01:10dboyan_: imirkin, I heard you talking about using gnurou's tree. I assume his nouveau repo is directly buildable out of kernel's source tree?
01:33gnurou: dboyan_: nope - you want to use https://github.com/Gnurou/nouveau/tree/secboot
01:34gnurou: dboyan_: and build against a compatible kernel, i.e. 4.10
01:34dboyan_: gnurou: Thanks, got it
01:35gnurou: dboyan_: also take the firmware from https://github.com/Gnurou/linux-firmware/tree/staging
03:02nyef: Damn. Looks like I ended up with a Kepler, not a Maxwell.
03:02nyef: Still, an apparently-working discrete card for that machine isn't bad.
03:03imirkin: the 860M?
03:03nyef: Yeah, the 860M.
03:03nyef: GPU database claims that the Kepler version is a PCIe x16, while the Maxwell version is an MXM card.
03:03nyef: So, of course, I end up with a Kepler MXM card.
03:04imirkin: that'll teach you to believe databases
03:04imirkin: hope you're not blowing all your allowance on this :)
03:04nyef: No, I'm still good on the financial front, and the machine needed a new discrete card anyway, so it's more a disappointment than an actual problem.
11:31dpdocs21: imirkin: Turns out that lab system is not in woking condition currently. So currently, I don't have access of any such GPU. I was assuming the availability of that system while considering this poject.
14:55dboyan: imirkin, I've run through some test of my rcp and rsq. After a correction to rcp, I never saw error exceeding 2ulp with more than 650 million random attempts (plus some manually selected value) each.
14:55dboyan: of course I'm comparing the result with cpu
14:55dboyan: that was about 2 hours of running
14:57dboyan: imirkin_: https://github.com/dboyan/mesa/tree/fp64-rcprsq2
16:51renegadevi: Does anyone encounter this glichy problem? It's been a issue with nouveau for a long time, but I realized that it only happens whenever Chromium or chromium-based software is running. http://i.imgur.com/g6kFyd0.png
16:52renegadevi: for the last year I've just installed the official nvidia driver to solve the issue, but then I'm stuck with Xorg for now.
17:06martm: who are you threattening imirkin_ with your biscuit? because i have not wanting to come into that channel of morans!
17:06karolherbst: renegadevi: you could create a bug report
17:07martm: i just am just looking into basically silently into carry theory, and there appears to be a hack, when you look at weiner martin or someones code, there is a parallel carry hack
17:08martm: it's an isling physics random number generator theory, that based denomstrates this possibility in high level code
17:08martm: sorry ising
17:09martm: i am entirely drunk again
17:09karolherbst: renegadevi: I could imagine you have an old version of mesa or the DDX installed, which could cause this
17:09martm: i basically discovered alcohol agian
17:10renegadevi: it's a new fresh installation of Fedora 25 with Chromium and updated to latest..
17:10renegadevi: but that's been a issue in other distros as well
17:11karolherbst: renegadevi: what gpu?
17:12renegadevi: It's a Quadro, and I think that could be a big reason
17:12karolherbst: renegadevi: but distributions also ship sometimes outdated stuff
17:12karolherbst: doesn't matter if this is a quadro usually
17:13karolherbst: or at least it shouldn't regarding issues like this
17:14martm: i stared this high level ising monte carlo code for like two days, cause i was so cracked up, but i think i finally understood it, it's slim but entirely complex algorithm
17:19martm: the basic understanding that i learned was , that when you do a carry out inside a loop, then the following instruction just does the assignment via inverting substracting da stuff
17:24martm: it just does a parallel carry, algo in high level is a loop via overflow mask, which carries out, next instruction is is addition and the one that follows addition is subtraction
18:46martm: yes the loop will be unrolled , most case at least, that implies that you have some static bounds of the loop though which is kinda making sense most of the time
18:47martm: so the assembly is looking to be quite primitive once unrolled, but it was a headache to read to be honest
18:57martm: michael i love you, if that is michael maze...but no bscuit for retarded michael long
18:59mwk: pq xexaxo1 RSpliet marcheu skeggsb mupuf imirkin_ ?
19:00mwk: or do you have an ancient /ignore going on?
19:01imirkin_: apparently i do :)
19:04mwk: another two proxies...
19:05mwk: we're 2/3 of the way to the ban list capacity limit
19:05imirkin_: super. might have to ban web proxies in general =/
19:13mupuf:wonders why he feels the need to come to this particular channel all the time :s
19:13mwk: he also spammed ##asm for a while in the last month
19:14mwk: I kept banning the proxies and people finally convinced me to just +q $~a (ie. only registered users can speak)
19:14mwk: somehow this worked
19:15mwk: even though he's been known to register accounts before
19:15mupuf: I guess, at some point, registering new emails gets tiresome
19:15mwk: you'd think finding proxies would be the same...
19:19airlied: I think his rewards centers are wired different :-P
19:34karolherbst: mupuf: are you home?
19:35mupuf: not yet, why?
19:35karolherbst: I need a fermi GPU in reator
19:35mupuf: cool! Will come back home soon-ish, couple of hours max
19:35karolherbst: then I am in bed already
19:36karolherbst: you can swap out the gm107
19:37mupuf: ok, will do!
19:43karolherbst: wow, nice. My reclocking update stuff is working now, aka fixing reclocking while suspended and restoring clock levels!
19:43mupuf: sounds great!
19:44karolherbst: airlied, Lekensteyn: thanks for your pointers, it helped :)
19:52Lekensteyn: karolherbst: is that about the pm_runtime_get thingey? (what did you use in the end?)
19:53karolherbst: now I only have the problem that I need to set a flag for the task the worker thread does
23:46ssdh: imirkin_: does the GM107 and GM108 also have signature enforcement? Someone told me that those didnt require a signature verification
23:47imirkin_: that's right - they do not.
23:51ssdh: got it: https://phoronix.com/scan.php?page=news_item&px=GTX-750-Nouveau-Linux-4.1-Try