Re: Lazy FPU benchmark?

Adam J. Richter (linker@nightshade.ml.org)
Fri, 17 Apr 1998 05:48:06 -0400 (EDT)


On Fri, 17 Apr 1998, Adam J. Richter wrote:

> >Might I suggest just forking off Multiple copies of the Unix bench 'float'
> >program?
>
> The benefits of the lazy FPU optimization should only occur
> when one or more integer only processes is sharing the CPU with one
> floating point process, so having multiple floating point
> processes will not demonstrate the optmization.
>

The current state always does a save and restore.. Your's only restores if
it's really needed right?

So if you run two fpu eaters concurently, and they dont hop between CPUs
then it should not need to restore.. Thus being faster..

A further test would be: Two fpu eaters and two int eaters..

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu