Fast-math optimization: fold multiply by zero

Added in first optimization using fast-math flags to serve as an example for following optimizations. SimplifyInstruction will now try to optimize an fmul observing its FastMathFlags to see if it can fold multiply by zero when 'nnan' and 'nsz' flags are set.

llvm-svn: 168648

diff --git a/llvm/include/llvm/Analysis/InstructionSimplify.h b/llvm/include/llvm/Analysis/InstructionSimplify.h
index 6db400c..ce78181 100644 (file)
--- a/llvm/include/llvm/Analysis/InstructionSimplify.h
+++ b/llvm/include/llvm/Analysis/InstructionSimplify.h
@@ -25,6 +25,7 @@ namespace llvm {
   class DominatorTree;
   class Instruction;
   class DataLayout;
+  struct FastMathFlags;
   class TargetLibraryInfo;
   class Type;
   class Value;
@@ -43,6 +44,14 @@ namespace llvm {
                          const TargetLibraryInfo *TLI = 0,
                          const DominatorTree *DT = 0);
 
+  /// Given operands for an FMul, see if we can fold the result.  If not, this
+  /// returns null.
+  Value *SimplifyFMulInst(Value *LHS, Value *RHS,
+                          FastMathFlags FMF,
+                          const DataLayout *TD = 0,
+                          const TargetLibraryInfo *TLI = 0,
+                          const DominatorTree *DT = 0);
+
   /// SimplifyMulInst - Given operands for a Mul, see if we can
   /// fold the result.  If not, this returns null.
   Value *SimplifyMulInst(Value *LHS, Value *RHS, const DataLayout *TD = 0,

diff --git a/llvm/lib/Analysis/InstructionSimplify.cpp b/llvm/lib/Analysis/InstructionSimplify.cpp
index a76e5ad..e3fe924 100644 (file)
--- a/llvm/lib/Analysis/InstructionSimplify.cpp
+++ b/llvm/lib/Analysis/InstructionSimplify.cpp
@@ -886,6 +886,33 @@ Value *llvm::SimplifySubInst(Value *Op0, Value *Op1, bool isNSW, bool isNUW,
                            RecursionLimit);
 }
 
+/// Given the operands for an FMul, see if we can fold the result
+static Value *SimplifyFMulInst(Value *Op0, Value *Op1,
+                               FastMathFlags FMF,
+                               const Query &Q,
+                               unsigned MaxRecurse) {
+ if (Constant *CLHS = dyn_cast<Constant>(Op0)) {
+    if (Constant *CRHS = dyn_cast<Constant>(Op1)) {
+      Constant *Ops[] = { CLHS, CRHS };
+      return ConstantFoldInstOperands(Instruction::FMul, CLHS->getType(),
+                                      Ops, Q.TD, Q.TLI);
+    }
+ }
+
+ // Check for some fast-math optimizations
+ if (FMF.NoNaNs) {
+   if (FMF.NoSignedZeros) {
+     // fmul N S 0, x ==> 0
+     if (match(Op0, m_Zero()))
+       return Op0;
+     if (match(Op1, m_Zero()))
+       return Op1;
+   }
+ }
+
+ return 0;
+}
+
 /// SimplifyMulInst - Given operands for a Mul, see if we can
 /// fold the result.  If not, this returns null.
 static Value *SimplifyMulInst(Value *Op0, Value *Op1, const Query &Q,
@@ -951,6 +978,14 @@ static Value *SimplifyMulInst(Value *Op0, Value *Op1, const Query &Q,
   return 0;
 }
 
+Value *llvm::SimplifyFMulInst(Value *Op0, Value *Op1,
+                              FastMathFlags FMF,
+                              const DataLayout *TD,
+                              const TargetLibraryInfo *TLI,
+                              const DominatorTree *DT) {
+  return ::SimplifyFMulInst(Op0, Op1, FMF, Query (TD, TLI, DT), RecursionLimit);
+}
+
 Value *llvm::SimplifyMulInst(Value *Op0, Value *Op1, const DataLayout *TD,
                              const TargetLibraryInfo *TLI,
                              const DominatorTree *DT) {
@@ -2799,6 +2834,10 @@ Value *llvm::SimplifyInstruction(Instruction *I, const DataLayout *TD,
                              cast<BinaryOperator>(I)->hasNoUnsignedWrap(),
                              TD, TLI, DT);
     break;
+  case Instruction::FMul:
+    Result = SimplifyFMulInst(I->getOperand(0), I->getOperand(1),
+                              I->getFastMathFlags(), TD, TLI, DT);
+    break;
   case Instruction::Mul:
     Result = SimplifyMulInst(I->getOperand(0), I->getOperand(1), TD, TLI, DT);
     break;