14:03imirkin: perhaps some day someone explain to me what the deal is with that thing
14:10hakzsam: mupuf, cool
14:11hakzsam: imirkin, a steam account for nouveau
14:11imirkin: hakzsam: the layout of that nve4 image info structure is not required by the hw, right?
14:11imirkin: i.e. i can change it and no one will care (assuming i fix it in both places)
14:11hakzsam: imirkin, you mean the surfaces info area in the driver cb?
14:11hakzsam: not required by the hw
14:12imirkin: k cool
14:12hakzsam: what do you want to change?
14:12hakzsam: I guess it's for ms images?
14:12imirkin: eh, dunno. probably not worth it.
14:13imirkin: but the number of samples
14:13imirkin: stored directly in that constbuf
14:13imirkin: isntead of having to add the x/y shifts, and do 1 << sum
14:13hakzsam: oh okay
14:13hakzsam: you will need to update the thing for fermi as well
14:14hakzsam: (if you change the way we store the surfaces info)
14:14imirkin: hakzsam: right. could you test that change on fermi btw? the -layer and -semantics tests are relevant
14:14hakzsam: imirkin, sure, I'll have a look and test later today (don't have a fermi here)
14:15imirkin: cool thanks
14:15imirkin: as for the steam account ... what are the implications of sharing an account?
14:15imirkin: i.e. why isn't there just one steam account, "the_world", that everyone uses?
14:15hakzsam: no idea
14:17hakzsam: but sharing the same account for a lot of people is not really good because only one can use it at the same time
14:17hakzsam: I have never used that steam nouveau account though
14:17imirkin: ok, so that's an implication of sharing :)
14:17imirkin: how is "use" defined?
14:17hakzsam: usually I do use mupuf's one :)
14:17hakzsam: connected I think
14:17imirkin: so like "logged into the steam client"?
14:18hakzsam: I think but I'm not sure
14:18mupuf: imirkin: 2 people cannot game at the same time with the same account
14:19hakzsam: mupuf, but they can be logged?
14:19mupuf: and we can only have up to 10 computers using the account ... but you are free to delete some computers
14:19mupuf: hakzsam: sure
14:19mupuf: we discussed this with karol before setting it up
14:19hakzsam: well, 10 is large enough for nouveau developers :)
14:20imirkin: "we"? i def know i wasn't part of any such discussion...
14:20mupuf: karol and I
14:21mupuf: that was because feral was asking for an account and karol and I thouight that it was a bad idea to share his personal account
14:22imirkin: ok ... sounds like there's more going on than i'm aware of. happy to stay ignorant. never mind that i asked.
14:22mupuf: imirkin: I am however surprised you were never given the credentials
14:23imirkin: perhaps i was and forgot?
14:23mupuf: let me fix this is in private
14:23imirkin: nah, don't worry about it
14:23imirkin: if i really want to run some game, i'll get the developer to give me a key
14:24mupuf: well, the account already has some games, so you may not have to
14:24imirkin: [it's also rare that i want to run some particular game]
14:24mupuf: same here
14:25mupuf: with that said, I am back from "vacation"
14:25hakzsam: welcome back :)
21:42hakzsam: imirkin_, tested and replied to your MS images patch
21:42imirkin_: cool thanks
21:45hakzsam: [assuming that you will fix your oops]
21:46imirkin_: should be fairly easy
21:47hakzsam: yeah, in handleSUQ() I assume?
22:13hakzsam: imirkin_, you might also want to fix up the out of bound access in adjustCoordinatesMS()
22:14imirkin_: hakzsam: that thought had crossed my mind ;)
22:14hakzsam: and hope that tex.r == 0 :)
22:17hakzsam: well, I'm off, see you
23:15Prince781: So I'm running nouveau with a GT 750M (discrete card). I'm trying reclocking on kernel 4.6. When I try to change pstate from sysfs, certain things hangup and then dmesg shows a backtrace with the error "unable to find matching pll values"
23:15Prince781: Any idea why this is happening?
23:16imirkin: what gpu is that? GK107?
23:16imirkin: GDDR5 vram? or DDR3?
23:17imirkin: actually irrespective of the answer to that question
23:17Prince781: Not sure. Think it might be gddr5 but how can I tell?
23:17imirkin: file a bug at bugs.freedesktop.org xorg -> Driver/nouveau
23:17imirkin: and include a full dmesg (up to and including the error)
23:17imirkin: as well as you vbios (/sys/kernel/debug/dri/0/vbios.rom)
23:17imirkin: er, probably /1/... for you
23:18imirkin: more questions and/or suggestions are likely to follow from people more in tune with this whole situation
23:19imirkin: btw, pstate is in debugfs in kernel 4.6 - i assume that was a typo?
23:19Prince781: it's debugfs
23:20imirkin: there's a set of patches which might help you
23:20imirkin: if you're up for compiling stuff
23:20imirkin: but i dunno if it'll address your specific issue
23:20Prince781: It's worth a try I suppose
23:20imirkin: check that out, cd drm; make
23:21imirkin: that will build an out-of-tree kernel module
23:21imirkin: which you'll need to copy over into your regular setup
23:23imirkin: [make sure you get the branch i pointed you at]
23:23Prince781: yup, stable_reclocking_kepler_v5
23:23Prince781: this isn't in 4.6?
23:25imirkin: there's like 30 patches in there
23:26imirkin: skeggsb: fyi, your email appears to have gotten associated with 'theif' on github
23:26imirkin: skeggsb: or at least that's the simplest explanation to what github is showing right now
23:27skeggsb: imirkin: ?!
23:27imirkin: skeggsb: https://github.com/skeggsb/nouveau/commits/master
23:27imirkin: i see all the commits as being done by 'theif'
23:29imirkin: [do you not?]
23:29skeggsb: no, i see it
23:29imirkin: i leave it in your capable hands
23:29imirkin: [good thing you moved off of fd.o...]
23:32Prince781: So I ran make and then sudo make install. It put nouveau.ko.gz in /usr/lib/modules/`uname -r`/extra
23:33imirkin: make install rarely does anything good
23:33imirkin: definitely not a good idea to run in conjunction with sudo without first carefully checking wtf it all does [general advice, not nouveau-specific]
23:34RSpliet: skeggsb: speaking of: is there a public document explaining the workaround for DVI on G9x/MCP7x?
23:35skeggsb: RSpliet: that version of the commit was a quick hack so i can backport it, i'll do a proper version (including the comments from nv) when i get a chance to fix a heap of other stuff there
23:35RSpliet: skeggsb: oh that's okay, just wondering whether there was a github issue or open-gpu-doc document associated with it, or whether it was private communication :-)
23:35skeggsb: they provided the code from RM to handle it, under MIT license
23:36RSpliet: does that contain other information on calculating mode parameters (like a less pessimistic vblank time)?
23:36skeggsb: no, it's just the code that does basically exactly what i just pushed
23:38skeggsb: i figured out half the fix myself, but apparently the full workaround was somewhat more extensive :)
23:38skeggsb: so, pretty nice of them to provide it