07:38 jani: simple review, anyone? https://lore.kernel.org/r/20240408104230.3191827-1-jani.nikula@intel.com
07:43 javierm: jani: done, on the ML
07:48 jani: javierm: thanks!
08:12 sima: jfalempe, acked the entire thing, thanks a lot for your work
08:12 jfalempe: sima: you're welcome, and thanks for taking time to review it.
10:31 tzimmermann: jfalempe, do you have comments on https://patchwork.freedesktop.org/series/131982/ ?
12:14 Matombo: Hi all
12:15 Matombo: I'm task to evaluate a better way to handle GPU offloading then it's currently done on Linux with environment variables. The Goal is to have some kind of GUI editable database/config file where the user can add and remove binaries/programs.
12:16 Matombo: For this I would like to push https://gitlab.freedesktop.org/glvnd/libglvnd/-/merge_requests/224 and https://gitlab.freedesktop.org/glvnd/libglvnd/-/merge_requests/228 as it seems to be a good basis for exactly this. However the discussion on them is stale, not because of technical difficulties, but because of lack of interest it seems.
12:16 Matombo: Are there other people here interested in this topic?
12:17 jfalempe: tzimmermann: they look good to me, I will reply on the ml shortly.
12:17 tzimmermann: ok, thanks
12:33 pq: Matombo, maybe ask in the XDG mailing list? This seems very desktop oriented rather than plumbing/drivers. They already have something with .desktop files AFAIU.
12:34 swick[m]: vsyrjala_ hwentlan_ can you both take a look at https://lists.freedesktop.org/archives/dri-devel/2024-March/444707.html at some point?
12:34 Matombo: pq: .desktop file based solutions don't cut it i'm afraid because auf launchers like steam and lutris which don't necessarily use desktop files
12:35 pq: Matombo, of course, but that is how you find actually interested parties.
12:35 Company: the desktop situation for that is a complete mess because there's like 3 different ways to do things in Gnome alone
12:35 pq: they have already had the same problem, and they have tried to solve it their way
12:36 Company: I was looking into adding GPU support to GTK, so we can properly expose it for dmabuf integration, and it's all a mess that should be thrown away and done properly
12:37 Company: and I think the desktop file just says "run on primary gpu"?
12:37 Company: which doesn't even work for me, because I use the primary default and want an option to *not* run on primary gpu
12:38 pq: I didn't say it's good, but that the people involved in it would be the audience to reach out to
12:38 Matombo: Company, i think the option is called run on primary or something, but it actually does run on whatever is the first non default one
12:39 pq: performance, not primary?
12:39 Matombo: iirc there was a discussion in the freedesktop gitlab i read through some time ago
12:40 Matombo: pq, i assume you mean this list? xdg@lists.freedesktop.org
12:40 pq: yes
12:40 pq: Matombo, if you advertise this problem on the XDG mailing list, I think you have much bigger chances of getting the attention of the interested people than on any mesa list.
12:41 pq: because, it's UI and you want a cross-desktop standard
12:42 Matombo: well I want low level support to built a UI on top
12:44 Matombo: the person that did the poc for glvnd wanted input from mesa ppl, tbh that's the main reason why i started there
12:46 pq: ok
12:47 Matombo: Hope someone on the xdg list replies ^^
12:48 pq: desktop people might have opinions about how and where any kind of association database is stored, and how do you deal with e.g. concurrent write access to it, how to identify apps, and how to work with sandboxes.
12:52 pq: Matombo, did you subscribe to the xdg list before sending?
12:53 Company: Matombo: when I tried it, it definitely ran on the primary
12:54 Company: I've always only managed to get non-primary via env vars, and I use DRI_PRIME usually because it works on GL and Vulkan
12:56 pq: DRI_PRIME never worked for me on the EGL Surfaceless platform in the past.
12:56 Company: though really, I'd like something that has software rendering and stuff like zink as options, too
12:56 pq: oh yeah
12:57 Company: and I'd like something that allows me to match Vulkan and GL devices
12:57 Matombo: pq, yes I did
12:57 Company: because then I can seamlessly switch between Vulkan and GL inside GTK
12:58 Company: while applications can use the same code for their dmabuf stuff
12:58 Matombo: yeah: that's my point: the current way it's handled is not really useable
12:58 pq: Matombo, ah, now I've received your email, Werner.
