[00:01] (I love that Firefox tells me it is an OLD file.) [00:04] virtuald: would you mind reporting it on bugs.freedesktop.org ? [00:05] virtuald: you should try to get a "full backtrace", see https://wiki.ubuntu.com/X/Backtracing [00:10] i really should sleep [01:01] lets try something [02:30] https://bugs.freedesktop.org/show_bug.cgi?id=21535 [02:30] Freedesktop bug 21535 in Driver/Radeon "X crashes with SIGSEGV in xf86_reload_cursors() with SWcursor" [Normal,New] [14:13] bryce: ping? [19:11] is this supposed to work or am I doing something wrong? --> https://bugs.launchpad.net/xserver-xorg-video-intel/+bug/371774 [19:11] Launchpad bug 371774 in xserver-xorg-video-intel "[G45] i915.modeset=1 causes black screen instead of gdm (UXA)" [Undecided,New] === ripps_ is now known as ripps [19:22] I have the same issue [19:23] but I read somewhere that there was some things to setup with kernel modules before it could work [19:23] ah, can you try to find the url again? [19:23] i really would like to try out kms [19:26] debian fixed that in initramfs-tools [19:26] if you're building your own kernel, just enable KMS by default and boot with nomodeset when you dont want to use it [19:27] mnemo: I searched about it few days ago, and I don't really remember, but it has to do with initramfs yes [19:27] it works fine on the ubuntu 2.6.30-2-3-generic just by adding options i915 modeset=1 to /etc/modprobe.d/whatever.conf [19:27] Sarvatt: or by adding i915.modeset=1 to kernel parameter line, in grub/menu.lst? [19:28] no that doesnt work until initramfs-tools gets updated.. [19:28] at least for me [19:28] Sarvatt: so how can I try it today on ubuntu? [19:30] install 2.6.30-2-3, sudo nano /etc/modprobe.d/i915.conf [19:30] options i915 modeset=1 [19:30] put that in it, and reboot [19:31] Sarvatt: thanks man [19:31] or you can just build KMS by default in your own kernel [19:31] and if you dont want to use it for some reason boot with nomodeset on the grub command line [19:32] i also found this --> http://wiki.debian.org/KernelModesetting [19:32] would it be sufficient to run this: [19:32] update-initramfs -k `uname -r` -u [19:32] or do I need a new kernel for sure? [19:32] that just rebuilds an initramfs.. [19:32] ok [19:33] ah sorry just opened it up [19:33] I'm pretty sure that requires the updated initramfs-tools [19:34] ok, well I guess they have that in debian then [19:34] http://git.debian.org/?p=kernel/initramfs-tools.git;a=summary [19:34] we're on 0.92 [19:35] right and we need this commit? --> http://git.debian.org/?p=kernel/initramfs-tools.git;a=commit;h=652423c6f5b636f95899254aba213d417caff158 [19:38] dont think thats it, have you tried whats on that wiki? that might end up working without any updates [19:39] i'll try it [19:39] it didnt work for me last i tried it but that was in jaunary, i've been enabling KMS by default every kernel since [19:43] well it sort of worked actually [19:43] it brings up gdm and X in 4:3 on my widescreen [19:44] if I vt switch to F1 then I get graphics corruption [19:44] then if I go back to CTRL-ALT-F7 it goes back to normal again [19:47] ah nice, sorry about that, i saw the update in initramfs-tools but didnt actually read what it did lol [19:47] well not nice you get VT corruption [19:48] what intel packages are you using? [19:48] x-updates [19:48] try xorg-edgers, there were a bunch of fixes for that since 2.7.0 [19:48] ah good idea [19:53] get the drm too if you arent adding the repo, one of the post 2.4.9 libdrm updates was related to VT corruption [19:53] ok [19:56] the latest libdrm/mesa/-intel stack on edgers is a real winner for me, i havent been able to use UXA stable since december until 3 days ago [19:59] in jaunty final UXA is better than EXA for me (on G45)... in EXA glxgears leaves repaint dirt when you move the window and also EXA is slower etc [20:00] Sarvatt: whats the difference between drm-snapshot "2.4.9+git20090502.68103b27-0ubuntu0sarvatt" and the one with "r1" in the version number? [20:01] wait, one of them is for karmic... [20:01] is that the only difference then? [20:01] get the karmic one if you're on karmic, the jaunty one replaces linux-libc-dev drm headers because jaunty has old ones [20:02] ok [20:05] exa is 2x faster than UXA for me on a 945GME, theres no more EXA though now in intel drivers :( UXA used to crash whenever I used firefox or ran glxgears in the background of another app [20:05] thats just glxgears speed though, actual 3d benchmarks arent that much different at all [20:06] Sarvatt: but you fixed your problem with glxgears/firefox/UXA, right? [20:06] yup! [20:08] plus mesa 7.6 gets me 120fps more in glxgears in uxa vs 7.4.1, 7.4.x really was a dud lol [20:09] nice [20:14] Sarvatt: glxgears? [20:16] yeah I know, just swap buffer performance :D openarena went up 7 FPS too [20:19] ah, good [20:20] the 120fps glxgears increase is a more than 50% increase for me is why I was saying its good [20:34] yeah, but even increases in glxgears numbers does not always correlate with better performance [20:34] and decreases don't always correlate with reduced performance [20:34] I should just disable the fps output of that command [20:34] we should give people a real benchmarking tool which is as easy to use as glxgears [20:35] * bryce nods [20:35] I would think games would be easy enough... [20:38] different commands to enable FPS printing in different games... also different levels have different complexity etc so its less comparable... [20:38] true [20:38] it should be super repeatable... and it should have windows version so we can benchmark against that later on [20:38] phoronix? [20:38] what's 'windows'? [20:39] I mean we should benchmark our progress compared