00:00 imirkin: this would narrow down the list of potential sources
00:00 imirkin: but would be a lot of compilation/testing for you -- probably 20 kernels to build
00:00 imirkin: [and test]
00:00 kernel-3xp: like 13 yeah
00:02 kernel-3xp: then i make my own patches?
00:04 imirkin: hm?
00:04 imirkin: well just identifying the commit would be revealing
00:04 imirkin: it could be, like you say, some behavioral change related to displaying frames
00:04 imirkin: or it could be something totally unrelated which added or removed 1ns of latency somewhere
00:04 imirkin: thus jiggering the whole system
00:05 kernel-3xp: no it was exactly 60 fps
00:05 imirkin: yeah i get that
00:05 kernel-3xp: and vblank thing could disable from glxgears
00:05 imirkin: but if e.g. previously it always JUST MISSED something, vs now it doesn't, etc
00:05 imirkin: timing can be weird.
00:05 imirkin: anyways - more likely than not - it's some behavioral change
00:06 imirkin: but without knowing the commit - hard to say
00:06 kernel-3xp: vsync was on, everything i tried stable 60 fps and mouse lag everywhere
00:09 imirkin: probably hasn't hit you, but i've also seen some *weird shit* relating to vsync every so often where it feels like the vsync counter stops and everything that's supposed to be vsync'd just stops displaying
00:09 imirkin: it seems to only happen after a few months of uptime
00:09 imirkin: i think some counter overflows
00:09 imirkin: or something
00:09 imirkin: but then it weirdly starts working
00:09 kernel-3xp: if it was just timing, should be all over the place on different workloads, also fresh reboot
00:10 imirkin: kernel-3xp: without a specific commit, you're unlikely to be able to track it down
00:10 imirkin: but good luck!
00:10 kernel-3xp: yeah, thx :)
00:38 Lyude: Someone mind approving the email I just sent to nouveau@lists.freedesktop.org? just got rejected because cc_cmd added too many people to it
00:39 imirkin: i don't think anyone here can approve
00:39 imirkin: maybe airlied
00:39 imirkin: wow, that's a long list.
00:40 Lyude: yeah-it touched every driver using MST so
09:22 sigod: karolherbst, are you there?
09:27 sigod: anyone know about the status of the nouveau firmware for NVE0?
10:14 pendingchaos: karolherbst, imirkin: does this change look good: https://lists.freedesktop.org/archives/mesa-dev/2018-August/203202.html ?
12:38 imirkin: pendingchaos: you can just push those without rb
12:38 imirkin: pendingchaos: but you can have my r-b if you like
12:38 pendingchaos: I wasn't sure about the GF100+ thing
12:38 imirkin: i'm not either ;)
12:38 pendingchaos: also patches for mesa-stable need to R-b
12:39 imirkin: iirc kepler has them
12:39 imirkin: not sure about fermi
12:39 imirkin: should be easy to check the code
12:40 pendingchaos: IIRC, after looking at the code, Fermi does support ARB_shader_image_load_store
12:40 pendingchaos: though I wanted to be a bit more sure before claiming support in the release notes
12:42 imirkin: it definitely supports images
12:42 imirkin: question is whether it supports MS images
12:42 pendingchaos: I don't see anything in nvc0_screen_is_format_supported that would suggest otherwise
12:44 imirkin: then it is :)
12:45 pendingchaos: I'll ask tanty to push it to the 18.2 stable branch then
13:24 imirkin: skeggsb: might want to fix those s-o-b's :)
13:26 skeggsb: imirkin: didn't mean to push those patches yet :P still had more to pull
13:27 skeggsb: i'll fix them before i send them somewhere else
13:28 imirkin: ah
13:30 karolherbst: skeggsb: nice for fixing the stuff. Currently I am a bit playing around by disabling all subdevs/engines and I get all kind of funny errors
13:30 karolherbst: and crashes
13:33 imirkin: skeggsb: btw, dunno if you saw, but Riastradh is trying to bring up more modern nouveau on netbsd, and is running into display issues
13:33 imirkin: i tried to look, but i just don't know enough about it
13:34 imirkin: perhaps you can have a gander
13:41 karolherbst: skeggsb, Lyude: ha! this "fixes" runpm: https://github.com/karolherbst/nouveau/commit/eb65ca81033764cdfe60efb530f1ac95e881ac84
13:41 karolherbst: at least it is good to know that nouveau itself doesn't do stuff even without any engines/subdevs
13:42 karolherbst: hopefully I will be able to enable them one by one to get to the root cause
13:43 karolherbst: I guess I can remove those other NV_ERROR replacements though
14:02 Lyude: pendingchaos: do you need something tested on Fermi?
14:03 pendingchaos: I guess running glxinfo | grep "GL_ARB_image_load_store" would be nice if it's not too difficult
14:03 pendingchaos: though that doesn't test for multisampled images
14:03 pendingchaos: I think there's a piglit test that would show whether they are supported or not
14:03 pendingchaos:looks
14:05 pendingchaos: piglit's arb_shader_image_load_store-max-size test would show it
14:05 pendingchaos: if it runs a "image2DMS max size test/..." then it supports multisampled images
14:15 Lyude: pendingchaos: if you are around in a few hours I can test it when I get in the office, I've got some Fermi cards
14:15 pendingchaos: I should be around
14:15 Lyude: Cool! I'll set a reminder for when I get in then
14:15 pendingchaos: it's not that important btw
14:17 pendingchaos: you don't have to do it if it's not already set up
15:08 sigod_: so can anyone tell me about the status of "open source" firmware for kepler?
16:30 adya: Hi imirkin, nikk_ and I were working on the fbo-blending-formats task from trello board. sorry for the very late response, we had our exams going on. We have shared our piglit results in a doc : https://docs.google.com/document/d/19ifBXylDF0LyrzoB06rQ7gsBqkida08__aZqVEZUNV0/edit?usp=sharing We are unable to find 10X2 results in this . What would you suggest? Also we have thought of a line up of code in state.c file, as the else cond
16:39 adya: We have thought of a line up of code in the state.c file, as the else condition corresponding to if(blend_en), the pseudo code of which is also in the doc. Please guide us on these. Thanks!
17:11 Lyude: Someone mind unblocking the messages I sent to the nouveal ml again?
17:47 Lyude: pendingchaos: almost forgot to ask! you can give me whatever you needed to get tested now
18:07 pendingchaos: Lyude: the master branch of mesa3d should be fine
18:10 Lyude: pendingchaos: GF110 good?
18:12 pendingchaos: yeah, should be
18:12 Lyude: right, setting up now
18:23 Lyude: pendingchaos: got local mesa built-what do you need to test?
18:23 pendingchaos: running piglit's arb_shader_image_load_store-max-size test and seeing if a image2DSMS subtest is run
18:33 Lyude: pendingchaos: https://paste.fedoraproject.org/paste/hT3n4JLBqcXpXBH4GrU0Rg
18:33 pendingchaos: thanks!
18:34 Lyude: np
20:33 Riastradh: ...This can't be right:
20:33 Riastradh: chan = rcu_dereference_protected(fence->channel, lockdep_is_held(&fctx->lock));
20:33 Riastradh: if (nouveau_fence_update(fence->channel, fctx))
21:07 Riastradh: Oh, I see, rcu_dereference_protected, so it is safe here since we're under the writer lock.