[opaque pointer types] [NFC] GEP: replace get(Pointer)ElementType uses with get{Source,Result}ElementType.

[eddyb]

Summary

GEPOperator: provide getResultElementType alongside getSourceElementType.
This is made possible by adding a result element type field to GetElementPtrConstantExpr, which GetElementPtrInst already has.

GEP: replace get(Pointer)ElementType uses with get{Source,Result}ElementType.

[eddyb]

Fixed missing PtrTy in ScalarReplAggregates.cpp's CanConvertToScalar.

[eddyb]

Simplified the first commit and took out the second one to shorten this revision.

[eddyb]

Remove change that depends on other patches (and does nullptr deref right now).

[mjacob]

lib/Analysis/ScalarEvolution.cpp

4090 ↗

(On Diff #45190)

You could move this variable closer to its only consumer or even inline it.

lib/CodeGen/SelectionDAG/FastISel.cpp

517 ↗

(On Diff #45198)

Please add a comment why this is correct here, i.e. that Ty can be a pointer type only on the first iteration.

lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp

3022 ↗

(On Diff #45198)

Same here.

lib/IR/ConstantFold.cpp

2045 ↗

(On Diff #45198)

Maybe this code was wrong, because Ptr->getScalarType() returns Ptr in any case. However I couldn't find a way to trigger a bug. I'll investigate this. Your changes look OK.

lib/IR/ConstantsContext.h

228 ↗

(On Diff #45198)

You should mention in the commit message that you add a field to this class.

lib/Target/AArch64/AArch64FastISel.cpp

4829 ↗

(On Diff #45198)

Please add a comment - same as above.

lib/Transforms/InstCombine/InstCombineLoadStoreAlloca.cpp

697–698 ↗

(On Diff #45198)

Can we use getResultElementType() here?

lib/Transforms/InstCombine/InstructionCombining.cpp

1421 ↗

(On Diff #45198)

I think CurTy is implied by J > 1. IMHO the extra check creates confusion.

[eddyb]

lib/Transforms/InstCombine/InstCombineLoadStoreAlloca.cpp

697–698 ↗

(On Diff #45198)

If Ops were to have all of GEPI's indices, then if (Idx == GEPI->getNumOperands()) return false; above would've handled it (by returning out of this function).
In other words, AllocTy will always stop short of the ResultElementType, as less indices will be applied.

[eddyb]

Address review comments.

[eddyb]

Rebased on top of rL258134.

[mjacob]

LGTM.

lib/Analysis/ValueTracking.cpp

3255 ↗

(On Diff #45270)

I think this variable should be inlined to its only user.

[dblaikie]

llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp

516

This seems like a slightly indirect test of the property you describe in the comment.

Do you think it might be nicer to check if the iterator is == I->op_begin() + 1? Or to add a flag of some kind before the loop for this first case? Or perhaps to use a gep type iterator in parallel with the operands? (or an assert for any of these conditions)

Not necessary, just throwing some ideas around in case any of them seem like an improvement.

llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp

3021

Same as above

llvm/trunk/lib/Target/AArch64/AArch64FastISel.cpp

4828

Same as above

(starting to see a pattern - perhaps there's a common API/abstraction we could use for these?)

[eddyb]

llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp

516

I think I would use gep_type_iterator, but I would have to rewrite each of these loops (and maybe make gep_type_iterator nicer to use, with C++11 range-for sugar perhaps).

Although, now that I've gotten myself set with a fast rebase/rebuild cycle, it shouldn't be such an ordeal anymore, I might just try it out.

llvm/trunk/include/llvm/IR/Operator.h

public:
   }
 
   /// Method to return the pointer operand as a PointerType.
   Type *getPointerOperandType() const {
     return getPointerOperand()->getType();
   }
 
   Type *getSourceElementType() const;
+  Type *getResultElementType() const;
 
   /// Method to return the address space of the pointer operand.
   unsigned getPointerAddressSpace() const {
     return getPointerOperandType()->getPointerAddressSpace();
   }
 
   unsigned getNumIndices() const {  // Note: always non-negative
     return getNumOperands() - 1;

llvm/trunk/lib/Analysis/BasicAliasAnalysis.cpp

if (!GEPOp) {
           V = Simplified;
           continue;
         }
 
       return V;
     }
 
     // Don't attempt to analyze GEPs over unsized objects.
-    if (!GEPOp->getOperand(0)->getType()->getPointerElementType()->isSized())
+    if (!GEPOp->getSourceElementType()->isSized())
       return V;
 
     unsigned AS = GEPOp->getPointerAddressSpace();
     // Walk the indices of the GEP, accumulating them into BaseOff/VarIndices.
     gep_type_iterator GTI = gep_type_begin(GEPOp);
     for (User::const_op_iterator I = GEPOp->op_begin() + 1, E = GEPOp->op_end();
          I != E; ++I) {
       const Value *Index = *I;

llvm/trunk/lib/Analysis/ScalarEvolution.cpp

const SCEV *ScalarEvolution::createNodeForSelectOrPHI(Instruction *I,
 
   return getUnknown(I);
 }
 
 /// createNodeForGEP - Expand GEP instructions into add and multiply
 /// operations. This allows them to be analyzed by regular SCEV code.
 ///
 const SCEV *ScalarEvolution::createNodeForGEP(GEPOperator *GEP) {
-  Value *Base = GEP->getOperand(0);
   // Don't attempt to analyze GEPs over unsized objects.
-  if (!Base->getType()->getPointerElementType()->isSized())
+  if (!GEP->getSourceElementType()->isSized())
     return getUnknown(GEP);
 
