[PATCH 00/20] New glamor core rendering code (v2)
michel at daenzer.net
Mon Apr 7 22:44:47 PDT 2014
On Die, 2014-04-01 at 16:01 -0700, Keith Packard wrote:
> Michel Dänzer <michel at daenzer.net> writes:
> > Sounds great, but...
> > Patch 11 breaks Xorg for me with radeonsi. It looks like solid fills are
> > basically not rendered at all anymore, so e.g. only the text is visible
> > in an xterm window. Then later patches make it even worse, resulting in
> > just a black screen instead of the root weave and an xterm window. Also,
> > e.g. x11perf -shmput500 gives ridiculously high results, so apparently
> > nothing is actually rendered.
> Ok, I found a pervasive mistake in the new glamor code -- my _nf
> functions were not falling back to fb/mi when the driver would be unable
> to access the pixmaps -- they weren't calling
> glamor_ddx_fallback_check_pixmap ever.
> I've fixed that, which makes Eric's patched xf86-video-intel 'just work'
> with the new acceleration code.
> I'd love to know if this also fixes radeonsi, but I don't have any
> hardware to try it on.
The radeon driver doesn't use the glamor_*_nf() entrypoints, it leaves
all the rendering and fallback handling to glamor.
> > Xephyr works mostly fine at patch 20, but e.g. gtkperf leaves artifacts
> > (it's fast though :).
> Can you isolate which tests are causing problems? I'm not seeing this; I
> assume that's because of the difference in GL drivers.
E.g. the GtkComboBox test leaves artifacts outside of the gtkperf
window, or also just scrolling the gtkperf widget containing the test
I think at least part of the problem is using glCopyPixels for
overlapping copies. 3D cores just cannot do that; if any versions of the
OpenGL spec do not explicitly say this produces undefined results,
that's probably just an oversight in the spec, as Markus pointed out.
Earthling Michel Dänzer | http://www.amd.com
Libre software enthusiast | Mesa and X developer
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 173 bytes
Desc: This is a digitally signed message part
More information about the xorg-devel