00:23porterfrank: The expression for \(\log_{i}(x)\) converted to base 10 is:\(\log _{i}(x)=\frac{\log _{10}(x)}{\log _{10}(i)}\)Using the principal value of the complex logarithm, this simplifies further to:\(\log _{i}(x)=\frac{-2i\ln (x)}{\pi }\) also with natural one, that all comes from eulers imaginary unit wikipedia article
00:25porterfrank: Nothing that computer systems required for any real performance was found in the mathematics, physics or computer programming papers, not sure if they were hiding all that
00:25porterfrank: so i invented this my own kind of
00:51porterfrank: however those papers carry significance to break ethereums encryption algorithm and hence find some way to financial freedom, same with RSA-1024 but the code to execute performance ways, was not found from the web, nor was data packing and access, but all that was already given to everyone, it's in the end very simple compiler.
00:52porterfrank: i am modifying nvidia quadro RTX 4000 driver
00:53porterfrank: i understand that the highest card today is 5090 gforce rtx
00:54porterfrank: but the amperes for running 4nm transistors are lower imo, and dependencies come in a way to take advantage of such massive card for modern games
01:05user8473: Well, that works, thanks. I just thought I had a distinct issue since I couldn't reproduce that one, but I guess I should have checked the patch first
01:16mhenning[d]: Oh, I just typed that patch up right now 🙂
01:17mhenning[d]: but yeah, we definitely need to figure out a proper fix. not sure if any of the developers have managed to reproduce it yet
01:20karolherbst[d]: I still believe it would be enough to simply disable it for buffer views
01:20karolherbst[d]: maybe I should check it on my turing tomorrow
01:26user8473: Well, I could have tested it with https://gitlab.freedesktop.org/mesa/mesa/-/commit/f3c53cf66bfcc38f3f54de01a16e40db0e7f79b6
01:31artemconca: They say on google AI, that 4nm works with higher clock frequencies on ADA, so Turing has larger transistors driving through larger voltages actually amperes as told, but generate more heat? This can't be true pals, you can put higher frequency pulser on smaller nodes to artificially show progress or linear growth in performance, but that is pretty much forcing or faking an angle. Smaller nodes leak more energy obviously, and this isn't behind any
01:31artemconca: clock cycles however
02:34ferencpolka: i remember this radeon way of telling one wavefront per wave/work-item i am sure you have not implemented this as of yet, this is where the real performance comes, i simulated this on GCN miaow. It was long time ago suggested by thomas stellard. But amd did not seem to like that idea of inhibiting the sales of graphics cards this way :P Well i suppose you can't even optimize any performance for single game, not to mention multiple parallel
02:34ferencpolka: contexts of graphics. pipeline of 3d is so bad and stagnated , it's not worth to waste time on those, one can't split the same shader to run on different SMs, so i am entirely unsure why they manufacture such gpus, i only noticed point based rendering talks, technically it should have the raw power to run such indeed on ada. Other than that i am sure that shader can not be split between sms, gpu register file does not work that way, global memory
02:34ferencpolka: is used to carry that between sms, and local and shared memory ocl terms between warps and work-items. Directx12/vulkan/metal may have some command queues to dynamically fill from vram but it ain't gonna fly well. 3d is messed up i stagnated. it's an interrupt that calls a new SMPE when last was ready and new one depends on the last i noted also. MSAA FXAA and such also use shared memory, it's fast cache
06:44visnigajevandrei: i would expect the SM's to be with so big resources at the moment, so if you are not rendering points and even if you are at 16K, you still can not that much parallelization opportunity from the shaders, it bottlenecks before turing already and this is clearer than clear, if you raise the compute unit resources yet more, than it no longer scales due to there is not so much async independent stream.
07:27Thinreyesfarm: So, for the record if anyone has not understood as to why i have to change the names to clear all the scams you are left with, there are genetically ill born tyrants at OFTC and stalkers in my life, those i call aids heros collaboratively, it is known that this is what they are, and abuse to me costs much less than earning money on their own, most of my friends have been under the same umbrella and family also, my cells to sell cost lot more than
07:27Thinreyesfarm: you as ill born people are ever capable to earn on your own, all my fans have been notified about this, but the war is going to be very complex , we will output some arsenal of real people continuously against such tyrants, i am not very highly optimistic, but worldwide conflict is in favor of getting this done without going to jail.
