Hi Marco, > I have several card cages at my disposal, PCI/PCI-X and AGP & PCI/PCI-X. > They work both fine in my rx2600 and rx2620 systems. (Under HP-UX and > Windows at least, with proper acceleration at least.)
Interesting. Do you think that these cages will also work in rx2620? > Thank you for the information and pointers! You're welcome! Huh, I just forgot about an important thing in my previous message! KMS, even with current kernel, still complains about AGP bind memory problem. (As a result?) GPU acceleration is disabled and AGP mode forced to PCI. This issue has been fixed in agp/amd-k7 for the AMD platform. I've simply adapted this patch for the zx1-derivatives and the problem seems to go away (see http://marc.info/?l=linux-ia64&m=130263778331960&w=2 for details). I never got any feedback on this patch port, so I don't know (i) if it's correct and (ii) if it really enables back GPU acceleration or simply get rid off the error messages! Indeed, with or without this patch applied to kernel source, glxgears still reports extremely low ~85 FPS with r300g. If a kernel developer is reading this, I would be happy to know his/her thoughts about my patch port and what's the process to submit it in order to make it available in official kernel source. > How do you mean exactly? Wouldn't in software be handled by the CPU? Yes, software rendering is handled by the CPU. >> With Mesa switching from r300 classic (r300c) driver to new Gallium 3D >> (r300g) one, DRI2/AIGLX is now successfully initialized and enabled. >> But GL rendering is completely garbaged [...]. I didn't try Mesa git >> for a while. Maybe everything is now working fine! I have to check this. > > Thanks again, that sounds good; well, it's a start. I'll definitely look > into this and give it a try, at least! I have good news: I just recompiled current Git master Mesa 7.12-devel (git-ccecc08) and it's working fine :-) I've updated https://bugs.freedesktop.org/show_bug.cgi?id=36608 accordingly. I hope that forthcoming libgl1-mesa-* packages in Debian Testing will be based (at least) on this revision. So that http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=622299 can be closed safely. Best regards, Émeric -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected] Archive: http://lists.debian.org/caa9xbm7nvtzp7ob3tt-0ldcwyeak9ycfgkapfrs7nub3r5u...@mail.gmail.com
