00:38Company: mort_, lina: re the discussion about cursor latency on reddit: Is it possible to update the cursor plane independent from the rest of the planes?
00:38Company: then you could do a compositor that does vsync for the main image and low latency for the cursor
00:44zamundaaa[m]: Company: KWin already does that, as far as KMS allows
00:44zamundaaa[m]: An AMEND flag to allow you to amend the last atomic state even after the pageflip has happened (whenever possible) has been talked about before, but never went anywhere unfortunately
00:45Company: would that mean that kwin should have lower latency in mort_'s test?
00:46Company: or do the drivers eat that benefit?
01:41zamundaaa[m]: Company: possibly. I haven't looked into it too far yet
01:42zamundaaa[m]: I did build myself a latency measurement thing for a different reason though, so I may test legacy vs atomic cursor movement soon-ish
01:45Company: I'd be interested in that
05:11DemiMarie: Does Linux have any analog to D3D12’s concept of reserved resources? The Windows kernel implementation is apparently designed to have the GPU update its own page tables, which seems rather strange and tricky because it requires the kernel driver to generate GPU commands to do that.
08:04mlankhorst: DemiMarie: Xe does that.. O:)
08:11phasta: agd5f, thx. Is there a list of AMD GPUs that should run on Linux?
08:26DemiMarie: phasta: It is safe to assume that Linux supports an AMD GPU unless you have a reason to believe otherwise.
08:27DemiMarie: mlankhorst: Interesting! Does it actually have a kernel-mode JIT or does it just use the DMA engine?
08:28MrCooper: yep, the list is basically "every released AMD GPU"
08:37phasta: Just yesterday I discovered this crazy Vulkan issue on AMD, where you have to use vulkan-radeon instead of amdvlk. Seems the usermode driver is broken at times? https://wiki.archlinux.org/title/AMDGPU#Installation
08:40HdkR: phasta: Most people here would recommend radv over amdvlk anyway
08:41HdkR: Which is what the vulkan-radeon package is
08:41phasta: HdkR, but why? And why are they even interchangeable?
08:42HdkR: phasta: Because there are two Vulkan drivers for Radeon hardware. AMD's proprietary driver (amdvlk) and the mesa based one (radv)
08:42phasta: He. Open 1:0 Proprietary
08:43HdkR: The proprietary one is also technically open-source :)
08:43psykose: they're both 'usermode' (in userspace)
08:43HdkR: That too
08:43glehmann: but throw-over-the-wall opensource
08:44glehmann: and there is also the proprietary amdvlk-pro (which shares everything but the compiler with amdvlk non-pro)
08:54phasta: Is there still an amdgpu-pro? Or maybe I confuse it with amdvlk-pro
08:57MrCooper: what do you mean by "amdgpu-pro"?
09:02berluccigringoof: i remember the fraud that Estonians comitted with with david airlie very well, the sexpolice crankgangsters harassed my fans in 2011 first time in cambodia where as in masses by british scumbags until a bomb was detonated on my way back to estonia where i got informed about it by agency of unknown works that annulated by status if the ticket, where the scumbag airied bragged here
09:02berluccigringoof: back to pills schizophrenic now last attacker from england is known, my carpender has made a coffin to him, and there is no long party for dave nor dwfreed either similarly to known estonian trash that all gets brutally and vulgarly handled by my delegation.
09:02berluccigringoof: of the ticket
09:35mlankhorst: DemiMarie: DMA engine + fencing, it's asynchronous so if you want to wait for completion you have to wait on the fence.
09:37DemiMarie: glehmann: does the amdvlk driver have any advantages at all?
09:37DemiMarie: mlankhorst: Makes sense!
09:40glehmann: DemiMarie: better ray tracing maybe, and for native Vulkan games it's more likely to work on release because game devs test on windows (which has basically amdvlk-pro)
09:41DemiMarie: glehmann: what is different about the amdvlk-pro compiler?
09:42glehmann: afaiu it's shared with windows gl/d3d9-12 instead of the LLVM based stack
09:44DemiMarie: Where does ACO fit in?
