InvokeInliningInfo::forwardResume(ResumeInst *RI,
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,
@@ -228,7 +227,7 @@ static void HandleCallsInBlockInlinedThroughInvoke(BasicBlock *BB,
}
}
-/// 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
@@ -279,8 +278,8 @@ static void HandleInlinedInvoke(InvokeInst *II, BasicBlock *FirstNewBlock,
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
+/// 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
@@ -391,12 +390,12 @@ static void CloneAliasScopeMetadata(CallSite CS, ValueToValueMapTy &VMap) {
}
}
-/// 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, AliasAnalysis *AA) {
if (!EnableNoAliasConversion)
return;
@@ -622,8 +621,9 @@ static void AddAliasScopeMetadata(CallSite CS, ValueToValueMapTy &VMap,
/// 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.
@@ -645,20 +645,20 @@ static void AddAlignmentAssumptions(CallSite CS, InlineFunctionInfo &IFI) {
// 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,
+ if (getKnownAlignment(Arg, DL, CS.getInstruction(),
&IFI.ACT->getAssumptionCache(*CalledFunc),
- CS.getInstruction(), &DT) >= Align)
+ &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,
@@ -693,8 +693,15 @@ static void UpdateCallGraphAfterInlining(CallSite CS,
// 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);
@@ -726,11 +733,7 @@ static void HandleByValArgumentInit(Value *Dst, Value *Src, Module *M,
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
@@ -738,7 +741,7 @@ static void HandleByValArgumentInit(Value *Dst, Value *Src, Module *M,
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,
@@ -759,11 +762,13 @@ static Value *HandleByValArgument(Value *Arg, Instruction *TheCall,
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
@@ -771,10 +776,9 @@ static Value *HandleByValArgument(Value *Arg, Instruction *TheCall,
}
// 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).
@@ -789,8 +793,7 @@ static Value *HandleByValArgument(Value *Arg, Instruction *TheCall,
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)) {
@@ -805,7 +808,7 @@ static bool isUsedByLifetimeMarker(Value *V) {
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();
@@ -862,7 +865,7 @@ updateInlinedAtInfo(DebugLoc DL, MDLocation *InlinedAtNode,
return DebugLoc::get(DL.getLine(), DL.getCol(), DL.getScope(Ctx), 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) {
@@ -920,10 +923,9 @@ static void fixupLineNumbers(Function *Fn, Function::iterator FI,
}
}
-/// 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
@@ -1008,6 +1010,8 @@ bool llvm::InlineFunction(CallSite CS, InlineFunctionInfo &IFI,
// 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!");
@@ -1042,9 +1046,9 @@ bool llvm::InlineFunction(CallSite CS, InlineFunctionInfo &IFI,
// 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;
@@ -1065,7 +1069,7 @@ bool llvm::InlineFunction(CallSite CS, InlineFunctionInfo &IFI,
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.
@@ -1173,18 +1177,17 @@ bool llvm::InlineFunction(CallSite CS, InlineFunctionInfo &IFI,
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);
}
}
@@ -1445,7 +1448,8 @@ bool llvm::InlineFunction(CallSite CS, InlineFunctionInfo &IFI,
// 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();
generated by cgit v1.1 at 2024-12-22 11:47:34 +0000