10:58 Mis012[m]: why do all examples like https://gitlab.freedesktop.org/wayland/weston/-/blob/master/clients/simple-dmabuf-egl.c#L1484 hardcode the path...?
11:02 soreau: in this program, it's merely an initial value. It can be changed with `-d`
11:10 humblehumming: And there was no beef, english people are fraudulent arrogant and stupid cockblockers who suicidally block others ways on vacation to terrorise hansom people away from regions near me, they went so far that they tried to kill me countless of times from 2011 biggest event was nasty bombing on my way back to home which estonian defense police and interpol was notifying me to swap my
11:10 humblehumming: ticket , but as the bomb was still exploding carrying 165 people with it, though not me, i got sentenced to meds and sanctioned financially for 10 years. So the beef is jews and english shithoses , they have no clue what they do though, those are not programmers those are mentally retarded genetically ill suicidal terrorists.
11:10 humblehumming: *exploded
11:27 humblehumming: When the covid and war started they were kinda pushed to corner and had to finally release me from sanctions, however i still get the meds and often my lower back hurts so i can not play temporarily then. Once again i had no such beef, but i just got abused pre-explosion and after that in too deep manners, so i can not make up a face that nothing ever happened, i am the most abused
11:27 humblehumming: person in the world after all.
11:31 humblehumming: what's cumbersome about such people until the world order is not forced to make them feel scared or fire emotions like covid-19 managed to do, day by day they are suicidal otherwise , there is some blockage in their brain no normal activity, i have been in immense trouble to handle their terror acts, cause they without issues can not enjoy life, do not want to live, and do not want to
11:31 humblehumming: let others to live...and by others i mean me and my true friends too.
12:29 Mis012[m]: soreau: but any real world app really ought to get this information by itself and not rely on the user to supply it
12:30 alice: it's not a real world app though
12:30 soreau: that's why these are called 'examples'
12:34 Mis012[m]: it seems to me like "it would be really annoying for me to have to do this properly, but I don't see that as a reason for an example to exist"
12:34 Mis012[m]: at least in the other example where it was specifially called out as a hack
12:35 Mis012[m]: obviously it might make sense to not have it in every example, but I did't see it in any
12:36 Mis012[m]: *didn't
12:36 soreau: the world is your oyster - it's FOSS
12:37 Mis012[m]: well, if I knew how to do it, I wouldn't need an example would I
12:37 soreau: how do you expect apps to know what the user wants, anyway?
12:38 Mis012[m]: well, since this is a wayland example, presumably you'd want to use the same render node that's being used for other wayland surfaces
12:39 karolherbst: maybe it would make sense to use the lowest renderer node by default, becuase on some systems 128 might not even exist in the first place, but yeah....
12:39 lumag: What are the guidelines for uprev'ing the IGT? We'd like to up it to 19838a9ff320 ("tests/kms_writeback: Set atomic and writeback_connector early"), but maybe there are other targets and/or requirements
12:40 soreau: Mis012[m]: that doesn't really sound like the best heuristic, to me
12:40 Mis012[m]: karolherbst: on my system 128 is the lowest, so that doesn't really fix anything
12:41 karolherbst: well.. the application doesn't know which device to use, but any should be working anyway
12:41 Mis012[m]: soreau: well, in my case I need to redo some rendering so I can get a dmabuf and use GtkOffload to display the result
12:41 soreau: Mis012[m]: with gtk4?
12:42 Mis012[m]: yes
12:43 soreau: in that case, Company here or in #gtk on libera chat, might have some ideas for you
12:44 Mis012[m]: currently I have my own hacks for having a wayland subsurface, but the GtkOffload is special cased to allow for having it under the Gtk window with a punchout
12:47 Mis012[m]: I think I saw somewhere that you're supposed to be able to get a render node for a wayland surface, which is probably what I'd want (GtkOffload doesn't work on X11 anyway so having to specify it manually wouldn't be the main issue)
12:47 Mis012[m]: but I can't find any examples so maybe I was wrong
12:48 Mis012[m]: soreau: the Gtk part is pretty much just a matter of giving Gtk a dmabuf, but as I understand it in order to get that for an EGL window surface I need to use gbm
12:56 Company: GTK doesn't have any API to get at render paths (yet?)
12:56 Company: but I think you can query EGL, no?
12:57 soreau: Mis012[m]: you might have a look at "Some libraries that could be used to make sharing easier" here - https://eleni.mutantstargoat.com/hikiko/egl-dma-1/
12:58 Company: ie get the egl display from GTK and query the default device
12:58 Company: but yes, this is all quite hidden inside mesa drivers
12:59 Mis012[m]: also, would you happen to know what's the deal with pbuffers and pixmaps
12:59 soreau: pixmaps are an X concept
12:59 Mis012[m]: since gbm doesn't support those
12:59 Company: pbuffers are a pretty outdated thing, no?