08:39nastyvaporcond: Originally there was nothing to do with the knee or muscle any everyone really knows that. Human leg is upper part has one bone and lower consists of two bones, and joint connects them together, i had osgood-schlatter on this lower bones upper so called knuckle, i am not developing financial problems but sports they did not allow me to play, and i loved sports more than anything, they took a very sophisticated victory against me, btw mark allen
08:39nastyvaporcond: queued the best i had ever seen him do and all that to miss a final black for yize to clear up, i had doubts on mark, but i see this very nice man is starting to play like a real deal now, previously i very much liked how he kicks out of snookers, now allround game is at marks reportoire, but luck did not favor him to become world champion, why i am talking about this? wst seemed to have played a tribute on me, at some point quarters were at 8:8
08:39nastyvaporcond: 8:8 8:8 3:5 where as john higgins and ronnie were at odds 88 144 lol, world has recognized my work. I graduate my missions without much trouble, i am mentally long time ago or have been nearly always stable.
14:23phomes_[d]: yesterday I did some tests of X4 at different resolutions:
14:23phomes_[d]: ` prop nvk %
14:23phomes_[d]: 1920x1080 84 73 87
14:23phomes_[d]: 3200x1800 63 31 49
14:23phomes_[d]: 3840x2160 48 22 46`
14:28phomes_[d]: I now wonder what the most relevant resoltion to do benchmarking at is? It would be great to try a few for each game, but then it becomes a really big task to update it
14:32karolherbst[d]: yeah....
14:32karolherbst[d]: I mean..
14:32karolherbst[d]: 4K@165 is the goal anyway 😛
14:33karolherbst[d]: ~~~and totally not because that's what my display is at~~
14:35marysaka[d]: ... I should grab a 4K display for my test bench
14:36marysaka[d]: or one of the small portable ones
15:01chikuwad[d]: marysaka[d]: gamescope
15:02chikuwad[d]: gamescope can advertise 4k to the game and downsample it to your monitor res
15:02marysaka[d]: Ooh
15:02chikuwad[d]: it's very handy
15:06mohamexiety[d]: phomes_[d]: 4k in general is the most gpu bound, but i really dont know why the % changes so much :thonk:
15:06mohamexiety[d]: could be going out of memory maybe?
15:06mohamexiety[d]: x4 shouldnt be that demanding though
15:07mohamexiety[d]: bandwidth requirements go up massively at 4k so it could be indicating that we're missing some special sauce here
15:07mohamexiety[d]: something like compression
15:11phomes_[d]: Compression did nothing for X4. Not at 4k or FHD
15:15karolherbst[d]: I wonder if it's compute related stuff.. but then why would it hit so hard on high res...
15:15karolherbst[d]: maybe postprocessing?
15:15karolherbst[d]: I've seen.... something strange in the dumps
15:16karolherbst[d]: nvidia is launching 691 compute shaders, we 37 times
15:16karolherbst[d]: so I wonder if the compute shaders they launch are memory copies?
15:16karolherbst[d]: if we use dma-copy for memory copies, then compression won't matter
15:16karolherbst[d]: it's going to be slow either way
15:17karolherbst[d]: phomes_[d]: mind sharing your gfxreconstruct as well so I can check out what API calls it's doing
15:20phomes_[d]: Vram use (on prop at least) is 5gb
15:31phomes_[d]: karolherbst[d]: gfxr at 3200x1800 is here: https://drive.google.com/file/d/1OOG6OD0g03pYObsoRJydEaVP-g5RIqpz/view?usp=sharing
15:31phomes_[d]: gfxr errors at replay when I do 4k
15:32karolherbst[d]: 1.4 GiB 🥲
15:34phomes_[d]: push dump from a replay of that in case anyone wants a look: https://drive.google.com/file/d/10M77UkDp5_XiIIbpDYOTjMmvxc9wCv6I/view?usp=sharing
16:01karolherbst[d]: https://cdn.discordapp.com/attachments/1034184951790305330/1500890036785254622/message.txt?ex=69fa1445&is=69f8c2c5&hm=bbaa97c588545936ee212a51247555db66c1c0a667536ebeaece256d796947a6&
16:01karolherbst[d]: not really sure what's up with the calls without a name..
