Use ConstantExpr::getExtractElement when constant-folding vectors

instead of getAggregateElement. This has the advantage of being
more consistent and allowing higher-level constant folding to
procede even if an inner extract element cannot be folded.

Make ConstantFoldInstruction call ConstantFoldConstantExpression
on the instruction's operands, making it more consistent with 
ConstantFoldConstantExpression itself. This makes sure that
ConstantExprs get TargetData-aware folding before being handed
off as operands for further folding.

This causes more expressions to be folded, but due to a known
shortcoming in constant folding, this currently has the side effect
of stripping a few more nuw and inbounds flags in the non-targetdata
side of constant-fold-gep.ll. This is mostly harmless.

This fixes rdar://11324230.

llvm-svn: 155682
This commit is contained in:
Dan Gohman 2012-04-27 00:54:36 +00:00
parent 641a87b983
commit 90f3798f26
4 changed files with 74 additions and 35 deletions

View File

@ -788,6 +788,10 @@ Constant *llvm::ConstantFoldInstruction(Instruction *I,
CommonValue = C;
}
// Fold the PHI's operands.
if (ConstantExpr *NewCE = dyn_cast<ConstantExpr>(CommonValue))
CommonValue = ConstantFoldConstantExpression(NewCE, TD, TLI);
// If we reach here, all incoming values are the same constant or undef.
return CommonValue ? CommonValue : UndefValue::get(PN->getType());
}
@ -795,12 +799,18 @@ Constant *llvm::ConstantFoldInstruction(Instruction *I,
// Scan the operand list, checking to see if they are all constants, if so,
// hand off to ConstantFoldInstOperands.
SmallVector<Constant*, 8> Ops;
for (User::op_iterator i = I->op_begin(), e = I->op_end(); i != e; ++i)
if (Constant *Op = dyn_cast<Constant>(*i))
Ops.push_back(Op);
else
for (User::op_iterator i = I->op_begin(), e = I->op_end(); i != e; ++i) {
Constant *Op = dyn_cast<Constant>(*i);
if (!Op)
return 0; // All operands not constant!
// Fold the Instruction's operands.
if (ConstantExpr *NewCE = dyn_cast<ConstantExpr>(Op))
Op = ConstantFoldConstantExpression(NewCE, TD, TLI);
Ops.push_back(Op);
}
if (const CmpInst *CI = dyn_cast<CmpInst>(I))
return ConstantFoldCompareInstOperands(CI->getPredicate(), Ops[0], Ops[1],
TD, TLI);

