[Mesa-dev] Improving ralloc performance for the GLSL compiler

Ernst Sjöstrand ernstp at gmail.com
Thu Sep 1 18:52:00 UTC 2016


Haven't I read in many places that the Linux port of NS2 is extremely
crappy?
Perhaps you shouldn't pay too much attention to that... Then again, maybe
it's a good torture test. :-)

Regards
//Ernst

2016-09-01 18:11 GMT+02:00 Marek Olšák <maraeo at gmail.com>:

> On Aug 31, 2016 11:31 AM, "Eero Tamminen" <eero.t.tamminen at intel.com>
> wrote:
> >
> > Hi,
> >
> >
> > On 31.08.2016 08:03, Kenneth Graunke wrote:
> >>
> >> On Tuesday, August 30, 2016 3:52:12 PM PDT Ian Romanick wrote:
> >> [snip]
> >>>
> >>> I don't recall any native Linux games encountering this, but my memory
> >>> may have faded.  There were some Windows DirectX games that would
> >>> exhaust VMA when run inside VMware.  Once we got to the point where the
> >>> GLSL IR was freed after linking, I believe pretty much all of these
> >>> problems went away.
> >>>
> >>> If we might do something that would dramatically increase our peak
> >>> memory usage, we'll need to test it in that environment.
> >>
> >>
> >> Dota 2 (32-bit) used to hit this a long time ago, but hasn't in quite
> >> some time.
> >
> >
> > Dota 2 is nowadays 64-bit.
> >
> > I think Natural Selection 2 (which is still 32-bit) also used to crash
> to memory issues.
> >
> > I don't think I saw other games with this issue, but DOTA2 & NS2 were
> probably the games that compiled largest number of shaders (many thousands
> of them
>
> While I agree with your earlier post, the number of shaders that is
> compiled is mostly irrelevant assuming all drivers release GLSL IR at link
> time. What remains to test is memory usage on 32-bit.
>
> Marek
>
> _______________________________________________
> mesa-dev mailing list
> mesa-dev at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/mesa-dev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20160901/d373c1e3/attachment.html>


More information about the mesa-dev mailing list