   SmallVector<const SCEV *, 4> IndexExprs;
   for (auto Index = GEP->idx_begin(); Index != GEP->idx_end(); ++Index)
     IndexExprs.push_back(getSCEV(*Index));
-  return getGEPExpr(GEP->getSourceElementType(), getSCEV(Base), IndexExprs,
-                    GEP->isInBounds());
+  return getGEPExpr(GEP->getSourceElementType(),
+                    getSCEV(GEP->getPointerOperand()),
+                    IndexExprs, GEP->isInBounds());
 }
 
 /// GetMinTrailingZeros - Determine the minimum number of zero bits that S is
 /// guaranteed to end in (at every loop iteration).  It is, at the same time,
 /// the minimum number of times S is divisible by 2.  For example, given {4,+,8}
 /// it returns 2.  If S is guaranteed to be 0, it returns the bitwidth of S.
 uint32_t
 ScalarEvolution::GetMinTrailingZeros(const SCEV *S) {

llvm/trunk/lib/Analysis/ValueTracking.cpp

bool llvm::getConstantStringInfo(const Value *V, StringRef &Str,
   // If the value is a GEP instruction or constant expression, treat it as an
   // offset.
   if (const GEPOperator *GEP = dyn_cast<GEPOperator>(V)) {
     // Make sure the GEP has exactly three arguments.
     if (GEP->getNumOperands() != 3)
       return false;
 
     // Make sure the index-ee is a pointer to array of i8.
-    PointerType *PT = cast<PointerType>(GEP->getOperand(0)->getType());
-    ArrayType *AT = dyn_cast<ArrayType>(PT->getElementType());
+    ArrayType *AT = dyn_cast<ArrayType>(GEP->getSourceElementType());
     if (!AT || !AT->getElementType()->isIntegerTy(8))
       return false;
 
     // Check to make sure that the first operand of the GEP is an integer and
     // has value 0 so that we are sure we're indexing into the initializer.
     const ConstantInt *FirstIdx = dyn_cast<ConstantInt>(GEP->getOperand(1));
     if (!FirstIdx || !FirstIdx->isZero())
       return false;
if (const Argument *A = dyn_cast<Argument>(V))
     if (A->hasByValAttr())
       return isAligned(V, Align, DL);
 
   if (isDereferenceableFromAttribute(V, DL, CtxI, DT, TLI))
     return isAligned(V, Align, DL);
 
   // For GEPs, determine if the indexing lands within the allocated object.
   if (const GEPOperator *GEP = dyn_cast<GEPOperator>(V)) {
-    Type *VTy = GEP->getType();
-    Type *Ty = VTy->getPointerElementType();
+    Type *Ty = GEP->getResultElementType();
     const Value *Base = GEP->getPointerOperand();
 
     // Conservatively require that the base pointer be fully dereferenceable
     // and aligned.
     if (!Visited.insert(Base).second)
       return false;
     if (!isDereferenceableAndAlignedPointer(Base, Align, DL, CtxI, DT, TLI,
                                             Visited))
       return false;
 
-    APInt Offset(DL.getPointerTypeSizeInBits(VTy), 0);
+    APInt Offset(DL.getPointerTypeSizeInBits(GEP->getType()), 0);
     if (!GEP->accumulateConstantOffset(DL, Offset))
       return false;
 
     // Check if the load is within the bounds of the underlying object
     // and offset is aligned.
     uint64_t LoadSize = DL.getTypeStoreSize(Ty);
-    Type *BaseType = Base->getType()->getPointerElementType();
+    Type *BaseType = GEP->getSourceElementType();
     assert(isPowerOf2_32(Align) && "must be a power of 2!");
     return (Offset + LoadSize).ule(DL.getTypeAllocSize(BaseType)) && 
            !(Offset & APInt(Offset.getBitWidth(), Align-1));
   }
 
   // For gc.relocate, look through relocations
   if (const GCRelocateInst *RelocateInst = dyn_cast<GCRelocateInst>(V))
     return isDereferenceableAndAlignedPointer(

llvm/trunk/lib/Analysis/VectorUtils.cpp

 }
 
 /// \brief Find the operand of the GEP that should be checked for consecutive
 /// stores. This ignores trailing indices that have no effect on the final
 /// pointer.
 unsigned llvm::getGEPInductionOperand(const GetElementPtrInst *Gep) {
   const DataLayout &DL = Gep->getModule()->getDataLayout();
   unsigned LastOperand = Gep->getNumOperands() - 1;
-  unsigned GEPAllocSize = DL.getTypeAllocSize(
-      cast<PointerType>(Gep->getType()->getScalarType())->getElementType());
+  unsigned GEPAllocSize = DL.getTypeAllocSize(Gep->getResultElementType());
 