to Windows 7 drivers etc [20:39] the target needs to be to beat them of course [20:40] phoronix is a good step forward, but I'd like something simpler, more repeatable still.. [20:41] Sarvatt: vt switch gives me corruption on 2.6.99 as well :( and I did the full xorg-egers so I got the libdrm fix as well I guess [20:46] dang, was worth a shot at least. [20:46] mnemo: do you have the fbcon module loaded? [20:47] albert23: no why do you wonder? I got these loaded --> http://pastebin.com/f6c4707c7 [20:47] mnemo: then make sure to load fbcon [20:48] ok thanks I will try it [20:52] uff, heh, just ran into another report of performance change evidenced only by glxgears numbers [20:53] k, that's gotta go. [20:59] gone [21:01] heya michaellarabel [21:01] Hi bryce [21:10] i threw in fbcon in /etc/modules now and it works... I get a KMS vt on CTRL-ALT-F1 ... however, a few seconds after that xorg SEGV'd --> http://pastebin.com/m72073d94 [21:11] mnemo: dunno, a full backtrace would be more interesting [21:12] not sure what aperture space is, but sounds like it doesn't have enough of it or something [21:12] aperture is the memory that intel shares with the CPU i think [21:13] anyway, I will see if I can repro this and file a good bug on it [21:13] thanks [21:13] bryce: it would be neat with an official guide on how to test KMS on ubuntu... right now I felt like I did some stuff without really understanding what I was doing... so im not entirely sure its "supposed" to work on this config [21:13] mnemo: btw, I notice a lot of people use ppracer as a second choice for fps measurements... think that's more sane? [21:14] maybe we could just recommend that to everyone, to get more consistent measurements [21:14] bryce: oh, i've never tried that.. sounds better than glxgears though [21:14] yea [21:14] mnemo: yes, that's a really good idea, would you be willing to help me draft it? [21:14] yey -ati finally has exa by default upstream! [21:14] bryce: sure.. I will post it mailing list later [21:15] ok cool, I'll figure out a good spot in the wiki to put it [21:16] ppracer? I tried etracer and it was much slower - I guess it uses more features [21:18] what's etracer? [21:19] ultimately I suspect we may really need multiple tools that exercise different aspects of performance [21:19] but coding such a thing is well beyond my time availability [21:19] "when you're bored with ppracer, use OA" [21:21] "[note: this is an automated message] dear bug reporter, please collect 10 fish in tuxracer and then attach your xorg.log to this bug report. thank you." [21:22] :-) [21:23] etracer = extreme tux racer, next generation of planet penguin racer [21:23] I think ppracer has stalled, and only etracer is worked on [21:23] wonder if it'd be feasible to automate the 10 fish collection [21:23] :) [21:23] does ppracer have a benchmark mode? I know some games have that, and it'd make results more comparable [21:24] tormod: actually that's good... if we're going to use it for benchmarking we don't want the fps to improve just because of game code changes [21:25] bryce: true. just that etracer makes the driver/card work harder (I think) [21:36] bryce: ppracer isn't much of a good test since it will not build on x86_64 and has other problems [21:38] not build on x86_64? [21:38] IIRC, the last released version of ppracer will not build on x86_64 or had other x86_64 issues. [21:39] sounds fixable.. [21:39] true, though etracer seems to replace ppracer and that has proper x86_64 support. [21:50] hmm, I was able to install and launch it on amd64 [21:50] although I wouldn't say it worked. also messed up dual-head. but it played music and stuff [21:51] libsdl probably talks xf86vidmode, which doesn't quite like randr. [21:52] someone really ought to fix that ;-) [21:53] World of Padman is one of my favorite tests that uses ioquake3 and works with the current Mesa stack. [21:53] michaellarabel: short of phoronix, do you know of a simple glxgears-class tool/game/test people could run? [21:54] bryce: what are you looking to indicate? [21:54] relative 3d performance? [21:54] or just the presence of accelleration at all? [21:54] There's lots of good ones... On Jaunty you can just do: sudo apt-get install phoronix-test-suite and experiment :) say then, phoronix-test-suite benchmark ioquake3-games to see how well all of the different ioquake3 games run [21:55] pwnguin: relative 3d perf [21:55] ioquake3 is at least a close approximation of modern games usage [21:55] ie, has textures [21:55] and geometry [21:55] something short would probably be good, darn openarena benchmark takes a good 10 minutes per run [21:56] yeah, I imagine the users we need to convert have an attention span of maybe 30 sec [21:56] you'd be surprised [21:57] windows gamers slave over framerates [21:57] well it runs 3 times by default in phoronix so it adds up :D [21:57] ok, certainly they're obsessive... [21:57] Sarvatt: It runs multiple times in PTS for accuracy [21:57] 3dmark takes easily 10 minutes [22:01] Unigine Tropics is beautiful, but good luck getting that to work with Mesa. [23:14] note to self: monitors didn't wake standby, gdm showed flying donuts after reboot [23:31] jcristau: Did you notice that the xserver-xorg-input-synaptics source was auto-decrufted from unstable a week after you uploaded it? [23:31] wgrant: i didn't, but dato did. there's an ftp.d.o bug about that. [23:32] bugs.debian.org/526078 [23:32] jcristau: Thanks. That had me very confused for a while. [23:33] * jcristau just pinged that bug [23:33] thanks for the reminder :) [23:33] jcristau: I'm glad you finally did the rename. [23:35] we just got xf86-video-omapfb to compensate for taht [23:35] that, even [23:36] Heh.