//
//===----------------------------------------------------------------------===//
-#include "InstCombine.h"
+#include "InstCombineInternal.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/Analysis/MemoryBuiltins.h"
#include "llvm/IR/CallSite.h"
#include "llvm/IR/Statepoint.h"
#include "llvm/Transforms/Utils/BuildLibCalls.h"
#include "llvm/Transforms/Utils/Local.h"
+#include "llvm/Transforms/Utils/SimplifyLibCalls.h"
using namespace llvm;
using namespace PatternMatch;
}
Instruction *InstCombiner::SimplifyMemTransfer(MemIntrinsic *MI) {
- unsigned DstAlign = getKnownAlignment(MI->getArgOperand(0), DL, AT, MI, DT);
- unsigned SrcAlign = getKnownAlignment(MI->getArgOperand(1), DL, AT, MI, DT);
+ unsigned DstAlign = getKnownAlignment(MI->getArgOperand(0), DL, AC, MI, DT);
+ unsigned SrcAlign = getKnownAlignment(MI->getArgOperand(1), DL, AC, MI, DT);
unsigned MinAlign = std::min(DstAlign, SrcAlign);
unsigned CopyAlign = MI->getAlignment();
}
Instruction *InstCombiner::SimplifyMemSet(MemSetInst *MI) {
- unsigned Alignment = getKnownAlignment(MI->getDest(), DL, AT, MI, DT);
+ unsigned Alignment = getKnownAlignment(MI->getDest(), DL, AC, MI, DT);
if (MI->getAlignment() < Alignment) {
MI->setAlignment(ConstantInt::get(MI->getAlignmentType(),
Alignment, false));
break;
case Intrinsic::uadd_with_overflow: {
Value *LHS = II->getArgOperand(0), *RHS = II->getArgOperand(1);
- IntegerType *IT = cast<IntegerType>(II->getArgOperand(0)->getType());
- uint32_t BitWidth = IT->getBitWidth();
- APInt LHSKnownZero(BitWidth, 0);
- APInt LHSKnownOne(BitWidth, 0);
- computeKnownBits(LHS, LHSKnownZero, LHSKnownOne, 0, II);
- bool LHSKnownNegative = LHSKnownOne[BitWidth - 1];
- bool LHSKnownPositive = LHSKnownZero[BitWidth - 1];
-
- if (LHSKnownNegative || LHSKnownPositive) {
- APInt RHSKnownZero(BitWidth, 0);
- APInt RHSKnownOne(BitWidth, 0);
- computeKnownBits(RHS, RHSKnownZero, RHSKnownOne, 0, II);
- bool RHSKnownNegative = RHSKnownOne[BitWidth - 1];
- bool RHSKnownPositive = RHSKnownZero[BitWidth - 1];
- if (LHSKnownNegative && RHSKnownNegative) {
- // The sign bit is set in both cases: this MUST overflow.
- // Create a simple add instruction, and insert it into the struct.
- return CreateOverflowTuple(II, Builder->CreateAdd(LHS, RHS), true,
- /*ReUseName*/true);
- }
-
- if (LHSKnownPositive && RHSKnownPositive) {
- // The sign bit is clear in both cases: this CANNOT overflow.
- // Create a simple add instruction, and insert it into the struct.
- return CreateOverflowTuple(II, Builder->CreateNUWAdd(LHS, RHS), false);
- }
- }
+ OverflowResult OR = computeOverflowForUnsignedAdd(LHS, RHS, II);
+ if (OR == OverflowResult::NeverOverflows)
+ return CreateOverflowTuple(II, Builder->CreateNUWAdd(LHS, RHS), false);
+ if (OR == OverflowResult::AlwaysOverflows)
+ return CreateOverflowTuple(II, Builder->CreateAdd(LHS, RHS), true);
}
// FALL THROUGH uadd into sadd
case Intrinsic::sadd_with_overflow:
}
case Intrinsic::umul_with_overflow: {
Value *LHS = II->getArgOperand(0), *RHS = II->getArgOperand(1);
- unsigned BitWidth = cast<IntegerType>(LHS->getType())->getBitWidth();
-
- APInt LHSKnownZero(BitWidth, 0);
- APInt LHSKnownOne(BitWidth, 0);
- computeKnownBits(LHS, LHSKnownZero, LHSKnownOne, 0, II);
- APInt RHSKnownZero(BitWidth, 0);
- APInt RHSKnownOne(BitWidth, 0);
- computeKnownBits(RHS, RHSKnownZero, RHSKnownOne, 0, II);
-
- // Get the largest possible values for each operand.
- APInt LHSMax = ~LHSKnownZero;
- APInt RHSMax = ~RHSKnownZero;
-
- // If multiplying the maximum values does not overflow then we can turn
- // this into a plain NUW mul.
- bool Overflow;
- LHSMax.umul_ov(RHSMax, Overflow);
- if (!Overflow) {
+ OverflowResult OR = computeOverflowForUnsignedMul(LHS, RHS, II);
+ if (OR == OverflowResult::NeverOverflows)
return CreateOverflowTuple(II, Builder->CreateNUWMul(LHS, RHS), false);
- }
+ if (OR == OverflowResult::AlwaysOverflows)
+ return CreateOverflowTuple(II, Builder->CreateMul(LHS, RHS), true);
} // FALL THROUGH
case Intrinsic::smul_with_overflow:
// Canonicalize constants into the RHS.
case Intrinsic::ppc_altivec_lvx:
case Intrinsic::ppc_altivec_lvxl:
// Turn PPC lvx -> load if the pointer is known aligned.