   // Walk backwards and try to peel off zeros.
   while (LastOperand > 1 && match(Gep->getOperand(LastOperand), m_Zero())) {
     // Find the type we're currently indexing into.
     gep_type_iterator GEPTI = gep_type_begin(Gep);
     std::advance(GEPTI, LastOperand - 1);
 
     // If it's a type with the same allocation size as the result of the GEP we

llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp

if (auto *StTy = dyn_cast<StructType>(Ty)) {
           if (!N) // Unhandled operand. Halt "fast" selection and bail.
             return false;
           NIsKill = true;
           TotalOffs = 0;
         }
       }
       Ty = StTy->getElementType(Field);
     } else {
+      if (Ty->isPointerTy()) {
+        // The only pointer type is for the very first index,
+        // therefore the next type is the source element type.
+        Ty = cast<GEPOperator>(I)->getSourceElementType();
+      } else {
         Ty = cast<SequentialType>(Ty)->getElementType();
+      }
 
       // If this is a constant subscript, handle it quickly.
       if (const auto *CI = dyn_cast<ConstantInt>(Idx)) {
         if (CI->isZero())
           continue;
         // N = N + Offset
         uint64_t IdxN = CI->getValue().sextOrTrunc(64).getSExtValue();
         TotalOffs += DL.getTypeAllocSize(Ty) * IdxN;

llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp

if (StructType *StTy = dyn_cast<StructType>(Ty)) {
           Flags.setNoUnsignedWrap(true);
 
         N = DAG.getNode(ISD::ADD, dl, N.getValueType(), N,
                         DAG.getConstant(Offset, dl, N.getValueType()), &Flags);
       }
 
       Ty = StTy->getElementType(Field);
     } else {
+      if (Ty->isPointerTy()) {
+        // The only pointer type is for the very first index,
+        // therefore the next type is the source element type.
+        Ty = cast<GEPOperator>(&I)->getSourceElementType();
+      } else {
         Ty = cast<SequentialType>(Ty)->getElementType();
+      }
+
       MVT PtrTy =
           DAG.getTargetLoweringInfo().getPointerTy(DAG.getDataLayout(), AS);
       unsigned PtrSize = PtrTy.getSizeInBits();
       APInt ElementSize(PtrSize, DL->getTypeAllocSize(Ty));
 
       // If this is a scalar constant or a splat vector of constants,
       // handle it quickly.
       const auto *CI = dyn_cast<ConstantInt>(Idx);

llvm/trunk/lib/IR/ConstantFold.h

namespace llvm {
   Constant *ConstantFoldExtractValueInstruction(Constant *Agg,
                                                 ArrayRef<unsigned> Idxs);
   Constant *ConstantFoldInsertValueInstruction(Constant *Agg, Constant *Val,
                                                ArrayRef<unsigned> Idxs);
   Constant *ConstantFoldBinaryInstruction(unsigned Opcode, Constant *V1,
                                           Constant *V2);
   Constant *ConstantFoldCompareInstruction(unsigned short predicate, 
                                            Constant *C1, Constant *C2);
-  Constant *ConstantFoldGetElementPtr(Constant *C, bool inBounds,
-                                      ArrayRef<Constant *> Idxs);
-  Constant *ConstantFoldGetElementPtr(Constant *C, bool inBounds,
-                                      ArrayRef<Value *> Idxs);
   Constant *ConstantFoldGetElementPtr(Type *Ty, Constant *C, bool inBounds,
                                       ArrayRef<Constant *> Idxs);
   Constant *ConstantFoldGetElementPtr(Type *Ty, Constant *C, bool inBounds,
                                       ArrayRef<Value *> Idxs);
 } // End llvm namespace
 
 #endif

llvm/trunk/lib/IR/ConstantFold.cpp

static Constant *ConstantFoldGetElementPtrImpl(Type *PointeeTy, Constant *C,
                                                ArrayRef<IndexTy> Idxs) {
   if (Idxs.empty()) return C;
   Constant *Idx0 = cast<Constant>(Idxs[0]);
   if ((Idxs.size() == 1 && Idx0->isNullValue()))
     return C;
 
   if (isa<UndefValue>(C)) {
     PointerType *PtrTy = cast<PointerType>(C->getType()->getScalarType());
-    Type *Ty = GetElementPtrInst::getIndexedType(PtrTy->getElementType(), Idxs);
+    Type *Ty = GetElementPtrInst::getIndexedType(PointeeTy, Idxs);
     assert(Ty && "Invalid indices for GEP!");
     Type *GEPTy = PointerType::get(Ty, PtrTy->getAddressSpace());
     if (VectorType *VT = dyn_cast<VectorType>(C->getType()))
       GEPTy = VectorType::get(GEPTy, VT->getNumElements());
     return UndefValue::get(GEPTy);
   }
 
   if (C->isNullValue()) {
     bool isNull = true;
     for (unsigned i = 0, e = Idxs.size(); i != e; ++i)
       if (!cast<Constant>(Idxs[i])->isNullValue()) {
         isNull = false;
         break;
       }
     if (isNull) {
       PointerType *PtrTy = cast<PointerType>(C->getType()->getScalarType());
-      Type *Ty =
-          GetElementPtrInst::getIndexedType(PtrTy->getElementType(), Idxs);
+      Type *Ty = GetElementPtrInst::getIndexedType(PointeeTy, Idxs);
+
       assert(Ty && "Invalid indices for GEP!");
       Type *GEPTy = PointerType::get(Ty, PtrTy->getAddressSpace());
       if (VectorType *VT = dyn_cast<VectorType>(C->getType()))
         GEPTy = VectorType::get(GEPTy, VT->getNumElements());
       return Constant::getNullValue(GEPTy);
     }
   }
 
