20:30cosurgi: imirkin: I bought my daughter a radeon (rx5600xt) for use, to get acquianted with radeons. I saw some insability problems. First impression was that, if stability is most important for me, then nouveau wins. Surprisingly. This might have changed however after I updated to latest kernel on her computer :)
20:31cosurgi: so far I got upime 65 days, and I disabled SMT completely in biose, so that I am not tempted to experiment anymore with turning it on or off ;)
20:31cosurgi: As recentely these experiments were the only cause of hangups.
20:31imirkin: can't speak to details, but the amd team tends to be responsive to issues
20:32cosurgi: this is nice :)
20:32cosurgi: someday I'll switch, time premissing :)
20:32cosurgi: lol.
20:32cosurgi: *permissing.
20:33imirkin: i think `permitting`, but i get what you mean :)
20:33cosurgi: ah, right. of course.
20:34cosurgi: also, compton did not cause a crash. Though I still remember to kill it after I don't need an inverse color (or semi-transparent) window
20:35cosurgi: (hm, maybe such kill compton in cron would be useful)
20:36imirkin: if you stick to not doing anything too crazy, nouveau tends to not completely suck
20:36cosurgi: the lack of 3d games on steam is only a minor problem, I get more time to work, thanks to this :)
20:36cosurgi: yeah :)
20:38cosurgi: btw, the SMT on/off experiments also cause problems with SSD nvme0n1 drives. So it seems that such problems of interrupts not handled (because CPU is off) are a general kernel problem. Not jusr nouveau.
20:40cosurgi: and these interrupts not handled are thousandfold more frequent after a hibernation-resume cycle.
20:41imirkin: i'm still rocking a PC from 2010, so ... none of these problems for me!
20:41cosurgi: heheh :)
20:43karolherbst: imirkin: did you see my email in regards to the wiki? Would be interested in your thoughts about it
20:43imirkin: eh
20:44imirkin: i spent some time cleaning it up about 5-6 years ago
20:44imirkin: do whatever you like :)
20:44karolherbst: ahh
20:44karolherbst: ohh one thing: there are a couple of broken img tags
20:44karolherbst: and I think the image files are mostly lost
20:44karolherbst: any idea?
20:44imirkin: it's git-backed
20:44imirkin: but they could have been external resources
20:44imirkin: dunno
20:44karolherbst: some are external URLs but some are internal and I didn't saw the images
20:45imirkin: https://cgit.freedesktop.org/wiki/nouveau/
20:45karolherbst: but yeah.. maybe they are in git _somewhere_
20:45karolherbst: imirkin: I already moved stuff to test gitlab pages here: https://gitlab.freedesktop.org/nouveau/wiki btw. It's quite nice to be in full control of the ikiwiki pipeline.. even though it has some drawbacks :/
20:45karolherbst: but I really couldn't find the images in the git anyway
20:46karolherbst: mhhh.. let me find an example
20:47karolherbst: imirkin: https://nouveau.freedesktop.org/wiki/Nouveau_Companion_27-fr/
20:47imirkin: i'd recommend porting as little content as possible
20:47karolherbst: there are broken img tags, but they also depend on the ikiwki CGI
20:47imirkin: hah.
20:47karolherbst: imirkin: it's all ported as it was jut copy paste anyway :p
20:47imirkin: yeah ... there's something odd in there. i don't remember precisely.
20:47karolherbst: it is running the same software
20:47imirkin: ok, well you should cut a lot of it.
20:47karolherbst: just as a gitlab CI pipeline then
20:48imirkin: most of it is beyond useless, except maybe interesting archives
20:48karolherbst: yeah..
20:48imirkin: like the nouveau companion stuff
20:48imirkin: from the bad old early days
20:48karolherbst: but I really just wanted to solve the technical issues first I wanted to solve
20:48karolherbst: and try out if it even works as I expect
20:48karolherbst: there are some annoying "regressions" though
20:49karolherbst: editing is a bit annoying as you need to wait for the CI pipeline to finish, and pages are aggressively cached as well :/
20:49karolherbst: but on the good site, anybody can create MRs now
20:49imirkin: honestly, i'd just chuck the whole thing and start over.
20:49imirkin: but whatever
20:49karolherbst: yeah.. that's probably the best idea
20:49imirkin: gtg
20:56Lyude: hm, does anyone know why the DRM_NOUVEAU_GETPARAM ioctl is supposedly deprecated? I'm looking for an easy way to retrieve the model information about a GPU from nouveau (I want it for igt so I can implement block linear tiling), but it looks like the only way of doing it is through this deprecated ioctl
20:57karolherbst: Lyude: well, that was the plan but there is no replacement really :p
20:57karolherbst: I think the idea was to move all over to nvif but that never worked out or something?
20:57karolherbst: dunno
20:58karolherbst: Lyude: ohh actually...
20:59karolherbst: nvif allows you to do that
20:59Lyude: karolherbst: have any examples?
20:59karolherbst: Lyude: libdrm/nouveau/nouveau.c
20:59karolherbst: NV_DEVICE_V0_INFO
20:59Lyude: cool, thanks :)
20:59karolherbst: but I think there were other things which don't have one... oh well