[Mesa-dev] Mesa (master): r600g: implement instanced drawing support

Christian König deathsimple at vodafone.de
Wed Mar 2 15:39:26 PST 2011


Am Mittwoch, den 02.03.2011, 08:18 +0100 schrieb Michel Dänzer:
> On Mit, 2011-03-02 at 00:06 +0100, Christian König wrote: 
> > Am Dienstag, den 01.03.2011, 19:36 +0100 schrieb Michel Dänzer:
> > > On Son, 2011-02-27 at 17:20 -0800, Christian KXXnig wrote: 
> > > > Module: Mesa
> > > > Branch: master
> > > > Commit: 96bbc627f369c0100b950f81531b1fe9ef586c34
> > > > URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=96bbc627f369c0100b950f81531b1fe9ef586c34
> > > > 
> > > > Author: Christian König <deathsimple at vodafone.de>
> > > > Date:   Mon Feb 28 02:00:01 2011 +0100
> > > > 
> > > > r600g: implement instanced drawing support
> > > 
> > > This commit breaks running Windows 7 guests with Aero in VMware
> > > Workstation. It doesn't make any difference if
> > > PIPE_CAP_INSTANCED_DRAWING is advertised or not, so apparently the
> > > problem isn't with actual instanced drawing.
> > > 
> > > 
> > > The piglit test draw-instanced-divisor is failing as well, though
> > > draw-instanced seems to work fine.
> > Let me guess: You working on an evergreen hardware, right?
> 
> No, RV635 (should have mentioned that, sorry).
> 
> 
> > The problem is that I don't have an evergreen hardware at hand to
> > reproduce the problem, so please set the R600_DUMP_SHADERS=1 environment
> > variable and let me take a look at the generated shader output.
> 
> I'll try to get that later.
I've just checked in a possible fix to mainline, please checkout and try
again.

I fixed the mega_fetch_count, but still don't have an idea why this
should only affect some chipsets and not my RV710 for example.

So it could also be something else, which I haven't figured out jet.

Christian.




More information about the mesa-dev mailing list