00:08 imirkin: tobijk: oh, yeah, qt + kde + whatever are fucked on nouveau
00:09 imirkin: wow, it recovers though
00:09 imirkin: or at least thinks it does
00:27 tobijk: imirkin: yeah it kinda did this time: the mouse cursor was movable, yet everything else gone
00:27 tobijk: so only half broken :D
04:45 gnurou: imirkin: mmm that one is a bit confusing, especially since that we did quite a few other reads before that one
04:46 gnurou: why would this one timeout and not the others
04:47 imirkin: coz it's coming out of resume?
04:47 imirkin: this was probably also the first time GR was being initialized
04:47 imirkin: [this was on an optimus laptop]
05:25 RAOF: Huh.
05:26 RAOF: Is nvbo->valid_domains means to be the set of domains that a bo can be migrated to?
05:26 RAOF: And if so, why does nouveau_bo_pin (or, indeed, anything in nouveau_bo) not check it when trying to pin to VRAM?
05:26 RAOF: (Or indeed at all)
09:44 pmoreau: RSpliet: Where do your patches live? I could have a try on the GFxxxs I have. I need to get a MMIO-trace for karolherbst as well.
11:34 dboyan: karolherbst: Things goes really funny with hitman on my laptop. It is crashing seems even only with tgsi cache on your "hitman" branch.
11:34 dboyan: karolherbst: It is showing artifacts sometimes with cache disabled
12:01 karolherbst: dboyan: do you use my "hitman" branch?
12:03 dboyan: karolherbst: yes, exactly your branch
12:03 karolherbst: mhh odd
12:04 karolherbst: is it fine if you clear the cache and disable it?
12:04 karolherbst: and yes, I am aware that there are still some crashes, it needs a lot of work still
12:04 karolherbst: dboyan: which GPU by the way? and is your dmesg filled with OOR_ADDR errors?
12:05 dboyan: karolherbst: GT 740M, and I also see OOR_ADDR errors
12:05 karolherbst: k
12:05 karolherbst: it's kepler, I thought you had a maxwell or so
12:05 karolherbst: dboyan: but you also notice the much better loading time?
12:07 dboyan: yeah, loading time is much better with cache on
12:08 karolherbst: could you reproduce the crash I got with your branch?
12:09 karolherbst: by the way: currently I try to fix that OOR_ADDR error and got no luck finding what really causes that. I have some clues, but couldn't find the cause
12:10 dboyan: well, I'll make a debug build and run it again to see how it crashes. But apparently I don't have the time today, maybe during the weekends
12:11 karolherbst: imirkin: ohhh maybe it is something silly like this: I think they actually use threading to offload certain things and maybe something just goes out of order and nouveau can't handle it. Like c2 is mapped inside g and then another buffer gets remaped and things go south? I am not good with the details, but maybe it is something silly like this
12:11 karolherbst: dboyan: also if you want to see the crash, you need to run the game through gdb, otherwise the game crash handler will catch it
12:12 dboyan: karolherbst: How do I run it through gdb?
12:12 karolherbst: dboyan: from command line
12:12 karolherbst: just go inside the game folder/bin
12:12 karolherbst: and launch the executable
12:12 karolherbst: HitmanPro
12:12 dboyan: okay, this seems really easy
12:13 karolherbst: yeah
12:13 karolherbst: some games are smart and let you run them through cli
14:26 imirkin: RAOF: those domains are a little fudgeable, and imho, misdesigned.
14:26 imirkin: RAOF: i think they made more sense in the pre-nv50 days
14:26 imirkin: RAOF: but it basically lets userspace allocate a "vram" or "gart" buffer
14:26 imirkin: and ensuring that ttm only ever places it in those (or sysmem)
14:43 dboyan: imirkin: I'm writing the gsoc project proposal, but I'm a little bit unsure about the "project timeline" section, mainly about how I should allocate time for different policies, without full knowledge of their details
14:48 karolherbst: afaik it doesn't matter, just assume you need all the time
14:48 karolherbst: and if it happens that you need more, your mentor writes something, why it takes longer aka why the scoped has to be reduced
14:48 karolherbst: *scope
14:49 karolherbst: just do a timeline which looks like it makes sense to you and your mentor (ilia I assume?) and then it's fine
14:55 dboyan: okay, i think i'll submit a draft version this weekend and wait for comments then
14:55 imirkin: dboyan: if you could find out what i have to do, if anything, that'd probably be good too :)
15:00 dboyan: imirkin: how much time do you think we want to devote to pre-RA and post-RA scheduling respectively?
