03:26alexandernst_: I'm looking at the FeatureMatrix page, but I'm not really sure to which column belongs my card. It's a GT610
03:33glennk: alexandernst_, wikipedia says its a rebranded GT 520, which is a nvd9 (GF119)
03:33Tom^: alexandernst_: http://nouveau.freedesktop.org/wiki/CodeNames/
03:34alexandernst_: Thank you glennk Tom^ :)
03:34alexandernst_: But I don't see "NVD9" or "GF119" in the FeatureMatrix page
03:35alexandernst_: an I searching for the wrong thing?
03:38glennk: alexandernst_, its from the fermi family (nvc0)
03:39alexandernst_: oh, cool :D
03:39alexandernst_: I see Multicard is implemented \o/
03:40alexandernst_: so, my next question. I'm using nouveau (yes, I installed it without knowing if it's supported, I love risky things!), and I have 2 GT610 cards (not in SLI), but only 2 of my 4 monitors work.
03:40alexandernst_: Is there a chance I could get all 4 monitors?
03:43alexandernst_: (get them working, I mean)
03:44mupuf: alexandernst_: not without disabling the acceleration IIRC
03:45mupuf: IIRC, it would require shatter to work
03:45mupuf: or reverse prime could possibly work, never tried it
03:46glennk: a 610 would be rather weak at driving more than one display as it is
03:46alexandernst_: glennk: I'm currently with 2 monitors that work, and things are fine :)
03:46glennk: works for displaying mostly still images :-)
03:46alexandernst_: I use my pc for work, I don't play games on it, so I guess I'm fine without acceleration as long as I'm able to watch some youtube videos
03:47pq: isn't separate SCREEN an option? That you get by default?
03:48alexandernst_: Right now 2 monitors are working and I'm able to drag windows from one to another, so I guess that is the same "SCREEN", right?
03:49alexandernst_: I'd like the combination of 4 monitors to work the same way
03:49pq: then it is what mupuf said.
03:50alexandernst_: Ok. How bad will be the performance with acceleration disabled? Will I still be able to work and watch some youtube videos ocasionally?
03:50alexandernst_: Also, what is "shatter" ?
03:50alexandernst_: And also, "reverse prime"
06:24mwk: and now I'll have to reject that.
06:25mwk: it creates a second copy of rnndb
06:25hakzsam: oh yes
06:25hakzsam: this should be merged then
07:11pmoreau: Cool! If I return NULL in evo_wait to avoid setting element 1.10^9 of dmac, I get interrupt 0x00000002 on PMC
07:12pmoreau: Maybe some clues at what's going wrong
08:38pmoreau: It is disturbing to have sometimes low first, sometimes high first in rnndb.
08:38pmoreau: Couldn't we set some ordering convention?
08:47imirkin: pmoreau: hm?
08:47imirkin: you mean in a bitfield? yeah, that's annoying
08:50pmoreau: Would need to change G200 to GT200 and G80 to NV50 if we use the same names as NVidia
14:06pmoreau: Quite done with incorporating EVO doc on the core channel into rnndb; it will need some reviewing afterwards (mostly on the form :D )
14:13imirkin: pmoreau: ... evo should have already been there
14:13pmoreau: Not complete compared to what NVidia released :-)
14:13imirkin: pmoreau: https://github.com/envytools/envytools/blob/master/rnndb/display/nv_evo.xml
14:13pmoreau: Or did I missed some recent commit?
14:14imirkin: ah yeah, probably not compared to that
14:16pmoreau: There were some erros on width/height fields for example: we had width 15:0 and height 31:16 whereas it should be width 14:0 and height 30:16.
14:20imirkin: meh, that's pretty minor
14:20pmoreau: Sure :)
14:20imirkin: diff chips have diff limits
15:08jjohn: Evening, guys. I am trying to get the firmware of the proprietary nvidia driver for the nouveau driver via mmio tracing, but need some help because I'd like to use Optimus/Prime ...
15:09imirkin: jjohn: what hardware are you on?
15:09jjohn: ... and the instructions on http://nouveau.freedesktop.org/wiki/MmioTrace/ and https://wiki.ubuntu.com/X/MMIOTracing do not really help.
15:09jjohn: "01:00.0 3D controller: NVIDIA Corporation GM107M [GeForce GTX 850M] (rev a2)"
15:09imirkin: ah, eternal sadness
15:09jjohn: The maxwell architecture. The one for that there is no firmware yet.
15:09imirkin: what are you looking to achieve?
15:10jjohn: Getting the files "nouveau/nv117_fuc409c" and "nouveau/fuc409c", to include them into my kernel built, so that they may be loaded in by the nouveau driver.
