On Wed, Mar 02, 2005 at 11:21:06PM +0100, Andries Brouwer wrote:
> On Wed, Mar 02, 2005 at 01:45:43PM +0000, Alan Cox wrote:
> > On Mer, 2005-03-02 at 08:02, Dave Jones wrote:
> > > If there are any of them still being used out there, I'd be even
> > > more surprised if they're running 2.6. Then again, there are
> > > probably loonies out there running it on 386/486's. 8-)
> >
> > I have one here running 2.4 still. I can test a 2.6 fix for the mtrr
> > init happily enough.
>
> Good. If I understand things correctly - you or davej or someone will
> correct me otherwise - failing to initialise mtrr does not break anything,
> it would just mean slower access to certain kinds of memory for certain
> kinds of access patterns. (Would you test using an X benchmark?)
The winchips had a funky feature where you could mark system ram
writes as out-of-order. This led to something like a 25% speedup iirc
on benchmarks that did lots of memory copying. lmbench showed
significant wins iirc, but any results I had saved are long since
wiped out in hard disk failures/cruft removal over the years.
> Below roughly speaking the same patch as before, but with calls
> to the cyrix and centaur mtrr init routines inserted.
Looks ok on a quick eyeball.
Dave
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