09:44glehmann: ACO is what radv uses now, it's part of mesa
09:45glehmann: -pro has faster compile time than LLVM (not hard), higher ones than ACO. Tends to produce better asm than LLVM, sometimes also better than aco
09:46glehmann: but ACO also does somethings a lot better than either of the other two, mainly control flow and divergence things
09:46glehmann: I'm also biased as I work on ACO :P
09:47DemiMarie: Is LLVM only used for amdvlk (non-pro), ROCm, and OpenGL?
09:49glehmann: I think so, radv has an debug option to use LLVM but that's mostly unmaintained and just for testing.
09:49pixelcluster: alyssa: yeah pretty much exactly what you said, "scratch" is really just space in per-function stack
09:50pixelcluster: that's what the ACO implementation of "real" function calls does too, it just adds the current shader->scratch_size to each function's stack
12:44logsoutofbound: they based their teamscam last time on a finnish set of analwhores or sluts i do not care who are collectively as gangstalkers still begging and harassing money out of me, they thought cause the whore got into anus by scumbags is enough of a reason for this, which is entire absurd as those aids people all are. Only jews that i should push more to get into relation with such , even
12:44logsoutofbound: though from my territory all they were disallowed to come, now the english cunts are still gangstalking me playing scouts to their girlfriends talking fairytailes how injusticed person called mart threattens their lives ad i wrote to jack, that if you do not turn yourself in this time, you are over i.e dead soon which in the end applies to all of those aburd gangstalkers who committed
12:44logsoutofbound: level of crimes perpendicular to higher penalties if not death penalties alike for estonians in the past.
12:52logsoutofbound: You do not hear geunine truth out of such devils, not their labs either, as both all the sports echivements pre injuries as well as code nowdays that makes also sense to anyone who has brain attached is in the end pushed by me and my friends around the world.
12:56logsoutofbound: and i said the scientific reason for this, i have little bit stronger scull that protects my brain cells out of evil events, it holds my neurons in bigger vacum, so the assaults failed on this structure, even though they were successful as delibarate evil on the rest of my body, that brought me into programming overall, which i am still happy with and proud, to have that opportunity as
12:56logsoutofbound: the minimum at the moment.
12:57logsoutofbound: all the sports was actually very complex and big weight on my shoulders anyhow, that none besides me was able to carry at youth, it was cancerous anyways.
12:57logsoutofbound: it ruined my life due to devils always choosing me for any terror.
12:59logsoutofbound: i am a twin human, with a fluke or higher immune protection, which indeed is very rare in the world of human births, i am aware of this.
13:00logsoutofbound: i technically was a snatch in my mothers belly against my twin sister, and had another very rare accident or disturbance to my gene alleles.
13:02logsoutofbound: which is to say i do all the works demanding wisdom or physical strength at my sisters company, but i am not ok with british scumbags attacking me in response.
15:19alyssa: thinking these must be trace flakes? https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33242#note_2754521
15:20alyssa: I don't see how that MR can possibly affect any driver other than intel & asahi
15:20logsoutofbound: I did all the implementation of the compression specification hdd storing routines, linkage of the format and execution as well as decoder encoder, which remained so that out of it, we use a small lookup table to fasten the routine cause of IO writes need full value, and encoder uses a loop or dma. I studied processors for long time, and i customize the routines to performance
15:20logsoutofbound: considering reordering and operand forwarding or safety and security, however the compilers skeleton was pushed by other people to the net, that i ended up using, cause i am bit lagging behind. Io digged out loops to insert my routines in the wasmati kit, and this is my last interaction before we already have the compiler functional, cause the specification for execution is bit simpler
15:20logsoutofbound: to understand, as said the system is designed for it naturally due to powers of twos presentation, it can selecitively target banks and elminate the reverse banks with two indexes that get shared. There are many methods but i had already spoken about very intiuitive ones, which is shared index and contiguous base etc.
