InnerLoopVectorizer(OrigLoop, SE, LI, DT, DL, TLI, 1, UnrollFactor) { }
private:
- virtual void scalarizeInstruction(Instruction *Instr, bool IfPredicateStore = false);
- virtual void vectorizeMemoryInstruction(Instruction *Instr);
- virtual Value *getBroadcastInstrs(Value *V);
- virtual Value *getConsecutiveVector(Value* Val, int StartIdx, bool Negate);
- virtual Value *reverseVector(Value *Vec);
+ void scalarizeInstruction(Instruction *Instr,
+ bool IfPredicateStore = false) override;
+ void vectorizeMemoryInstruction(Instruction *Instr) override;
+ Value *getBroadcastInstrs(Value *V) override;
+ Value *getConsecutiveVector(Value* Val, int StartIdx, bool Negate) override;
+ Value *reverseVector(Value *Vec) override;
};
/// \brief Look for a meaningful debug location on the instruction or it's
BlockFrequency ColdEntryFreq;
- virtual bool runOnFunction(Function &F) {
+ bool runOnFunction(Function &F) override {
SE = &getAnalysis<ScalarEvolution>();
DataLayoutPass *DLP = getAnalysisIfAvailable<DataLayoutPass>();
DL = DLP ? &DLP->getDataLayout() : 0;
return true;
}
- virtual void getAnalysisUsage(AnalysisUsage &AU) const {
+ void getAnalysisUsage(AnalysisUsage &AU) const override {
AU.addRequiredID(LoopSimplifyID);
AU.addRequiredID(LCSSAID);
AU.addRequired<BlockFrequencyInfo>();
}
}
+/// \brief Adds a 'fast' flag to floating point operations.
+static Value *addFastMathFlag(Value *V) {
+ if (isa<FPMathOperator>(V)){
+ FastMathFlags Flags;
+ Flags.setUnsafeAlgebra();
+ cast<Instruction>(V)->setFastMathFlags(Flags);
+ }
+ return V;
+}
+
void InnerLoopVectorizer::vectorizeLoop() {
//===------------------------------------------------===//
//
setDebugLocFromInst(Builder, ReducedPartRdx);
for (unsigned part = 1; part < UF; ++part) {
if (Op != Instruction::ICmp && Op != Instruction::FCmp)
- ReducedPartRdx = Builder.CreateBinOp((Instruction::BinaryOps)Op,
- RdxParts[part], ReducedPartRdx,
- "bin.rdx");
+ // Floating point operations had to be 'fast' to enable the reduction.
+ ReducedPartRdx = addFastMathFlag(
+ Builder.CreateBinOp((Instruction::BinaryOps)Op, RdxParts[part],
+ ReducedPartRdx, "bin.rdx"));
else
ReducedPartRdx = createMinMaxOp(Builder, RdxDesc.MinMaxKind,
ReducedPartRdx, RdxParts[part]);
"rdx.shuf");
if (Op != Instruction::ICmp && Op != Instruction::FCmp)
- TmpVec = Builder.CreateBinOp((Instruction::BinaryOps)Op, TmpVec, Shuf,
- "bin.rdx");
+ // Floating point operations had to be 'fast' to enable the reduction.
+ TmpVec = addFastMathFlag(Builder.CreateBinOp(
+ (Instruction::BinaryOps)Op, TmpVec, Shuf, "bin.rdx"));
else
TmpVec = createMinMaxOp(Builder, RdxDesc.MinMaxKind, TmpVec, Shuf);
}
if (VecOp && isa<PossiblyExactOperator>(VecOp))
VecOp->setIsExact(BinOp->isExact());
+ // Copy the fast-math flags.
+ if (VecOp && isa<FPMathOperator>(V))
+ VecOp->setFastMathFlags(it->getFastMathFlags());
+
Entry[Part] = V;
}
break;