View File

@ -55,13 +55,12 @@ static Constant *BitCastConstantVector(Constant *CV, VectorType *DstTy) {
Type *DstEltTy = DstTy->getElementType();
// Check to verify that all elements of the input are simple.
SmallVector<Constant*, 16> Result;
Type *Ty = IntegerType::get(CV->getContext(), 32);
for (unsigned i = 0; i != NumElts; ++i) {
Constant *C = CV->getAggregateElement(i);
if (C == 0) return 0;
Constant *C =
ConstantExpr::getExtractElement(CV, ConstantInt::get(Ty, i));
C = ConstantExpr::getBitCast(C, DstEltTy);
if (isa<ConstantExpr>(C)) return 0;
Result.push_back(C);
}
@ -553,9 +552,12 @@ Constant *llvm::ConstantFoldCastInstruction(unsigned opc, Constant *V,
SmallVector<Constant*, 16> res;
VectorType *DestVecTy = cast<VectorType>(DestTy);
Type *DstEltTy = DestVecTy->getElementType();
for (unsigned i = 0, e = V->getType()->getVectorNumElements(); i != e; ++i)
res.push_back(ConstantExpr::getCast(opc,
V->getAggregateElement(i), DstEltTy));
Type *Ty = IntegerType::get(V->getContext(), 32);
for (unsigned i = 0, e = V->getType()->getVectorNumElements(); i != e; ++i) {
Constant *C =
ConstantExpr::getExtractElement(V, ConstantInt::get(Ty, i));
res.push_back(ConstantExpr::getCast(opc, C, DstEltTy));
}
return ConstantVector::get(res);
}
@ -696,12 +698,13 @@ Constant *llvm::ConstantFoldSelectInstruction(Constant *Cond,
// If the condition is a vector constant, fold the result elementwise.
if (ConstantVector *CondV = dyn_cast<ConstantVector>(Cond)) {
SmallVector<Constant*, 16> Result;
Type *Ty = IntegerType::get(CondV->getContext(), 32);
for (unsigned i = 0, e = V1->getType()->getVectorNumElements(); i != e;++i){
ConstantInt *Cond = dyn_cast<ConstantInt>(CondV->getOperand(i));
if (Cond == 0) break;
Constant *Res = (Cond->getZExtValue() ? V1 : V2)->getAggregateElement(i);
if (Res == 0) break;
Constant *V = Cond->isNullValue() ? V2 : V1;
Constant *Res = ConstantExpr::getExtractElement(V, ConstantInt::get(Ty, i));
Result.push_back(Res);
}
@ -760,16 +763,16 @@ Constant *llvm::ConstantFoldInsertElementInstruction(Constant *Val,
const APInt &IdxVal = CIdx->getValue();
SmallVector<Constant*, 16> Result;
Type *Ty = IntegerType::get(Val->getContext(), 32);
for (unsigned i = 0, e = Val->getType()->getVectorNumElements(); i != e; ++i){
if (i == IdxVal) {
Result.push_back(Elt);
continue;
}
if (Constant *C = Val->getAggregateElement(i))
Result.push_back(C);
else
return 0;
Constant *C =
ConstantExpr::getExtractElement(Val, ConstantInt::get(Ty, i));
Result.push_back(C);
}
return ConstantVector::get(Result);
@ -801,11 +804,15 @@ Constant *llvm::ConstantFoldShuffleVectorInstruction(Constant *V1,
Constant *InElt;
if (unsigned(Elt) >= SrcNumElts*2)
InElt = UndefValue::get(EltTy);
else if (unsigned(Elt) >= SrcNumElts)
InElt = V2->getAggregateElement(Elt - SrcNumElts);
else
InElt = V1->getAggregateElement(Elt);
if (InElt == 0) return 0;
else if (unsigned(Elt) >= SrcNumElts) {
Type *Ty = IntegerType::get(V2->getContext(), 32);
InElt =
ConstantExpr::getExtractElement(V2,
ConstantInt::get(Ty, Elt - SrcNumElts));
} else {
Type *Ty = IntegerType::get(V1->getContext(), 32);
InElt = ConstantExpr::getExtractElement(V1, ConstantInt::get(Ty, Elt));
}
Result.push_back(InElt);
}
@ -1130,16 +1137,17 @@ Constant *llvm::ConstantFoldBinaryInstruction(unsigned Opcode,
} else if (VectorType *VTy = dyn_cast<VectorType>(C1->getType())) {
// Perform elementwise folding.
SmallVector<Constant*, 16> Result;
Type *Ty = IntegerType::get(VTy->getContext(), 32);
for (unsigned i = 0, e = VTy->getNumElements(); i != e; ++i) {
Constant *LHS = C1->getAggregateElement(i);
Constant *RHS = C2->getAggregateElement(i);
if (LHS == 0 || RHS == 0) break;
Constant *LHS =
ConstantExpr::getExtractElement(C1, ConstantInt::get(Ty, i));
Constant *RHS =
ConstantExpr::getExtractElement(C2, ConstantInt::get(Ty, i));
Result.push_back(ConstantExpr::get(Opcode, LHS, RHS));
}
if (Result.size() == VTy->getNumElements())
return ConstantVector::get(Result);
return ConstantVector::get(Result);
}
if (ConstantExpr *CE1 = dyn_cast<ConstantExpr>(C1)) {
@ -1697,17 +1705,18 @@ Constant *llvm::ConstantFoldCompareInstruction(unsigned short pred,
// If we can constant fold the comparison of each element, constant fold
// the whole vector comparison.
SmallVector<Constant*, 4> ResElts;
Type *Ty = IntegerType::get(C1->getContext(), 32);
// Compare the elements, producing an i1 result or constant expr.
for (unsigned i = 0, e = C1->getType()->getVectorNumElements(); i != e;++i){
Constant *C1E = C1->getAggregateElement(i);
Constant *C2E = C2->getAggregateElement(i);
if (C1E == 0 || C2E == 0) break;
Constant *C1E =
ConstantExpr::getExtractElement(C1, ConstantInt::get(Ty, i));
Constant *C2E =
ConstantExpr::getExtractElement(C2, ConstantInt::get(Ty, i));
ResElts.push_back(ConstantExpr::getCompare(pred, C1E, C2E));
}
if (ResElts.size() == C1->getType()->getVectorNumElements())
return ConstantVector::get(ResElts);
return ConstantVector::get(ResElts);
}
if (C1->getType()->isFloatingPointTy()) {

