[llvm] [Attributor] Change allocation size and load/store offsets using AAPointerInfo for Alloca instructions (PR #72029)
Vidush Singhal via llvm-commits
llvm-commits at lists.llvm.org
Fri Jun 21 10:17:06 PDT 2024
https://github.com/vidsinghal updated https://github.com/llvm/llvm-project/pull/72029
>From 3bbdf3e089e6117c964fd5dbce32f0d3f0c7e2e5 Mon Sep 17 00:00:00 2001
From: Vidush Singhal <singhal2 at ruby964.llnl.gov>
Date: Tue, 4 Jun 2024 16:51:06 -0700
Subject: [PATCH] [Attributor]: Change allocation size and load/store offsets
using AAPointerInfo for Alloca instructions
---
llvm/include/llvm/Transforms/IPO/Attributor.h | 152 +++++++
.../Transforms/IPO/AttributorAttributes.cpp | 430 ++++++++++++++----
.../Attributor/ArgumentPromotion/crash.ll | 6 +-
.../live_called_from_dead.ll | 3 +-
.../live_called_from_dead_2.ll | 3 +-
.../nonzero-address-spaces.ll | 3 +-
.../Attributor/IPConstantProp/pthreads.ll | 4 +-
llvm/test/Transforms/Attributor/allocator.ll | 193 ++++++--
.../Attributor/call-simplify-pointer-info.ll | 28 +-
.../Transforms/Attributor/heap_to_stack.ll | 3 +-
.../Attributor/heap_to_stack_gpu.ll | 3 +-
llvm/test/Transforms/Attributor/liveness.ll | 8 +-
llvm/test/Transforms/Attributor/nodelete.ll | 5 +-
.../Transforms/Attributor/pointer-info.ll | 6 +-
.../Attributor/value-simplify-pointer-info.ll | 18 +-
15 files changed, 689 insertions(+), 176 deletions(-)
diff --git a/llvm/include/llvm/Transforms/IPO/Attributor.h b/llvm/include/llvm/Transforms/IPO/Attributor.h
index 6ba04dbc31db3..f3b506e8eecf9 100644
--- a/llvm/include/llvm/Transforms/IPO/Attributor.h
+++ b/llvm/include/llvm/Transforms/IPO/Attributor.h
@@ -5751,8 +5751,10 @@ struct AANonConvergent : public StateWrapper<BooleanState, AbstractAttribute> {
/// An abstract interface for struct information.
struct AAPointerInfo : public AbstractAttribute {
+protected:
AAPointerInfo(const IRPosition &IRP) : AbstractAttribute(IRP) {}
+public:
/// See AbstractAttribute::isValidIRPositionForInit
static bool isValidIRPositionForInit(Attributor &A, const IRPosition &IRP) {
if (!IRP.getAssociatedType()->isPtrOrPtrVectorTy())
@@ -6106,6 +6108,56 @@ struct AAPointerInfo : public AbstractAttribute {
Type *Ty;
};
+ /// A helper containing a list of offsets computed for a Use. Ideally this
+ /// list should be strictly ascending, but we ensure that only when we
+ /// actually translate the list of offsets to a RangeList.
+ struct OffsetInfo {
+ using VecTy = SmallVector<int64_t>;
+ using const_iterator = VecTy::const_iterator;
+ VecTy Offsets;
+
+ const_iterator begin() const { return Offsets.begin(); }
+ const_iterator end() const { return Offsets.end(); }
+
+ bool operator==(const OffsetInfo &RHS) const {
+ return Offsets == RHS.Offsets;
+ }
+
+ bool operator!=(const OffsetInfo &RHS) const { return !(*this == RHS); }
+
+ void insert(int64_t Offset) { Offsets.push_back(Offset); }
+ bool isUnassigned() const { return Offsets.empty(); }
+
+ bool isUnknown() const {
+ if (isUnassigned())
+ return false;
+ if (Offsets.size() == 1)
+ return Offsets.front() == AA::RangeTy::Unknown;
+ return false;
+ }
+
+ void setUnknown() {
+ Offsets.clear();
+ Offsets.push_back(AA::RangeTy::Unknown);
+ }
+
+ void addToAll(int64_t Inc) {
+ for (auto &Offset : Offsets)
+ Offset += Inc;
+ }
+
+ /// Copy offsets from \p R into the current list.
+ ///
+ /// Ideally all lists should be strictly ascending, but we defer that to the
+ /// actual use of the list. So we just blindly append here.
+ void merge(const OffsetInfo &R) {
+ Offsets.append(R.Offsets);
+ // ensure elements are unique.
+ sort(Offsets.begin(), Offsets.end());
+ Offsets.erase(std::unique(Offsets.begin(), Offsets.end()), Offsets.end());
+ }
+ };
+
/// Create an abstract attribute view for the position \p IRP.
static AAPointerInfo &createForPosition(const IRPosition &IRP, Attributor &A);
@@ -6120,6 +6172,9 @@ struct AAPointerInfo : public AbstractAttribute {
virtual const_bin_iterator begin() const = 0;
virtual const_bin_iterator end() const = 0;
virtual int64_t numOffsetBins() const = 0;
+ virtual void dumpState(raw_ostream &O) const = 0;
+ virtual const Access &getBinAccess(unsigned Index) const = 0;
+ virtual const DenseMap<Value *, OffsetInfo> &getOffsetInfoMap() const = 0;
/// Call \p CB on all accesses that might interfere with \p Range and return
/// true if all such accesses were known and the callback returned true for
@@ -6149,6 +6204,9 @@ struct AAPointerInfo : public AbstractAttribute {
return (AA->getIdAddr() == &ID);
}
+ /// Offsets Info Map
+ DenseMap<Value *, OffsetInfo> OffsetInfoMap;
+
/// Unique ID (due to the unique address)
static const char ID;
};
@@ -6285,12 +6343,106 @@ struct AAAllocationInfo : public StateWrapper<BooleanState, AbstractAttribute> {
return AbstractAttribute::isValidIRPositionForInit(A, IRP);
}
+ // A helper function to check if simplified values exists for the current
+ // instruction.
+ // TODO: handle case for a similified value
+ // Right now we don't change the value and give up
+ // on modifying the size and offsets of the allocation
+ // this may be sub-optimal
+ bool checkIfSimplifiedValuesExists(Attributor &A, Instruction *LocalInst) {
+
+ // If there are potential values that replace the accessed instruction, we
+ // should use those instead
+ bool UsedAssumedInformation = false;
+ SmallVector<AA::ValueAndContext> Values;
+ if (A.getAssumedSimplifiedValues(IRPosition::inst(*LocalInst), *this,
+ Values, AA::AnyScope,
+ UsedAssumedInformation))
+
+ for (auto &ValAndContext : Values)
+ // don't modify instruction if any simplified value exists
+ if (ValAndContext.getValue() && ValAndContext.getValue() != LocalInst)
+ return true;
+
+ return false;
+ }
+
+ // A helper function to back track the pointer operand to
+ // a GEP that calculates the pointer operand from the original allocation.
+ // So that we can correct the type of the GEP.
+ // If such a GEP is not found, we return nullptr.
+ Instruction *
+ backTrackPointerOperandToGepFromAllocation(Instruction *PointerOperand,
+ Instruction *Allocation) {
+
+ SmallVector<Instruction *> ReadyList;
+ ReadyList.push_back(PointerOperand);
+ while (!ReadyList.empty()) {
+ Instruction *Back = ReadyList.back();
+ ReadyList.pop_back();
+
+ if (Back == Allocation)
+ continue;
+
+ for (auto *It = Back->op_begin(); It != Back->op_end(); It++) {
+ Instruction *OperandInstruction = dyn_cast<Instruction>(It);
+
+ if (!OperandInstruction)
+ continue;
+
+ // This is the GEP instruction whose type and offsets we can change
+ // without any illegal memory access or poison result.
+ if (Back->getOpcode() == Instruction::GetElementPtr &&
+ OperandInstruction == Allocation)
+ return Back;
+
+ ReadyList.push_back(OperandInstruction);
+ }
+ }
+ return nullptr;
+ }
+
+ bool checkIfAccessChainUsesMultipleBins(
+ Attributor &A, Instruction *LocalInst,
+ const DenseMap<Value *, AAPointerInfo::OffsetInfo> &OffsetInfoMap) {
+
+ // BackTrack and check if there are multiple bins for instructions in
+ // the
+ // chain
+ std::vector<Instruction *> ReadyList;
+ DenseMap<Instruction *, bool> Visited;
+ ReadyList.push_back(LocalInst);
+ while (!ReadyList.empty()) {
+ Instruction *GetBack = ReadyList.back();
+ ReadyList.pop_back();
+ // check if the Instruction has multiple bins, if so give up
+ // for calls it is okay to have multiple bins
+ // TODO: handle when one instruction has multiple bins
+ auto OffsetsVecArg = OffsetInfoMap.lookup(GetBack).Offsets;
+ if (GetBack->getOpcode() != Instruction::Call && OffsetsVecArg.size() > 1)
+ return true;
+
+ for (auto *It = GetBack->op_begin(); It != GetBack->op_end(); It++) {
+ if (Instruction *Ins = dyn_cast<Instruction>(*It)) {
+ if (!Visited[Ins])
+ ReadyList.push_back(Ins);
+ }
+ }
+ Visited[GetBack] = true;
+ }
+
+ return false;
+ }
+
/// Create an abstract attribute view for the position \p IRP.
static AAAllocationInfo &createForPosition(const IRPosition &IRP,
Attributor &A);
virtual std::optional<TypeSize> getAllocatedSize() const = 0;
+ using NewOffsetsTy = DenseMap<AA::RangeTy, AA::RangeTy>;
+ virtual const NewOffsetsTy &getNewOffsets() const = 0;
+
/// See AbstractAttribute::getName()
const std::string getName() const override { return "AAAllocationInfo"; }
diff --git a/llvm/lib/Transforms/IPO/AttributorAttributes.cpp b/llvm/lib/Transforms/IPO/AttributorAttributes.cpp
index 57579bbca00ee..9a324091f71c6 100644
--- a/llvm/lib/Transforms/IPO/AttributorAttributes.cpp
+++ b/llvm/lib/Transforms/IPO/AttributorAttributes.cpp
@@ -1002,54 +1002,9 @@ ChangeStatus AA::PointerInfo::State::addAccess(
namespace {
-/// A helper containing a list of offsets computed for a Use. Ideally this
-/// list should be strictly ascending, but we ensure that only when we
-/// actually translate the list of offsets to a RangeList.
