[llvm-commits] [llvm] r51268 - in /llvm/trunk: lib/Transforms/Scalar/InstructionCombining.cpp test/Transforms/InstCombine/sitofp.ll

[Neil Booth]

> +  // See if the FP constant is top large for the integer.  For example,

Typo.

> +  // Okay, now we know that the FP constant fits in the range [SMIN, SMAX] but
> +  // it may still be fractional.  See if it is fractional by casting the FP
> +  // value to the integer value and back, checking for equality.  Don't do this
> +  // for zero, because -0.0 is not fractional.

Perhaps this is better done as an APFloat method if I understand what
you're trying to do?  It should be able to determine this reasonably
easily without introducing allocations / conversions, though I've not
thought about it a lot.

>    // Handle fcmp with constant RHS
>    if (Constant *RHSC = dyn_cast<Constant>(Op1)) {
> +    // If the constant is a nan, see if we can fold the comparison based on it.
> +    if (ConstantFP *CFP = dyn_cast<ConstantFP>(RHSC)) {
> +      if (CFP->getValueAPF().isNaN()) {
> +        if (FCmpInst::isOrdered(I.getPredicate()))   // True if ordered and...
> +          return ReplaceInstUsesWith(I, ConstantInt::get(Type::Int1Ty, 0));
> +        if (FCmpInst::isUnordered(I.getPredicate())) // True if unordered or...
> +          return ReplaceInstUsesWith(I, ConstantInt::get(Type::Int1Ty, 1));
> +        if (FCmpInst::isUnordered(I.getPredicate())) // Undef on unordered.
> +          return ReplaceInstUsesWith(I, UndefValue::get(Type::Int1Ty));
> +    

Duplicate test?

Neil.