Phoronix: xf86-video-intel 2.7.1 Released With Crash Fixes
The focus of Intel's X.Org driver developers is on getting ready the xf86-video-intel 2.8 driver for this quarter's update that will remove the EXA acceleration architecture in favor of UXA acceleration and it completely strips away the DRI1 support. The Intel X.Org team released the xf86-video-intel 2.7.0 driver nearly a month ago, but they have cherry-picked a few fixes from their current development efforts to provide a 2.7.1 release. The xf86-video-intel 2.7.1 driver hooks up RandR properties support under a KMS-enabled environment and also five other bug fixes that hopefully will fix a few crashes and other problems that have come about...
I'm not getting this whole "performance regression" thing. Even under 2.6.28, I could play $GAME_I_FORGOT with real-time lighting, specular mapping, and bump mapping at a decent framerate at 1440x900 inside Compiz (without fullscreen unredirected).
Did you try manually specifying the virtual screen size as the bug report suggested.
Not yet. I prefer to use Ubuntu as a "regular user" would, so I'm waiting for the official release from Canonical that will do it for me. Either way, that's not really a fix. It's a workaround that makes the crashes less frequent. I'm wondering if this will actually fix the crashes.
With my Core2duo@2.8 and G33 (and kernel 2.6.29), I can play The Matrix@1080p perfectly smoothly with VLC, without using complex instructions in a console (which were mandatory up to now, and I could only use mplayer that way...so I could not pause a video, only abort it).
I can also now play Torcs smoothly at 1280*1024 on some circuits that were unplayable until now. A year ago no circuit was playable even at 800*600.
Intel and Xorg programmers (and the VLC ones too) are really doing a fine job!
Last edited by DebianAroundParis; 05-27-2009 at 04:43 AM.