00:16alyssa: robclark: no worries, it blends in with all of steam's error spam ;P
00:16alyssa: uriah: cheers :)
00:17soreau: alyssa: grep -v spam ;)
00:20uriah: :)
00:23alyssa: soreau: real
07:19pepp: uriah: nice
07:33DavidHeidelberg: ci_run_n_monitor is recently brutally slow, I feel like clicking on the jobs by hand is faster, anyone else having this issue?
07:34DavidHeidelberg: I wondering if it's GitLab instance or the script itself got slower
08:05mupuf: The instance definitely got slower
08:12DavidHeidelberg: it usually takes 1-2 minutes until I got to the cancelling jobs
08:12DavidHeidelberg: it feels like a bit too much
10:22K900: Hey folks, I've got a weird bug on my 680M laptop, it seemingly randomly just becomes VERY SLOW and I get a bunch of this in dmesg
10:22K900: Sep 18 13:13:20 akane kernel: amdgpu 0000:77:00.0: [drm] *ERROR* dc_dmub_srv_log_diagnostic_data: DMCUB error - collecting diagnostic data
10:22K900: * `Sep 18, * diagnostic data`
10:22K900: And then no diagnostic data actually ever comes up
10:23K900: The machine is completely operational except the display is running at like seconds per frame
10:27K900: I think it started with 6.11, or maybe a linux-firmware update?
10:32K900: Oh great it's logged with DC_LOG_DEBUG
10:32K900: * Oh great it's logged with `DC_LOG_DEBUG`
10:43psykose: i have a 780m and i get a kernel hang on 6.11 upon a specific application gesture happening so i wouldn't be surprised if amdgpu is just broken on the first iteration
10:43K900: The weird thing is
10:43K900: It's not hanging
10:44K900: Like, everything works
10:44K900: Even the display technically _works_
10:44K900: It just does so very
10:44K900: very
10:44K900: VERY
10:44K900: slowly
10:44psykose: nvidia stole the fps
11:30K900: On an unrelated note, I did some more meson cleanup: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31223
13:58Ermine: Can minimum viable driver get around without TTM?
13:59K900: Depends on what you consider viable
14:00K900: simpledrm doesn't use TTM
14:00K900: (though it does use GEM)
14:04Ermine: If it can show linux console that would be great. If it can run wlroots with pixman renderer, that would be even greater
14:04K900: Are you trying to write your own driver?
14:05K900: For what hardware?
14:05Ermine: Yes, trying to revive r128
14:06K900: That had a DRM driver at one point, you might want to start by reverting that
14:06K900: And getting it to compile
14:07Ermine: It was UMS driver which got removed in 6.3 (and it didn't work at that moment)
14:07Ermine: and all UMS infrastructure is already deleted as well
14:10K900: Looking at that driver honestly there isn't much to it if you just want to get a framebuffer
14:11K900: And you're not getting hardware rendering from it in a way that's any useful
14:11agd5f: Ermine, r128 is similar enough to old radeons, that it would probably be easiest to add support to radeon.
14:11agd5f: you can re-use most of the modesetting code for example and a lot of the CP set up is the same
14:13jnoorman: K900: your recent GBM work (!31074) seems to have broken my `meson devenv` setup:
14:13jnoorman: ```
14:13jnoorman: MESA-LOADER: failed to open dri: /usr/local/lib/gbm/dri_gbm.so: cannot open shared object file: No such file or directory (search paths /usr/local/lib/gbm, suffix _gbm)
14:13jnoorman: ERROR: Couldn't create gbm device
14:13jnoorman: ```
14:13jnoorman: I can work around it by setting `GBM_BACKENDS_PATH=/path/to/build/src/gbm/backends/dri` but I wonder if we could teach `meson devenv` about this somehow?
14:14K900: Uhhhh
14:14K900: Yeah probably
14:14K900: Let me look
14:23Ermine: agd5f: sound idea
14:33Ermine: Though r128 seems to have an irq for vblanks, and radeons do not
14:41K900: @_oftc_jnoorman:matrix.org try https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31223
14:51agd5f: Ermine, radeons do. I suspect it's even the same registers
14:56Ermine: Ah, my bad. They didn't use what drm core for that, so I've missed it
15:10jnoorman: K900: that fixes it, thanks!
15:11K900: I'll poke some people about getting it merged then
15:11jnoorman: Great :)
17:51zmike: jenatali: how does va work on win32? I'm trying to piece it together from the msdn docs and libva
17:51zmike: it looks like you don't get a HWND or anything?
17:51jenatali: Pretty sure it's all offscreen
17:51zmike: ah I see
17:51zmike: so then the LUID is just for color profiles and such?
17:51jenatali: https://github.com/intel/libva/blob/master/va/win32/va_win32.h#L61 might help?
