| kklimonda | evening.. | 00:11 |
|---|---|---|
| kklimonda | any idea why I'm not seeing splash with nouveau? I get a scrambled screen - mostly black, with some random colors | 00:12 |
| kklimonda | as soon X loads everything works fine | 00:13 |
| kklimonda | ah, it's bug 723477 | 00:14 |
| ubot4` | Launchpad bug 723477 in xserver-xorg-video-nouveau (Ubuntu) "screen corruption instead of pylmouth on boot with nouveau (affects: 1) (heat: 159)" [Undecided,New] https://launchpad.net/bugs/723477 | 00:14 |
| RAOF | bryceh: Ok. So, I too see a (smaller) memory delta (on amd64); my steady-state numbers change from ~500MiB to ~645MiB. | 00:40 |
| RAOF | Hm. And it looks like the mere act of loading nvidia's libGL dirties in excess of 4MiB of memory, doubling the dirty memory of my gnome-terminal. | 01:06 |
| RAOF | bryceh: Looks like the mere act of linking to nvidia's libGL is going to increase the memory of the process by ~5MiB. If cairo links to libGL, that's a lot of memory. | 01:38 |
| bjsnider | why shouldn't it? it's not using the mesa libgl at all. | 01:40 |
| bryceh | RAOF, could we prevent that from occurring with nvidia? wayland won't work with nvidia anyway | 01:42 |
| RAOF | bjsnider: You don't understand; that's 5MiB of *unsharable* memory. If you have 30 processes linking to GTK you've uselessly eaten 150MiB of memory. | 01:43 |
| RAOF | bryceh: We could get cairo-gl to dynamically load libGL when first constructing a GL context, I guess. | 01:44 |
| bjsnider | RAOF, have you talked to nvidia about that? | 01:44 |
| bryceh | anyway, need to get off the computer, bbl | 01:44 |
| RAOF | bjsnider: No, I've only just noticed it. | 01:44 |
| RAOF | bryceh: It's probably pretty easy to dynamically load libgl; I'll check it out. | 01:44 |
| bjsnider | RAOF, surely that's not how nvidia _wants_ things to work | 01:47 |
| RAOF | I can imagine reasons why nvidia would do that. | 01:47 |
| bjsnider | RAOF, you're more qualified to imagine that kind of thing than i are | 01:48 |
| RAOF | Specifically, runtime patching of code will cause that, and it's not unreasonable to do that in libGL (mesa did on i386, until I stopped it). | 01:48 |
| bjsnider | RAOF, i'm sure if you notify aplattner he can either do something about it or explain why it works that way | 01:50 |
| kklimonda | RAOF: hmm.. in my case it was more than 5MB I think.. I do remember that some basic Qt app I wrote used 45MB of ram, and 20MB came from one of nvidia libraries.. | 01:53 |
| kklimonda | but I've assumed I don't know how to read /proc/pid/smaps | 01:54 |
| RAOF | kklimonda: Was that dirtied memory or a sharable library text? nvidia's libGL is pretty big, and will show up as such, but should be sharable across all the processes linking to it. | 01:55 |
| kklimonda | RAOF: right, I can't really remember right now - I'll check it out when I go back to nvidia. | 01:55 |
| RAOF | Unless they've done something crazy like make a non-PIC libGL on i386. | 01:55 |
| kklimonda | btw, is there something wrong with nvidia drivers that makes unity so slow on it? | 01:56 |
| kklimonda | running dmesg takes seconds for example | 01:56 |
| kklimonda | pretty much everything I do in terminal gets slowed down considerably the moment I start looking at it (it works fine when terminal is hidden) | 01:57 |
| RAOF | Ah. | 02:00 |
| RAOF | I'd guess you're hitting a font-rendering fallbakc. | 02:00 |
| kklimonda | yeah, it's quite possible - at some point changing to "Best shape" gave me a lot of performance boost | 02:02 |
