Go to the source code of this file.
|
| namespace | llvm |
| | This is an optimization pass for GlobalISel generic memory operations.
|
| namespace | llvm::slpvectorizer |
| | A private "module" namespace for types and utilities used by this pass.
|
|
| static bool | llvm::slpvectorizer::matchRdxBop (Instruction *I, Value *&V0, Value *&V1) |
| Instruction * | llvm::slpvectorizer::getNonPhiOperand (Instruction *I, PHINode *Phi) |
| bool | llvm::slpvectorizer::isReductionCandidate (Instruction *I) |
| Type * | llvm::slpvectorizer::getBoolReduxWideTy (RecurKind RdxKind, Type *RootTy, Type *LeafTy) |
| Value * | llvm::slpvectorizer::tryEmitBoolReduxBitcastCmp (IRBuilderBase &Builder, const TargetTransformInfo &TTI, RecurKind RdxKind, Value *Vec, const Value *Root, FastMathFlags FMF, TargetTransformInfo::TargetCostKind CostKind) |
| | Emits the booleanized logical and/or reduction of Vec with the i1 root Root as trunc+bitcast+cmp (all-ones comparison for And, zero for Or) if it is cheaper than the wide reduction plus the result trunc.
|