-struct OffsetInfo {
- using VecTy = SmallVector<int64_t>;
- using const_iterator = VecTy::const_iterator;
- VecTy Offsets;
-
- const_iterator begin() const { return Offsets.begin(); }
- const_iterator end() const { return Offsets.end(); }
-
- bool operator==(const OffsetInfo &RHS) const {
- return Offsets == RHS.Offsets;
- }
-
- bool operator!=(const OffsetInfo &RHS) const { return !(*this == RHS); }
-
- void insert(int64_t Offset) { Offsets.push_back(Offset); }
- bool isUnassigned() const { return Offsets.size() == 0; }
-
- bool isUnknown() const {
- if (isUnassigned())
- return false;
- if (Offsets.size() == 1)
- return Offsets.front() == AA::RangeTy::Unknown;
- return false;
- }
-
- void setUnknown() {
- Offsets.clear();
- Offsets.push_back(AA::RangeTy::Unknown);
- }
-
- void addToAll(int64_t Inc) {
- for (auto &Offset : Offsets) {
- Offset += Inc;
- }
- }
-
- /// Copy offsets from \p R into the current list.
- ///
- /// Ideally all lists should be strictly ascending, but we defer that to the
- /// actual use of the list. So we just blindly append here.
- void merge(const OffsetInfo &R) { Offsets.append(R.Offsets); }
-};
-
#ifndef NDEBUG
-static raw_ostream &operator<<(raw_ostream &OS, const OffsetInfo &OI) {
+static raw_ostream &operator<<(raw_ostream &OS,
+ const AAPointerInfo::OffsetInfo &OI) {
ListSeparator LS;
OS << "[";
for (auto Offset : OI) {
@@ -1084,6 +1039,15 @@ struct AAPointerInfoImpl
return State::numOffsetBins();
}
+ virtual const Access &getBinAccess(unsigned Index) const override {
+ return getAccess(Index);
+ }
+
+ virtual const DenseMap<Value *, OffsetInfo> &
+ getOffsetInfoMap() const override {
+ return OffsetInfoMap;
+ }
+
bool forallInterferingAccesses(
AA::RangeTy Range,
function_ref<bool(const AAPointerInfo::Access &, bool)> CB)
@@ -1430,7 +1394,7 @@ struct AAPointerInfoImpl
void trackPointerInfoStatistics(const IRPosition &IRP) const {}
/// Dump the state into \p O.
- void dumpState(raw_ostream &O) {
+ virtual void dumpState(raw_ostream &O) const override {
for (auto &It : OffsetBins) {
O << "[" << It.first.Offset << "-" << It.first.Offset + It.first.Size
<< "] : " << It.getSecond().size() << "\n";
@@ -1464,6 +1428,7 @@ struct AAPointerInfoFloating : public AAPointerInfoImpl {
std::optional<Value *> Content, AccessKind Kind,
SmallVectorImpl<int64_t> &Offsets, ChangeStatus &Changed,
Type &Ty) {
+
using namespace AA::PointerInfo;
auto Size = AA::RangeTy::Unknown;
const DataLayout &DL = A.getDataLayout();
@@ -1596,7 +1561,7 @@ ChangeStatus AAPointerInfoFloating::updateImpl(Attributor &A) {
const DataLayout &DL = A.getDataLayout();
Value &AssociatedValue = getAssociatedValue();
- DenseMap<Value *, OffsetInfo> OffsetInfoMap;
+ OffsetInfoMap.clear();
OffsetInfoMap[&AssociatedValue].insert(0);
auto HandlePassthroughUser = [&](Value *Usr, Value *CurPtr, bool &Follow) {
@@ -12663,6 +12628,11 @@ struct AAAllocationInfoImpl : public AAAllocationInfo {
return AssumedAllocatedSize;
}
+ const NewOffsetsTy &getNewOffsets() const override {
+ assert(isValidState() && "the AA is invalid");
+ return NewComputedOffsets;
+ }
+
std::optional<TypeSize> findInitialAllocationSize(Instruction *I,
const DataLayout &DL) {
@@ -12703,46 +12673,56 @@ struct AAAllocationInfoImpl : public AAAllocationInfo {
const DataLayout &DL = A.getDataLayout();
const auto AllocationSize = findInitialAllocationSize(I, DL);
- // If allocation size is nullopt, we give up.
+ // If allocation size is nullopt, we give up
if (!AllocationSize)
return indicatePessimisticFixpoint();
- // For zero sized allocations, we give up.
+ // For zero sized allocations, we give up
// Since we can't reduce further
if (*AllocationSize == 0)
return indicatePessimisticFixpoint();
- int64_t BinSize = PI->numOffsetBins();
-
- // TODO: implement for multiple bins
- if (BinSize > 1)
- return indicatePessimisticFixpoint();
-
- if (BinSize == 0) {
+ int64_t NumBins = PI->numOffsetBins();
+ if (NumBins == 0) {
auto NewAllocationSize = std::optional<TypeSize>(TypeSize(0, false));
if (!changeAllocationSize(NewAllocationSize))
return ChangeStatus::UNCHANGED;
return ChangeStatus::CHANGED;
}
- // TODO: refactor this to be part of multiple bin case
- const auto &It = PI->begin();
+ // For each access bin
+ // Compute its new start Offset and store the results in a new map
+ // (NewOffsetBins)
+ unsigned long PrevBinEndOffset = 0;
+ bool ChangedOffsets = false;
+ for (AAPointerInfo::OffsetBinsTy::const_iterator It = PI->begin();
+ It != PI->end(); It++) {
+ const AA::RangeTy &OldRange = It->getFirst();
- // TODO: handle if Offset is not zero
- if (It->first.Offset != 0)
- return indicatePessimisticFixpoint();
+ // If any range has an unknown offset or size, we should leave the
+ // allocation unmodified
+ if (OldRange.offsetOrSizeAreUnknown())
+ return indicatePessimisticFixpoint();
- uint64_t SizeOfBin = It->first.Offset + It->first.Size;
+ unsigned long NewStartOffset = PrevBinEndOffset;
+ unsigned long NewEndOffset = NewStartOffset + OldRange.Size;
+ PrevBinEndOffset = NewEndOffset;
- if (SizeOfBin >= *AllocationSize)
- return indicatePessimisticFixpoint();
+ ChangedOffsets |= setNewOffsets(OldRange, OldRange.Offset, NewStartOffset,
+ OldRange.Size);
+ }
+ // Set the new size of the allocation, the new size of the Allocation should
+ // be the size of PrevBinEndOffset * 8, in bits
auto NewAllocationSize =
- std::optional<TypeSize>(TypeSize(SizeOfBin * 8, false));
+ std::optional<TypeSize>(TypeSize(PrevBinEndOffset * 8, false));
if (!changeAllocationSize(NewAllocationSize))
return ChangeStatus::UNCHANGED;
+ if (!ChangedOffsets)
+ return ChangeStatus::UNCHANGED;
+
return ChangeStatus::CHANGED;
}
@@ -12752,39 +12732,284 @@ struct AAAllocationInfoImpl : public AAAllocationInfo {
assert(isValidState() &&
"Manifest should only be called if the state is valid.");
- Instruction *I = getIRPosition().getCtxI();
+ bool Changed = false;
+ const IRPosition &IRP = getIRPosition();
+ Instruction *I = IRP.getCtxI();
- auto FixedAllocatedSizeInBits = getAllocatedSize()->getFixedValue();
+ // check if simplified values exist
+ if (checkIfSimplifiedValuesExists(A, I))
+ return ChangeStatus::UNCHANGED;
- unsigned long NumBytesToAllocate = (FixedAllocatedSizeInBits + 7) / 8;
+ if (getAllocatedSize() == HasNoAllocationSize)
+ return ChangeStatus::UNCHANGED;
- switch (I->getOpcode()) {
- // TODO: add case for malloc like calls
- case Instruction::Alloca: {
+ const AAPointerInfo *PI =
+ A.getOrCreateAAFor<AAPointerInfo>(IRP, *this, DepClassTy::REQUIRED);
- AllocaInst *AI = cast<AllocaInst>(I);
+ if (!PI)
+ return ChangeStatus::UNCHANGED;
- Type *CharType = Type::getInt8Ty(I->getContext());
+ if (!PI->getState().isValidState())
+ return ChangeStatus::UNCHANGED;
- auto *NumBytesToValue =
- ConstantInt::get(I->getContext(), APInt(32, NumBytesToAllocate));
+ // Store a map where each instruction maps to a set of bins accessed by that
+ // instruction
+ DenseMap<Instruction *, DenseMap<AA::RangeTy, AA::RangeTy>>
+ AccessedInstructionsToBinsMap;
- BasicBlock::iterator insertPt = AI->getIterator();
- insertPt = std::next(insertPt);
- AllocaInst *NewAllocaInst =
- new AllocaInst(CharType, AI->getAddressSpace(), NumBytesToValue,
- AI->getAlign(), AI->getName(), insertPt);
+ auto AddBins =
+ [](DenseMap<Instruction *, DenseMap<AA::RangeTy, AA::RangeTy>> &Map,
+ Instruction *LocalInst, const AA::RangeTy &OldRange,
+ const AA::RangeTy &NewRange) {
+ DenseMap<AA::RangeTy, AA::RangeTy> &NewBinsForInstruction =
+ Map.getOrInsertDefault(LocalInst);
- if (A.changeAfterManifest(IRPosition::inst(*AI), *NewAllocaInst))
- return ChangeStatus::CHANGED;
+ NewBinsForInstruction.insert(std::make_pair(OldRange, NewRange));
+ };
+
+ const auto &NewOffsetsMap = getNewOffsets();
+ const auto &OffsetInfoMap = PI->getOffsetInfoMap();
+
+ // Map access causing instructions to tuple of (Old, New) bins.
