> > You most certainly should be doign page zeroing in the idle loop, too.
> > 
> > And I agree with you on this one. Has anyone picked up the Linux/PPC hack for
> > doing pagezero's bypassing the cache?
> 
> Hmm, what's the real problem here - why does the cache need to be disabled?

To not pollute the cache with zeroes.

> You may know that in i386 port, there was a problem that Cyrix CPU's kept
> crashing with the idle pagezero code.  The problem is said to be fixed (I
> had to switch CPU's long before the fix).  Are these two problems (Cyrix
> and Microsparc 1) related?  (No, I haven't got a Microsparc, I'm just
> curious..)
> 
>   -jm
>