15:21zmike: alyssa: at least some of those are https://gitlab.freedesktop.org/mesa/mesa/-/issues/12465
15:21alyssa: zmike: ok, makes sense
15:21zmike: v3d didn't render anything at all
15:21alyssa: the v3d is the one that killed the merge
15:22alyssa: but like
15:22alyssa: i don't think that MR could have possibly affected v3d
15:22zmike: it didn't
15:22zmike: the runner just fucked up and didn't render anything
15:22zmike: look at the output
15:22alyssa: yeehaw
15:22logsoutofbound: i continuously push code for graphics, and i have d3d12 backup trees as well, there is a big batch of graphics and video endocing work too pending, i also did libinput code in the past, and feedback on this will be asked in near future.
15:22alyssa:again wants to kill premerge trace jobs but is too exhausted to care
15:22alyssa: back into the marge queue it goes
15:23zmike: I think this is probably a different issue than the typical fuzzy rendering
15:27daniels: 'error: drawable failed to resize: expected 640x480, got 32x32'
15:30logsoutofbound: i am being blacked out by some community members who demand reports on the net of all my heroics, but i am avoiding that all , and leave all the powers to the power users, i actually know that prime gaps has big mathematical community behind the research, but all of it i already had in the works too.
15:45tzimmermann: jfalempe, about the ast series: you mention the switch in patch 13, but don't complain about the one in patch 10. shall i replace both of them?
16:01zmike: kusma: on that mipmap issue, if you're satisfied with the answer can you close the ticket
16:15grandmanback: The rocket engines rocket trouble as to forward as much data as possible in milliseconds has been solved by me already, however it much seems that though elon musk achieved the rocket engine of first generation spacecrafts almost with three initial accidents, well happens i do not see a devil there in Musk accidents happen. Despite of that i think europeans are capable of building a last
16:15grandmanback: generation and very last generation animatter rocket also, nuclear annihillation based engine, though rocket engine operation i do not so specifically know off hand, cause i am only a programmer, they now send me the recordings and videos of that state of spaceX on my facebook page, others had gotten to that point earlier for some reason.
16:24jfalempe: tzimmermann: yes you can replace both of them. The switch makes it more complex to read (at least to my eyes).
16:26tzimmermann: jfalempe, ok
17:44jenatali: Do folks here have anything to do with apitrace? I suspect there's some overlap
17:45zmike: somewhat
17:46jenatali: I was hoping to use it on my arm64 laptop but it looks like the injector doesn't handle Windows arm64 nicely
17:46jenatali: I guess I can probably just file an issue, nevermind :)
18:02zmike: dcbaker: here's one for you https://github.com/mesonbuild/meson/issues/14197
18:02zmike: mesa cross-compile is broken
18:02dcbaker: zmike: you too have hit a problem that I have been trying to solve for the better part of two years :)
18:03zmike: ah good
18:03zmike: glad we're finally on the same page
18:05dcbaker: Meson's subprojects are always built for the host machine, which is because the people who drove that basically wanted to be able to build a GTK/Gnome SDK. I have been slowly landing the refactors so that Meson can get to build subprojects for both the host and the build machine at the same time, which is what rust needs
18:05dcbaker: Hopefully I'll eventually get everything in for that, we're getting there!
18:08zmike: 😬
18:08zmike: dcbaker: is there any kind of quick hack to just make it build with a set compiler?
18:08zmike: something I could do locally...
18:10dcbaker: zmike: It's unfortunately a design issue with Meson... Let me see if I have a branch somewhere that works, you might be able to take that + some hacks
18:10zmike: no rush, I gotta clock it for today
18:11dcbaker: Okay. It really is on my list of things to do, it's just a really deep change to the core of Meson, and there's a bunch of stuff that has to happen.
18:13zmike: sure
18:15grandmanback_: is there also some confusion as to how the powers would be taken according to the Cornell formula? i took them from 32 then 34 which is smallest non colliding distribution where i encoded second operand differently, however technically any contiguous distribution can be taken and the formula makes them work through eulers imaginary numbers that get rounded up, since we want to present
18:15grandmanback_: them in fixed integers format, so logic imply collision on first 8powers due to gaps being too small, i do not know what is ideal distribution from previous papers, but 15 is more than enough for first negative logarithm to avoid float based format. so i put 15to79 for example to the calculator of binary distribution.
18:16alyssa: dcbaker: is this related to the clc cross build mess?
