[llvm-commits] CVS: llvm/lib/Transforms/Scalar/InstructionCombining.cpp

[Chris Lattner]

> For PR1248: http://llvm.org/PR1248 :
> * Fix some indentation and comments in InsertRangeTest
> * Add an "IsSigned" parameter to AddWithOverflow and make it handle  
> signed
>   additions. Also, APIntify this function so it works with any  
> bitwidth.
> * For the icmp pred ([us]div %X, C1), C2 transforms, exit early if the
>   div instruction's RHS is zero.
> * Finally, for icmp pred (sdiv %X, C1), -C2, fix an off-by-one  
> error. The
>   HiBound needs to be incremented in order to get the range test  
> correct.

Thanks Reid, testcase please!

>  /// AddWithOverflow - Compute Result = In1+In2, returning true if  
> the result
>  /// overflowed for this type.
>  static bool AddWithOverflow(ConstantInt *&Result, ConstantInt *In1,
> -                            ConstantInt *In2) {
> +                            ConstantInt *In2, bool IsSigned =  
> false) {
>    Result = cast<ConstantInt>(ConstantExpr::getAdd(In1, In2));
>
> -  return cast<ConstantInt>(Result)->getZExtValue() <
> -         cast<ConstantInt>(In1)->getZExtValue();
> +  if (IsSigned)
> +    if (In2->getValue().isNegative())
> +      return Result->getValue().sgt(In1->getValue());
> +    else
> +      return Result->getValue().slt(In1->getValue());

I don't understand this logic.  Shouldn't it always be slt?

-Chris