15:11jjohn: /var/log/kernel.log shows that they are missing. I hope that if I provide them, that I can use the additional GPU power for rendering and video en/decoding.
15:11imirkin: jjohn: that's a means... what's the end? are you hoping to use nouveau for acceleration, or merely powering displays that are exclusively hooked up to the nvidia card?
15:12jjohn: See my last sentence. :)
15:12jjohn: "so that I can use the additional GPU power for rendering and video en/decoding."
15:12imirkin: video decoding isn't supported on maxwell yet
15:12imirkin: video encoding isn't supported on any gpu's by nouveau
15:13imirkin: and without reclocking, it's unlikely that you'll get faster speeds from the maxwell chip than the intel gpu.
15:13jjohn: "any gpu's by nouveau (yet)?"
15:13imirkin: it should all be relatively simple. the ubuntu wiki is pretty good
15:13imirkin: what step are you having issues with?
15:14jjohn: After I installed the driver and made sure the nvidia.ko files was loaded in, I tried to start xinit -e sh -c "glxgears & sleep 10".
15:15imirkin: did you start mmiotrace before loading nvidia.ko?
15:15jjohn: No, I didn't.
15:15imirkin: ok, well, that's how mmiotrace works
15:16jjohn: I installed the OpenGL userspace sos into a separate directory, because the last time I tried to do that it destroyed the Mesa OpenGL files.
15:16imirkin: it can be a bit tricky to get X going with the nvidia stuff installed off by the side
15:16jjohn: They are overwritten/backupped by the nvidia driver installation routine, so I can simply back them up. But I cannot use both, it seems.
15:16imirkin: i do a combination of adding a modulepath and adding all sorts of ld_library_path things
15:17imirkin: and carefully ldconfigs
15:18jjohn: So, what do you suggest?
15:19imirkin: you have to do the mmiotrace...
15:20imirkin: skeggsb_: so... what's required for nouveau to have open ctxsw there? have you looked at it at all?
15:20jjohn: Oh, damn. Sorry, I misunderstood your question. Uhm, yes, I started the mmiotrace BEFORE loading in the kernel module.
15:21jjohn: English is not my native language, and sometimes I am just stupid. Yes, I did the mmiotrace. Shall I upload it to pastebin?
15:22pmoreau: jjohn: Pastebin limit is 600k, I hope your mmiotrace is bigger than that!
15:23jjohn: You mean "smaller"? And as you can see, it is.
15:23pmoreau: Did you started X while running the mmiotrace? The blob only sets the card when X is run.
15:23imirkin: jjohn: looks like it didn't actually use nvidia
15:24jjohn: I tried to use xinit -e sh -c "glxgears & sleep 10", but since it wouldn't find the OpenGL files, it wouldn't even start up.
15:24imirkin: jjohn: yeah, it's a bit tricky
15:24imirkin: jjohn: first try to get it to work without the mmiotrace
15:24imirkin: and then once you work it all out
15:24imirkin: redo it with the mmiotrace
15:26jjohn: OK. Can you give me a pointer or two how to set it up? What parameters should I pass to the nvidia installer script? Or do you have a link for me?
15:26imirkin: jjohn: wellll... if you don't want to screw up your whole system, the nvidia installer script is out
15:26jjohn: Wonderful. :)
15:26imirkin: jjohn: what i usually do is just run the extractor bit of it (iirc -x)
15:27imirkin: and then create an xorg.conf that adds the relevant place as a ModulePath
15:27imirkin: hold on, let me see if i can find it
15:29jjohn: Just take your time. I'll be waiting.
15:32imirkin: jjohn: see pm
16:09jjohn: Hello again.
16:10jjohn: imirkin: So, I made the kernel load in the firmware files, but when my system tries to start X, it hangs completly.
16:11jjohn: Keyboard is not responding at all. All that I can do is hard-resetting.
16:11chdrsto: hi guys ... im trying to build mesa from source on Debian Wheezy ... and having issue with dependencies
16:11imirkin: jjohn: sad. what version of xf86-video-nouveau do you have? you need at least 1.0.10 iirc for maxwell
16:11chdrsto: Im getting the message ... : Requested 'libdrm_nouveau >= 2.4.33' but version of libdrm_nouveau is 0.6
16:11imirkin: jjohn: oh, and you need mesa 10.3.x at least
16:12imirkin: chdrsto: yeah, 0.6 is a prehistoric version
16:12imirkin: chdrsto: turns out nouveau has improved in the past 4 years or so since libdrm_nouveau 0.6
16:12chdrsto: imirkin : might be, but its on stable Wheezy
16:13jjohn: Shucks. Mesa is 10.2.8 ...
16:13jjohn: I want to scream.