- if (getOrEnforceKnownAlignment(II->getArgOperand(0), 16,
- DL, AT, II, DT) >= 16) {
+ if (getOrEnforceKnownAlignment(II->getArgOperand(0), 16, DL, AC, II, DT) >=
+ 16) {
Value *Ptr = Builder->CreateBitCast(II->getArgOperand(0),
PointerType::getUnqual(II->getType()));
return new LoadInst(Ptr);
case Intrinsic::ppc_altivec_stvx:
case Intrinsic::ppc_altivec_stvxl:
// Turn stvx -> store if the pointer is known aligned.
- if (getOrEnforceKnownAlignment(II->getArgOperand(1), 16,
- DL, AT, II, DT) >= 16) {
+ if (getOrEnforceKnownAlignment(II->getArgOperand(1), 16, DL, AC, II, DT) >=
+ 16) {
Type *OpPtrTy =
PointerType::getUnqual(II->getArgOperand(0)->getType());
Value *Ptr = Builder->CreateBitCast(II->getArgOperand(1), OpPtrTy);
case Intrinsic::x86_sse2_storeu_pd:
case Intrinsic::x86_sse2_storeu_dq:
// Turn X86 storeu -> store if the pointer is known aligned.
- if (getOrEnforceKnownAlignment(II->getArgOperand(0), 16,
- DL, AT, II, DT) >= 16) {
+ if (getOrEnforceKnownAlignment(II->getArgOperand(0), 16, DL, AC, II, DT) >=
+ 16) {
Type *OpPtrTy =
PointerType::getUnqual(II->getArgOperand(1)->getType());
Value *Ptr = Builder->CreateBitCast(II->getArgOperand(0), OpPtrTy);
case Intrinsic::arm_neon_vst2lane:
case Intrinsic::arm_neon_vst3lane:
case Intrinsic::arm_neon_vst4lane: {
- unsigned MemAlign = getKnownAlignment(II->getArgOperand(0), DL, AT, II, DT);
+ unsigned MemAlign = getKnownAlignment(II->getArgOperand(0), DL, AC, II, DT);
unsigned AlignArg = II->getNumArgOperands() - 1;
ConstantInt *IntrAlign = dyn_cast<ConstantInt>(II->getArgOperand(AlignArg));
if (IntrAlign && IntrAlign->getZExtValue() < MemAlign) {
Instruction *InstCombiner::tryOptimizeCall(CallInst *CI, const DataLayout *DL) {
if (!CI->getCalledFunction()) return nullptr;
- if (Value *With = Simplifier->optimizeCall(CI)) {
+ auto InstCombineRAUW = [this](Instruction *From, Value *With) {
+ ReplaceInstUsesWith(*From, With);
+ };
+ LibCallSimplifier Simplifier(DL, TLI, InstCombineRAUW);
+ if (Value *With = Simplifier.optimizeCall(CI)) {
++NumSimplified;
return CI->use_empty() ? CI : ReplaceInstUsesWith(*CI, With);
}
dyn_cast<Function>(CS.getCalledValue()->stripPointerCasts());
if (!Callee)
return false;
+ // The prototype of thunks are a lie, don't try to directly call such
+ // functions.
+ if (Callee->hasFnAttribute("thunk"))
+ return false;
Instruction *Caller = CS.getInstruction();
const AttributeSet &CallerPAL = CS.getAttributes();
if (NewRetTy->isStructTy())
return false; // TODO: Handle multiple return values.
- if (!CastInst::isBitCastable(NewRetTy, OldRetTy)) {
+ if (!CastInst::isBitOrNoopPointerCastable(NewRetTy, OldRetTy, DL)) {
if (Callee->isDeclaration())
return false; // Cannot transform this return value.
unsigned NumActualArgs = CS.arg_size();
unsigned NumCommonArgs = std::min(FT->getNumParams(), NumActualArgs);
+ // Prevent us turning:
+ // declare void @takes_i32_inalloca(i32* inalloca)
+ // call void bitcast (void (i32*)* @takes_i32_inalloca to void (i32)*)(i32 0)
+ //
+ // into:
+ // call void @takes_i32_inalloca(i32* null)
+ if (Callee->getAttributes().hasAttrSomewhere(Attribute::InAlloca))
+ return false;
+
CallSite::arg_iterator AI = CS.arg_begin();
for (unsigned i = 0, e = NumCommonArgs; i != e; ++i, ++AI) {
Type *ParamTy = FT->getParamType(i);
Type *ActTy = (*AI)->getType();
- if (!CastInst::isBitCastable(ActTy, ParamTy))
+ if (!CastInst::isBitOrNoopPointerCastable(ActTy, ParamTy, DL))
return false; // Cannot transform this parameter value.
if (AttrBuilder(CallerPAL.getParamAttributes(i + 1), i + 1).
if ((*AI)->getType() == ParamTy) {
Args.push_back(*AI);
} else {
- Args.push_back(Builder->CreateBitCast(*AI, ParamTy));
+ Args.push_back(Builder->CreateBitOrPointerCast(*AI, ParamTy));
}
// Add any parameter attributes.
Value *NV = NC;
if (OldRetTy != NV->getType() && !Caller->use_empty()) {
if (!NV->getType()->isVoidTy()) {
- NV = NC = CastInst::Create(CastInst::BitCast, NC, OldRetTy);
+ NV = NC = CastInst::CreateBitOrPointerCast(NC, OldRetTy);
NC->setDebugLoc(Caller->getDebugLoc());
// If this is an invoke instruction, we should insert it after the first
projects / oota-llvm.git / blobdiff