02:36RSpliet: pmoreau: I probably did have... but I forgot :-D
05:06Cloudef: I have program that seems to crash on nouveau but works with llvmpipe. It worked before however, should I apitrace it or is there more useful dumps?
05:11Cloudef: Of course
05:11Cloudef: I'll do that when I get home
06:10imirkin: gnurou: ah ok, thanks for the update. there is some render fail in mesa on maxwell that i was semi-hoping you guys would take care of :)
06:23hakzsam: mlankhorst, Hi, you have a nvd7?
06:52imirkin: Cloudef: also if you're using libdrm-2.4.60, that's known to introduce some issues for nouveau -- downgrade to 2.4.59
06:57Cloudef: imirkin: that could be it
07:24mlankhorst: hakzsam: no longer
07:24mlankhorst: I had to return it. :(
07:25Cloudef: imirkin: indeed, downgrading libdrm seems to make it work
07:30hakzsam: mlankhorst, bad news :/
07:30hakzsam: does someone else have a nvd7? :)
07:36mlankhorst: i have a d9 though, why do you ask?
07:41hakzsam: I have d9 too
07:41hakzsam: because I need to trace the blob for performance counters
07:41hakzsam: I traced all Fermi except c3 and d7
07:49imirkin_: gnurou: any bets on the chances of me getting a reply to a question about a weird tessellation failure on the gpu doc list? basically some tris occasionally don't get filled, even though it should be perfectly deterministic
07:49imirkin_: gnurou: it'd take some time to put a cogent question together, and i don't want to go to the effort if the list is abandoned
07:54mlankhorst: hakzsam: sadly I don't have anything :(
09:39imirkin_: fun fact -- tess totally doesn't work on GK208 :(
09:42imirkin_: ah. that might be my bad.
09:52imirkin_: turns out the the "patch" and "output" flags aren't just for show, we have to actually emit them :)
10:14imirkin_: gnurou: fwiw these are the tess fails i'm talking about: http://imgur.com/DqJ82Kd,iw2icq3#1
10:34imirkin_: anyone here have bioshock infinity which requires GL 4.1?
10:34imirkin_: [and a fermi or later GPU]
10:34imirkin_: if so, would be very interested to know what happens with https://github.com/imirkin/mesa/commits/gl4-integration
11:21pmoreau: RSpliet: Maybe you did, but I can't see any new commits in your tree. ;)
11:21pmoreau: I'll give it a try right now
11:59imirkin_: hm interesting. the tess fails are accompanied by nouveau E[ PGRAPH][0000:01:00.0] GPC0/TPC1/MP trap: MEM_OUT_OF_BOUNDS
11:59imirkin_: so i must not be setting something somewhere...
12:01imirkin_: which quads (and heaven) triggers, but nop and sanity don't...
12:03imirkin_: one big difference is that # patch vertices != # of tcs output vertices... probably related
13:00pmoreau: RSpliet: Meh... Are you sure you don't want to give nv50_ram_gpio a pointer to pfb? That would be way easier!!
13:11RSpliet: pmoreau: i don't know... what you see is a copy-paste booboo
13:24voxadam: Nouveau is going to be the death of me.
13:25xexaxo: voxadam: you're welcome :)
13:26imirkin_: voxadam: downgrade to libdrm 2.4.59
13:26imirkin_: i bet you hvae 2.4.60
13:28voxadam: Crap. I had downgraded but apparently I forgot to lock the version.
13:31hakzsam: imirkin_, I think you need to add this GM108 chipset to the CodeNames page, right?
13:31imirkin_: dunno about *need*
13:31imirkin_: but i guess i should
13:38imirkin_: hakzsam: how's that
13:47voxadam: imirkin_: How do I verify which version of libdrm is running?
13:48imirkin_: dunno. the way i verify it is by running 'equery l libdrm'
13:49imirkin_: on the off chance that you're not using gentoo, it's likely that whatever package management system your distro uses also provides this sort of functionality
13:51voxadam: Well, I grabbed 2.4.59-4.fc22 from koji, installed it manually, and excluded it from my package management system. I'll just trust myself.
13:51voxadam: Not always the best idea.
13:53RSpliet: there exists such a thing as
13:53RSpliet: "yum downgrade"
13:59voxadam: RSpliet: I'm using dnf on F22. For some reason it refused to pickup the old package. I just went ahead and grabbed 2.4.59 of of Koji, installed it manually, and excluded it in dnf.conf.
13:59voxadam: That being said, after rebooting it only took a few minutes to trigger a GPU lockup.
13:59imirkin_: with 2.4.59?
14:00imirkin_: same messages in dmesg?
14:01voxadam: No messages of use, just 'nouveau E[ DRM] GPU lockup - switching to software fbcon'
14:03imirkin_: "you might not have noticed that your screen stopped updating, so just fyi, your screen has stopped updating"
14:04voxadam: Hey, more than half the time it doesn't even let me know that, it just locks.
14:07voxadam: I find it odd that while it's locked the screen does eventually blank. Not only that but it will turn back on if I hit a key.
14:08imirkin_: dpms is separate
14:08imirkin_: sounds like fb updates get broken
14:08imirkin_: but everything else is fine
14:38voxadam: I'm buying an AMD card next time.
14:43a1fa: dont bother
14:44a1fa: i think intel is where its at
14:55voxadam: If only Intel made discreet GPUs.
14:55imirkin_: they did
14:55imirkin_: Intel i740 iirc
14:55voxadam: er i740.
14:56imirkin_: AGP and everything :)
14:56voxadam: I'd have to devise some sort of PCIe to AGP hack.
14:56voxadam: Not that AGP wasn't a hack unto itself.
15:05glennk:is reminded of VLB and its HIF (hammer insertion force) sockets
15:06imirkin_: he he
15:07imirkin_: perhaps you'll bring up MCA as well?
15:08voxadam: Let's bring back VESA Local Bus.
15:09voxadam: VLB was a fscking pain.
19:50gnurou: imirkin_: if things are broken beyond basic functionality, then chances are that we will get involved (at least I will push in that direction)
19:51gnurou: imirkin_: regarding questions, I believe any question is acceptable :)
19:52gnurou: the email channel is not ideal for these though, so I am thinking of opening a new one... let's see if the idea sinks internally first
19:52gnurou: well email *is* ideal for this actually, but the problem is that engineers don't have time dedicated for this
19:56imirkin: gnurou: i think a bug tracker or stack overflow style thing would be the ideal channel