|
Bugzilla – Full Text Bug Listing |
| Summary: | enable COMPOSITE extension | ||
|---|---|---|---|
| Product: | [openSUSE] SUSE Linux 10.1 | Reporter: | Adrian Schröter <adrian.schroeter> |
| Component: | X.Org | Assignee: | Marcus Schaefer <ms> |
| Status: | RESOLVED FIXED | QA Contact: | Stefan Dirsch <sndirsch> |
| Severity: | Enhancement | ||
| Priority: | P5 - None | CC: | aritger, coolo, dmueller, eich, sndirsch |
| Version: | unspecified | ||
| Target Milestone: | --- | ||
| Hardware: | Other | ||
| OS: | All | ||
| Whiteboard: | |||
| Found By: | Other | Services Priority: | |
| Business Priority: | Blocker: | --- | |
| Marketing QA Status: | --- | IT Deployment: | --- |
|
Description
Adrian Schröter
2005-09-23 14:29:52 UTC
activate Composite extensions now by default We see problems when enabling Composite extensions, when the nvidia driver runs in a dualhead setup with different resolutions. With this configuration the Xserver crashes immediately during startup. This might happen to other drivers as well, but has not been investigated. This problem does not occur when using the same resolution. Andy might want to comment on enabling Composite extension by default for the nvidia driver. Last we investigated this, there were still some interaction problems with the NVIDIA driver and the Composite extension. It looked like a gc validation problem in the cw, but we didn't get to root cause it at the time. I think if you find any bugs with the NVIDIA driver + Composite (like the one mentioned above), please file bugs in the NVIDIA bug database and we'll investigate. Also, note that NVIDIA does not yet support OpenGL when the Composite extension is enabled (we're working on it, though); current NVIDIA drivers disable GLX if the Composite extension is enabled. Matthias, could you please make a bugreport in NVIDIA's bugreport system? For now I think it doesn't make much sense to enable Composite extension for NVIDIA users when it disables OpenGL support. I don't think we want to enable it - and probably break it - with option "AllowGLXWithComposite". And since Marcus don't want to set/unset this option depending on the driver we shouldn't set it at all for now. Marcus, could you revert the change for 10.1 again? Thanks. Thanks, Stefan. We'll work hard to make GLX work with Composite for a future driver release (and to fix any NVIDIA+Composite bugs) so that Composite can be enabled by default in a future release. not so quick: the point about enabling it during the preview/alpha phase was finding these kind of problems. We were aware of the shortcomings. And if Andy says, NVIDIA works on the problems, there is hope, no? I mean, 10.1 is still some months away - so don't rush into anything! And we should seriously consider to enable COMPOSITE for working drivers. First we can also debug the general COMPOSITE bugs in this way and secondly we would have at least the shadow and kompose support for these. This would be a big Box feature in opinion, it has been requested since years. Meanwhile we did some testing with Exa/Composite: Simple Test scenario: --------------------- Enable Composite extension and AccelMethod "EXA" in xorg.conf. Start xcompmgr and two xterms A and B. Set transparency for xterm A with transset. Move xterm A over xterm B. Results: -------- - radeon * R300/R480 (Radeon 9700/Radeon X850 XT): slow * RV200/R200 (Radeon 7500/FireGL 8800): fast - sis * XGI V3XT: slow Exa Status: ----------- http://xorg.freedesktop.org/wiki/ExaStatus Hmm, is this a pro for composite extension or should I disable it ? IMHO we should disable it Currently Composite/EXA makes only sense with Radeon RV200/R200 boards. This still needs to be discussed. Actually, not only Composite/EXA makes sense only for R200, but also EXA alone. Because the other EXA implementations are clearly only at the beginning of implementing acceleration structures. So selecting EXA will always degrade performance, perhaps except for the R200 which seems to be pretty advanced. Don't know about stability, though, there have been several mails WRT that on the mailing list. (In reply to comment #4) > Matthias, could you please make a bugreport in NVIDIA's bugreport system? It took a while, but I finally filed a bugreport. Happy to hear about updates. NVIDIA bugtracking id #197993. cool, good news :) composite will be disabled for fglrx and nvidia binary driver > Also, note that NVIDIA does not yet support OpenGL when the Composite
> extension is enabled (we're working on it, though); current NVIDIA drivers
> disable GLX if the Composite extension is enabled.
Oops we need to check this as well.
> (EE) GLX is not supported with the Composite extension
Ok. Behaviour has not changed for 1.0-8174. Nothing to be done for now.
Our nvr80-series drivers will not support GLX + Composite; we are currently investigating and hope to support GLX + Composite in our nvr85-series drivers. Thanks. nvr85-series is nothing you plan to release in the next 2 months, right? > nvr85-series is nothing you plan to release in the next 2 months, right?
Unfortunately, I do not believe production nvr85 drivers will be available in the next 2 months.
Ok. We'll consider again enabling Composite also for the nvidia driver once nvr85 drivers become available. Has composite been enabled for the nv driver? There seem to be patches for Exa + nv yes is enabled for the nv driver > There seem to be patches for Exa + nv. Yes, this are experimental driver patches. --> http://wiki.x.org/wiki/ExaStatus [...] Work in progress: (Not included in 6.9/7.0) [...] nv (you'll also need the DRM module from CVS if you want to use the Download/Upload hooks) --> Kernel patch No, I don't want to maintain a experimental patch for the next 7+ years ... |