[InstSimplify] Don't lose inbounds when simplifying a GEP

[arichardson]

Summary

I noticed this while working on a (ptrtoint (gep null, x)) -> x fold.

[nikic]

LGTM

llvm/lib/Transforms/Scalar/NewGVN.cpp

1197–1200

Legit warning.

llvm/include/llvm/Analysis/InstructionSimplify.h

 Value *SimplifyFCmpInst(unsigned Predicate, Value *LHS, Value *RHS,
                         FastMathFlags FMF, const SimplifyQuery &Q);
 
 /// Given operands for a SelectInst, fold the result or return null.
 Value *SimplifySelectInst(Value *Cond, Value *TrueVal, Value *FalseVal,
                           const SimplifyQuery &Q);
 
 /// Given operands for a GetElementPtrInst, fold the result or return null.
-Value *SimplifyGEPInst(Type *SrcTy, ArrayRef<Value *> Ops,
+Value *SimplifyGEPInst(Type *SrcTy, ArrayRef<Value *> Ops, bool InBounds,
                        const SimplifyQuery &Q);
 
 /// Given operands for an InsertValueInst, fold the result or return null.
 Value *SimplifyInsertValueInst(Value *Agg, Value *Val, ArrayRef<unsigned> Idxs,
                                const SimplifyQuery &Q);
 
 /// Given operands for an InsertElement, fold the result or return null.
 Value *SimplifyInsertElementInst(Value *Vec, Value *Elt, Value *Idx,

llvm/lib/Analysis/InstructionSimplify.cpp

 static Value *SimplifyCmpInst(unsigned, Value *, Value *, const SimplifyQuery &,
                               unsigned);
 static Value *SimplifyICmpInst(unsigned Predicate, Value *LHS, Value *RHS,
                                const SimplifyQuery &Q, unsigned MaxRecurse);
 static Value *SimplifyOrInst(Value *, Value *, const SimplifyQuery &, unsigned);
 static Value *SimplifyXorInst(Value *, Value *, const SimplifyQuery &, unsigned);
 static Value *SimplifyCastInst(unsigned, Value *, Type *,
                                const SimplifyQuery &, unsigned);
-static Value *SimplifyGEPInst(Type *, ArrayRef<Value *>, const SimplifyQuery &,
-                              unsigned);
+static Value *SimplifyGEPInst(Type *, ArrayRef<Value *>, bool,
+                              const SimplifyQuery &, unsigned);
 static Value *SimplifySelectInst(Value *, Value *, Value *,
                                  const SimplifyQuery &, unsigned);
 
 static Value *foldSelectWithBinaryOp(Value *Cond, Value *TrueVal,
                                      Value *FalseVal) {
   BinaryOperator::BinaryOps BinOpCode;
   if (auto *BO = dyn_cast<BinaryOperator>(Cond))
     BinOpCode = BO->getOpcode();
if (auto *B = dyn_cast<BinaryOperator>(I))
                                                NewOps[1], Q, MaxRecurse - 1));
 
     if (CmpInst *C = dyn_cast<CmpInst>(I))
       return PreventSelfSimplify(SimplifyCmpInst(C->getPredicate(), NewOps[0],
                                                  NewOps[1], Q, MaxRecurse - 1));
 
     if (auto *GEP = dyn_cast<GetElementPtrInst>(I))
       return PreventSelfSimplify(SimplifyGEPInst(GEP->getSourceElementType(),
-                                                 NewOps, Q, MaxRecurse - 1));
+                                                 NewOps, GEP->isInBounds(), Q,
+                                                 MaxRecurse - 1));
 
     if (isa<SelectInst>(I))
       return PreventSelfSimplify(
           SimplifySelectInst(NewOps[0], NewOps[1], NewOps[2], Q,
                              MaxRecurse - 1));
     // TODO: We could hand off more cases to instsimplify here.
   }
 

 
 Value *llvm::SimplifySelectInst(Value *Cond, Value *TrueVal, Value *FalseVal,
                                 const SimplifyQuery &Q) {
   return ::SimplifySelectInst(Cond, TrueVal, FalseVal, Q, RecursionLimit);
 }
 
 /// Given operands for an GetElementPtrInst, see if we can fold the result.
 /// If not, this returns null.
-static Value *SimplifyGEPInst(Type *SrcTy, ArrayRef<Value *> Ops,
+static Value *SimplifyGEPInst(Type *SrcTy, ArrayRef<Value *> Ops, bool InBounds,
                               const SimplifyQuery &Q, unsigned) {
   // The type of the GEP pointer operand.
   unsigned AS =
       cast<PointerType>(Ops[0]->getType()->getScalarType())->getAddressSpace();
 