+ // The access causing instruction contains the pointer operand
+ // which comes from the allocation
+ // We may want to backtrack that pointer operand, there are 2 cases that may
+ // arise:
+ // A) A GEP exists that calculates the pointer operand from the original
+ // allocation instruction: I
+ // B) A GEP does not exists
+ // in which case we need to insert a GEP just before the access causing
+ // instruction with the shift value from the original offset.
+ for (AAPointerInfo::OffsetBinsTy::const_iterator It = PI->begin();
+ It != PI->end(); It++) {
+ const auto &OldOffsetRange = It->getFirst();
+ // If the OldOffsetRange is not in the map, offsets for that bin did not
+ // change. We should just continue and skip changing the offsets in that
+ // case
+ if (!NewOffsetsMap.contains(OldOffsetRange))
+ continue;
+
+ const auto &NewOffsetRange = NewOffsetsMap.lookup(OldOffsetRange);
+ for (const auto AccIndex : It->getSecond()) {
+ const auto &AccessInstruction = PI->getBinAccess(AccIndex);
+ Instruction *LocalInst = AccessInstruction.getLocalInst();
+
+ if (checkIfSimplifiedValuesExists(A, LocalInst))
+ return ChangeStatus::UNCHANGED;
+
+ if (checkIfAccessChainUsesMultipleBins(A, LocalInst, OffsetInfoMap))
+ return ChangeStatus::UNCHANGED;
+
+ // check if we can backtrack the access causing instruction to a GEP
+ // from the original allocation, if yes, then we prefer to change the
+ // GEP rather than the access causing instruction
+ switch (LocalInst->getOpcode()) {
+ case Instruction::Load: {
+ LoadInst *Load = cast<LoadInst>(LocalInst);
+ Instruction *PointerOperand =
+ cast<Instruction>(Load->getPointerOperand());
+ Instruction *BackTrackedGEP =
+ backTrackPointerOperandToGepFromAllocation(PointerOperand, I);
+
+ if (!BackTrackedGEP) {
+ AddBins(AccessedInstructionsToBinsMap, LocalInst, OldOffsetRange,
+ NewOffsetRange);
+ break;
+ }
+
+ if (checkIfSimplifiedValuesExists(A, BackTrackedGEP))
+ return ChangeStatus::UNCHANGED;
+ AddBins(AccessedInstructionsToBinsMap, BackTrackedGEP, OldOffsetRange,
+ NewOffsetRange);
+ break;
+ }
+ case Instruction::Store: {
+ StoreInst *Store = cast<StoreInst>(LocalInst);
+ Instruction *PointerOperand =
+ cast<Instruction>(Store->getPointerOperand());
+ Instruction *BackTrackedGEP =
+ backTrackPointerOperandToGepFromAllocation(PointerOperand, I);
+
+ if (!BackTrackedGEP) {
+ AddBins(AccessedInstructionsToBinsMap, LocalInst, OldOffsetRange,
+ NewOffsetRange);
+ break;
+ }
+
+ if (checkIfSimplifiedValuesExists(A, BackTrackedGEP))
+ return ChangeStatus::UNCHANGED;
+
+ AddBins(AccessedInstructionsToBinsMap, BackTrackedGEP, OldOffsetRange,
+ NewOffsetRange);
+ break;
+ }
+ case Instruction::Call: {
+ CallInst *CallInstruction = cast<CallInst>(LocalInst);
+ for (auto *It = CallInstruction->op_begin();
+ It != CallInstruction->op_end(); It++) {
+ if (Instruction *OperandInstruction = dyn_cast<Instruction>(It)) {
+ // Operand does not have any accessed offsets
+ if (!OffsetInfoMap.contains(OperandInstruction))
+ continue;
+
+ // Find the old offset and the corresponding new offset for the
+ // call argument
+ auto OffsetsVecArg =
+ OffsetInfoMap.lookup(OperandInstruction).Offsets;
+ int64_t OldOffsetArg = OffsetsVecArg.front();
+ int NewOffsetArg = 0;
+ for (auto OldToNewRange : NewOffsetsMap) {
+ auto Old = OldToNewRange.getFirst();
+ if (Old.Offset == OldOffsetArg)
+ NewOffsetArg = OldToNewRange.getSecond().Offset;
+ }
+
+ // If the offsets did not change, no need to change the offsets.
+ if (NewOffsetArg == OldOffsetArg)
+ continue;
+
+ // We don't have access to the size of the offset here but its ok
+ // since we don't need it here.
+ AA::RangeTy CallArgOldRange = AA::RangeTy(OldOffsetArg, -1);
+ AA::RangeTy CallArgNewRange = AA::RangeTy(NewOffsetArg, -1);
+ Instruction *BackTrackedGEP =
+ backTrackPointerOperandToGepFromAllocation(OperandInstruction,
+ I);
+
+ if (!BackTrackedGEP) {
+ AddBins(AccessedInstructionsToBinsMap, OperandInstruction,
+ CallArgOldRange, CallArgNewRange);
+ continue;
+ }
+
+ if (checkIfSimplifiedValuesExists(A, BackTrackedGEP))
+ return ChangeStatus::UNCHANGED;
+
+ AddBins(AccessedInstructionsToBinsMap, BackTrackedGEP,
+ CallArgOldRange, CallArgNewRange);
+ }
+ }
+ break;
+ }
+ default: {
+ AddBins(AccessedInstructionsToBinsMap, LocalInst, OldOffsetRange,
+ NewOffsetRange);
+ break;
+ }
+ }
+ }
+ }
+
+ unsigned long FixedAllocatedSizeInBits =
+ getAllocatedSize()->getFixedValue();
+ unsigned long NumBytesToAllocate = (FixedAllocatedSizeInBits + 7) / 8;
+ Type *NewAllocationType = nullptr;
+ switch (I->getOpcode()) {
+ // TODO: add case for malloc like calls
+ case Instruction::Alloca: {
+ AllocaInst *OldAllocaInst = cast<AllocaInst>(I);
+ const DataLayout &DL = A.getDataLayout();
+ auto OriginalAllocationSize = OldAllocaInst->getAllocationSizeInBits(DL);
+
+ if (*OriginalAllocationSize <= FixedAllocatedSizeInBits)
+ return ChangeStatus::UNCHANGED;
+
+ Type *CharType = Type::getInt8Ty(I->getContext());
+ Type *CharArrayType = ArrayType::get(CharType, NumBytesToAllocate);
+ NewAllocationType = CharArrayType;
+ BasicBlock::iterator InsertPt = OldAllocaInst->getIterator();
+ InsertPt = std::next(InsertPt);
+ Instruction *NewAllocationInstruction =
+ new AllocaInst(CharArrayType, OldAllocaInst->getAddressSpace(),
+ OldAllocaInst->getName(), InsertPt);
+
+ Changed |= A.changeAfterManifest(IRPosition::inst(*I),
+ *NewAllocationInstruction);
+ A.deleteAfterManifest(*I);
break;
}
default:
break;
}
- return ChangeStatus::UNCHANGED;
+ for (auto &It : AccessedInstructionsToBinsMap) {
+ Instruction *LocalInst = It.first;
+ // Get a hold of a map, mapping old to new bins
+ DenseMap<AA::RangeTy, AA::RangeTy> &OldToNewBins = It.second;
+ IntegerType *Int64TyInteger =
+ IntegerType::get(LocalInst->getContext(), 64);
+ switch (LocalInst->getOpcode()) {
+ case Instruction::Load: {
+ // The number of bytes to shift the load/store by
+ int64_t OffsetOld = OldToNewBins.begin()->getFirst().Offset;
+ int64_t OffsetNew = OldToNewBins.begin()->getSecond().Offset;
+ LoadInst *OldLoadInst = cast<LoadInst>(LocalInst);
+ Instruction *PointerOperand =
+ cast<Instruction>(OldLoadInst->getPointerOperand());
+ Type *PointeeTy = OldLoadInst->getPointerOperandType();
+ int64_t ShiftValue = OffsetNew - OffsetOld;
+ Value *IndexList[1] = {ConstantInt::get(Int64TyInteger, ShiftValue)};
+ Value *GepToNewAddress = GetElementPtrInst::Create(
+ PointeeTy, PointerOperand, IndexList, "NewGep", OldLoadInst);
+
+ LoadInst *NewLoadInst = new LoadInst(
+ OldLoadInst->getType(), GepToNewAddress, OldLoadInst->getName(),
+ false, OldLoadInst->getAlign(), OldLoadInst);
+
+ Changed |=
+ A.changeAfterManifest(IRPosition::inst(*OldLoadInst), *NewLoadInst);
+
+ A.deleteAfterManifest(*OldLoadInst);
+ break;
+ }
+ case Instruction::Store: {
+ // The number of bytes to shift the load/store by
+ int64_t OffsetOld = OldToNewBins.begin()->getFirst().Offset;
+ int64_t OffsetNew = OldToNewBins.begin()->getSecond().Offset;
+ int64_t ShiftValue = OffsetNew - OffsetOld;
+ StoreInst *OldStoreInst = cast<StoreInst>(LocalInst);
+ Instruction *PointerOperand =
+ cast<Instruction>(OldStoreInst->getPointerOperand());
+ Type *PointeeTy = OldStoreInst->getPointerOperandType();
+ Value *IndexList[1] = {ConstantInt::get(Int64TyInteger, ShiftValue)};
+ Value *GepToNewAddress = GetElementPtrInst::Create(
+ PointeeTy, PointerOperand, IndexList, "NewGep", OldStoreInst);
+
+ StoreInst *NewStoreInst =
+ new StoreInst(OldStoreInst->getValueOperand(), GepToNewAddress,
+ false, OldStoreInst->getAlign(), OldStoreInst);
+
+ Changed |= A.changeAfterManifest(IRPosition::inst(*OldStoreInst),
+ *NewStoreInst);
+
+ A.deleteAfterManifest(*OldStoreInst);
+ break;
+ }
+ case Instruction::GetElementPtr: {
+ GetElementPtrInst *OldGEP = cast<GetElementPtrInst>(LocalInst);
+ int64_t OffsetNew = OldToNewBins.begin()->getSecond().Offset;
+ Value *IndexList[1] = {ConstantInt::get(Int64TyInteger, OffsetNew)};
+ Value *OldPointerOperand = OldGEP->getPointerOperand();
+ Value *GepToNewAddress = GetElementPtrInst::Create(
+ NewAllocationType, OldPointerOperand, IndexList, "NewGep", OldGEP);
+
+ Changed |=
+ A.changeAfterManifest(IRPosition::inst(*OldGEP), *GepToNewAddress);
+
+ A.deleteAfterManifest(*OldGEP);
+ break;
+ }
+ default:
+ static_assert(true, "[AAallocationInfo]: Changing offsets not "
+ "implemented for instruction!");
+ break;
+ }
+ }
+
+ if (!Changed)
+ return ChangeStatus::UNCHANGED;
+ return ChangeStatus::CHANGED;
}
/// See AbstractAttribute::getAsStr().