View File

@ -263,10 +263,10 @@ define i1* @hoo1() nounwind {
; OPT: ret i64 ptrtoint (double* getelementptr ({ i1, double }* null, i64 0, i32 1) to i64)
; OPT: }
; OPT: define i64 @fc() nounwind {
; OPT: ret i64 mul nuw (i64 ptrtoint (double* getelementptr (double* null, i32 1) to i64), i64 2)
; OPT: ret i64 mul (i64 ptrtoint (double* getelementptr (double* null, i32 1) to i64), i64 2)
; OPT: }
; OPT: define i64 @fd() nounwind {
; OPT: ret i64 mul nuw (i64 ptrtoint (double* getelementptr (double* null, i32 1) to i64), i64 11)
; OPT: ret i64 mul (i64 ptrtoint (double* getelementptr (double* null, i32 1) to i64), i64 11)
; OPT: }
; OPT: define i64 @fe() nounwind {
; OPT: ret i64 ptrtoint (double* getelementptr ({ double, float, double, double }* null, i64 0, i32 2) to i64)
@ -433,7 +433,7 @@ define i64* @fO() nounwind {
; PLAIN: ret i32* %t
; PLAIN: }
; OPT: define i32* @fZ() nounwind {
; OPT: ret i32* getelementptr inbounds (i32* getelementptr inbounds ([3 x { i32, i32 }]* @ext, i64 0, i64 1, i32 0), i64 1)
; OPT: ret i32* getelementptr (i32* getelementptr inbounds ([3 x { i32, i32 }]* @ext, i64 0, i64 1, i32 0), i64 1)
; OPT: }
; TO: define i32* @fZ() nounwind {
; TO: ret i32* getelementptr inbounds ([3 x { i32, i32 }]* @ext, i64 0, i64 1, i32 1)

View File

@ -0,0 +1,20 @@
; RUN: opt -sccp -S < %s | FileCheck %s
target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128-n8:16:32-S128"
; CHECK: store volatile <2 x i64> zeroinitializer, <2 x i64>* %p
; rdar://11324230
define void @foo(<2 x i64>* %p) nounwind {
entry:
br label %while.body.i
while.body.i: ; preds = %while.body.i, %entry
%vWorkExponent.i.033 = phi <4 x i32> [ %sub.i.i, %while.body.i ], [ <i32 939524096, i32 939524096, i32 939524096, i32 939524096>, %entry ]
%sub.i.i = add <4 x i32> %vWorkExponent.i.033, <i32 -8388608, i32 -8388608, i32 -8388608, i32 -8388608>
%0 = bitcast <4 x i32> %sub.i.i to <2 x i64>
%and.i119.i = and <2 x i64> %0, zeroinitializer
store volatile <2 x i64> %and.i119.i, <2 x i64>* %p
br label %while.body.i
}