13:00 Company: like, today you create a texture and render to it, no need to go via pbuffers
13:02 Company: Mis012[m]: what are you actually trying to do?
13:03 Mis012[m]: I need to support pbuffers in https://gitlab.com/android_translation_layer/android_translation_layer/, it seems that adding a wrapper that makes a window instead works but idk if that's necessarily guaranteed
13:03 Mis012[m]: and using X11 won't be a viable workaround if I switch to GtkOffload
13:06 Company: you could try implementing GtkOffload on X11 if you're into that - it's not like that's impossible to make work, but we think X11 is outdated and it's not worth spending much time on
13:06 Mis012[m]: If Gtk exposed subsurfaces that should also work to get the benefits that GtkOffload has
13:06 Company: that said, offloading is optional anyway, so if stuff works with offloading, it works without
13:07 Mis012[m]: copying the dmabuf with cpu has zero chance to have usable peformance outside maybe x86
13:08 Company: dmabufs are imported via EGL
13:09 Mis012[m]: so, I have two separate problems, one of which I don't know if it's a problem
13:09 Mis012[m]: the first is that I need to get the render node path, which I guess should be possible to do like you suggested
13:10 Mis012[m]: the second is that I need pbuffers on Wayland and idk if there's anything that will break if I just give the app a window surface instead
13:10 Mis012[m]: s/wayland/gbm/ if I change to that
13:12 soreau: Mis012[m]: https://gitlab.freedesktop.org/mesa/drm/-/issues/41
13:21 MrCooper: hmm, debian-testing is one of the slowest CI build jobs now, which is bad as many test jobs depend on it
13:44 radulamehona: I still think my code works as addition and subtraction when only used the order of those operations does not matter as long as the length is sufficiently large for polderan, so as you remove the constant once from value+index+lengthfromconst+lengthfromconst that yields value+index+length.. at poi and value+index+length+length at remaining non selected elements, now you remove the value
13:44 radulamehona: it yields index+length and value+index+length+length, remove all the lenghts now you got index ay poi constants on non-poi, remove x times constants and the index is there, now it needs to ripple combine with another daughter cell, which becomes a mother again, and that might be achieved with having one time constant removed from that offset, that yields length+index but now the bank is
13:44 radulamehona: swapped in the compilation phase, removing all indexes yields another index or value etc. so forth, the idea i am after is a super computer and third generation vaccines, where i work with people that lock up devils who do such shit to me as well as to others with terroristic lies and acts of fraud, that injects fear only then to those dickheads when they know the public will punish them
13:44 radulamehona: according to rules, so that they are transparent and equal in terms of justice to others, to punish people for their fradulent terror acts. We have such technology to make it happen. Am i smarter than MrPooper, of course i am and all others cause it was me who brought this super computer to light. I designed it and invented, of course i am wiser. But having said that, i have miles of
13:44 radulamehona: code still to type, i just started my coding phase, list of modules is large, the intrinsics depend on encoder that encoders the packed operands as needed , it makes the allocation upfront at the moment to be in range of 6k per alu, that is why i base all powers at 64, cause it's merely allocation for the future reserved fields to implement polderan logic at. it just filters the answers
13:44 radulamehona: and the way i see, it can do it in batched, so it can achieve million of alus executed per word with 20 to more add sub compact intrinsic ops executed. If i see any of the terrorists here i make the responsible for their actions later, and you gonna get tortured for terrorising.
14:47 highupintherank: it was otherwise consistent and it seems to function when certain order is still preserved for those , but one type, in midsentence remove all lengths not indexes in the later remark. I shared it and i was suppose to, anyways some criminals cracked my neck up and inserted another chip or something. I have found many before, they harvest my blood and cells against laws such leeches i
14:47 highupintherank: fight against with my coding, cause sports is not possible with the set of injuries arranged on decent level.
15:13 zmike: mareko: is it expected that tc can do [ bind_vbo(dst) -> draw -> replace_buffer_storage(src, dst) -> subdata(src) -> draw ] and expect the subdata to update vbo dst for draw?
15:20 zmike: I guess so
15:59 zmike: hm no, I think this is a bug
16:08 rgallaispou: Hi,
16:08 rgallaispou: as a maintainer I would like to merge several patches that are piling in the DRM/STM merge backlog.
16:08 rgallaispou: I tried to push one of the changes onto the drm-misc-next according to the maintainer-tools documentation, but it failed.
16:08 rgallaispou: Since I've never pushed patches it seems that I need to get some permissions: "remote: You are not allowed to push code to this project."
16:08 rgallaispou: Do I need to do anything on gitlab.freedesktop.org in order to be able to push changes ?