15:00 dboyan: I assume they use different policies are somewhat independent
15:21 karolherbst: imirkin: you mainly have to write reports :p
15:22 karolherbst: imirkin: I am quite sure that as a mentor you would have to write the first report in the first third of the gsoc, so that the student gets money. Something like this
15:33 mupuf: karolherbst: there are two reports now
15:33 mupuf: but they take no more than 10 minutes to fill out
15:33 mupuf: so, the cost really is on the mentoring side
15:33 karolherbst: oh, so there are templates
15:34 mupuf: yes, there are questions you need to answer
15:34 karolherbst: yeah, I was refering to the mentor side anyway
15:34 mupuf: like, how much did you spend on gsoc, how often do you interact
15:34 mupuf: how happy are you with the progress
15:34 karolherbst: sure
15:34 mupuf: and whether you pass the student or not
15:34 karolherbst: but ilia asked what he needs to do, so ;)
15:35 karolherbst: ohh, now I understand
15:35 karolherbst: k
15:35 karolherbst: yeah I see
15:35 mupuf: good
15:36 mupuf: so yeah, as I said, the cost is really on how much time you spend with ythe student
15:37 mupuf: we have been trying to get students to interact more with the community, to spread the load
15:37 mupuf: the mentor is of course responsible for the student though
15:37 mupuf: students will have to write weekly reports though!
15:38 karolherbst: :O
15:38 karolherbst: 12 reports then
15:39 karolherbst: it's 3 months, right? or did I missunderstood anything
15:45 mupuf: yes, 3 months
15:45 mupuf: karolherbst: yes, 12 reports
15:45 mupuf: so as we can see if they are stuck
15:45 mupuf: because these damn students do not like telling they are stuck!
15:45 mupuf: :D
15:45 karolherbst: :D
15:46 mupuf: think about me though, I will have to make 12 reports .... to summarize what the students are doing
16:33 karolherbst: mupuf: :O
16:33 karolherbst: mupuf: but I am sure you are allowed do to it at your working time?
17:08 mupuf: well, you would not be entirely right. But yeah, Intel values this work. Less of course if there are no projects that would benefit intel, but that's the game!
17:33 imirkin_: dboyan_: pre-ra: 100%. post-ra: 0%
17:34 imirkin_: mupuf: do i have to do something to register as a mentor though? or does dboyan_ just name me as a mentor and all's well?
17:34 mupuf: imirkin_: give me your email address and you'll be ready
17:34 mupuf: has to be gmail
17:34 mupuf: :s
17:35 imirkin_: see pm.
17:37 mupuf: ok, you should be set now :)
17:38 pmoreau: Wait, you have to have a gmail address to be a mentor? O.O
17:38 imirkin_: received.
17:39 imirkin_: probably a google account, not necessary gmail
17:39 pmoreau: Kinda makes sense since Google is organising it
17:44 imirkin_: yay. i am now a mentor.
17:46 pmoreau: Congrats! :-)
17:46 nyef: What projects are being considered for GSoC this year?
17:46 imirkin_: nyef: whatever students propose
17:47 nyef: And I ask more out of curiousity than any other reason: I'm not about to ask to be a mentor, wouldn't expect to be accepted if I did, and while I study things I don't fit the usual definition of a "student". (-:
17:47 imirkin_: i think 'student' means you are a full time student in an accredited institution of higher learning
17:48 nyef: See?
17:49 imirkin_: Were you ever convicted of a felony, misdemeanor, that sort of thing?; Convicted? No, never convicted. https://www.youtube.com/watch?v=aOnHi7KOxCA
17:49 nyef: Heh.
18:00 mupuf: imirkin_: welcome to the program ;)
22:09 random_dousche: any status on working gfx 1050