Subject: Re: RFC: est.c driver synced with OpenBSD.
To: Quentin Garnier <firstname.lastname@example.org>
From: Simon Burge <simonb@NetBSD.org>
Date: 09/01/2006 13:04:13
Quentin Garnier wrote:
> On Thu, Aug 31, 2006 at 10:17:18PM +0000, Michael van Elst wrote:
> > email@example.com (Juan RP) writes:
> > >Anybody could please review it before committing? at least this version
> > >of est.c will detect the highest and lowest frequency in CPUs that we
> > >don't know the table and will work in more CPUs than before.
> > On a T2300 it works somewhat. I have to run sysctl several times
> > to get the speed changed. I guess the speed needs to be set on
> > both cores indvidually.
> This is actually the behaviour I expected! It actually depends on the
> state of the other core. Thanks for confirming I wasn't reading the
> datasheet wrong :)
The spec said that on a Core Duo that the greater of the two target
speeds would be detected. On my T2400, the second core always had the
lowest frequency selected, and any changes made with sysctl only seemed
to affect the first core. Note that I haven't verified which core the
sysctl was running on yet - I'll add some instrumentation to check that.
I'm not sure if the behaviour I've seen so far blind luck or not...