On Sat, Jul 24, 2004 at 11:37:05AM +1000, Paul Ripke wrote:
> My only bit of benchmarking thus far is timing a kernel build. HT
> reliably increased the time from 116s to 119s. Warm cache, multiple
> runs, with -pipe. Guessing this is due to increase in various kernel
> overheads (locking, IPIs, etc).

This is what I noticed with linux, but not with NetBSD. I have to re-run one
test though, to make sure. I'll post the results here once I've all of them.

> 
> Knowing what I know about HT, I can see it isn't always going to be a
> benefit... unless maybe you're talking IBM Power5 chips? Those I'm
> curious to see.

No, I'm talking about the P4-HT. The Xeon-HP may be better in this area,
but I don't have any to test.

-- 
Manuel Bouyer <bouyer@antioche.eu.org>
     NetBSD: 26 ans d'experience feront toujours la difference
--