| kklimonda | but when I tried doing that in natty it didn't help.. at least I don't remember the same boost :) | 02:02 |
| RAOF | It's possible that switching your fonts will do somethnig, too. | 02:07 |
| kklimonda | RAOF: http://paste.ubuntu.com/583614/ | 02:40 |
| kklimonda | Private Dirty: 16396 kB | 02:40 |
| RAOF | Owch. What process was that? | 02:44 |
| kklimonda | gnome-terminal | 02:44 |
| wgrant | RAOF: Hi. | 02:46 |
| RAOF | wgrant: Howdie. | 03:14 |
| RAOF | wgrant: After somethnig? | 03:14 |
| RAOF | Other than a chat after a long period of no chatting? :) | 03:15 |
| RAOF | Bah. Dear Telstra: unless you're willing to buy out my existing phone contract you *definitely* won't be able to offer me a ‘fantastic deal on my mobile phone’. | 04:49 |
| wgrant | RAOF: Sorry, back. | 06:09 |
| wgrant | RAOF: Your 113_tls.diff in mesa has made it into xorg-edgers, and it breaks unity pretty badly (at least on r600g). | 06:09 |
| wgrant | Is it meant to do that? | 06:09 |
| RAOF | It is not. | 06:09 |
| RAOF | What does it do? | 06:09 |
| wgrant | Well, Unity basically doesn't render. It occasionally flickers up a very corrupt version of itself when hovered. | 06:10 |
| wgrant | Normal compiz mostly works, except that some shadows are broken. | 06:10 |
| RAOF | Hm. | 06:11 |
| RAOF | And you're sure this is a tls problem? | 06:11 |
| wgrant | (at least in the 20110318 xorg-edgers mesa build... I reverted the patch from that and it's all good again) | 06:11 |
| RAOF | Oh. Curses. | 06:12 |
| wgrant | I might try building it with the patch locally, just in case. But the 20110316-20110318 diff is pretty minimal, apart from that patch, some 965 stuff, and some tiny refactorings. | 06:13 |
| RAOF | wgrant: Are you on i386 or x86-64? | 06:16 |
| wgrant | RAOF: x86-64 | 06:16 |
| RAOF | That makes it easier. I *have* an x86-64 r600g system :) | 06:16 |
| RAOF | Although I guess I could also break open the esprimo and throw in the radeon card. | 06:17 |
| wgrant | It's a Radeon 5700 of some variety, FWIW. | 06:17 |
| RAOF | Which, conveniently, is also what I have. | 06:18 |
| wgrant | Excellent. | 06:20 |
| RAOF | Oh, balls. Something's broke in my monitor, and it's now sending invalid EDIDs. | 06:55 |
| wgrant | :( | 06:58 |
| wgrant | Sure it's not just X being really broken? | 06:58 |
| RAOF | Nah; the kenel's throwing “checksum invaliD!” into dmesg all the time. | 07:01 |
| bjsnider | http://arstechnica.com/open-source/guides/2011/03/the-linux-graphics-stack-from-x-to-wayland.ars | 14:20 |
| mvo | I get some really odd artifacts on my intel notebook with natty, looks like redraw problems, oddly enough with the unaccelerated desktop. | 14:26 |
| === JanC_ is now known as JanC | ||
| mvo | bryceh: are there any report on redraw artifacts on intel and noveau with the classic desktop? it happens for me on natty a lot when e.g. minimizing a window | 16:34 |
| bryceh | mvo, 622068 725333 737967 are visual corruption bugs reported on -intel. | 17:38 |
| mdeslaur | mvo: I see it, and kees sees it also | 18:26 |
| mdeslaur | bryceh: ^ | 18:26 |
| mdeslaur | mvo: are you using metacity? | 18:27 |
| bryceh | mdeslaur, hmm, didn't that get identified as a kernel bug and sorted out a couple weeks ago? | 18:31 |
| mdeslaur | bryceh: I don't know...but I'm up-to-date | 18:32 |
| mdeslaur | bryceh: it only happens with metacity, but it happened with my nvidia laptop and it happens with my new intel laptop this week | 18:32 |
| bryceh | bug 717114 | 18:32 |