16:13imirkin: chdrsto: ok. not sure how that relates to my comment.
16:15jjohn: OK. Would you advice me to build Mesa on my own as well? Because Gentoo is not Debian, but it sometimes uses the same old versions.
16:15chdrsto: imirkin : sorry! ... but debian is telling me that libdrm version is 2.4.40
16:15jjohn: xf86-video-nouveau is 1.0.10 ... me lucky one.
16:16imirkin: jjohn: there's a 10.4.3 ebuild in gentoo right now
16:16imirkin: but you'll have to keyword a few things with ~amd64 to get it to build
16:16imirkin: chdrsto: oh. 2.4.40 should be enough. do you have the relevant foo-dev packages installed too?
16:17jjohn: Well, that's a bit bad. Because I need to use Wine, and as far as I remember, it needs some libraries for x86 stuff.
16:18jjohn: And Windows has so much x86 that dropping support to favor x64 renders wine way less useful.
16:18imirkin: jjohn: ah yeah... that's a bit annoying. there's the abi_x86_32 stuff, although tbh, i haven't fully worked out how to use it effectively
16:19jjohn: imirkin I already said that up. I am not using the emul* packages anymore.
16:19jjohn: set that up*
16:19jjohn: Sorry. Bad english again.
16:20imirkin: jjohn: ah ok. so then the masked 10.4.3 should work fine, once you unmask its deps (eselect-opengl, glproto, xorg-server, and a few others)
16:20chdrsto: imirkin: maybe I have to add that Im trying to build it on powerpc since I get weird color issue ABGR instead RGBA
16:21imirkin: chdrsto: ah. well, you may be better off with an older mesa then... what gpu do you have?
16:21imirkin: nv34 or something?
16:22imirkin: chdrsto: either way, it sounds like you're missing headers
16:23chdrsto: imirkin: OpenGL renderer string: Gallium 0.4 on NV47
16:23imirkin: iirc the abgr vs rgba "issues" got introduced in mesa 9.1 or 9.2. it was actually a partial fixing of things, but not enough, so some things ended up swapped.
16:24imirkin: i _think_ there have been further fixes, but as i don't personally have the hw, i dunno what the current state is
16:25jjohn: imirkin: Uhm, stupid question ... but how do I find out which packages I have to unmask?
16:25jjohn: I am not entirely unfamiliar with portage, but also no expert.
16:25imirkin: jjohn: unmask mesa. try to build. see what it complains about. repeat. i think there's also a emerge --auto-do-everything-for-me but i don't trust it
16:28jjohn: So, I added media-libs/mesa to package.unmask ... but it still offers me 10.2.8 only.
16:28imirkin: unmask was probably the wrong term
16:31jjohn: Should I then unmask every package that would cause a block there (like x11-base/xorg-server)? Because that one is blocking app-admin/eselect-opengl.
17:51skeggsb_: imirkin: there's just a huge chunk of state we don't save/restore that nvidia does i think
18:28imirkin: skeggsb_: for maxwell you mean?
18:59jjohn: OK, /quit
19:25FunkyBob: morning all
19:27FunkyBob: if I've installed the nvidia driver, should I expect to still see nouveau during boot? because I do
19:51jjohn: Holy motherf****. OK, I updated my X server and all corresponding libraries. 3.0 Mesa 10.4.3, and the firmware loads ... but the system still hangs when the system tries to start X.
19:52jjohn: Took me some time because I still don't know much about portage, so I had to try out a lot.
19:53jjohn: And there are no error messages in the /var/log/kern.log except for "ACPI: \_SB_.PCI0.PEG0.PEGP: failed to evaluate _DSM".
19:54jjohn: Here is what the kernel has to say about the nouveau driver: http://pastebin.com/1tMJKNFG
20:22jjohn: Noone got an idea?
21:29imirkin: jjohn: that log looks largely fine...
21:29imirkin: don't worry about that failed to evaluate _DSM thing
21:30jjohn: Well, but my box keeps dying with the nouveau driver. :(
21:30imirkin: [ 3.977703] nouveau E[ PIBUS][0000:01:00.0] HUB0: 0x10ecc0 0xffffffff (0x1a40822c)
21:30imirkin: can't say i'm thrilled about that...
21:30jjohn: I really appriciate your help. But at this point I don't even know what to do. I want to get this running, but I am at a total loss here.
21:31jjohn: What is about that?
21:48jjohn: OK, imirkin, I just looked into the sources ...
21:48jjohn: ... and how. I cannot believe that Red Hat dared to ship something like that. And I cannot believe that those responsible for the kernel signed it.
23:00jjohn: Does anyone else have more knowledge about what the error message means exactly.