Re: ABS() in calc record

["Johnson, Andrew N." <anj@aps.anl.gov>]

Hi Mike,

On May 21, 2016, at 12:43 PM, Laznovsky Michael (PSI) <michael.laznovsky@psi.ch> wrote:
> 
>         case ABS_VAL:
>             if (*ptop < 0.0) *ptop = - *ptop;
>             break;
>  
> In my case, the CALC is "ABS(A*B)", and e.g. A=0 and B<0 results in ABS(-0), which does 
> not satisfy the if() in the code because -0 is NOT less than zero, so the sign is not flipped 
> back to positive.
>  
> Using fabs() instead fixes it (and probably also saves a branch).

Interesting effect; I just committed your fabs() change to the 3.14 branch.

Thanks!

- Andrew