17:52jenatali: LUID is for GPU selection
17:52zmike: yeah that's what I've been looking at
17:53zmike: this is gonna be gross isn't it
17:53jenatali: What's the problem?
17:53zmike: no problem
17:53zmike: just trying to comprehend the brilliance of win32 api
17:55jenatali: So, what's going to be gross then?
17:55zmike: I'm zinking
17:59jenatali: The LUID's the same one you get from Vulkan physical device properties on Windows
18:03zmike: jenatali: I assume you mean deviceLUID ? I see this field is sized to 8 bytes, but struct LUID is {DWORD, LONG}
18:03zmike: which would be 12 bytes?
18:03jenatali: Yep. LONG is 4 bytes on Windows
18:03jenatali: So is plain old long
18:03zmike: oh right
18:03zmike: 🤕
18:03zmike: I didn't want to remember any of this
18:34nowallchanged: -66+174+16−26=98 −69+174+16−26=95 443−95−256−115−43=-66 446−105−256−16=69 those are the key arithmetic of compressed access, i arrived at sigma proof, in other words your contribution to the world of science/scientology of performant computer systems is zero, you contributed only tyranny, any real logics as well as any sanity is committed by me.
18:35K900: ~~who's sigma~~
18:36nowallchanged: sigma is a science term, you need a test to pass some number of repetitions that it would count as science breaktrhough.
18:37K900: ~~must... resist~~
18:37Ermine: rizz science
18:37soreau: dwfreed: ^^
18:44nowallchanged: this i got today , it is a real breakthrough number of mine where i lost count, only minus and plus arithmetic is used, so happy hacking on r300 and nv34 era of hw.
19:06nowallchanged: my dream was to defend my theory, and make this access and supercompute possible on any pocket calculator until it has some memory wired, and it looks like i worked long on this, but success entire success was happening today.
19:10nowallchanged: dudes it is a war going on, me as estonian am afraid there is battle for life coming in, and we need this tech badly, that is why i worked so passiontly on this.
19:20nowallchanged: there is not so much time left but i need couple of months more to get it all done, so far ukranians have done decent job on battlefields, it's a matter of time when they fail eventually and russian federation is attacking us -
19:32DemiMarie: Could someone get admin on the matrix side?
19:38nowallchanged: even jewish have proved from country Israel that if enough investements go to military and enough thought it's possible to defend , but the israel invests 13-15 or more percent of cross domestic product of the country to military, we are on 4% only or so, it needs to be at 20% for us but this is not possible so every electronics of cheap kind has to be pimped up against the brutes.
19:44nowallchanged: gotta be prepared and i have enough of those trolls here, it's getting serious very serious , if the defense systems have brilliant reaction times, they do not land much i would say, but so far one artillery ammunition costs 100k dollars.
19:55nowallchanged: DemiMarie: you are not on the correct path or track if i describe this more humanly, interact with me, if you get on the former.
19:56bl4ckb0ne: hi joss, long time no see
19:56soreau: airlied: ^^
19:58nowallchanged: alice feeds me aforisms, damn i can not read so quick, so alice i remember is another estonian as well as lynne.
19:59nowallchanged: it looks like i have received the whole book worth of those already and it's going on.
20:01nowallchanged: i have pretty fast computer at the moment, and goguma does this text spam fairly well so now they all try to ddos me
20:02nowallchanged: emersion: you are a genius dude, however there are some bugs in your client, but they do not distrurb me -- bugs that was.
20:02nowallchanged: this chacha auth token stuff was fantastic idea, i read your blogs
20:06nowallchanged: i wonder how much does it take to torn oftc , i am pretty sure that freenode splits happened cause of distributed denial of service attacks alike
20:06nowallchanged: as well as i am pretty sure that alice did not wire that bot
20:06nowallchanged: russians do such things, i am sure about this
20:06nowallchanged: there is a hacker group called youth Russia.
20:09nowallchanged: they are been encouraged by Vladislav Surkov and such politicians
20:21nowallchanged: alice and alike estonians who spammed their scam to the world, are the first one to die in the war anyways, absolute nonsense scammers, they do wanker tunes. Very proud of their fucker culture , i have given up on those.
20:23nowallchanged: they have been beaten up not once btw. and they continue with their nonsesense, i have not been able to live for ages, and every worldwide visitor has embargod these.
20:25nowallchanged: but they have their porn clan still present just it shrinks every day, businesses of these are dry etc. But that is what scammers have to fece.
20:25nowallchanged: face.
20:35nowallchanged: okey, shm fills up pretty quick with alice the real deal fuckers messages :D
20:35nowallchanged: they arranged an implant to me, i have several i picked up a signal many times
20:36nowallchanged: trust me, those are biggest human garbages on earth , not only in estonia, and yeah estonian clans started that scam.
