diff options
Diffstat (limited to 'llvm/lib')
| -rw-r--r-- | llvm/lib/Analysis/IPA/InlineCost.cpp | 8 | ||||
| -rw-r--r-- | llvm/lib/Analysis/InstructionSimplify.cpp | 30 | ||||
| -rw-r--r-- | llvm/lib/Transforms/Scalar/LoopUnrollPass.cpp | 7 |
3 files changed, 43 insertions, 2 deletions
diff --git a/llvm/lib/Analysis/IPA/InlineCost.cpp b/llvm/lib/Analysis/IPA/InlineCost.cpp index c180f36b923..64f78f408fd 100644 --- a/llvm/lib/Analysis/IPA/InlineCost.cpp +++ b/llvm/lib/Analysis/IPA/InlineCost.cpp @@ -601,7 +601,13 @@ bool CallAnalyzer::visitBinaryOperator(BinaryOperator &I) { if (!isa<Constant>(RHS)) if (Constant *SimpleRHS = SimplifiedValues.lookup(RHS)) RHS = SimpleRHS; - Value *SimpleV = SimplifyBinOp(I.getOpcode(), LHS, RHS, DL); + Value *SimpleV = nullptr; + if (auto FI = dyn_cast<FPMathOperator>(&I)) + SimpleV = + SimplifyFPBinOp(I.getOpcode(), LHS, RHS, FI->getFastMathFlags(), DL); + else + SimpleV = SimplifyBinOp(I.getOpcode(), LHS, RHS, DL); + if (Constant *C = dyn_cast_or_null<Constant>(SimpleV)) { SimplifiedValues[&I] = C; return true; diff --git a/llvm/lib/Analysis/InstructionSimplify.cpp b/llvm/lib/Analysis/InstructionSimplify.cpp index 588d625910f..d02bb2b8f52 100644 --- a/llvm/lib/Analysis/InstructionSimplify.cpp +++ b/llvm/lib/Analysis/InstructionSimplify.cpp @@ -61,6 +61,8 @@ struct Query { static Value *SimplifyAndInst(Value *, Value *, const Query &, unsigned); static Value *SimplifyBinOp(unsigned, Value *, Value *, const Query &, unsigned); +static Value *SimplifyFPBinOp(unsigned, Value *, Value *, const FastMathFlags &, + const Query &, unsigned); static Value *SimplifyCmpInst(unsigned, Value *, Value *, const Query &, unsigned); static Value *SimplifyOrInst(Value *, Value *, const Query &, unsigned); @@ -3465,6 +3467,25 @@ static Value *SimplifyBinOp(unsigned Opcode, Value *LHS, Value *RHS, } } +/// SimplifyFPBinOp - Given operands for a BinaryOperator, see if we can +/// fold the result. If not, this returns null. +/// In contrast to SimplifyBinOp, try to use FastMathFlag when folding the +/// result. In case we don't need FastMathFlags, simply fall to SimplifyBinOp. +static Value *SimplifyFPBinOp(unsigned Opcode, Value *LHS, Value *RHS, + const FastMathFlags &FMF, const Query &Q, + unsigned MaxRecurse) { + switch (Opcode) { + case Instruction::FAdd: + return SimplifyFAddInst(LHS, RHS, FMF, Q, MaxRecurse); + case Instruction::FSub: + return SimplifyFSubInst(LHS, RHS, FMF, Q, MaxRecurse); + case Instruction::FMul: + return SimplifyFMulInst(LHS, RHS, FMF, Q, MaxRecurse); + default: + return SimplifyBinOp(Opcode, LHS, RHS, Q, MaxRecurse); + } +} + Value *llvm::SimplifyBinOp(unsigned Opcode, Value *LHS, Value *RHS, const DataLayout *DL, const TargetLibraryInfo *TLI, const DominatorTree *DT, AssumptionCache *AC, @@ -3473,6 +3494,15 @@ Value *llvm::SimplifyBinOp(unsigned Opcode, Value *LHS, Value *RHS, RecursionLimit); } +Value *llvm::SimplifyFPBinOp(unsigned Opcode, Value *LHS, Value *RHS, + const FastMathFlags &FMF, const DataLayout *DL, + const TargetLibraryInfo *TLI, + const DominatorTree *DT, AssumptionCache *AC, + const Instruction *CxtI) { + return ::SimplifyFPBinOp(Opcode, LHS, RHS, FMF, Query(DL, TLI, DT, AC, CxtI), + RecursionLimit); +} + /// SimplifyCmpInst - Given operands for a CmpInst, see if we can /// fold the result. static Value *SimplifyCmpInst(unsigned Predicate, Value *LHS, Value *RHS, diff --git a/llvm/lib/Transforms/Scalar/LoopUnrollPass.cpp b/llvm/lib/Transforms/Scalar/LoopUnrollPass.cpp index bd9da14f669..87237c62299 100644 --- a/llvm/lib/Transforms/Scalar/LoopUnrollPass.cpp +++ b/llvm/lib/Transforms/Scalar/LoopUnrollPass.cpp @@ -311,7 +311,12 @@ class UnrollAnalyzer : public InstVisitor<UnrollAnalyzer, bool> { if (!isa<Constant>(RHS)) if (Constant *SimpleRHS = SimplifiedValues.lookup(RHS)) RHS = SimpleRHS; - Value *SimpleV = SimplifyBinOp(I.getOpcode(), LHS, RHS); + Value *SimpleV = nullptr; + if (auto FI = dyn_cast<FPMathOperator>(&I)) + SimpleV = + SimplifyFPBinOp(I.getOpcode(), LHS, RHS, FI->getFastMathFlags()); + else + SimpleV = SimplifyBinOp(I.getOpcode(), LHS, RHS); if (SimpleV && CountedInsns.insert(&I).second) NumberOfOptimizedInstructions += TTI.getUserCost(&I); |
