On Sun, Dec 15, 2002 at 12:44:43PM -0600, Dan Debertin wrote:
> I think Manuel B. was trying to attribute it to that problem last month
> when he asked whether this machine was hw or sw flush. That problem is
> apparently with IP[CX] and 1/1+ (and 2? not sure) that use sw flush,
> whereas this model has hw flush. This instance might invalidate that
> assumption, though, for all I know.

Yes, until now this problem had only been reproduced on systems with
sw cache flush. Your experience shows that maybe it's not cache-related,
but maybe something else. Maybe timing related ...

-- 
Manuel Bouyer <bouyer@antioche.eu.org>
     NetBSD: 23 ans d'experience feront toujours la difference
--