20:37x512: What is happening here?
20:37nowallchanged: but i have my own friends to defend
20:37x512: It is now about politics?
20:38nowallchanged: i still have many, it's not politics those are estonian horny porn backed scammers
20:38nowallchanged: this is how you heard about me wordlwide
20:38Ermine: x512: a spammer which evades bans for a lot of time
20:39nowallchanged: Ermine: how you call yourself, you live others lives or not?
20:39nowallchanged: I provide technical details, at every public place you spam your garbage
20:40nowallchanged: your wank garbage and everyone has banned you
20:41Ermine: dwfreed: requesting banhammer
20:41nowallchanged: i know those scammers yes, who do such things, yes they are very big freaks in life
20:42nowallchanged: the wanker tunes are pretty much a smokecover about the fixings and tyranny they commit
20:51nowallchanged: those scumbags suck so bad, that they stalk so bad, that final thing is always everyone has to leave
20:52nowallchanged: suicidal terrorists so to speak
20:55llyyr: is there really no way to permanently ban that guy
20:55bl4ckb0ne: he always comes back
20:55nowallchanged: i ask for the permission to kill them off and so we go, military permission needed , everyone relevant gives it to me, but i need to tell you, that not every estonian is on bar with our suicidal terrorist leftovers, typically they cooperate on certain stereotypes where they collaborate
20:56bl4ckb0ne: registers a new account, and spams
20:56nowallchanged: i told you, you have spammed 25years every radio station about your fraud
20:56nowallchanged: you never did any sports and you never had any brain, and your only results are fixings seen in my body
20:56nowallchanged: and on my body
20:57nowallchanged: you are the worst spammers ever seen
20:58nowallchanged: they want to show something, but human being who never commited anything to any practicing can not possibly reborn or show anything
20:58nowallchanged: yes, you have your gang stalkers and you do your porn , i understand , i am not interested you see.
20:59nowallchanged: i had one girl overseas who manipulated me, i banned her for life and never reconsider really
21:00nowallchanged: she has broken bones and no teeth today, cause she was monster scammer, went on to streets yelling how i am wanker
21:00nowallchanged: karma is a bitch
21:01nowallchanged: but they are so obnoxious threatteners and scammers and tyran parasites, that people can not go to vaccations
21:01nowallchanged: as simple as that
21:01nowallchanged: vacations
21:02Ermine: bl4ckb0ne: also finds a new ip
21:03bl4ckb0ne: thats some dedicated shitpost
21:03Ermine: extremely dedicated
21:04Ermine: I'd be very well-versed in amdgpu code if I had such dedication
21:07nowallchanged: those ips are offered for free, to combat your scam, i never paid any dollars for those
21:08nowallchanged: the world is disgusted by such gang stalkers
21:10Ermine: freeloader
21:37uriah: What just happened
22:02heat: i guess he hasn't popped up in here in a while?
23:06mareko: alyssa: do you know whether NIR already has something like this? https://gitlab.freedesktop.org/mesa/mesa/-/issues/11910
23:06mareko: Venemo: ^^
23:09Venemo: mareko: I'm not aware of anything like it exactly, though I was thinking about it. either it could be implemented like the pass that moves discard up, or similarly to how we do ngg culling, that is clone the CF of the shader and delete everything else from it, then reinsert
23:13jenatali: mareko: That's more or less how the lowering we do for DXIL works
23:13jenatali: DXIL describes a TCS (hull shader) as 2 functions, one that computes patch constants and another that computes control points
23:14jenatali: That means we don't have the luxury of: "If memory stores are present, the pass will give up." As a result our tess support for GL is buggy. Nothing I can really do about that I think...
23:20jenatali: x512 Yeah. You need to authenticate with NickServ
23:20jenatali: You can DM it. It was a pain to set up :(
23:54alyssa: mareko: yeah, dxil is the prior art
23:54alyssa: I am curious though - is all of that expected to be .. worth it ?
23:55alyssa: I assume you have a particular workload in mind
23:55alyssa: my operating assumption is that the cost of VS+TCS is small relative to TES+FS since tessellation tends to amplify work
23:56alyssa: but maybe that's not the whole story ?
23:57alyssa: > It's assumed that all VS resources (VBOs, UBOs, etc.) are available in TCS. The intrinsics accessing those resources will need a flag indicating that they should use VS bindings, not TCS bindings
23:57alyssa: this is GL-specific weirdness, VK isn't affected by this fortunately
23:57alyssa: but it's also not a TCS specific problem, you get the same problem if you try to e.g. inline TES into GS
23:58alyssa: might make sense to add a gl_shader_stage to any intrinsic that accesses GL bindings
23:58alyssa: (I have nir_shader_info::tes_agx for something related, if we're going to be adding flags everywhere I'd like the sol'n to cover that too.)