[llvm] [InstCombine] Ensure Safe Handling of Flags in foldFNegIntoConstant (PR #94148)

[Matt Arsenault via llvm-commits]

================
@@ -2653,8 +2653,15 @@ static Instruction *foldFNegIntoConstant(Instruction &I, const DataLayout &DL) {
   // Fold negation into constant operand.
   // -(X * C) --> X * (-C)
   if (match(FNegOp, m_FMul(m_Value(X), m_Constant(C))))
-    if (Constant *NegC = ConstantFoldUnaryOpOperand(Instruction::FNeg, C, DL))
+    if (Constant *NegC = ConstantFoldUnaryOpOperand(Instruction::FNeg, C, DL)) {
+      if (match(C, m_AnyZeroFP()) && I.getFastMathFlags().noInfs()) {
+        return BinaryOperator::CreateFMulFMF(
+            X, NegC,
+            I.getFastMathFlags() &
+                cast<FPMathOperator>(I.getOperand(0))->getFastMathFlags());
----------------
arsenm wrote:

I think ninf is the only problematic flag, since ninf on the fneg does not imply fneg on fmul, and fmul inf * 0 = nan. 

So you need something like

Flags = FnegFlags | FMulFlags
Flags.ninf &= FmulFlags.ninf 

https://github.com/llvm/llvm-project/pull/94148