#panfrost on 2023-12-24 — irc logs at oftc.irclog.whitequark.org

2023-04-06 13:40 alyssa changed the topic of #panfrost to: Panfrost - FLOSS Mali Midgard + Bifrost + Valhall - Logs https://oftc.irclog.whitequark.org/panfrost

00:20 Leopold_ has quit [Remote host closed the connection]

01:56 Leopold_ has joined #panfrost

02:14 hanetzer has quit [Read error: Connection reset by peer]

08:17 <urja> Umm, i have a question ... kicad these days (i think it's because it was built with wxWidgets EGL now) is asking for EGL version 1.5 or else you get the software renderer

08:18 <urja> reading through https://gitlab.freedesktop.org/mesa/mesa/-/blob/main/src/egl/main/eglapi.c?ref_type=heads#L636 and comparing to an eglinfo output on my C201 says the problem is ...

08:18 <urja> a lack of EGL_EXT_create_context_robustness

08:18 <urja> is that a hard one to have or just missed? (or something that panfrost/T760 fundamentally cant have?)

08:21 <urja> (unfortunately there doesnt seem to be a "just fudge it" env var for the EGL version lol... because, i bet kicad is not asking for that one to be able to run...)

09:43 <urja> ... apparently that boils down to having PIPE_CAP_DEVICE_RESET_STATUS & get_device_reset_status

09:57 <urja> which just involves knowing if the GPU was reset (or just claim innocent perfection like nvc0 lol)

10:05 warpme has joined #panfrost

10:53 Googulator has quit [Remote host closed the connection]

10:54 Googulator has joined #panfrost

11:50 pH5 has joined #panfrost

12:03 Leopold_ has quit [Remote host closed the connection]

12:08 tchebb has quit [Ping timeout: 480 seconds]

12:26 warpme has quit []

12:31 warpme has joined #panfrost

13:26 <urja> Yeah, a dummy impl makes kicad just work(TM)

13:29 <urja> I think i'll make a PR with that - not sure if it honestly is worth the effort to pipe the reset info (and like, I'm not sure exactly what kind of reset we should report) ... and it's a kernel-related thing so I'm even less interested :P

13:29 <urja> *MR :P

13:38 <urja> Oh yeah, one of the example commit summaries in the mesa Submitting Patches do is ...

13:38 <urja> gallium: add PIPE_CAP_DEVICE_RESET_STATUS_QUERY

13:38 <urja> I wondered why my search results for it were littered with that document (all over the internet) :P

13:57 Leopold has joined #panfrost

14:12 <urja> https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/26811 lol

15:02 warpme has quit []

15:20 Leopold has quit [Remote host closed the connection]

15:20 Leopold__ has joined #panfrost

16:01 Leopold__ has quit [Remote host closed the connection]

16:23 rasterman has joined #panfrost

16:42 Googulator has quit [Remote host closed the connection]

16:55 Googulator has joined #panfrost

17:03 Leopold_ has joined #panfrost

17:11 camus has quit []

17:16 tchebb has joined #panfrost

17:28 Googulator has quit [Remote host closed the connection]

17:29 Googulator has joined #panfrost

17:31 tchebb_ has joined #panfrost

17:35 tchebb has quit [Ping timeout: 480 seconds]

18:19 Leopold_ has quit [Remote host closed the connection]

18:31 tchebb has joined #panfrost

18:36 tchebb_ has quit [Ping timeout: 480 seconds]

19:11 cphealy has joined #panfrost

19:13 <robclark> urja: the drm driver should have (maybe not implemented yet for panfrost) some uabi to expose reset count

19:15 <urja> yeah i didn't find anything super obvious (freedreno had a parameter request ioctl of some sort, didn't find the same for panfrost...)

19:18 <urja> (i mean, panfrost has a similar parameter request but that's only for GPU ID and features etc, not reset count ... and i'm not even sure if that's the best way to do it?)

19:18 <robclark> maybe look at some of the other drivers.. in the end you want a way to query # of global faults and # of context specific faults (although you could do a bare metal robustness impl with just # of global faults)

19:18 <robclark> maybe extend the existing param request?

19:21 <urja> I did figure from my skim of the kernel driver that the kernel side would somehow report failed jobs (which is what happens if it does a fault), but I'm not enough of a drm developer to even be sure (fences? signalling? uh)

19:24 <urja> but yeah my reading of the other drivers just said that their implementation also varies (from a stub, to just reporting _some kind of fault_, to having the two counters...)

19:27 <urja> But yeah, if i need to add something to the kernel uabi, well, my C201 will likely never see that (unless someone sees it worthy of a backport to 6.1 ... i mean, i could do that to my kernel, but)

19:28 <urja> (6.2 didn't boot so i told my kernel autobuilder to instead follow the 6.1 LTS from then on; the laptop will likely be dead before it runs out of LTS)

19:31 tchebb_ has joined #panfrost

19:35 tchebb has quit [Ping timeout: 480 seconds]

19:58 f_ has quit [autokilled: Possible spambot. Mail support@oftc.net if you think this is in error. (2023-12-24 19:58:10)]

19:58 f_ has joined #panfrost

19:58 f_ has quit [autokilled: Possible spambot. Mail support@oftc.net if you think this is in error. (2023-12-24 19:58:10)]

19:58 f_ has joined #panfrost

19:58 f_ has quit [autokilled: Possible spambot. Mail support@oftc.net if you think this is in error. (2023-12-24 19:58:10)]

19:58 f_ has joined #panfrost

19:58 f_ has quit [autokilled: Possible spambot. Mail support@oftc.net if you think this is in error. (2023-12-24 19:58:10)]

19:58 f_ has joined #panfrost

19:58 f_ has quit [autokilled: Possible spambot. Mail support@oftc.net if you think this is in error. (2023-12-24 19:58:10)]

19:59 f_ has joined #panfrost

19:59 f_ has quit [autokilled: Possible spambot. Mail support@oftc.net if you think this is in error. (2023-12-24 19:58:10)]

19:59 f_ has joined #panfrost

19:59 f_ has quit [autokilled: Possible spambot. Mail support@oftc.net if you think this is in error. (2023-12-24 19:58:10)]

19:59 f_ has joined #panfrost

19:59 f_ has quit [autokilled: Possible spambot. Mail support@oftc.net if you think this is in error. (2023-12-24 19:58:10)]

19:59 f_ has joined #panfrost

19:59 f_ has quit [autokilled: Possible spambot. Mail support@oftc.net if you think this is in error. (2023-12-24 19:58:10)]

19:59 f_ has joined #panfrost

19:59 f_ has quit [autokilled: Possible spambot. Mail support@oftc.net if you think this is in error. (2023-12-24 19:58:10)]

20:00 f_ has joined #panfrost