16:01karolherbst[d]: but...
16:02karolherbst[d]: 691 >= vkCmdCopyBuffer + vkCmdDispatch
16:02karolherbst[d]: but 691 buffer copies? impressive
16:02karolherbst[d]: hopefully they aren't huge ones
16:09karolherbst[d]: sooo maybe doing copybuffer as meta or something would help? Would be a good opportunity to verify how fast/slow dma-copy is
16:11PhilAlbano: hey
16:11karolherbst[d]: https://cdn.discordapp.com/attachments/1034184951790305330/1500892679125602364/message.txt?ex=69fa16bb&is=69f8c53b&hm=c5a836bd65ee8bb933207668c5eb8a278007577ec932d452d4af318a061288e9&
16:11karolherbst[d]: sizes of those copies:
16:12PhilAlbano: how it is possible that 1050Ti provides h264 acceleration on novueau in firefox?
16:16karolherbst: PhilAlbano: it shouldn't...
16:16karolherbst: except it's software
16:19PhilAlbano: weird
16:19PhilAlbano: nouveau put same cpu load as prioprietary driver does
16:21karolherbst: well decoding h264 isn't that expensive
16:28PhilAlbano: yeah, about:support says that all hw decoders failure because of gfxInfo
17:54nastyvaporcond: It should be quite easy to rewrite the TCP/UDP in terms, of hanging up the connection as well as securing the communication, but myself i found something to work on again, i am putting allwinner arm boards on steroids for the sake of my prototype, i am participating alone in the driver development too. But my friend who uses the nvidia board does not use linux, so i am looking if i can install the driver on wsl2's linux that you had developed, or
17:54nastyvaporcond: directly to windows through some kind of winsys.
18:22pooraaltojouha: karolherbst, cause you were outsider in your life such as totally unskilled human being with the rest of your toxic ponies, you actually never knew what is real pain consistently committed at you, since no one is ever even hurt you, you put emphasis on word pain, and there is nothing painful in development of drivers at all. But i hope that cannibals and terrorists who enjoyed their success over causing pain to me, and
18:22pooraaltojouha: i almost know that already will get handled more and more with team work alike spawned against them. It wasn't that i collapsed out from athletics or sports which i was about to dominate years in a row, they committed terrible crimes against me. We are going to make them who were involved pay for this.
18:22pooraaltojouha: is/has
18:50pooraaltojouha: Well really i have not been sexually abused the way i had known, i got first class terror , making it probably the most complex case ever seen in human history, so as i speak the terrible fact around all of it was determined and fixed with major battle upfront, they made long lasting plan how in different years to injure me out of qualities in life. This battle still lasts, as i pushed from 13-18 years of age to
18:50pooraaltojouha: outplay this clan of real toxic people, so linux and systems have been even with persistent pain that i pose, a very positive landscape to me. This is where i fight my position in life back through wisdom and work.
19:28pooraaltojouha: genetic mutant aka survivor, i have gene's encoded like as for top two most fearless animals, wolverine and honey badger, i got too much help from god and basically nothing from human beings, only tyranny from there. So that is a question whether i am god king or not, i am survivor. And childhood traumas i do not have, it was yesterday just seeming when my muscle was butchered to mcdonalds at my own sight, but yes the
19:28pooraaltojouha: complex part was the battle from 13-18years old, when i saw this being happened, this was decided by my family against me and they planned it for years in a row as for every detail howto injure me, cannibals so to speak, unforgettable scenes similar to shikatilo andrei case sexual humiliation through words everyday i get but none will handle me accoring to rules, i am so much faster even with injuries.
19:39pooraaltojouha: As for network vulnerabilities , yeah aware of all of them, i do not expect anything from web3 either, piles garbage, so voltage shield here is under others control, i saw defcon faraday cage infiltration, and x86 reverse engineered ring bypass hide instructions, i mean it never bothers me much, i did not line up against united states though, but i am anyways at another side in the upcoming conflict, aka Russia, China,
19:39pooraaltojouha: India, Iran etc.
19:44_lyude[d]: airlied[d]: tbf - i don't feel like it needs to be mentioned, as long as you didn't actually generate code with it. I mostly mention that because if AI isn't actually writing the code maybe it's not a great idea to give it credit and encourage others to try to contribute with it?