@@ -12798,8 +13023,28 @@ struct AAAllocationInfoImpl : public AAAllocationInfo {
")";
}
+ void dumpNewOffsetBins(raw_ostream &O) {
+
+ O << "Printing Map from [OldOffsetsRange] : [NewOffsetsRange] if the "
+ "offsets changed."
+ << "\n";
+ const auto &NewOffsetsMap = getNewOffsets();
+ for (auto It = NewOffsetsMap.begin(); It != NewOffsetsMap.end(); It++) {
+
+ const auto &OldRange = It->getFirst();
+ const auto &NewRange = It->getSecond();
+
+ O << "[" << OldRange.Offset << "," << OldRange.Offset + OldRange.Size
+ << "] : ";
+ O << "[" << NewRange.Offset << "," << NewRange.Offset + NewRange.Size
+ << "]";
+ O << "\n";
+ }
+ }
+
private:
std::optional<TypeSize> AssumedAllocatedSize = HasNoAllocationSize;
+ NewOffsetsTy NewComputedOffsets;
// Maintain the computed allocation size of the object.
// Returns (bool) weather the size of the allocation was modified or not.
@@ -12811,6 +13056,21 @@ struct AAAllocationInfoImpl : public AAAllocationInfo {
}
return false;
}
+
+ // Maps an old byte range to its new Offset range in the new allocation.
+ // Returns (bool) weather the old byte range's offsets changed or not.
+ bool setNewOffsets(const AA::RangeTy &OldRange, int64_t OldOffset,
+ int64_t NewComputedOffset, int64_t Size) {
+
+ if (OldOffset == NewComputedOffset)
+ return false;
+
+ AA::RangeTy &NewRange = NewComputedOffsets.getOrInsertDefault(OldRange);
+ NewRange.Offset = NewComputedOffset;
+ NewRange.Size = Size;
+
+ return true;
+ }
};
struct AAAllocationInfoFloating : AAAllocationInfoImpl {
diff --git a/llvm/test/Transforms/Attributor/ArgumentPromotion/crash.ll b/llvm/test/Transforms/Attributor/ArgumentPromotion/crash.ll
index 595cb37c6c93e..f0efa2a0ae3c1 100644
--- a/llvm/test/Transforms/Attributor/ArgumentPromotion/crash.ll
+++ b/llvm/test/Transforms/Attributor/ArgumentPromotion/crash.ll
@@ -106,10 +106,8 @@ define i32 @test_inf_promote_caller(i32 %arg) {
; CGSCC-LABEL: define {{[^@]+}}@test_inf_promote_caller
; CGSCC-SAME: (i32 [[ARG:%.*]]) #[[ATTR3:[0-9]+]] {
; CGSCC-NEXT: bb:
-; CGSCC-NEXT: [[TMP:%.*]] = alloca [[S:%.*]], align 8
-; CGSCC-NEXT: [[TMP3:%.*]] = alloca i8, i32 0, align 8
-; CGSCC-NEXT: [[TMP1:%.*]] = alloca [[S]], align 8
-; CGSCC-NEXT: [[TMP14:%.*]] = alloca i8, i32 0, align 8
+; CGSCC-NEXT: [[TMP3:%.*]] = alloca [0 x i8], align 1
+; CGSCC-NEXT: [[TMP14:%.*]] = alloca [0 x i8], align 1
; CGSCC-NEXT: ret i32 0
;
bb:
diff --git a/llvm/test/Transforms/Attributor/ArgumentPromotion/live_called_from_dead.ll b/llvm/test/Transforms/Attributor/ArgumentPromotion/live_called_from_dead.ll
index 1c34fff8dd755..63dbc4da7da37 100644
--- a/llvm/test/Transforms/Attributor/ArgumentPromotion/live_called_from_dead.ll
+++ b/llvm/test/Transforms/Attributor/ArgumentPromotion/live_called_from_dead.ll
@@ -36,8 +36,7 @@ define internal i32 @caller(ptr %B) {
; CGSCC: Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(none)
; CGSCC-LABEL: define {{[^@]+}}@caller
; CGSCC-SAME: () #[[ATTR0]] {
-; CGSCC-NEXT: [[A:%.*]] = alloca i32, align 4
-; CGSCC-NEXT: [[A1:%.*]] = alloca i8, i32 0, align 4
+; CGSCC-NEXT: [[A1:%.*]] = alloca [0 x i8], align 1
; CGSCC-NEXT: ret i32 0
;
%A = alloca i32
diff --git a/llvm/test/Transforms/Attributor/ArgumentPromotion/live_called_from_dead_2.ll b/llvm/test/Transforms/Attributor/ArgumentPromotion/live_called_from_dead_2.ll
index b42647840f7cf..956fa0e88b028 100644
--- a/llvm/test/Transforms/Attributor/ArgumentPromotion/live_called_from_dead_2.ll
+++ b/llvm/test/Transforms/Attributor/ArgumentPromotion/live_called_from_dead_2.ll
@@ -53,8 +53,7 @@ define internal i32 @caller(ptr %B) {
; CGSCC: Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(argmem: write)
; CGSCC-LABEL: define {{[^@]+}}@caller
; CGSCC-SAME: (ptr noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR0]] {
-; CGSCC-NEXT: [[A:%.*]] = alloca i32, align 4
-; CGSCC-NEXT: [[A1:%.*]] = alloca i8, i32 0, align 4
+; CGSCC-NEXT: [[A1:%.*]] = alloca [0 x i8], align 1
; CGSCC-NEXT: [[C:%.*]] = call i32 @test(ptr noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) [[B]]) #[[ATTR2:[0-9]+]]
; CGSCC-NEXT: ret i32 0
;
diff --git a/llvm/test/Transforms/Attributor/ArgumentPromotion/nonzero-address-spaces.ll b/llvm/test/Transforms/Attributor/ArgumentPromotion/nonzero-address-spaces.ll
index b588a399e5bd9..7b5e1276ac212 100644
--- a/llvm/test/Transforms/Attributor/ArgumentPromotion/nonzero-address-spaces.ll
+++ b/llvm/test/Transforms/Attributor/ArgumentPromotion/nonzero-address-spaces.ll
@@ -29,8 +29,7 @@ define internal i32 @foo(ptr) {
; CHECK-LABEL: define {{[^@]+}}@foo
; CHECK-SAME: () addrspace(1) #[[ATTR0:[0-9]+]] {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
-; CHECK-NEXT: [[RETVAL1:%.*]] = alloca i8, i32 0, align 4
+; CHECK-NEXT: [[RETVAL1:%.*]] = alloca [0 x i8], align 1
; CHECK-NEXT: call addrspace(0) void asm sideeffect "ldr r0, [r0] \0Abx lr \0A", ""()
; CHECK-NEXT: unreachable
;
diff --git a/llvm/test/Transforms/Attributor/IPConstantProp/pthreads.ll b/llvm/test/Transforms/Attributor/IPConstantProp/pthreads.ll
index 490894d129023..af2d1ef1eabba 100644
--- a/llvm/test/Transforms/Attributor/IPConstantProp/pthreads.ll
+++ b/llvm/test/Transforms/Attributor/IPConstantProp/pthreads.ll
@@ -34,8 +34,8 @@ target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
define dso_local i32 @main() {
; TUNIT-LABEL: define {{[^@]+}}@main() {
; TUNIT-NEXT: entry:
-; TUNIT-NEXT: [[ALLOC11:%.*]] = alloca i8, i32 0, align 8
-; TUNIT-NEXT: [[ALLOC22:%.*]] = alloca i8, i32 0, align 8
+; TUNIT-NEXT: [[ALLOC11:%.*]] = alloca [0 x i8], align 1
+; TUNIT-NEXT: [[ALLOC22:%.*]] = alloca [0 x i8], align 1
; TUNIT-NEXT: [[THREAD:%.*]] = alloca i64, align 8
; TUNIT-NEXT: [[CALL:%.*]] = call i32 @pthread_create(ptr noundef nonnull align 8 dereferenceable(8) [[THREAD]], ptr noundef align 4294967296 null, ptr noundef nonnull @foo, ptr nofree readnone align 4294967296 undef)
; TUNIT-NEXT: [[CALL1:%.*]] = call i32 @pthread_create(ptr noundef nonnull align 8 dereferenceable(8) [[THREAD]], ptr noundef align 4294967296 null, ptr noundef nonnull @bar, ptr noalias nocapture nofree nonnull readnone align 8 dereferenceable(8) undef)
diff --git a/llvm/test/Transforms/Attributor/allocator.ll b/llvm/test/Transforms/Attributor/allocator.ll
index f2d9ecd1d8fa4..693386f053615 100644
--- a/llvm/test/Transforms/Attributor/allocator.ll
+++ b/llvm/test/Transforms/Attributor/allocator.ll
@@ -13,8 +13,8 @@ define dso_local void @positive_alloca_1(i32 noundef %val) #0 {
; CHECK-LABEL: define dso_local void @positive_alloca_1
; CHECK-SAME: (i32 noundef [[VAL:%.*]]) {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[VAL_ADDR1:%.*]] = alloca i8, i32 4, align 4
-; CHECK-NEXT: [[F2:%.*]] = alloca i8, i32 4, align 4
+; CHECK-NEXT: [[VAL_ADDR1:%.*]] = alloca [4 x i8], align 1
+; CHECK-NEXT: [[F2:%.*]] = alloca [4 x i8], align 1
; CHECK-NEXT: store i32 [[VAL]], ptr [[VAL_ADDR1]], align 4
; CHECK-NEXT: store i32 10, ptr [[F2]], align 4
; CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[F2]], align 4
@@ -164,37 +164,52 @@ entry:
;TODO: The allocation can be reduced here.