   // getelementptr P -> P.
   if (Ops.size() == 1)
     return Ops[0];
if (!IsScalableVec && Q.DL.getTypeAllocSize(LastType) == 1 &&
     }
   }
 
   // Check to see if this is constant foldable.
   if (!all_of(Ops, [](Value *V) { return isa<Constant>(V); }))
     return nullptr;
 
   auto *CE = ConstantExpr::getGetElementPtr(SrcTy, cast<Constant>(Ops[0]),
-                                            Ops.slice(1));
+                                            Ops.slice(1), InBounds);
   return ConstantFoldConstant(CE, Q.DL);
 }
 
-Value *llvm::SimplifyGEPInst(Type *SrcTy, ArrayRef<Value *> Ops,
+Value *llvm::SimplifyGEPInst(Type *SrcTy, ArrayRef<Value *> Ops, bool InBounds,
                              const SimplifyQuery &Q) {
-  return ::SimplifyGEPInst(SrcTy, Ops, Q, RecursionLimit);
+  return ::SimplifyGEPInst(SrcTy, Ops, InBounds, Q, RecursionLimit);
 }
 
 /// Given operands for an InsertValueInst, see if we can fold the result.
 /// If not, this returns null.
 static Value *SimplifyInsertValueInst(Value *Agg, Value *Val,
                                       ArrayRef<unsigned> Idxs, const SimplifyQuery &Q,
                                       unsigned) {
   if (Constant *CAgg = dyn_cast<Constant>(Agg))
static Value *simplifyInstructionWithOperands(Instruction *I,
   case Instruction::FCmp:
     Result = SimplifyFCmpInst(cast<FCmpInst>(I)->getPredicate(), NewOps[0],
                               NewOps[1], I->getFastMathFlags(), Q);
     break;
   case Instruction::Select:
     Result = SimplifySelectInst(NewOps[0], NewOps[1], NewOps[2], Q);
     break;
   case Instruction::GetElementPtr: {
-    Result = SimplifyGEPInst(cast<GetElementPtrInst>(I)->getSourceElementType(),
-                             NewOps, Q);
+    auto *GEPI = cast<GetElementPtrInst>(I);
+    Result = SimplifyGEPInst(GEPI->getSourceElementType(), NewOps,
+                             GEPI->isInBounds(), Q);
     break;
   }
   case Instruction::InsertValue: {
     InsertValueInst *IV = cast<InsertValueInst>(I);
     Result = SimplifyInsertValueInst(NewOps[0], NewOps[1], IV->getIndices(), Q);
     break;
   }
   case Instruction::InsertElement: {

llvm/lib/Analysis/PHITransAddr.cpp

for (unsigned i = 0, e = GEP->getNumOperands(); i != e; ++i) {
       AnyChanged |= GEPOp != GEP->getOperand(i);
       GEPOps.push_back(GEPOp);
     }
 
     if (!AnyChanged)
       return GEP;
 
     // Simplify the GEP to handle 'gep x, 0' -> x etc.
-    if (Value *V = SimplifyGEPInst(GEP->getSourceElementType(),
-                                   GEPOps, {DL, TLI, DT, AC})) {
+    if (Value *V = SimplifyGEPInst(GEP->getSourceElementType(), GEPOps,
+                                   GEP->isInBounds(), {DL, TLI, DT, AC})) {
       for (unsigned i = 0, e = GEPOps.size(); i != e; ++i)
         RemoveInstInputs(GEPOps[i], InstInputs);
 
       return AddAsInput(V);
     }
 
     // Scan to see if we have this GEP available.
     Value *APHIOp = GEPOps[0];

llvm/lib/Transforms/InstCombine/InstructionCombining.cpp

static Instruction *foldSelectGEP(GetElementPtrInst &GEP,
   return SelectInst::Create(Cond, NewTrueC, NewFalseC, "", nullptr, Sel);
 }
 
 Instruction *InstCombinerImpl::visitGetElementPtrInst(GetElementPtrInst &GEP) {
   SmallVector<Value *, 8> Ops(GEP.operands());
   Type *GEPType = GEP.getType();
   Type *GEPEltType = GEP.getSourceElementType();
   bool IsGEPSrcEleScalable = isa<ScalableVectorType>(GEPEltType);
-  if (Value *V = SimplifyGEPInst(GEPEltType, Ops, SQ.getWithInstruction(&GEP)))
+  if (Value *V = SimplifyGEPInst(GEPEltType, Ops, GEP.isInBounds(),
+                                 SQ.getWithInstruction(&GEP)))
     return replaceInstUsesWith(GEP, V);
 
