Re: [Xen-ia64-devel] [PATCH] Enable hash vtlb

[Tristan Gingold <Tristan.Gingold@xxxxxxxx>]

Le Vendredi 07 Avril 2006 20:02, Magenheimer, Dan (HP Labs Fort Collins) a 
écrit :
> > >individually.  For example, if your TLB mapping fix solves
> > >the "gcc segmentation fault" issue, even if there is a
> > >performance hit, the fix should be accepted.
> >
> > I got domU performance data now,
> > Kernel build time is about 1997s on native UP machine
> > Kernel build time is about 1990s on domU without this patch.
> > Kernel build time is about 1965s on domU with this patch.
> >
> > In short, though this patch loses 2% performance on dom0,
> > it gains >1% performance on domU.
> > As we know the performance of domU is more importance than
> > that of dom0.
> > Seems this patch can be checked in.:-)
>
> Could be, but your numbers raise questions about your
> measurement methodology.  Why
> is domU faster than native (with or without the patch)?
> And why is domU much faster than the previous numbers
> you posted for dom0 (with or without the patch)?
> I've never seen either of these to be true.
I have never made such tests, but here are a few suggestions:

* memory size (are they the same) ?

* domU also benefits from dom0 memory for caching files.

Tristan.


_______________________________________________
Xen-ia64-devel mailing list
Xen-ia64-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-ia64-devel