| ubot4` | Launchpad bug 717114 in xserver-xorg-video-intel (Ubuntu Natty) (and 4 other projects) "[i945gm] Screen Corruption with new Xorg stack with terminal programs (affects: 20) (dups: 3) (heat: 104)" [Medium,Triaged] https://launchpad.net/bugs/717114 | 18:32 |
| bryceh | set to fixed for linux on 3/18 | 18:33 |
| mdeslaur | bryceh: that's not the same issue | 18:33 |
| mdeslaur | bryceh: I'll try and get a couple of screenshots...but am not sure under what to file the bug | 18:34 |
| bryceh | mdeslaur, ok, mind digging up the bug report #? | 18:34 |
| mdeslaur | bryceh: can I just file it under xorg until someone figures out what's causing it? | 18:35 |
| bryceh | given this other one was a kernel issue, wouldn't you think yours is kernel too? | 18:36 |
| mdeslaur | bryceh: you're just trying to get rid of me :) | 18:38 |
| bryceh | hehe | 18:38 |
| bryceh | mdeslaur, but no, if you feel there is value to having me involved in looking at it, go ahead and file against X, but I'm like 90% certain it's going to end up being another kernel issue | 18:39 |
| bryceh | mdeslaur, I posted a list of other graphical corruption bugs to mvo up above, I'm going to forward them all upstream today, so if you file yours against xorg I'll send that one up too | 18:43 |
| bryceh | even if it ends up being a kernel bug, when I upstream them they often will point out the patch for us | 18:43 |
| mdeslaur | bryceh: does binary nvidia use the kernel? | 18:43 |
| mdeslaur | bryceh: ok, let me reproduce and get a couple of screenshots, and I'll file it | 18:43 |
| bryceh | it uses a lot less of the kernel than the foss drivers do | 18:46 |
| mdeslaur | bryceh: bug 740387 | 18:49 |
| ubot4` | Launchpad bug 740387 in xorg (Ubuntu) "graphical corruption with multiple drivers and classic desktop (affects: 1) (heat: 6)" [Undecided,New] https://launchpad.net/bugs/740387 | 18:49 |
| bryceh | mdeslaur, ah yeah that is quite different (and I think different from what mvo is mentioning) | 18:52 |
| bryceh | mdeslaur, I would guess even if it looks similar, the issue on -nvidia and -intel could have different root causes. Mind filing a second bug for the other machine? | 18:53 |
| mdeslaur | bryceh: sure | 18:53 |
| mdeslaur | bryceh: bug 740422 for issue on nvidia with nvidia screenshots | 19:26 |
| ubot4` | Launchpad bug 740422 in xorg (Ubuntu) "graphical corruption with classic desktop (affects: 1) (heat: 6)" [Undecided,New] https://launchpad.net/bugs/740422 | 19:26 |
| mvo | mdeslaur: yeah, I use metacity | 19:55 |
| mdeslaur | mvo: are 740387 and 740422 what you are seeing? | 19:55 |
| mvo | hold on a sec, let me check | 19:55 |
| mvo | mine looks like this: http://people.canonical.com/~mvo/tmp/Screenshot.png | 19:55 |
| mvo | except for the blur, I added that :) | 19:56 |
| mdeslaur | mvo: looks similar | 19:56 |
| mvo | indeed | 19:57 |
| mvo | I have it on intel and noveau | 19:57 |
| mvo | nouveau is not stable enough for me to run unity | 19:58 |
| mvo | but that is a different matter | 19:58 |
| mvo | mdz: thanks, I have a look at 740372 | 20:00 |
| bryceh | mdeslaur, mvo, do you recall about when you started seeing this regression? | 20:23 |
| bryceh | trying to get a feeling for what kernel (or X) version brought the issue | 20:23 |
| mdeslaur | bryceh: honestly, I'm investigating metacity now...I think it's a DX patch | 20:23 |
| mvo | my gut feeling is the 1.9 - > 1.10 switch, but I may be wrong | 20:24 |
| mvo | mdeslaur: interessting idea! | 20:25 |
| tjaalton | is it the same corruption we discussed the other day? | 20:25 |
| tjaalton | i'm seeing it with savage | 20:25 |
| tjaalton | when opening menus etc | 20:25 |