18:16dcbaker: yes it is
18:17dcbaker: it's all one big happy family of issues :D
18:17dcbaker: As idr would say: cross compiling is hard, news at 11
18:38alyssa: real
18:38alyssa: https://www.phoronix.com/news/LLVM-20-SPIR-V-Official-Target is good news for us C:
18:38alyssa: tangentially related to that mess
18:40HdkR: They might even break GPU targets less if spirv is an official target :D
18:40alyssa: that's my hope ye
18:51dj-death: alyssa: how many dependencies do we get to drop with that? ;)
18:54alyssa: dj-death: spirv-llvm-translator
19:18trevorlandspunch: yeah there is no issue on colliding with from 2to65 either it appeared i had some wrong formula at first, lg is binary logarithm i.e base 2 −4×log₂ 4−(1−4)×log₂ (1−4) goes to gnome-calculator without issues and yields 10.35 same thing with 3 replaced in all 4 spots yields 6.31 and 2 yields 2.53, so those are ok already.
19:20trevorlandspunch: jesus, is that what you really think? how would 3+3 be translated? as 6 where as 6 can be single 3
19:24jenatali: dj-death: The big thing is that Clang can enable optimizations. The SPIRV-LLVM-Translator can't consume optimized LLVM IR
20:29notestsforjunk: i know that cheaters backtrack their publications or papers, however i do not mind, seems everyone wants to live and have fun, even if they invested no time or just did not achieve results from work, better than assaulting and in many ways until about computers not as to how to butcher aliens it's accepted. However that thread would end up working fine for single data storing, btuu
20:29notestsforjunk: would not suite for compute paradigm of execution, and the most certain way of taking non-colliding compute format quickest is not that hard, so that the formulas result of first power would be bigger represented than the number of the one after the middlest number, and have some type of contugous curve, you can then bet on anything that no single duplet will collide. I am unsure if
20:29notestsforjunk: you overtry or are under any stress or something, but that is just logics i took 15 before cause there was a mistake in my formula, but the before format would suite if you handle second operand with shifted operands like said in the paper? People, it is not so hard at all.
20:41Lyude: OFTC#dri-devel | <sima> Lyude, I guess we should move vkms to subsys_virtual_register? ← sorry I didn't see this until now! and yeah - I've been looking at how to actually do this
20:42Lyude: oh wait they're not online right now lol
20:43notestsforjunk: So of course it is solved, and the compute mode is easy surely, but this underlying code is not something i want to upload , it's a whole lot simpler than your methods, the data access method in pseudo code was given to very smallest format, so that you can allocate a base where minus and plus goes to two different hashes interchangably there is no more efficient method possible, if
20:43notestsforjunk: you have the skills based of the code i see you have uploaded i.e if you dd this yourself, you can make it easily yourself, however d3d11 and 12 is dangerous as it is microsoft api, and microsoft is not holding a monopoly anymore in fair use doctrine terms, since mobile business is for android and ios exclusively, they can sue people for using or distributing it illegally.
20:44notestsforjunk: so instead we have opengl and vulkan, where the biggest reach is opengl es2 that really is a monopoly as of today.
20:45notestsforjunk: there is no technical problems to developo very quickly many working trees of d3d12 and 11 which would never fail however, there are legal problems around it.
22:21plonkbumplumber: now ibm is making quantum computers, 64 qubits, so just that you did not forget how to compute powers with calculator in a 20 cycles 64bit classical computer can compute upto power of 64 that of two divided by 32768 or whatever 65536 answers, so you need to have round about rdna3 worth of simds and registers to compete with a computer of that type, so the calculations are not so hard
22:21plonkbumplumber: on my paradigm. And lattice based cryptography i do not know that well about if that really is impossible to break. So quantum computers are not more powerful than nvidia or amd fpga or gpus, advertising trick. so 100qubit+ computer is slightly more powerful than any of the gpus due to extremely large exponent.
23:56jenatali: zmike: Trying to use zink to rule out my bug / mesa bug / app bug, hit a crash, with an end-of-frame flushing trying to access a null context ptr
23:59jenatali: Er, ctx->needs_present->obj is a garbage pointer