my crude benchmark results

Wed Jul 14 18:59:00 GMT 2004

Andrew Cowie wrote:
>From the sounds of it, --fast-math as an option to gcj is something else
>entirely. If so, 
>>a) that's very cool
>b) that's very confusing
>>
Yeah, I suspect --fast-math currently has the dual effect of enabling 
Math.* inlines in the front end and enabling potentially unsafe math 
optimizations in the back end. If so we'll want to separate this 
behaviour, and perhaps enable the inlines by default.
Regards
Bryce