;However, the offsets (load/store etc.) Need to be changed.
; Function Attrs: noinline nounwind uwtable
-define dso_local { i64, ptr } @positive_test_not_a_single_start_offset(i32 noundef %val) #0 {
-; CHECK: Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(none)
-; CHECK-LABEL: define dso_local { i64, ptr } @positive_test_not_a_single_start_offset
-; CHECK-SAME: (i32 noundef [[VAL:%.*]]) #[[ATTR0:[0-9]+]] {
+define dso_local void @positive_test_not_a_single_start_offset(i32 noundef %val) #0 {
+; CHECK-LABEL: define dso_local void @positive_test_not_a_single_start_offset
+; CHECK-SAME: (i32 noundef [[VAL:%.*]]) {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[RETVAL:%.*]] = alloca [[STRUCT_FOO:%.*]], align 8
; CHECK-NEXT: [[VAL_ADDR:%.*]] = alloca i32, align 4
+; CHECK-NEXT: [[F1:%.*]] = alloca [5 x i8], align 1
; CHECK-NEXT: store i32 [[VAL]], ptr [[VAL_ADDR]], align 4
-; CHECK-NEXT: store i32 2, ptr [[RETVAL]], align 8
-; CHECK-NEXT: [[FIELD3:%.*]] = getelementptr inbounds [[STRUCT_FOO]], ptr [[RETVAL]], i32 0, i32 2
-; CHECK-NEXT: store ptr [[VAL_ADDR]], ptr [[FIELD3]], align 8
-; CHECK-NEXT: [[TMP0:%.*]] = load { i64, ptr }, ptr [[RETVAL]], align 8
-; CHECK-NEXT: ret { i64, ptr } [[TMP0]]
+; CHECK-NEXT: [[MUL:%.*]] = mul nsw i32 2, [[VAL]]
+; CHECK-NEXT: store i32 [[MUL]], ptr [[F1]], align 4
+; CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[F1]], align 4
+; CHECK-NEXT: [[CALL:%.*]] = call i32 (ptr, ...) @printf(ptr noundef nonnull dereferenceable(17) @.str, i32 noundef [[TMP0]])
+; CHECK-NEXT: [[NEWGEP:%.*]] = getelementptr [5 x i8], ptr [[F1]], i64 4
+; CHECK-NEXT: [[CONV1:%.*]] = trunc i32 [[TMP0]] to i8
+; CHECK-NEXT: store i8 [[CONV1]], ptr [[NEWGEP]], align 4
+; CHECK-NEXT: [[NEWGEP2:%.*]] = getelementptr [5 x i8], ptr [[F1]], i64 4
+; CHECK-NEXT: [[TMP1:%.*]] = load i8, ptr [[NEWGEP2]], align 4
+; CHECK-NEXT: [[CONV:%.*]] = sext i8 [[TMP1]] to i32
+; CHECK-NEXT: [[CALL3:%.*]] = call i32 (ptr, ...) @printf(ptr noundef nonnull dereferenceable(17) @.str, i32 noundef [[CONV]])
+; CHECK-NEXT: ret void
;
entry:
- %retval = alloca %struct.Foo, align 8
%val.addr = alloca i32, align 4
+ %f = alloca %struct.Foo, align 4
store i32 %val, ptr %val.addr, align 4
- %field1 = getelementptr inbounds %struct.Foo, ptr %retval, i32 0, i32 0
- store i32 2, ptr %field1, align 8
- %field3 = getelementptr inbounds %struct.Foo, ptr %retval, i32 0, i32 2
- store ptr %val.addr, ptr %field3, align 8
- %0 = load { i64, ptr }, ptr %retval, align 8
- ret { i64, ptr } %0
+ %0 = load i32, ptr %val.addr, align 4
+ %mul = mul nsw i32 2, %0
+ %a = getelementptr inbounds %struct.Foo, ptr %f, i32 0, i32 0
+ store i32 %mul, ptr %a, align 4
+ %a1 = getelementptr inbounds %struct.Foo, ptr %f, i32 0, i32 0
+ %1 = load i32, ptr %a1, align 4
+ %call = call i32 (ptr, ...) @printf(ptr noundef @.str, i32 noundef %1)
+ %c = getelementptr inbounds %struct.Foo, ptr %f, i32 0, i32 2
+ %conv1 = trunc i32 %1 to i8
+ store i8 %conv1, ptr %c, align 4
+ %c2 = getelementptr inbounds %struct.Foo, ptr %f, i32 0, i32 2
+ %2 = load i8, ptr %c2, align 4
+ %conv = sext i8 %2 to i32
+ %call3 = call i32 (ptr, ...) @printf(ptr noundef @.str, i32 noundef %conv)
+ ret void
}
; Function Attrs: noinline nounwind uwtable
define dso_local void @positive_test_reduce_array_allocation_1() {
; CHECK-LABEL: define dso_local void @positive_test_reduce_array_allocation_1() {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[ARRAY1:%.*]] = alloca i8, i32 4, align 8
+; CHECK-NEXT: [[ARRAY1:%.*]] = alloca [4 x i8], align 1
; CHECK-NEXT: store i32 0, ptr [[ARRAY1]], align 8
; CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[ARRAY1]], align 8
; CHECK-NEXT: [[TMP1:%.*]] = add i32 [[TMP0]], 2
@@ -275,37 +290,37 @@ entry:
define dso_local void @positive_test_reduce_array_allocation_2() #0 {
; CHECK-LABEL: define dso_local void @positive_test_reduce_array_allocation_2() {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[ARRAY:%.*]] = alloca ptr, align 8
-; CHECK-NEXT: [[I:%.*]] = alloca i32, align 4
+; CHECK-NEXT: [[ARRAY1:%.*]] = alloca ptr, align 8
+; CHECK-NEXT: [[I2:%.*]] = alloca i32, align 4
; CHECK-NEXT: [[CALL:%.*]] = call noalias ptr @malloc(i64 noundef 40000)
-; CHECK-NEXT: store ptr [[CALL]], ptr [[ARRAY]], align 8
-; CHECK-NEXT: store i32 0, ptr [[I]], align 4
+; CHECK-NEXT: store ptr [[CALL]], ptr [[ARRAY1]], align 8
+; CHECK-NEXT: store i32 0, ptr [[I2]], align 4
; CHECK-NEXT: br label [[FOR_COND:%.*]]
; CHECK: for.cond:
-; CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I]], align 4
+; CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I2]], align 4
; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 [[TMP0]], 10000
; CHECK-NEXT: br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
; CHECK: for.body:
-; CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[I]], align 4
-; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[I]], align 4
+; CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[I2]], align 4
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[I2]], align 4
; CHECK-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP2]] to i64
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[CALL]], i64 [[IDXPROM]]
; CHECK-NEXT: store i32 [[TMP1]], ptr [[ARRAYIDX]], align 4
; CHECK-NEXT: br label [[FOR_INC:%.*]]
; CHECK: for.inc:
-; CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr [[I]], align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr [[I2]], align 4
; CHECK-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP3]], 2
-; CHECK-NEXT: store i32 [[ADD]], ptr [[I]], align 4
+; CHECK-NEXT: store i32 [[ADD]], ptr [[I2]], align 4
; CHECK-NEXT: br label [[FOR_COND]]
; CHECK: for.end:
-; CHECK-NEXT: store i32 0, ptr [[I]], align 4
+; CHECK-NEXT: store i32 0, ptr [[I2]], align 4
; CHECK-NEXT: br label [[FOR_COND1:%.*]]
; CHECK: for.cond1:
-; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr [[I]], align 4
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr [[I2]], align 4
; CHECK-NEXT: [[CMP2:%.*]] = icmp slt i32 [[TMP4]], 10000
; CHECK-NEXT: br i1 [[CMP2]], label [[FOR_BODY3:%.*]], label [[FOR_END9:%.*]]
; CHECK: for.body3:
-; CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[I]], align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[I2]], align 4
; CHECK-NEXT: [[IDXPROM4:%.*]] = sext i32 [[TMP5]] to i64
; CHECK-NEXT: [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, ptr [[CALL]], i64 [[IDXPROM4]]
; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr [[ARRAYIDX5]], align 4
@@ -313,28 +328,28 @@ define dso_local void @positive_test_reduce_array_allocation_2() #0 {
; CHECK-NEXT: store i32 [[ADD6]], ptr [[ARRAYIDX5]], align 4
; CHECK-NEXT: br label [[FOR_INC7:%.*]]
; CHECK: for.inc7:
-; CHECK-NEXT: [[TMP7:%.*]] = load i32, ptr [[I]], align 4