| bryceh | fwiw, I've seen something similar on -ati but it only shows up when launching URLs from xchat to firefox so dunno if it's the same issue. Similar effect though (invalidly drawn areas of the region) | 20:25 |
| bryceh | tjaalton, don't think so | 20:26 |
| mvo | bryceh: this is what triggers it most reliable for me as well | 20:26 |
| bryceh | that menu corruption was particular to -ati it seems (upstream has pointed to a kernel patch to revert) | 20:26 |
| tjaalton | ah, looked at the screenshot, and it's not | 20:26 |
| bryceh | I had been thinking perhaps the damage issue was a subset of the menu drawing issue since I started seeing both at the same time | 20:27 |
| bryceh | but now I'm pretty sure they're distinct bugs | 20:27 |
| bryceh | mdeslaur, should I hold off on upstreaming this bug? (about to pull the trigger on it to go to -intel) | 20:28 |
| mdeslaur | bryceh: yeah, you should probably wait a bit | 20:29 |
| bryceh | alrighty | 20:29 |
| mdeslaur | bryceh: I can't seem to reproduce it at the moment with the DX patches ripped out | 20:29 |
| bryceh | mdeslaur, 06_Add_UXD_shadows_and_borders.patch ? | 20:34 |
| mdeslaur | bryceh: maybe, I'm trying to figure out which one now | 20:35 |
| mdeslaur | but that,s the one that I suspect | 20:35 |
| bryceh | hmm, some of these metacity patches sound like they could have some performance implications | 20:36 |
| AlanBell | wondering if someone could help me get a little further with bug 738330 | 20:57 |
| ubot4` | Launchpad bug 738330 in virtualbox-ose (Ubuntu) (and 1 other project) "Today's Natty update means no guest additions in virtualbox (affects: 9) (heat: 58)" [Undecided,Invalid] https://launchpad.net/bugs/738330 | 20:57 |
| AlanBell | this is the oracle virtualbox drivers, not the Free Software OSE ones - however the source is available and gets compiled with dkms I think | 20:58 |
| AlanBell | the problem is an ABI issue with the new Natty xorg ABI 10 and the drivers | 20:59 |
| jcristau | dkms is for kernel module, not X drivers | 21:02 |
| tjaalton | so the guest-additions are opensource? | 21:02 |
| tjaalton | if not, there's nothing we can do other than wait for oracle to fix it | 21:03 |
| jcristau | tjaalton: virtualbox-guest-additions is in non-free in debian, so i guess not | 21:03 |
| tjaalton | right | 21:03 |
| AlanBell | interesting point about dkms, I was just reading up on that | 21:04 |
| AlanBell | so it does build a video related driver, but I think that is the kernel part, so the xorg driver itself would be separate | 21:05 |
| tjaalton | X.Org guests: support X.Org Server 1.10 pre-release and Ubuntu 11.04 Alpha. | 21:05 |
| tjaalton | from 4.0.4 changelog | 21:05 |
| tjaalton | which means abi 9, so they just need to fix it | 21:05 |
| jcristau | so they have both a closed and an open X driver? | 21:08 |
| jcristau | sigh | 21:08 |
| AlanBell | http://www.virtualbox.org/changeset/36301 does that look like the crucial change to you? | 21:08 |
| AlanBell | jcristau: the source is all available | 21:08 |
| AlanBell | the latest version comes prebuilt, the OSE version in Ubuntu lags behind | 21:09 |
| jcristau | this stuff is so fucking broken.. | 21:09 |
| AlanBell | and unity only runs on the bleeding edge prebuilt version | 21:09 |
| AlanBell | or did until recently :) | 21:09 |
| tjaalton | AlanBell: so where is the vboxvideo_drv_110rc.so built? | 21:11 |
| tjaalton | um, 110.so | 21:11 |
| jcristau | AlanBell: "comes prebuilt" means without source then? | 21:11 |
| AlanBell | jcristau: no, the stable release version is available built - just like anything else | 21:13 |