17:44 harimoto: Tallinn has the facilities to host XDC surely, but even though i have something special for this year a total turning point for computer programming history and i live quite abundant already now, and become very abundantly rich shortly I still do not know how to communicate with estonian academics to host that conference. I started new revolution in computing.
17:45 harimoto: I succeeded passing all the tests in multi bank compressed sets, it can index any set and access any element with index and bank offset, in any order stored.
18:05 harimoto: yeah i redid the babylonian method, it's only idea is to filter the banks and index the cells within the bank, and offer output where one cell becomes the smallest and then this is chosen above others by a method described on nouveau.
18:06 harimoto: this year i will be building the full alu corpus and compute in compressed format, and once ready i likely offer the world my own work too.
18:21 harimoto: ah more than five years occasional calculator work, like once a week, and now i hit everything, my papers need to be sorted and technically i am ready for appearing in conference to answer to questions and the plan is to lead the linux bloated state to real performance.
18:24 harimoto: it's likely that google advancements through chatgpt and alike settled me down nicely in place algorithms and definitions through googling , it accesses one hash, and builds a whole program from it, totally monstrous accomplishment from my side finally.
18:35 harimoto: by the way i had major headaches when i tried to access my thoughts, which are all gone cause it appears they modernized something, the organic chip melted off which they planted on eye lash, it destroyed itself some call it falsely tear channel blockage symptom, as soon as this got destroyed i no longer have hallucinations.
18:36 harimoto: also by eye does not twitch anymore
18:37 harimoto: it was exhibiting a twitching symptom before but no longer, like there was a pump or something, had a twitching interval.
18:41 harimoto: it was a intra occular pressure indicator, whatever was the reason well i got hit with a pan very heavily on the left hemisphere of the brain straight crushing the eyebrow, it was 30 years ago, but something in the eye started to malfunction in the same datum.
18:45 harimoto: so tasks that took my years before are solved in minutes , but this is the brain that has recovered so far nicely, some other injuries yet to treat.
18:57 harimoto: but some offspring i see china and russia expanding, taiwan is very dangerous subject, since computers are more capable than expected is known to me, however lithography or chemical vaporizing depositing units, must be sanctioned to such manufacturers, who are threatening the peace of the world.
18:59 harimoto: if we managed to take control of central Europe manufacturing facilities with new technology we'd win the war or avoid expansion.
19:02 harimoto: even if the defense costs or contributions in nato countries were not significant such as tanks, aerodynamic fighters, artillery, radar systems, only with drones it is possible to win
19:04 harimoto: there is no question that estonia is next, the same war simulation has been done near the narva border as was done in ukraine, we soon need everything from europe too all factories at fullest power.
19:05 harimoto: i do not have time to deal with anger management anymore, the war is devastating already.
19:08 harimoto: XDC opportunity in estonia is likely the last , cause ukraine is losing in the war, and sure putin expands it's range.
19:23 harimoto: if someone sent me letter from austria a marked letter , send it again my mom rejected it,i will receive it an put my signature, it was something important it seems.
19:25 harimoto: the technology can not be exposed to certain participants, i have had it before too, that is why i cipher all the content even in research in my home, dudes this is very dangerous to give guns to enemies.
19:25 airlied: dwfreed: another
22:03 DavidHeidelberg: Btw. do you think we can make some plan how to DEBUG and fix LTO?
22:03 DavidHeidelberg: Some distros shipping with it, in general works for them just fine... except some weird corner cases.
22:04 DavidHeidelberg: IMHO we losing like 5% CPU perf on whole mesa3d lib without LTO, so far it's work 99.99% reliably. as usual that 0.01% is tricky.. I know.
23:23 dwfreed: airlied: ugh
23:33 kode54: is custom edid supported by xe.ko?
23:33 kode54: seems some people are having trouble getting it to work with Intel dGPUs, and I don't know if that's the reason