+; CHECK-NEXT: [[TMP7:%.*]] = load i32, ptr [[I2]], align 4
; CHECK-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP7]], 2
-; CHECK-NEXT: store i32 [[ADD8]], ptr [[I]], align 4
+; CHECK-NEXT: store i32 [[ADD8]], ptr [[I2]], align 4
; CHECK-NEXT: br label [[FOR_COND1]]
; CHECK: for.end9:
-; CHECK-NEXT: store i32 0, ptr [[I]], align 4
+; CHECK-NEXT: store i32 0, ptr [[I2]], align 4
; CHECK-NEXT: br label [[FOR_COND10:%.*]]
; CHECK: for.cond10:
-; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr [[I]], align 4
+; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr [[I2]], align 4
; CHECK-NEXT: [[CMP11:%.*]] = icmp slt i32 [[TMP8]], 10000
; CHECK-NEXT: br i1 [[CMP11]], label [[FOR_BODY12:%.*]], label [[FOR_END18:%.*]]
; CHECK: for.body12:
-; CHECK-NEXT: [[TMP9:%.*]] = load i32, ptr [[I]], align 4
+; CHECK-NEXT: [[TMP9:%.*]] = load i32, ptr [[I2]], align 4
; CHECK-NEXT: [[IDXPROM13:%.*]] = sext i32 [[TMP9]] to i64
; CHECK-NEXT: [[ARRAYIDX14:%.*]] = getelementptr inbounds i32, ptr [[CALL]], i64 [[IDXPROM13]]
; CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr [[ARRAYIDX14]], align 4
; CHECK-NEXT: [[CALL15:%.*]] = call i32 (ptr, ...) @printf(ptr noundef nonnull dereferenceable(17) @.str, i32 noundef [[TMP10]])
; CHECK-NEXT: br label [[FOR_INC16:%.*]]
; CHECK: for.inc16:
-; CHECK-NEXT: [[TMP11:%.*]] = load i32, ptr [[I]], align 4
+; CHECK-NEXT: [[TMP11:%.*]] = load i32, ptr [[I2]], align 4
; CHECK-NEXT: [[ADD17:%.*]] = add nsw i32 [[TMP11]], 2
-; CHECK-NEXT: store i32 [[ADD17]], ptr [[I]], align 4
+; CHECK-NEXT: store i32 [[ADD17]], ptr [[I2]], align 4
; CHECK-NEXT: br label [[FOR_COND10]]
; CHECK: for.end18:
; CHECK-NEXT: ret void
@@ -426,7 +441,7 @@ define dso_local void @pthread_test(){
; TUNIT-NEXT: [[ARG1:%.*]] = alloca i8, align 8
; TUNIT-NEXT: [[THREAD:%.*]] = alloca i64, align 8
; TUNIT-NEXT: [[CALL1:%.*]] = call i32 @pthread_create(ptr noundef nonnull align 8 dereferenceable(8) [[THREAD]], ptr noundef align 4294967296 null, ptr noundef nonnull @pthread_allocation_should_remain_same, ptr noundef nonnull align 8 dereferenceable(1) [[ARG1]])
-; TUNIT-NEXT: [[F1:%.*]] = alloca i8, i32 4, align 4
+; TUNIT-NEXT: [[F1:%.*]] = alloca [4 x i8], align 1
; TUNIT-NEXT: [[CALL2:%.*]] = call i32 @pthread_create(ptr noundef nonnull align 8 dereferenceable(8) [[THREAD]], ptr noundef align 4294967296 null, ptr noundef nonnull @pthread_allocation_should_be_reduced, ptr noalias nocapture nofree nonnull readnone align 4 dereferenceable(12) undef)
; TUNIT-NEXT: [[F2:%.*]] = alloca [[STRUCT_FOO:%.*]], align 4
; TUNIT-NEXT: [[CALL3:%.*]] = call i32 @pthread_create(ptr noundef nonnull align 8 dereferenceable(8) [[THREAD]], ptr noundef align 4294967296 null, ptr noundef nonnull @pthread_check_captured_pointer, ptr noundef nonnull align 4 dereferenceable(12) [[F2]])
@@ -452,6 +467,46 @@ define dso_local void @pthread_test(){
ret void
}
+
+define dso_local void @select_case(i1 %cond){
+; CHECK: Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(write)
+; CHECK-LABEL: define dso_local void @select_case
+; CHECK-SAME: (i1 [[COND:%.*]]) #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT: [[A:%.*]] = alloca [100 x i8], align 1
+; CHECK-NEXT: [[B:%.*]] = getelementptr inbounds [100 x i8], ptr [[A]], i64 0, i64 3
+; CHECK-NEXT: [[C:%.*]] = getelementptr inbounds [100 x i8], ptr [[A]], i64 0, i64 1
+; CHECK-NEXT: [[SEL:%.*]] = select i1 [[COND]], ptr [[B]], ptr [[C]]
+; CHECK-NEXT: store i8 100, ptr [[SEL]], align 1
+; CHECK-NEXT: ret void
+;
+ %a = alloca [100 x i8], align 1
+ %b = getelementptr inbounds [100 x i8], ptr %a, i64 0, i64 3
+ %c = getelementptr inbounds [100 x i8], ptr %a, i64 0, i64 1
+ %sel = select i1 %cond, ptr %b, ptr %c
+ store i8 100, ptr %sel, align 1
+ ret void
+}
+
+define dso_local void @select_case_2(i1 %cond){
+; CHECK: Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(write)
+; CHECK-LABEL: define dso_local void @select_case_2
+; CHECK-SAME: (i1 [[COND:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[A:%.*]] = alloca [100 x i32], align 1
+; CHECK-NEXT: [[B:%.*]] = getelementptr inbounds [100 x i32], ptr [[A]], i64 0, i64 3
+; CHECK-NEXT: [[C:%.*]] = getelementptr inbounds [100 x i32], ptr [[A]], i64 0, i64 1
+; CHECK-NEXT: [[SEL:%.*]] = select i1 [[COND]], ptr [[B]], ptr [[C]]
+; CHECK-NEXT: store i8 100, ptr [[SEL]], align 1
+; CHECK-NEXT: ret void
+;
+ %a = alloca [100 x i32], align 1
+ %b = getelementptr inbounds [100 x i32], ptr %a, i64 0, i64 3
+ %c = getelementptr inbounds [100 x i32], ptr %a, i64 0, i64 1
+ %sel = select i1 %cond, ptr %b, ptr %c
+ %sel2 = getelementptr inbounds i32, ptr %sel, i64 0
+ store i8 100, ptr %sel2, align 1
+ ret void
+}
+
define internal ptr @pthread_allocation_should_remain_same(ptr %arg) {
; CHECK-LABEL: define internal noundef nonnull align 8 dereferenceable(1) ptr @pthread_allocation_should_remain_same
; CHECK-SAME: (ptr noundef nonnull returned align 8 dereferenceable(1) [[ARG:%.*]]) {
@@ -499,6 +554,58 @@ entry:
ret void
}
+define dso_local void @alloca_array_multi_offset(){
+; CHECK: Function Attrs: nofree norecurse nosync nounwind memory(none)
+; CHECK-LABEL: define dso_local void @alloca_array_multi_offset
+; CHECK-SAME: () #[[ATTR1:[0-9]+]] {
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[I:%.*]] = alloca i32, align 4
+; CHECK-NEXT: store i32 0, ptr [[I]], align 4
+; CHECK-NEXT: br label [[FOR_COND:%.*]]
+; CHECK: for.cond:
+; CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I]], align 4
+; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 [[TMP0]], 10
+; CHECK-NEXT: br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
+; CHECK: for.body:
+; CHECK-NEXT: br label [[FOR_INC:%.*]]
+; CHECK: for.inc:
+; CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[I]], align 4
+; CHECK-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP1]], 2
+; CHECK-NEXT: store i32 [[ADD]], ptr [[I]], align 4
+; CHECK-NEXT: br label [[FOR_COND]]
+; CHECK: for.end:
+; CHECK-NEXT: ret void
+;
+entry:
+ %arr = alloca i8, i32 10, align 4
+ %i = alloca i32, align 4
+ store i32 0, ptr %i, align 4
+ br label %for.cond
+
+for.cond:
+ %0 = load i32, ptr %i, align 4
+ %cmp = icmp slt i32 %0, 10
+ br i1 %cmp, label %for.body, label %for.end
+
+for.body:
+ %1 = load i32, ptr %i, align 4
+ %2 = load ptr, ptr %arr, align 8
+ %3 = load i32, ptr %i, align 4
+ %arrayidx = getelementptr inbounds i32, ptr %2, i32 %3
+ store i32 %1, ptr %arrayidx, align 4
+ br label %for.inc
+
+for.inc:
+ %4 = load i32, ptr %i, align 4
+ %add = add nsw i32 %4, 2
+ store i32 %add, ptr %i, align 4
+ br label %for.cond
+
+for.end:
+ ret void
+
+}
+
declare external void @external_call(ptr)
@@ -511,9 +618,11 @@ declare i32 @printf(ptr noundef, ...) #1
; Function Attrs: nounwind allocsize(0)
declare noalias ptr @malloc(i64 noundef) #1
;.