19:46_lyude[d]: (that message was further back then I thought it was oops, it just popped up in my search history as I was looking for something)
19:53trinidadman: <pooraaltojouha> with my personal case, my own doctor from united states would had done the minor task without any issues, but i had no money, since i was not allowed to go, by the decision of my family who blocked my own assets for implementing their cannibal world show.
19:53trinidadman: <pooraaltojouha> the bug is back as expected on my bootrom.
19:53trinidadman: <pooraaltojouha> but i do the last talking there, we will beat all those people
19:53trinidadman: <pooraaltojouha> odds are 1 to 10 against me
19:53trinidadman: <pooraaltojouha> for you that is
19:53trinidadman: SO what else do you expect from cannibals? And their mindfart at TCP/UDP betrayal.
19:56_lyude[d]: i might despise llms but if you keep this up i will literally train one just to repeatedly remove you from the channel
20:49HdkR: _lyude[d]: I mean, might be the only reasonable path forward. A decade of heckling getting solved with abysmal LLM technology.
20:50karolherbst[d]: finally a real use
20:55marysaka[d]: phomes_[d]: I have some still WIP changes that revamp query pools here if you want to poke at some changes https://gitlab.freedesktop.org/marysaka/mesa/-/commits/nvk-query-rework
20:55marysaka[d]: I *hope* it will help some games but in the worst case this reduce query pool memory usage and reduce overhead on copies/clears
20:56marysaka[d]: https://cdn.discordapp.com/attachments/1034184951790305330/1500964438935732365/image.png?ex=69fa5990&is=69f90810&hm=4bd5483db250b021f278b50f6f4063a44f059a37cac2265fa7c1235ef844510d&
20:56marysaka[d]: it at least reduce the amount of commands on XCOM2 capture I have drastically but seems to not change perfs at all on it sadly
21:23reneklandorf: I sell all my x86 hardware and stop dealing with Anglo-Saxon fuckers, though likely this illegal terror is managed by the orders of Estonian country, all their cannibal abuse was in court with such people there’s not much to do same as osdev kinda graphitemaster idiots or fecalbrecht, for my fans I need to figure out how to remove the implant, linux as software is also hugest crap not only
21:23reneklandorf: tcp/udp and their viruses. having had closer look its catastrophic, I get hw and sw from China now on. there’s nothing to research from my side I never pay any money for you extorting fuckers anyhow, all the Cambodian time gangstawannabees were sent out with their blackmailing and Laura tornado can fuck it’s lesbians and cranks bums however it likes for the rest of its life, I am not
21:23reneklandorf: affiliated with it nor its bum fuckers from south africa to Australia to Sweden, none of those people I ever got along with.
21:23_lyude[d]: FINE I'll make the damn llm
21:25Lyude: reneklandorf: if you could save me some effort stop coming back, otherwise you're getting automatically banned by an llm into infinity because i have almost an entire decade of logs of you speaking.
21:25Lyude: annoying
21:37karolherbst[d]: _lyude[d]: maybe should also feed it past names and see if it can have a good enough hit rate on the name alone 🙃
21:38_lyude[d]: i mean my idea is quite literally just to look for really long messages
21:39karolherbst[d]: sure but also would be better if nobody would have to see it in the first place
21:39karolherbst[d]: but..
21:39karolherbst[d]: that might be going too far, so I'm more curious about flagging and pinging an op
22:11_lyude[d]: I'm not sure how much they could do
22:11_lyude[d]: just from having once been an IRC op ages ago
22:15dwfreed: There are some tools that weren't common in the past
22:15_lyude[d]: hm
22:15_lyude[d]: i'll poke them then
22:15dwfreed: They have become substantially less effective here
22:15dwfreed: You're talking to one :)
22:15_lyude[d]: oh haha
22:20dwfreed: the challenge with really long messages is that regular people can also be rather verbose
22:20_lyude[d]: yeah - I mean, I meant getting long messages for training data
22:20_lyude[d]: not "as the thing that bans people"
22:20dwfreed: ah
23:38phomes_[d]: marysaka[d]: Most games stay at same perf. Serious Sam improved by 2 fps
23:51marysaka[d]: phomes_[d]: Hmmm I see, thanks for testing!