static Constant *ConstantFoldGetElementPtrImpl(Type *PointeeTy, Constant *C,
   if (!Unknown && !inBounds)
     if (auto *GV = dyn_cast<GlobalVariable>(C))
       if (!GV->hasExternalWeakLinkage() && isInBoundsIndices(Idxs))
         return ConstantExpr::getInBoundsGetElementPtr(PointeeTy, C, Idxs);
 
   return nullptr;
 }
 
-Constant *llvm::ConstantFoldGetElementPtr(Constant *C,
-                                          bool inBounds,
-                                          ArrayRef<Constant *> Idxs) {
-  return ConstantFoldGetElementPtrImpl(
-      cast<PointerType>(C->getType()->getScalarType())->getElementType(), C,
-      inBounds, Idxs);
-}
-
-Constant *llvm::ConstantFoldGetElementPtr(Constant *C,
-                                          bool inBounds,
-                                          ArrayRef<Value *> Idxs) {
-  return ConstantFoldGetElementPtrImpl(
-      cast<PointerType>(C->getType()->getScalarType())->getElementType(), C,
-      inBounds, Idxs);
-}
-
 Constant *llvm::ConstantFoldGetElementPtr(Type *Ty, Constant *C,
                                           bool inBounds,
                                           ArrayRef<Constant *> Idxs) {
   return ConstantFoldGetElementPtrImpl(Ty, C, inBounds, Idxs);
 }
 
 Constant *llvm::ConstantFoldGetElementPtr(Type *Ty, Constant *C,
                                           bool inBounds,
                                           ArrayRef<Value *> Idxs) {
   return ConstantFoldGetElementPtrImpl(Ty, C, inBounds, Idxs);
 }

llvm/trunk/lib/IR/Constants.cpp

 }
 
 GetElementPtrConstantExpr::GetElementPtrConstantExpr(
     Type *SrcElementTy, Constant *C, ArrayRef<Constant *> IdxList, Type *DestTy)
     : ConstantExpr(DestTy, Instruction::GetElementPtr,
                    OperandTraits<GetElementPtrConstantExpr>::op_end(this) -
                        (IdxList.size() + 1),
                    IdxList.size() + 1),
-      SrcElementTy(SrcElementTy) {
+      SrcElementTy(SrcElementTy),
+      ResElementTy(GetElementPtrInst::getIndexedType(SrcElementTy, IdxList)) {
   Op<0>() = C;
   Use *OperandList = getOperandList();
   for (unsigned i = 0, E = IdxList.size(); i != E; ++i)
     OperandList[i+1] = IdxList[i];
 }
 
 Type *GetElementPtrConstantExpr::getSourceElementType() const {
   return SrcElementTy;
 }
 
+Type *GetElementPtrConstantExpr::getResultElementType() const {
+  return ResElementTy;
+}
+
 //===----------------------------------------------------------------------===//
 //                       ConstantData* implementations
 
 void ConstantDataArray::anchor() {}
 void ConstantDataVector::anchor() {}
 
 /// getElementType - Return the element type of the array/vector.
 Type *ConstantDataSequential::getElementType() const {

llvm/trunk/lib/IR/ConstantsContext.h

static bool classof(const Value *V) {
     return isa<ConstantExpr>(V) && classof(cast<ConstantExpr>(V));
   }
 };
 
 /// GetElementPtrConstantExpr - This class is private to Constants.cpp, and is
 /// used behind the scenes to implement getelementpr constant exprs.
 class GetElementPtrConstantExpr : public ConstantExpr {
   Type *SrcElementTy;
+  Type *ResElementTy;
   void anchor() override;
   GetElementPtrConstantExpr(Type *SrcElementTy, Constant *C,
                             ArrayRef<Constant *> IdxList, Type *DestTy);
 
 public:
-  static GetElementPtrConstantExpr *Create(Constant *C,
-                                           ArrayRef<Constant*> IdxList,
-                                           Type *DestTy,
-                                           unsigned Flags) {
-    return Create(
-        cast<PointerType>(C->getType()->getScalarType())->getElementType(), C,
-        IdxList, DestTy, Flags);
-  }
   static GetElementPtrConstantExpr *Create(Type *SrcElementTy, Constant *C,
                                            ArrayRef<Constant *> IdxList,
                                            Type *DestTy, unsigned Flags) {
     GetElementPtrConstantExpr *Result = new (IdxList.size() + 1)
         GetElementPtrConstantExpr(SrcElementTy, C, IdxList, DestTy);
     Result->SubclassOptionalData = Flags;
     return Result;
   }
   Type *getSourceElementType() const;
+  Type *getResultElementType() const;
   /// Transparently provide more efficient getOperand methods.
   DECLARE_TRANSPARENT_OPERAND_ACCESSORS(Value);
 
   static bool classof(const ConstantExpr *CE) {
     return CE->getOpcode() == Instruction::GetElementPtr;
   }
   static bool classof(const Value *V) {
     return isa<ConstantExpr>(V) && classof(cast<ConstantExpr>(V));

llvm/trunk/lib/IR/Operator.cpp

 #include "llvm/IR/Operator.h"
 #include "llvm/IR/GetElementPtrTypeIterator.h"
 #include "llvm/IR/Instructions.h"
 #include "llvm/IR/Type.h"
 