-; TUNIT: attributes #[[ATTR0]] = { mustprogress nofree norecurse nosync nounwind willreturn memory(none) }
+; TUNIT: attributes #[[ATTR0]] = { mustprogress nofree norecurse nosync nounwind willreturn memory(write) }
+; TUNIT: attributes #[[ATTR1]] = { nofree norecurse nosync nounwind memory(none) }
;.
-; CGSCC: attributes #[[ATTR0]] = { mustprogress nofree norecurse nosync nounwind willreturn memory(none) }
+; CGSCC: attributes #[[ATTR0]] = { mustprogress nofree norecurse nosync nounwind willreturn memory(write) }
+; CGSCC: attributes #[[ATTR1]] = { nofree norecurse nosync nounwind memory(none) }
;.
; TUNIT: [[META0:![0-9]+]] = !{[[META1:![0-9]+]]}
; TUNIT: [[META1]] = !{i64 2, i64 3, i1 false}
diff --git a/llvm/test/Transforms/Attributor/call-simplify-pointer-info.ll b/llvm/test/Transforms/Attributor/call-simplify-pointer-info.ll
index 5bb795911ce40..b313500cf16bb 100644
--- a/llvm/test/Transforms/Attributor/call-simplify-pointer-info.ll
+++ b/llvm/test/Transforms/Attributor/call-simplify-pointer-info.ll
@@ -36,8 +36,8 @@ define i8 @call_simplifiable_1() {
; TUNIT-LABEL: define {{[^@]+}}@call_simplifiable_1
; TUNIT-SAME: () #[[ATTR0:[0-9]+]] {
; TUNIT-NEXT: entry:
-; TUNIT-NEXT: [[BYTES:%.*]] = alloca [1024 x i8], align 16
-; TUNIT-NEXT: [[I0:%.*]] = getelementptr inbounds [1024 x i8], ptr [[BYTES]], i64 0, i64 2
+; TUNIT-NEXT: [[BYTES1:%.*]] = alloca [1 x i8], align 1
+; TUNIT-NEXT: [[NEWGEP:%.*]] = getelementptr [1 x i8], ptr [[BYTES1]], i64 0
; TUNIT-NEXT: ret i8 2
;
; CGSCC: Function Attrs: mustprogress nofree nosync nounwind willreturn memory(none)
@@ -93,9 +93,9 @@ define i8 @call_simplifiable_2() {
; TUNIT-LABEL: define {{[^@]+}}@call_simplifiable_2
; TUNIT-SAME: () #[[ATTR0]] {
; TUNIT-NEXT: entry:
-; TUNIT-NEXT: [[BYTES:%.*]] = alloca [1024 x i8], align 16
-; TUNIT-NEXT: [[I0:%.*]] = getelementptr inbounds [1024 x i8], ptr [[BYTES]], i64 0, i64 2
-; TUNIT-NEXT: [[I1:%.*]] = getelementptr inbounds [1024 x i8], ptr [[BYTES]], i64 0, i64 3
+; TUNIT-NEXT: [[BYTES1:%.*]] = alloca [2 x i8], align 1
+; TUNIT-NEXT: [[NEWGEP2:%.*]] = getelementptr [2 x i8], ptr [[BYTES1]], i64 0
+; TUNIT-NEXT: [[NEWGEP:%.*]] = getelementptr [2 x i8], ptr [[BYTES1]], i64 1
; TUNIT-NEXT: ret i8 4
;
; CGSCC: Function Attrs: mustprogress nofree nosync nounwind willreturn memory(none)
@@ -125,8 +125,8 @@ define i8 @call_simplifiable_3() {
; TUNIT-LABEL: define {{[^@]+}}@call_simplifiable_3
; TUNIT-SAME: () #[[ATTR0]] {
; TUNIT-NEXT: entry:
-; TUNIT-NEXT: [[BYTES:%.*]] = alloca [1024 x i8], align 16
-; TUNIT-NEXT: [[I2:%.*]] = getelementptr inbounds [1024 x i8], ptr [[BYTES]], i64 0, i64 2
+; TUNIT-NEXT: [[BYTES1:%.*]] = alloca [1 x i8], align 1
+; TUNIT-NEXT: [[NEWGEP:%.*]] = getelementptr [1 x i8], ptr [[BYTES1]], i64 0
; TUNIT-NEXT: ret i8 2
;
; CGSCC: Function Attrs: mustprogress nofree nosync nounwind willreturn memory(none)
@@ -198,13 +198,13 @@ define i8 @call_partially_simplifiable_1() {
; TUNIT-LABEL: define {{[^@]+}}@call_partially_simplifiable_1
; TUNIT-SAME: () #[[ATTR0]] {
; TUNIT-NEXT: entry:
-; TUNIT-NEXT: [[BYTES:%.*]] = alloca [1024 x i8], align 16
-; TUNIT-NEXT: [[I2:%.*]] = getelementptr inbounds [1024 x i8], ptr [[BYTES]], i64 0, i64 2
-; TUNIT-NEXT: store i8 2, ptr [[I2]], align 2
-; TUNIT-NEXT: [[I3:%.*]] = getelementptr inbounds [1024 x i8], ptr [[BYTES]], i64 0, i64 3
-; TUNIT-NEXT: store i8 3, ptr [[I3]], align 1
-; TUNIT-NEXT: [[I4:%.*]] = getelementptr inbounds [1024 x i8], ptr [[BYTES]], i64 0, i64 4
-; TUNIT-NEXT: [[R:%.*]] = call i8 @sum_two_different_loads(ptr nocapture nofree noundef nonnull readonly align 2 dereferenceable(1022) [[I2]], ptr nocapture nofree noundef nonnull readonly dereferenceable(1021) [[I3]]) #[[ATTR3]]
+; TUNIT-NEXT: [[BYTES1:%.*]] = alloca [3 x i8], align 1
+; TUNIT-NEXT: [[NEWGEP2:%.*]] = getelementptr [3 x i8], ptr [[BYTES1]], i64 0
+; TUNIT-NEXT: store i8 2, ptr [[NEWGEP2]], align 2
+; TUNIT-NEXT: [[NEWGEP:%.*]] = getelementptr [3 x i8], ptr [[BYTES1]], i64 2
+; TUNIT-NEXT: store i8 3, ptr [[NEWGEP]], align 1
+; TUNIT-NEXT: [[NEWGEP3:%.*]] = getelementptr [3 x i8], ptr [[BYTES1]], i64 1
+; TUNIT-NEXT: [[R:%.*]] = call i8 @sum_two_different_loads(ptr nocapture nofree noundef nonnull readonly align 2 dereferenceable(1022) [[NEWGEP2]], ptr nocapture nofree noundef nonnull readonly dereferenceable(1021) [[NEWGEP]]) #[[ATTR3]]
; TUNIT-NEXT: ret i8 [[R]]
;
; CGSCC: Function Attrs: mustprogress nofree nosync nounwind willreturn memory(none)
diff --git a/llvm/test/Transforms/Attributor/heap_to_stack.ll b/llvm/test/Transforms/Attributor/heap_to_stack.ll
index 33ac066e43d09..846373e05be1a 100644
--- a/llvm/test/Transforms/Attributor/heap_to_stack.ll
+++ b/llvm/test/Transforms/Attributor/heap_to_stack.ll
@@ -502,8 +502,7 @@ define i32 @malloc_in_loop(i32 %arg) {
; CHECK-SAME: (i32 [[ARG:%.*]]) {
; CHECK-NEXT: bb:
; CHECK-NEXT: [[I:%.*]] = alloca i32, align 4
-; CHECK-NEXT: [[I1:%.*]] = alloca ptr, align 8
-; CHECK-NEXT: [[I11:%.*]] = alloca i8, i32 0, align 8
+; CHECK-NEXT: [[I11:%.*]] = alloca [0 x i8], align 1
; CHECK-NEXT: store i32 [[ARG]], ptr [[I]], align 4
; CHECK-NEXT: br label [[BB2:%.*]]
; CHECK: bb2:
diff --git a/llvm/test/Transforms/Attributor/heap_to_stack_gpu.ll b/llvm/test/Transforms/Attributor/heap_to_stack_gpu.ll
index 2a5b3e94291a2..70aace8100abd 100644
--- a/llvm/test/Transforms/Attributor/heap_to_stack_gpu.ll
+++ b/llvm/test/Transforms/Attributor/heap_to_stack_gpu.ll
@@ -452,8 +452,7 @@ define i32 @malloc_in_loop(i32 %arg) {
; CHECK-SAME: (i32 [[ARG:%.*]]) {
; CHECK-NEXT: bb:
; CHECK-NEXT: [[I:%.*]] = alloca i32, align 4
-; CHECK-NEXT: [[I1:%.*]] = alloca ptr, align 8
-; CHECK-NEXT: [[I11:%.*]] = alloca i8, i32 0, align 8
+; CHECK-NEXT: [[I11:%.*]] = alloca [0 x i8], align 1
; CHECK-NEXT: store i32 [[ARG]], ptr [[I]], align 4
; CHECK-NEXT: br label [[BB2:%.*]]
; CHECK: bb2:
diff --git a/llvm/test/Transforms/Attributor/liveness.ll b/llvm/test/Transforms/Attributor/liveness.ll
