[parisc-linux] Problem on floating calculation

[Matthew Wilcox]

On Thu, Mar 08, 2001 at 03:16:48PM +0000, Bruno IGLESIAS wrote:
> 	When I execute Seti@home client (32 Bits executable) on a 64 Bit     
> architecture ( HP900A500-A5X i.e ESIEE's A500) ,the expression :      
>                                                                       
>    "exp(-(0.693 * SQUARE((float)(i) - (float)(ul_TOffset))) /         
> (float)(f_SigmaSq))"                                                  
>                                                                       
> doesn't work with some specific value. For example exp(-89.790901)    
> result in an illegal instruction message.                             

i had trouble with a divide performed by a prefetch benchmark program.
there's definitely something wrong with our FP stuff.  I believe this is
bug #72 in the BTS.

-- 
Revolutions do not require corporate support.