State of bleeding edge Intel drivers and opengl?

[Eric Anholt]

On Thu, 2008-08-21 at 14:23 -0700, Philip Langdale wrote:
> This is a follow up to a question I asked a couple of months ago.
> 
> Right now, I've got master of drm/mesa/xf86-video-intel/xserver up
> and running with the kernel drm driver from the drm-merge branch
> of Eric's linux tree - but I believe it's the same as the drm driver
> on drm master.
> 
> Mesa is built without TTM support because that claims to require xf86mm.h
> which master of drm does not export anymore. Does this matter? It causes
> mesa to only claim OpenGL 1.4 support but the full set of ARB extensions
> for GLSL and FBOs are still published.

You should be missing ARB_occlusion_query if you're using my tree, which
means no OpenGL 1.5.  It's never been merged to the kernel, because the
API provided by the kernel modules was insufficient and needs to be
redesigned.

> Once up and running, the GLSL test programs seem to work but the FBO
> situation is much dicier. fbo_firecube locks the machine up and the simpler
> fbo tests show visual corruption and dmesg is filled with messages like:
> 
> [drm:i915_gem_execbuffer] *ERROR* Going to lose the write domain obj 11 size 1048576
> 
> I'm using EXA as X refuses to come up with UXA and I'm not using KMS.
> 
> Is this behaviour expected?

The lose write domain warning was something we didn't think would be
tripped up, but apparently is happening.  It could explain visual
corruption at least, and on small buffers could explain even more
serious issues.  Fixing it shouldn't be hard, just haven't got around to
it quite yet.

-- 
Eric Anholt
eric at anholt.net                         eric.anholt at intel.com


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 197 bytes
Desc: This is a digitally signed message part
URL: <http://lists.x.org/archives/xorg/attachments/20080822/0be007e0/attachment.pgp>