 #include "ConstantsContext.h"
 
 namespace llvm {
 Type *GEPOperator::getSourceElementType() const {
   if (auto *I = dyn_cast<GetElementPtrInst>(this))
     return I->getSourceElementType();
   return cast<GetElementPtrConstantExpr>(this)->getSourceElementType();
 }
 
+Type *GEPOperator::getResultElementType() const {
+  if (auto *I = dyn_cast<GetElementPtrInst>(this))
+    return I->getResultElementType();
+  return cast<GetElementPtrConstantExpr>(this)->getResultElementType();
+}
+
 bool GEPOperator::accumulateConstantOffset(const DataLayout &DL,
                                            APInt &Offset) const {
   assert(Offset.getBitWidth() ==
              DL.getPointerSizeInBits(getPointerAddressSpace()) &&
          "The offset must have exactly as many bits as our pointer.");
 
   for (gep_type_iterator GTI = gep_type_begin(this), GTE = gep_type_end(this);
        GTI != GTE; ++GTI) {

llvm/trunk/lib/Target/AArch64/AArch64FastISel.cpp

for (auto OI = std::next(I->op_begin()), E = I->op_end(); OI != E; ++OI) {
     const Value *Idx = *OI;
     if (auto *StTy = dyn_cast<StructType>(Ty)) {
       unsigned Field = cast<ConstantInt>(Idx)->getZExtValue();
       // N = N + Offset
       if (Field)
         TotalOffs += DL.getStructLayout(StTy)->getElementOffset(Field);
       Ty = StTy->getElementType(Field);
     } else {
+      if (Ty->isPointerTy()) {
+        // The only pointer type is for the very first index,
+        // therefore the next type is the source element type.
+        Ty = cast<GEPOperator>(I)->getSourceElementType();
+      } else {
         Ty = cast<SequentialType>(Ty)->getElementType();
+      }
       // If this is a constant subscript, handle it quickly.
       if (const auto *CI = dyn_cast<ConstantInt>(Idx)) {
         if (CI->isZero())
           continue;
         // N = N + Offset
         TotalOffs +=
             DL.getTypeAllocSize(Ty) * cast<ConstantInt>(CI)->getSExtValue();
         continue;

llvm/trunk/lib/Transforms/IPO/GlobalOpt.cpp

if (LoadInst *LI = dyn_cast<LoadInst>(U)) {
             ConstantFoldInstruction(GEP, DL, TLI));
         if (Init && CE && CE->getOpcode() == Instruction::GetElementPtr)
           SubInit = ConstantFoldLoadThroughGEPConstantExpr(Init, CE);
 
         // If the initializer is an all-null value and we have an inbounds GEP,
         // we already know what the result of any load from that GEP is.
         // TODO: Handle splats.
         if (Init && isa<ConstantAggregateZero>(Init) && GEP->isInBounds())
-          SubInit = Constant::getNullValue(GEP->getType()->getElementType());
+          SubInit = Constant::getNullValue(GEP->getResultElementType());
       }
       Changed |= CleanupConstantGlobalUsers(GEP, SubInit, DL, TLI);
 
       if (GEP->use_empty()) {
         GEP->eraseFromParent();
         Changed = true;
       }
     } else if (MemIntrinsic *MI = dyn_cast<MemIntrinsic>(U)) { // memset/cpy/mv

llvm/trunk/lib/Transforms/InstCombine/InstCombineLoadStoreAlloca.cpp

static bool canReplaceGEPIdxWithZero(InstCombiner &IC, GetElementPtrInst *GEPI,
   // type. See if the next index is not a constant.
   Idx = FirstNZIdx(GEPI);
   if (Idx == GEPI->getNumOperands())
     return false;
   if (isa<Constant>(GEPI->getOperand(Idx)))
     return false;
 
   SmallVector<Value *, 4> Ops(GEPI->idx_begin(), GEPI->idx_begin() + Idx);
-  Type *AllocTy = GetElementPtrInst::getIndexedType(
-      cast<PointerType>(GEPI->getOperand(0)->getType()->getScalarType())
-          ->getElementType(),
-      Ops);
+  Type *AllocTy =
+    GetElementPtrInst::getIndexedType(GEPI->getSourceElementType(), Ops);
   if (!AllocTy || !AllocTy->isSized())
     return false;
   const DataLayout &DL = IC.getDataLayout();
   uint64_t TyAllocSize = DL.getTypeAllocSize(AllocTy);
 