index f17bd5795a174..9eb79f8a46723 100644
--- a/llvm/test/Transforms/Attributor/liveness.ll
+++ b/llvm/test/Transforms/Attributor/liveness.ll
@@ -2587,8 +2587,8 @@ define void @bad_gep() {
; TUNIT-LABEL: define {{[^@]+}}@bad_gep
; TUNIT-SAME: () #[[ATTR13]] {
; TUNIT-NEXT: entry:
-; TUNIT-NEXT: [[N1:%.*]] = alloca i8, i32 0, align 1
-; TUNIT-NEXT: [[M2:%.*]] = alloca i8, i32 0, align 1
+; TUNIT-NEXT: [[N1:%.*]] = alloca [0 x i8], align 1
+; TUNIT-NEXT: [[M2:%.*]] = alloca [0 x i8], align 1
; TUNIT-NEXT: call void @llvm.lifetime.start.p0(i64 noundef 1, ptr noalias nocapture nofree noundef nonnull dereferenceable(1) [[N1]]) #[[ATTR18:[0-9]+]]
; TUNIT-NEXT: br label [[EXIT:%.*]]
; TUNIT: while.body:
@@ -2605,8 +2605,8 @@ define void @bad_gep() {
; CGSCC-LABEL: define {{[^@]+}}@bad_gep
; CGSCC-SAME: () #[[ATTR6]] {
; CGSCC-NEXT: entry:
-; CGSCC-NEXT: [[N1:%.*]] = alloca i8, i32 0, align 1
-; CGSCC-NEXT: [[M2:%.*]] = alloca i8, i32 0, align 1
+; CGSCC-NEXT: [[N1:%.*]] = alloca [0 x i8], align 1
+; CGSCC-NEXT: [[M2:%.*]] = alloca [0 x i8], align 1
; CGSCC-NEXT: call void @llvm.lifetime.start.p0(i64 noundef 1, ptr noalias nocapture nofree noundef nonnull dereferenceable(1) [[N1]]) #[[ATTR21:[0-9]+]]
; CGSCC-NEXT: br label [[EXIT:%.*]]
; CGSCC: while.body:
diff --git a/llvm/test/Transforms/Attributor/nodelete.ll b/llvm/test/Transforms/Attributor/nodelete.ll
index c28cb28379348..6357bf742bbf1 100644
--- a/llvm/test/Transforms/Attributor/nodelete.ll
+++ b/llvm/test/Transforms/Attributor/nodelete.ll
@@ -10,15 +10,14 @@ define hidden i64 @f1() align 2 {
; TUNIT-LABEL: define {{[^@]+}}@f1
; TUNIT-SAME: () #[[ATTR0:[0-9]+]] align 2 {
; TUNIT-NEXT: entry:
-; TUNIT-NEXT: [[REF_TMP1:%.*]] = alloca i8, i32 0, align 8
+; TUNIT-NEXT: [[REF_TMP1:%.*]] = alloca [0 x i8], align 1
; TUNIT-NEXT: ret i64 undef
;
; CGSCC: Function Attrs: mustprogress nofree nosync nounwind willreturn memory(none)
; CGSCC-LABEL: define {{[^@]+}}@f1
; CGSCC-SAME: () #[[ATTR0:[0-9]+]] align 2 {
; CGSCC-NEXT: entry:
-; CGSCC-NEXT: [[REF_TMP:%.*]] = alloca [[A:%.*]], align 8
-; CGSCC-NEXT: [[REF_TMP1:%.*]] = alloca i8, i32 0, align 8
+; CGSCC-NEXT: [[REF_TMP1:%.*]] = alloca [0 x i8], align 1
; CGSCC-NEXT: [[CALL2:%.*]] = call i64 @f2() #[[ATTR2:[0-9]+]]
; CGSCC-NEXT: ret i64 [[CALL2]]
;
diff --git a/llvm/test/Transforms/Attributor/pointer-info.ll b/llvm/test/Transforms/Attributor/pointer-info.ll
index 6afdbdaee317c..66dc0160a4e99 100644
--- a/llvm/test/Transforms/Attributor/pointer-info.ll
+++ b/llvm/test/Transforms/Attributor/pointer-info.ll
@@ -10,11 +10,11 @@ define void @foo(ptr %ptr) {
; TUNIT-LABEL: define {{[^@]+}}@foo
; TUNIT-SAME: (ptr nocapture nofree readnone [[PTR:%.*]]) #[[ATTR0:[0-9]+]] {
; TUNIT-NEXT: entry:
-; TUNIT-NEXT: [[TMP0:%.*]] = alloca [[STRUCT_TEST_A:%.*]], align 8
+; TUNIT-NEXT: [[TMP0:%.*]] = alloca [8 x i8], align 1
; TUNIT-NEXT: br label [[CALL_BR:%.*]]
; TUNIT: call.br:
-; TUNIT-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_TEST_A]], ptr [[TMP0]], i64 0, i32 2
-; TUNIT-NEXT: tail call void @bar(ptr noalias nocapture nofree noundef nonnull readonly byval([[STRUCT_TEST_A]]) align 8 dereferenceable(24) [[TMP0]]) #[[ATTR2:[0-9]+]]
+; TUNIT-NEXT: [[NEWGEP:%.*]] = getelementptr [8 x i8], ptr [[TMP0]], i64 0
+; TUNIT-NEXT: tail call void @bar(ptr noalias nocapture nofree noundef nonnull readonly byval([[STRUCT_TEST_A:%.*]]) align 8 dereferenceable(24) [[TMP0]]) #[[ATTR2:[0-9]+]]
; TUNIT-NEXT: ret void
;
; CGSCC: Function Attrs: mustprogress nofree nosync nounwind willreturn memory(none)
diff --git a/llvm/test/Transforms/Attributor/value-simplify-pointer-info.ll b/llvm/test/Transforms/Attributor/value-simplify-pointer-info.ll
index 7a35b5c856097..07b25f6232436 100644
--- a/llvm/test/Transforms/Attributor/value-simplify-pointer-info.ll
+++ b/llvm/test/Transforms/Attributor/value-simplify-pointer-info.ll
@@ -2666,18 +2666,18 @@ define dso_local void @test_nested_memory(ptr %dst, ptr %src) {
; TUNIT-SAME: (ptr nocapture nofree writeonly [[DST:%.*]], ptr nocapture nofree readonly [[SRC:%.*]]) {
; TUNIT-NEXT: entry:
; TUNIT-NEXT: [[CALL_H2S:%.*]] = alloca i8, i64 24, align 1
-; TUNIT-NEXT: [[LOCAL:%.*]] = alloca [[STRUCT_STY:%.*]], align 8
-; TUNIT-NEXT: [[INNER:%.*]] = getelementptr inbounds [[STRUCT_STY]], ptr [[LOCAL]], i64 0, i32 2
-; TUNIT-NEXT: store ptr @global, ptr [[INNER]], align 8
+; TUNIT-NEXT: [[LOCAL1:%.*]] = alloca [8 x i8], align 1
+; TUNIT-NEXT: [[NEWGEP:%.*]] = getelementptr [8 x i8], ptr [[LOCAL1]], i64 0
+; TUNIT-NEXT: store ptr @global, ptr [[NEWGEP]], align 8
; TUNIT-NEXT: store ptr [[DST]], ptr [[CALL_H2S]], align 8
; TUNIT-NEXT: [[SRC2:%.*]] = getelementptr inbounds i8, ptr [[CALL_H2S]], i64 8
; TUNIT-NEXT: store ptr [[SRC]], ptr [[SRC2]], align 8
-; TUNIT-NEXT: store ptr [[CALL_H2S]], ptr getelementptr inbounds ([[STRUCT_STY]], ptr @global, i64 0, i32 2), align 8
-; TUNIT-NEXT: [[TMP0:%.*]] = load ptr, ptr [[LOCAL]], align 8
-; TUNIT-NEXT: [[LOCAL_B8:%.*]] = getelementptr i8, ptr [[LOCAL]], i64 8
-; TUNIT-NEXT: [[TMP1:%.*]] = load ptr, ptr [[LOCAL_B8]], align 8
-; TUNIT-NEXT: [[LOCAL_B16:%.*]] = getelementptr i8, ptr [[LOCAL]], i64 16
-; TUNIT-NEXT: [[TMP2:%.*]] = load ptr, ptr [[LOCAL_B16]], align 8
+; TUNIT-NEXT: store ptr [[CALL_H2S]], ptr getelementptr inbounds ([[STRUCT_STY:%.*]], ptr @global, i64 0, i32 2), align 8
+; TUNIT-NEXT: [[TMP0:%.*]] = load ptr, ptr [[LOCAL1]], align 8
+; TUNIT-NEXT: [[LOCAL1_B8:%.*]] = getelementptr i8, ptr [[LOCAL1]], i64 8
+; TUNIT-NEXT: [[TMP1:%.*]] = load ptr, ptr [[LOCAL1_B8]], align 8
+; TUNIT-NEXT: [[LOCAL1_B16:%.*]] = getelementptr i8, ptr [[LOCAL1]], i64 16
+; TUNIT-NEXT: [[TMP2:%.*]] = load ptr, ptr [[LOCAL1_B16]], align 8
; TUNIT-NEXT: call fastcc void @nested_memory_callee(ptr [[TMP0]], ptr [[TMP1]], ptr [[TMP2]]) #[[ATTR21:[0-9]+]]
; TUNIT-NEXT: ret void
;
More information about the llvm-commits
mailing list