   // For vector geps, use the generic demanded vector support.
   // Skip if GEP return type is scalable. The number of elements is unknown at
   // compile-time.
   if (auto *GEPFVTy = dyn_cast<FixedVectorType>(GEPType)) {
     auto VWidth = GEPFVTy->getNumElements();
     APInt UndefElts(VWidth, 0);

llvm/lib/Transforms/Scalar/NewGVN.cpp

Value *V =
         SimplifyBinOp(E->getOpcode(), E->getOperand(0), E->getOperand(1), SQ);
     if (auto Simplified = checkExprResults(E, I, V))
       return Simplified;
   } else if (auto *CI = dyn_cast<CastInst>(I)) {
     Value *V =
         SimplifyCastInst(CI->getOpcode(), E->getOperand(0), CI->getType(), SQ);
     if (auto Simplified = checkExprResults(E, I, V))
       return Simplified;
-  } else if (isa<GetElementPtrInst>(I)) {
-    Value *V = SimplifyGEPInst(
-        E->getType(), ArrayRef<Value *>(E->op_begin(), E->op_end()), SQ);
+  } else if (auto *GEPI = dyn_cast<GetElementPtrInst>(I)) {
+    Value *V = SimplifyGEPInst(GEPI->getSourceElementType(),
+                               ArrayRef<Value *>(E->op_begin(), E->op_end()),
+                               GEPI->isInBounds(), SQ);
     if (auto Simplified = checkExprResults(E, I, V))
       return Simplified;
   } else if (AllConstant) {
     // We don't bother trying to simplify unless all of the operands
     // were constant.
     // TODO: There are a lot of Simplify*'s we could call here, if we
     // wanted to.  The original motivating case for this code was a
     // zext i1 false to i8, which we don't have an interface to

llvm/test/Transforms/InstSimplify/simplify-nested-bitcast.ll

entry:
   %3 = bitcast i32 (i32 addrspace(1)*) addrspace(4)* %1 to i8 addrspace(4)*
   %4 = bitcast i8* addrspace(4)* %2 to i32 (i8 addrspace(4)*, i32 addrspace(1)*)* addrspace(4)*
   %5 = load i32 (i8 addrspace(4)*, i32 addrspace(1)*)*, i32 (i8 addrspace(4)*, i32 addrspace(1)*)* addrspace(4)* %4, align 8
   %call = tail call i32 %5(i8 addrspace(4)* %3, i32 addrspace(1)* %0) #2
   ret i32 %call
 }
 
 ; CHECK-LABEL: define void @f1
-; CHECK: %1 = load i32 (i8 addrspace(4)*, i32 addrspace(1)*)*, i32 (i8 addrspace(4)*, i32 addrspace(1)*)* addrspace(4)* bitcast (i8* addrspace(4)* getelementptr (%__aaa_struct, %__aaa_struct addrspace(4)* addrspacecast (%__aaa_struct addrspace(1)* @__aaa_struct_ptr to %__aaa_struct addrspace(4)*), i64 0, i32 0, i32 3) to i32 (i8 addrspace(4)*, i32 addrspace(1)*)* addrspace(4)*), align 8
+; CHECK: %1 = load i32 (i8 addrspace(4)*, i32 addrspace(1)*)*, i32 (i8 addrspace(4)*, i32 addrspace(1)*)* addrspace(4)* bitcast (i8* addrspace(4)* getelementptr inbounds (%__aaa_struct, %__aaa_struct addrspace(4)* addrspacecast (%__aaa_struct addrspace(1)* @__aaa_struct_ptr to %__aaa_struct addrspace(4)*), i64 0, i32 0, i32 3) to i32 (i8 addrspace(4)*, i32 addrspace(1)*)* addrspace(4)*), align 8
 
 ; Function Attrs: alwaysinline nounwind
 define void @f1(i32 addrspace(1)*) #1 {
 entry:
   %call = tail call i32 @f0(i32 addrspace(1)* %0, i32 (i32 addrspace(1)*) addrspace(4)* addrspacecast (i32 (i32 addrspace(1)*) addrspace(1)* bitcast (%__aaa_struct addrspace(1)* @__aaa_struct_ptr to i32 (i32 addrspace(1)*) addrspace(1)*) to i32 (i32 addrspace(1)*) addrspace(4)*)) #3
   store i32 %call, i32 addrspace(1)* %0, align 4
   %call1 = tail call i32 @f0(i32 addrspace(1)* %0, i32 (i32 addrspace(1)*) addrspace(4)* addrspacecast (i32 (i32 addrspace(1)*) addrspace(1)* bitcast ({ i8**, i32, i32, i8*, %struct.__block_descriptor addrspace(1)* } addrspace(1)* getelementptr inbounds (%__aaa_struct, %__aaa_struct addrspace(1)* @__aaa_struct_ptr, i32 0, i32 3) to i32 (i32 addrspace(1)*) addrspace(1)*) to i32 (i32 addrspace(1)*) addrspace(4)*)) #3
   store i32 %call1, i32 addrspace(1)* %0, align 4