   // If there are more indices after the one we might replace with a zero, make
   // sure they're all non-negative. If any of them are negative, the overall
   // address being computed might be before the base address determined by the

llvm/trunk/lib/Transforms/InstCombine/InstructionCombining.cpp

Instruction *InstCombiner::visitGetElementPtrInst(GetElementPtrInst &GEP) {
   bool MadeChange = false;
   Type *IntPtrTy =
     DL.getIntPtrType(GEP.getPointerOperandType()->getScalarType());
 
   gep_type_iterator GTI = gep_type_begin(GEP);
   for (User::op_iterator I = GEP.op_begin() + 1, E = GEP.op_end(); I != E;
        ++I, ++GTI) {
     // Skip indices into struct types.
-    SequentialType *SeqTy = dyn_cast<SequentialType>(*GTI);
-    if (!SeqTy)
+    if (isa<StructType>(*GTI))
       continue;
 
     // Index type should have the same width as IntPtr
     Type *IndexTy = (*I)->getType();
     Type *NewIndexType = IndexTy->isVectorTy() ?
       VectorType::get(IntPtrTy, IndexTy->getVectorNumElements()) : IntPtrTy;
- 
     // If the element type has zero size then any index over it is equivalent
     // to an index of zero, so replace it with zero if it is not zero already.
-    if (SeqTy->getElementType()->isSized() &&
-        DL.getTypeAllocSize(SeqTy->getElementType()) == 0)
+    Type *EltTy = GTI.getIndexedType();
+    if (EltTy->isSized() && DL.getTypeAllocSize(EltTy) == 0)
       if (!isa<Constant>(*I) || !cast<Constant>(*I)->isNullValue()) {
         *I = Constant::getNullValue(NewIndexType);
         MadeChange = true;
       }
 
     if (IndexTy != NewIndexType) {
       // If we are using a wider index than needed for this platform, shrink
       // it to what we need.  If narrower, sign-extend it to what we need.
for (auto I = PN->op_begin()+1, E = PN->op_end(); I !=E; ++I) {
       if (!Op2 || Op1->getNumOperands() != Op2->getNumOperands())
         return nullptr;
 
       // As for Op1 above, don't try to fold a GEP into itself.
       if (Op2 == &GEP)
         return nullptr;
 
       // Keep track of the type as we walk the GEP.
-      Type *CurTy = Op1->getOperand(0)->getType()->getScalarType();
+      Type *CurTy = nullptr;
 
       for (unsigned J = 0, F = Op1->getNumOperands(); J != F; ++J) {
         if (Op1->getOperand(J)->getType() != Op2->getOperand(J)->getType())
           return nullptr;
 
         if (Op1->getOperand(J) != Op2->getOperand(J)) {
           if (DI == -1) {
             // We have not seen any differences yet in the GEPs feeding the
for (auto I = PN->op_begin()+1, E = PN->op_end(); I !=E; ++I) {
             // directly supports and would need to be broken into several
             // simpler instructions anyway.
             return nullptr;
           }
         }
 
         // Sink down a layer of the type for the next iteration.
         if (J > 0) {
-          if (CompositeType *CT = dyn_cast<CompositeType>(CurTy)) {
+          if (J == 1) {
+            CurTy = Op1->getSourceElementType();
+          } else if (CompositeType *CT = dyn_cast<CompositeType>(CurTy)) {
             CurTy = CT->getTypeAtIndex(Op1->getOperand(J));
           } else {
             CurTy = nullptr;
           }
         }
       }
     }
 
if (!Indices.empty())
                                              Src->getOperand(0), Indices,
                                              GEP.getName());
   }
 
   if (GEP.getNumIndices() == 1) {
     unsigned AS = GEP.getPointerAddressSpace();
     if (GEP.getOperand(1)->getType()->getScalarSizeInBits() ==
         DL.getPointerSizeInBits(AS)) {
-      Type *PtrTy = GEP.getPointerOperandType();
-      Type *Ty = PtrTy->getPointerElementType();
+      Type *Ty = GEP.getSourceElementType();
       uint64_t TyAllocSize = DL.getTypeAllocSize(Ty);
 
       bool Matched = false;
       uint64_t C;
       Value *V = nullptr;
       if (TyAllocSize == 1) {
         V = GEP.getOperand(1);
         Matched = true;
if (StrippedPtr != PtrOp) {
     // Transform: GEP (bitcast [10 x i8]* X to [0 x i8]*), i32 0, ...
     // into     : GEP [10 x i8]* X, i32 0, ...
     //
     // Likewise, transform: GEP (bitcast i8* X to [0 x i8]*), i32 0, ...
     //           into     : GEP i8* X, ...
     //
     // This occurs when the program declares an array extern like "int X[];"
     if (HasZeroPointerIndex) {
-      PointerType *CPTy = cast<PointerType>(PtrOp->getType());
       if (ArrayType *CATy =
-          dyn_cast<ArrayType>(CPTy->getElementType())) {
+          dyn_cast<ArrayType>(GEP.getSourceElementType())) {
         // GEP (bitcast i8* X to [0 x i8]*), i32 0, ... ?
         if (CATy->getElementType() == StrippedPtrTy->getElementType()) {
           // -> GEP i8* X, ...
           SmallVector<Value*, 8> Idx(GEP.idx_begin()+1, GEP.idx_end());
           GetElementPtrInst *Res = GetElementPtrInst::Create(
               StrippedPtrTy->getElementType(), StrippedPtr, Idx, GEP.getName());
           Res->setIsInBounds(GEP.isInBounds());
           if (StrippedPtrTy->getAddressSpace() == GEP.getAddressSpace())
if (HasZeroPointerIndex) {
           }
         }
       }
     } else if (GEP.getNumOperands() == 2) {
       // Transform things like:
       // %t = getelementptr i32* bitcast ([2 x i32]* %str to i32*), i32 %V
       // into:  %t1 = getelementptr [2 x i32]* %str, i32 0, i32 %V; bitcast
       Type *SrcElTy = StrippedPtrTy->getElementType();
-      Type *ResElTy = PtrOp->getType()->getPointerElementType();
+      Type *ResElTy = GEP.getSourceElementType();
       if (SrcElTy->isArrayTy() &&
           DL.getTypeAllocSize(SrcElTy->getArrayElementType()) ==
               DL.getTypeAllocSize(ResElTy)) {
         Type *IdxType = DL.getIntPtrType(GEP.getType());
         Value *Idx[2] = { Constant::getNullValue(IdxType), GEP.getOperand(1) };
         Value *NewGEP =
             GEP.isInBounds()
                 ? Builder->CreateInBoundsGEP(nullptr, StrippedPtr, Idx,

