01:34RSpliet: psaisanas1: MSI interrupts should work just fine on anything newer than NV50 (Geforce 8xx0)
01:35RSpliet: that doesn't answer your question I know, but you might want to be more specific about your problem before someone can help you :-P
01:36RSpliet: Weaselweb: ah... hmm..... log literally says "unplugged"
01:37RSpliet: have you already excluded the "faulty cable" case?
01:38Weaselweb: well, actually not :-/
01:40Weaselweb: RSpliet: I copied those messages directly from journalctl -b
01:43Weaselweb: I'll give a new DVI cable a try tomorrow
01:45RSpliet: if it's not too much trouble, please
02:55pmoreau: I've been running Nouveau on my different Tesla cards for some checking, and got this warning on G80-G96: "failed to create encoder 0/1/0: -19"
02:55pmoreau: And on G86 and G98: "M0203T not found" and "M0203E not matched!"
04:27pmoreau: Is there a file somewhere listing all the cards each dev has? If not, where would it be best to put it?
04:28mupuf: pmoreau: there is a wiki page
04:30pmoreau: mupuf: I'm aware of that one, I even added myself to it some time ago ;)
04:30pmoreau: However, this one lists hardware developers would like to have, not hardware they already have
04:30mupuf: you want pciids then?
04:31mupuf: in a way, the vbios repo should list what every dev has
04:31mupuf: as in, devs SHOULD add their data in the vbios repo
04:31mupuf: but it is often not the case
04:31pmoreau: Oops... O:-)
04:31mupuf: which is sad :s
04:31pmoreau: I only added one card...
04:31mupuf: I try to add my cards as I go
04:32mupuf: I would think that the vbios repo would be perfect for what you want
04:32pmoreau: It's more to have an easy way to find out if someone has a specific chipset+VRAM, and ping that person to get some additional data/run some tests
04:32pmoreau: Could be
04:33mupuf: well, vbios repo would be perfect for that, and it is not limited t odevs
04:34pmoreau: Ok, I'll update it. :)
05:45infinity0: anyone seen this yet https://1vwjbxf1wko0yhnr.wordpress.com/2015/08/10/overclocking-tools-for-nvidia-gpus-suck-i-made-my-own/
05:45infinity0: it was hilarious to scroll down that secret API list, i thought it would end after 2-3 screen but it just kept going and going
05:46infinity0: wonder if it would help with the reclocking efforts
06:02mupuf: infinity0: it won't help
06:02mupuf: and we already had this information because nvidia-settings is open source :)
06:02infinity0: oh :(
06:02mupuf: fun little project though
06:05infinity0: mupuf: why won't this work? is it because it's the application/driver interface, i.e. nouveau would have to implement this API to make the tool work, but that's not relevant to the driver/hardware interface
06:08mupuf: you are definitely right
06:08mupuf: what we want is to understand the hw and how to program it
11:20ebrasca: how to configure multiple graphics cards?
14:24whoopi_cat: All, I have a card which identifies itself as "GK110GL" through lspci. I wanted to make sure that this is (or is still) one of the cards developers would like to see MmioTrace output from as described on the "Testers Wanted" page.
14:35RSpliet: whoopi_cat: let me see
14:36RSpliet: ah hmm, that... page is mildly outdated I think
14:38RSpliet: I don't think that's really necessary any more, but thanks for the offer
14:38RSpliet: the best way to contribute with that card is probably just using it and file bug reports (if they don't already exists) for those tasks that fail :-)
14:38RSpliet: because from what I can tell, nouveau should support it just fine (albeit a bit slow given reclocking is still a WIP)
14:51whoopi_cat: RSpliet: thanks. actually nouveau really runs the card nicely. We use a product from NASA called WorldWind, and it runs great with nouveau. That's what got me interested in helping out.
14:53imirkin: whoopi_cat: wouldn't it running *poorly* on nouveau be a better incentive to help? :)
14:54imirkin: whoopi_cat: anyways, i don't think we need GK110 traces, unless there's something especially messed up in yours
14:57whoopi_cat: imirkin: I see your point. Better perhaps to say that it's smooth running got me thinking about a few sysadmins' blythely (and annoyingly) installing the proprietary drivers without talking to the programming staff. ;->
14:58imirkin: well, the GK110GL is quite a beefy card
14:58imirkin: nouveau can't even clock the memory up to full speed, not to speak of the extra years of optimization the proprietary driver has had
14:58imirkin: but if you're happy with nouveau, then great ;)
15:00whoopi_cat: It's certainly a nice card. But I don't really get to push it (I can't game at work!) and our customers use a disparate hardware set. If we could semi-standardize on nouveau, it'd be nice.
15:01imirkin: well, the biggest issue is that nouveau is unlikely to be able to get that card's memory to maximum speed
15:01imirkin: if that's not a concern, then nouveau should be fine
15:03whoopi_cat: Sounds great. Thanks.