16:09 meizuphone: all modules are ready but it does not hurt to say that offset is actually removed from index+value pair so that the constant would shift, in other words in case one needs it, i am unsure if it makes any difference cause it is all modulo based as to how much you remove it should technically correct itself, anyways i am going to leave near days anyhow from that irc maddness, i currently do
16:09 meizuphone: not anymore see where the offset is useful, but the formula was not correct when offset is involved, its index+value-offset+distance=constant+offset so 2*distance put to the same formula yields constant+distance+offset, or maybe i just do not know enough maths, but i won't be giving up neither i do not care what god or phone says to me. I just program that way, i abondened my earlier plan
16:09 meizuphone: for commodity systems to scale below single cycles via cross domain checked self-timing, i.e clockless, i even than saw lots of threats when things were executed this way, after i got beamed in excess i understood right away i need plan b for upstreamable code, but i did not have all the modules together at that time in cambodia, only beginners platform , early idea, but it seems i have it
16:09 meizuphone: worked out now, i see consistent vision with all modules being stable now. It's really time to code, hypersonic weapons under this paradigm is not possible but execution inside hw is.
16:10 karolherbst: rgallaispou: yeah, you need to request a developer account
16:11 karolherbst: rgallaispou: https://drm.pages.freedesktop.org/maintainer-tools/commit-access.html
16:12 rgallaispou: karolherbst: I read that. This is confusing this it points to https://www.freedesktop.org/wiki/AccountRequests/
16:12 rgallaispou: And in which it says that I do not need to
16:12 karolherbst: huh.. maybe the links didn't got updated
16:13 karolherbst: rgallaispou: "Legacy SSH accounts"
16:14 karolherbst: on the same page
16:14 rgallaispou: karolherbst: Alright, I thought this was.. well legacy. I'll apply there then.
16:14 karolherbst: yeah.. it mostly is, but the drm trees are still not on gitlab
16:17 daniels: they are on GitLab now!
16:17 daniels: it’s only amd that’s missing
16:18 karolherbst: ahh
16:18 karolherbst: I just checked and it still said issues only
16:18 karolherbst: so I thought it wasn't
16:23 daniels: hmm, no mripard
16:28 sima: uh yeah should probably update dim docs ... jani or demarchi maybe?
16:37 rgallaispou: I'll fill this when I can retrieve my GPG key back. Thanks for the help karolherbst :)
16:38 karolherbst: rgallaispou: yeah, but as others stated it _Should_ be on gitlab now, so you'll only need to grant access. I kinda forgot about that. anyway...
17:11 jani: sima: I think I'm quite ready to pass the torch on maintainer-tools if we could find new maintainers for it
17:53 daniels: rgallaispou: you no longer need a GPG key - just file an issue on freedesktop/freedesktop asking for access to drm-misc, and I'll assign it to mripard
18:38 rgallaispou: daniels: Okay, I'll close the one I did and file another one.
19:23 karolherbst: airlied: what are the alignment rules for llvmpipe on various address spaces (global, ubo, scratch, shared)? I need to fix unaligned load/stores via nir_lower_mem_access_bit_sizes
19:28 karolherbst: or rather I'd need to know what this should end up as in the end: @load_ubo (%0 (0x0), %7 (0x19)) (access=none, align_mul=256, align_offset=25, range_base=0, range=32)
19:35 airlied: karolherbst: I think we always load each channel separately and then vector them
19:36 airlied: look at lp_bld_nir_soa.c:build_gather
19:39 karolherbst: right...
19:40 karolherbst: but unaligned loads will be all messed up
19:40 karolherbst: I'm mostly curious what alignment the code assumes
19:40 karolherbst: is it vector component size?
19:40 karolherbst: or something else?
19:41 karolherbst: but that might be hard to lower nir_lower_mem_access_bit_sizes if special slicing is going on
19:42 airlied: it loads scalar
19:42 karolherbst: okay
19:42 karolherbst: yeah, that should be trivial to lower then
19:59 karolherbst: PASS :) that wasn't too hard
20:31 karolherbst: airlied: so now that load above becomes a 8x4 one :) https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29896/diffs?commit_id=50e8f06c154d22858b7fd27b324338778422325a
20:32 karolherbst: though I think we might have to optimize the lowering at some point
20:32 karolherbst: oh well...
20:33 karolherbst: don't use packed/unaligned memory
20:36 airlied: karolherbst: I wonder how different the llvm-ir/asm is in the end, we already break down the loads into scalars anyways
20:36 karolherbst: good question
20:36 karolherbst: how do I print it again?
20:38 karolherbst: ir and asm apparently
20:39 karolherbst: uhhh
20:39 karolherbst: I'm sure it's fine
20:40 karolherbst: https://gist.githubusercontent.com/karolherbst/122a7fe1ff31a791072de5b11e445709/raw/aa1bb9d6b92d096575f7ec770c392dbe9cf0c001/gistfile1.txt
22:27 emersion: would anyone have time to review this? https://gitlab.freedesktop.org/mesa/drm/-/merge_requests/363