llvm/trunk/lib/Transforms/Scalar/NaryReassociate.cpp

for (auto I = GEP->idx_begin(); I != GEP->idx_end(); ++I, ++GTI) {
     } else {
       StructType *STy = cast<StructType>(*GTI);
       uint64_t Field = cast<ConstantInt>(*I)->getZExtValue();
       BaseOffset += DL->getStructLayout(STy)->getElementOffset(Field);
     }
   }
 
   unsigned AddrSpace = GEP->getPointerAddressSpace();
-  return TTI->isLegalAddressingMode(GEP->getType()->getElementType(), BaseGV,
+  return TTI->isLegalAddressingMode(GEP->getResultElementType(), BaseGV,
                                     BaseOffset, HasBaseReg, Scale, AddrSpace);
 }
 
 Instruction *NaryReassociate::tryReassociateGEP(GetElementPtrInst *GEP) {
   // Not worth reassociating GEP if it is foldable.
   if (isGEPFoldable(GEP, TTI, DL))
     return nullptr;
 
GetElementPtrInst *NaryReassociate::tryReassociateGEPAtIndex(
   // Candidate does not necessarily have the same pointer type as GEP. Use
   // bitcast or pointer cast to make sure they have the same type, so that the
   // later RAUW doesn't complain.
   Candidate = Builder.CreateBitOrPointerCast(Candidate, GEP->getType());
   assert(Candidate->getType() == GEP->getType());
 
   // NewGEP = (char *)Candidate + RHS * sizeof(IndexedType)
   uint64_t IndexedSize = DL->getTypeAllocSize(IndexedType);
-  Type *ElementType = GEP->getType()->getElementType();
+  Type *ElementType = GEP->getResultElementType();
   uint64_t ElementSize = DL->getTypeAllocSize(ElementType);
   // Another less rare case: because I is not necessarily the last index of the
   // GEP, the size of the type at the I-th index (IndexedSize) is not
   // necessarily divisible by ElementSize. For example,
   //
   // #pragma pack(1)
   // struct S {
   //   int a[3];

llvm/trunk/lib/Transforms/Scalar/RewriteStatepointsForGC.cpp

if (CastInst *CI = dyn_cast<CastInst>(Instr)) {
       assert(CI->isNoopCast(CI->getModule()->getDataLayout()) &&
              "non noop cast is found during rematerialization");
 
       Type *SrcTy = CI->getOperand(0)->getType();
       Cost += TTI.getCastInstrCost(CI->getOpcode(), CI->getType(), SrcTy);
 
     } else if (GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(Instr)) {
       // Cost of the address calculation
-      Type *ValTy = GEP->getPointerOperandType()->getPointerElementType();
+      Type *ValTy = GEP->getSourceElementType();
       Cost += TTI.getAddressComputationCost(ValTy);
 
       // And cost of the GEP itself
       // TODO: Use TTI->getGEPCost here (it exists, but appears to be not
       //       allowed for the external usage)
       if (!GEP->hasAllConstantIndices())
         Cost += 2;
 

llvm/trunk/lib/Transforms/Scalar/ScalarReplAggregates.cpp

if (BitCastInst *BCI = dyn_cast<BitCastInst>(UI)) {
         IsNotTrivial = true;  // Can't be mem2reg'd.
       if (!CanConvertToScalar(BCI, Offset, NonConstantIdx))
         return false;
       continue;
     }
 