| AlanBell | they don't have daily builds from SVN | 21:13 |
| AlanBell | and the version in Ubuntu is older than the upstream stable release | 21:13 |
| AlanBell | it really isn't much more insane than anything else out there tbh | 21:14 |
| tjaalton | 4.0.4-dfsg-1ubuntu3 | 21:14 |
| tjaalton | don't see anything newer upstream | 21:14 |
| tjaalton | 4.0.4 is the latest release | 21:14 |
| AlanBell | ah, ok sorry, the latest is in Natty, but I am using a Maverick host | 21:15 |
| tjaalton | virtualbox-ose-guest-x11 should depend on the correct abi | 21:19 |
| tjaalton | what it supports | 21:19 |
| tjaalton | not what it happens to be "built" against | 21:20 |
| AlanBell | so for instant gratification I guess I compile from SVN and install those drivers in the guest Natty | 21:29 |
| tjaalton | something like that | 21:29 |
| speakman | Hi folks. Is there any tools to trace what's causing X to be extremely CPU intensive? | 21:53 |
| speakman | I'm on a w3680 (6-core @ 3.33GHz) but I'm getting flashbacks from the mid -90. | 21:53 |
| tjaalton | kill apps one by one, when it stops you have your culprit :) | 21:55 |
| speakman | I've tried (of course) but it's still laggy no matter which apps are running | 21:56 |
| speakman | I'm on two GPUs with four monitors if it may have anything to do with it. Both are Nvidia with the propretary drivers instaleld. | 21:56 |
| tjaalton | likely does | 21:57 |
| speakman | But anyway; is there a way to trace what's causing the X processes the CPU hogging? | 21:57 |
| tjaalton | try a profiler | 21:58 |
| speakman | There's no way to attach to the running X process? | 21:58 |
| tjaalton | yes, but nvidia.. | 21:58 |
| tjaalton | besides, what would you benefit by hooking gdb to it? | 21:59 |
| speakman | I've not mentioned gdb. But I thought there might be tools for profiling X. | 22:00 |
| speakman | This can't be the only computer with X hogging the CPU? | 22:00 |
| tjaalton | i said try a profiler | 22:00 |
| tjaalton | there are tools | 22:00 |
| tjaalton | i just don't have any docs handy | 22:00 |
| speakman | is there any X profilers available? | 22:01 |
| tjaalton | or try with a single gpu & monitor first | 22:01 |
| tjaalton | why would x need a profiler of it's own? | 22:02 |
| AlanBell | well virtualbox is now compiling away on my natty VM | 22:08 |
| tjaalton | AlanBell: could you tell felix geyer about the issue? | 22:16 |
| AlanBell | tjaalton: where do I find him? | 22:18 |
| speakman | I just ran sysprof and it seems like nvidia_drv.so is the trouble | 22:20 |
| AlanBell | nvm, devfx | 22:20 |
| tjaalton | speakman: surprise :) | 22:20 |
| tjaalton | it's also an unusual setup | 22:21 |
| tjaalton | i guess | 22:21 |
| speakman | tjaalton: not really... but now how to fix it :/ | 22:22 |
| bryceh | mdeslaur, good work narrowing down to patch 16; are you or someone you know going to disable that patch? | 22:22 |
| bryceh | mdeslaur, if not I can take care of doing it | 22:23 |
| bryceh | mdeslaur, have you also verified this was the cause of the -nvidia issue as well? | 22:25 |
| mdeslaur | bryceh: I've pinged the author of the patch, let's see what he says | 22:26 |
| mdeslaur | bryceh: I only tested it on -nvidia | 22:26 |
| mdeslaur | bryceh: but will change the gconf key on my -intel laptop tomorrow | 22:27 |
| AlanBell | yay, I have unity in 3d again, the virtualbox-ose-guest-x11 package in natty works just fine | 22:41 |
| === Amaranth_ is now known as Amaranth | ||
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!