16:28psaisanas: RSpliet: this is in regards to NV47 on ppc64
16:28imirkin: psaisanas: reports of trouble with NV46 + MSI, wouldn't be surprised if NV47 also had issues
16:28imirkin: the trouble was with vblank i think
16:30psaisanas: imirkin: NV47 on x86 seems to work fine with MSI.
16:31imirkin: ok, well there's no great reason for MSI to not work, but i suspect there are some timing issues
16:31imirkin: wrt when it's rearmed, etc
16:33psaisanas: imirkin: with ppc64, msi interrupts seems to have drm gpu lockup. i.e. one and only one msi interrupt is generated, get logs "DRM GPU lockup - switching to software fbcon"
16:33imirkin: makes sense
16:33imirkin: i'd solve it by disabling msi and move on to the next show-stopping problem
16:33imirkin: psaisanas: btw, someone just pushed a patch that should fix the fbconfigs thing
16:34imirkin: i take that back
16:34imirkin: someone sent a patch. but not pushed yet.
16:34imirkin: psaisanas: http://patchwork.freedesktop.org/patch/56756/
16:37phillipsjk256: would it be possible to use the mimotrace tools developed for the nouveau project on ATI hardware? Apparently the Free driver still needs an ATI-supplied binary blob.
16:38imirkin: but mmiotrace won't help you with that blob
16:38imirkin: (beyond being able to figure out what the bytes of that blob are... but you already know, it's in your fs)
16:38imirkin: mmiotrace traces cpu <-> pci interactions (well, mmio reads/writes). that firmware blob runs *on the gpu*
16:38imirkin: (in the command processor)
16:39phillipsjk256: does nouveau use similar binary blobs?
16:39psaisanas: imirkin: strangely, switching to legacy interrupt seems to work fine, but seem to be getting spurious interrupts with legacy interrupts, could DRM vblank interrupt be a possible cause?
16:40imirkin: phillipsjk256: sure, there's the ctxsw firmware, video decoding firmwares, etc
16:41phillipsjk256: So I am not "more free" running nouveau over the radeon driver then?
16:41imirkin: phillipsjk256: there's open ctxsw firmware included in nouveau, but not video decoding, for which we rely on extracted firmware
16:41imirkin: phillipsjk256: no, you're still running on an intel or amd cpu, which has tons of microcode inside, not to mention the bios with its smi stuff, etc
16:42imirkin: under some *extremely* contrived definitions of 'free', nouveau is "more free" than radeon
16:42phillipsjk256: imirkin, I have mostly avoided such CPUs since 2006 ;)
16:42imirkin: errr... 2006 you already had p4 netburst no
16:43imirkin: which is heavily microcoded
16:45phillipsjk256: My Pentium II has microcode; introduced after the infamous Pentium bug.
16:46psaisanas: imirkin: mesa patch looks very interesting for big endian, thanks for the link.
16:46imirkin: the fdiv bug was on the original p60/p90, there were tons of pentiums after that but before p2
16:46imirkin: psaisanas: there's also been lots of general BE talk on mesa recently
16:47imirkin: psaisanas: i've merely pointed out that it's totally fubar'd for hw drivers afaik, but i have no direct experience with any hw. feel free to chime in with test results/etc
16:49psaisanas: imirkin: ill definately try it out with nouveau... i mentioned earlier regarding NV47 "strangely, switching to legacy interrupt seems to work fine, but seem to be getting spurious interrupts with legacy interrupts, could DRM vblank interrupt be a possible cause?"
16:49imirkin: yeah, no idea. "DRM vblank interrupt" isn't really a thing btw
16:50imirkin: interrupts come from the hardware
16:51imirkin: there might be an issue in "software" interrupt handling wrt msi in nouveau, dunno. (i.e. when there's a hw interrupt on the sw object)
16:57psaisanas: imirkin: true i agree. since nv47 seemed to work on x86, i sort of initially blamed the particular ppc64 hw platform. however, the gpu seemed to be set up as you would expect according to lspci. Also msi ints seemed to work for a pcie lan card that i placed in that particular pci bridge slot.... dunno, just trying to figure out where to start...
21:55Hoolootwo: I'm having issues with my NVS 3100M, where occasionally when I get back from work, X either is dead and shows nouveau errors on the screen, reboots completely, or restarts X to lightdm
21:55Hoolootwo: I can't find anything specific which always causes it to happen though
21:56Hoolootwo: The error message which I got in my dmesg is quite a few messages like "[352817.889798] nouveau E[ PFIFO][0000:01:00.0] CACHE_ERROR - ch 3 [Xorg] subc 3 mthd 0x1aa8 data 0x00d07aa8"
23:27karolherbst: so, now I am back :)