     if (GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(UI)) {
       // If this is a GEP with a variable indices, we can't handle it.
-      PointerType* PtrTy = dyn_cast<PointerType>(GEP->getPointerOperandType());
-      if (!PtrTy)
-        return false;
-
       // Compute the offset that this GEP adds to the pointer.
       SmallVector<Value*, 8> Indices(GEP->op_begin()+1, GEP->op_end());
       Value *GEPNonConstantIdx = nullptr;
       if (!GEP->hasAllConstantIndices()) {
-        if (!isa<VectorType>(PtrTy->getElementType()))
+        if (!isa<VectorType>(GEP->getSourceElementType()))
           return false;
         if (NonConstantIdx)
           return false;
         GEPNonConstantIdx = Indices.pop_back_val();
         if (!GEPNonConstantIdx->getType()->isIntegerTy(32))
           return false;
         HadDynamicAccess = true;
       } else
         GEPNonConstantIdx = NonConstantIdx;
-      uint64_t GEPOffset = DL.getIndexedOffset(PtrTy,
+      uint64_t GEPOffset = DL.getIndexedOffset(GEP->getPointerOperandType(),
                                                Indices);
       // See if all uses can be converted.
       if (!CanConvertToScalar(GEP, Offset+GEPOffset, GEPNonConstantIdx))
         return false;
       IsNotTrivial = true;  // Can't be mem2reg'd.
       HadNonMemTransferAccess = true;
       continue;
     }

llvm/trunk/lib/Transforms/Scalar/SeparateConstOffsetFromGEP.cpp

bool SeparateConstOffsetFromGEP::splitGEP(GetElementPtrInst *GEP) {
   // the addressing mode, we can still do optimizations to other lowered parts
   // of variable indices. Therefore, we don't check for addressing modes in that
   // case.
   if (!LowerGEP) {
     TargetTransformInfo &TTI =
         getAnalysis<TargetTransformInfoWrapperPass>().getTTI(
             *GEP->getParent()->getParent());
     unsigned AddrSpace = GEP->getPointerAddressSpace();
-    if (!TTI.isLegalAddressingMode(GEP->getType()->getElementType(),
+    if (!TTI.isLegalAddressingMode(GEP->getResultElementType(),
                                    /*BaseGV=*/nullptr, AccumulativeByteOffset,
                                    /*HasBaseReg=*/true, /*Scale=*/0,
                                    AddrSpace)) {
       return Changed;
     }
   }
 
   // Remove the constant offset in each sequential index. The resultant GEP
bool SeparateConstOffsetFromGEP::splitGEP(GetElementPtrInst *GEP) {
   //   ... %new.gep ...
   Instruction *NewGEP = GEP->clone();
   NewGEP->insertBefore(GEP);
 
   // Per ANSI C standard, signed / unsigned = unsigned and signed % unsigned =
   // unsigned.. Therefore, we cast ElementTypeSizeOfGEP to signed because it is
   // used with unsigned integers later.
   int64_t ElementTypeSizeOfGEP = static_cast<int64_t>(
-      DL->getTypeAllocSize(GEP->getType()->getElementType()));
+      DL->getTypeAllocSize(GEP->getResultElementType()));
   Type *IntPtrTy = DL->getIntPtrType(GEP->getType());
   if (AccumulativeByteOffset % ElementTypeSizeOfGEP == 0) {
     // Very likely. As long as %gep is natually aligned, the byte offset we
     // extracted should be a multiple of sizeof(*%gep).
     int64_t Index = AccumulativeByteOffset / ElementTypeSizeOfGEP;
     NewGEP = GetElementPtrInst::Create(GEP->getResultElementType(), NewGEP,
                                        ConstantInt::get(IntPtrTy, Index, true),
                                        GEP->getName(), GEP);

llvm/trunk/lib/Transforms/Scalar/StraightLineStrengthReduce.cpp

for (auto I = GEP->idx_begin(); I != GEP->idx_end(); ++I, ++GTI) {
     } else {
       StructType *STy = cast<StructType>(*GTI);
       uint64_t Field = cast<ConstantInt>(*I)->getZExtValue();
       BaseOffset += DL->getStructLayout(STy)->getElementOffset(Field);
     }
   }
 
   unsigned AddrSpace = GEP->getPointerAddressSpace();
-  return TTI->isLegalAddressingMode(GEP->getType()->getElementType(), BaseGV,
+  return TTI->isLegalAddressingMode(GEP->getResultElementType(), BaseGV,
                                     BaseOffset, HasBaseReg, Scale, AddrSpace);
 }
 
 // Returns whether (Base + Index * Stride) can be folded to an addressing mode.
 static bool isAddFoldable(const SCEV *Base, ConstantInt *Index, Value *Stride,
                           TargetTransformInfo *TTI) {
   return TTI->isLegalAddressingMode(Base->getType(), nullptr, 0, true,
                                     Index->getSExtValue());
Value *StraightLineStrengthReduce::emitBump(const Candidate &Basis,
   APInt Idx = C.Index->getValue(), BasisIdx = Basis.Index->getValue();
   unifyBitWidth(Idx, BasisIdx);
   APInt IndexOffset = Idx - BasisIdx;
 
   BumpWithUglyGEP = false;
   if (Basis.CandidateKind == Candidate::GEP) {
     APInt ElementSize(
         IndexOffset.getBitWidth(),
-        DL->getTypeAllocSize(
-            cast<GetElementPtrInst>(Basis.Ins)->getType()->getElementType()));
+        DL->getTypeAllocSize(cast<GetElementPtrInst>(Basis.Ins)->getResultElementType()));
     APInt Q, R;
     APInt::sdivrem(IndexOffset, ElementSize, Q, R);
     if (R.getSExtValue() == 0)
       IndexOffset = Q;
     else
       BumpWithUglyGEP = true;
   }