00:18 mangodev[d]: mhenning[d]: really nice improvement only by a little plumbing
00:18 mangodev[d]: -# how did NVK not just start this way
00:21 julionitra: So this is evil what you do, and you get charged by this. We might argue that taking cells and substances from another host illegally with betrayal should not or could not by the help of real science injure almost fatally the host donor, are we that far with real science, yes long time ago probably, yet every Estonian real sportsman complained and showed that this is western world treating
00:21 julionitra: us as cows, however the injurless procedures were not even tried there, making it as devilish evil as possible, and later on when it became clear that muscle was only left with varicosis state after butchering which i saw in live, same with all junctions most dangerous to be ever touched not only was i silenced, none gave me treaatment upuntil today, and there were biggest born leftovers
00:21 julionitra: picketing on the streets and parties, who never were capable of accomplishing anything but just drink others blood as well as munch/consume steroids, today i am diagnosed with brain illness and go to work doing their lifting jobs, come to work doing their programming jobs, and am not excepted to paraolympics team either, so what am i missing here? I do not care who kills such devils, are
00:21 julionitra: they yanks in the end or Russians, such need to be killed if authorities do not do the job conspiredly enough. But there is one thing to argue should there be a net loss, that host cow gets injured? Let me explain those birth leftovers who were tried to be cured never turn into a champion state , cause they were not born to do such things and have the resilience to train and their brain was
00:21 julionitra: not coorinating anything from youth ages. Yes i know i was again chipped, they said upfront that there will be kimaroid in my ass and cancer caused on prostate gland, so that destroyed me with neck pain, the metal hand knee how big of a shit can they do more? Yes we kill those and their devil childrens the same, one hundred percent. But again should the science be at the level of not
00:21 julionitra: injuring the host, cause you do not make bread from shit, absolutely no, it was just pure evil.
00:22 julionitra: and every estonianian earlier spirtsmen said similar stuff about europes monkeys.
00:23 mhenning[d]: mangodev[d]: It didn't start this way because getting things working at all is a big enough task on its own. We haven't spent much time yet on improving shader compile times - we've been more focused on correctness and improving runtime performance
00:24 mangodev[d]: mhenning[d]: what was used before for hashing? custom function, or rust stdlib?
00:24 mhenning[d]: rust stdlib
00:25 mangodev[d]: i'd think the default would be something already in mesa, since mesa is in part about reusing code and all
00:25 mangodev[d]: like nir or the Vulkan runtime
00:25 mangodev[d]: what dictates whether code is reused or made custom in a mesa project?
00:30 mhenning[d]: There isn't an exact rule for what ends up in common code.
00:31 mhenning[d]: Figuring out where code reuse makes sense is one of the hard parts of writing software
01:11 mangodev[d]: mhenning[d]: fair
01:11 mangodev[d]: didn't know if there was anything different in mesa or not
01:11 mangodev[d]: good to know though
01:17 gfxstrand[d]: mhenning[d]: Okay, I've typed up the rest of the changes I intended for https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34994 and rebased it on your `FxHashMap` changes to avoid conflicts. It also contains a small phi rework which I think makes things more ergonamic and easier to understand.
01:19 gfxstrand[d]: I've meaning to do that one for a while...
01:59 gfxstrand[d]: mhenning[d]: Are the trace fails you're hitting with !34865 fixed somewhere or are you just throwing it at Marge again?
01:59 mhenning[d]: I'm just throwing it at marge again
01:59 mhenning[d]: I think they're flakes
02:02 mhenning[d]: mhh maybe not. maybe ci is broken at baseline
02:02 mhenning[d]: I don't think they're a result of that MR though
02:08 gfxstrand[d]: Looks ilke it passed this time
02:19 gfxstrand[d]: Ooes noes! Mine is kicking off the entire CI, too. 🙈
02:19 gfxstrand[d]: We really should make src/compiler/rust only kick off drivers with Rust compilers.
02:32 mangodev[d]: https://cdn.discordapp.com/attachments/1034184951790305330/1372763532999004170/image.png?ex=6827f52a&is=6826a3aa&hm=12f722b640586024eb9a0715a9d452d283ce259632b5880394db8d14ac88f55c&
02:32 mangodev[d]: any idea on what changes to my build script can fix this?
02:59 mhenning[d]: mangodev[d]: Either install your distro's rustc-hash package or allow it to use the fallback
03:01 mangodev[d]: mhenning[d]: arch doesn't have `rustc-hash`
03:01 mangodev[d]: but wouldn't i still have to add the cargo package to the build script in order for it to fallback?
03:03 mhenning[d]: right, on arch there isn't a package for it. you'll want to follow what arch does for the other crates
03:04 mhenning[d]: so add it to the _crates array and add checksums https://gitlab.archlinux.org/archlinux/packaging/packages/mesa/-/blob/main/PKGBUILD?ref_type=heads#L103
03:06 mangodev[d]: mhenning[d]: i'll be honest
03:06 mangodev[d]: i just add `SKIP` because i don't know how to add checksums
03:06 mangodev[d]: but even adding an extra `SKIP` still is too few? says the amount of checksums mismatches the packages, meaning they import in a different order than in the script 🫠
04:10 gfxstrand[d]: snowycoder[d]: Pushed you some SM30 `sustga` and `suldga` encodings. If we ever care about SM20, they use different ops.
04:15 gfxstrand[d]: But SM20 is Fermi so everything is different there with surfaces and textures
10:01 snowycoder[d]: gfxstrand[d]: Storage image lowering seems easier on fermi, but if we only support bindless it's not really useful.
12:10 asdqueerfromeu[d]: https://bugzilla.kernel.org/show_bug.cgi?id=220130 :nouveau:
12:50 snowycoder[d]: gfxstrand[d]: Where should we lower kepler shared atomics, in NIR or in NAK?
12:50 snowycoder[d]: for newer gens we do it in NIR but it doesn't know about load-and-lock/store-and-unlock, maybe it's simpler to do it in NAK directly.
13:19 gfxstrand[d]: NIR. You don't want to be building loops directly in NAK if you can avoid it.
15:40 snowycoder[d]: r0 p0 = ldslk.b32 [rZ]
15:40 snowycoder[d]: r2 = sel !p0 rZ 0x1
15:40 snowycoder[d]: p0 = isetp.ne.i32 r2 rZ
15:40 snowycoder[d]: Guess who stopped working? -.-
15:49 dwfreed: me when I tried to read that
15:49 dwfreed: :D
16:01 snowycoder[d]: lol
16:56 pavlo_kozlenko[d]: asdqueerfromeu[d]: Artem S. Tashkinov
17:05 ermine1716[d]: Bugzilla supervisor (self-proclaimed)
17:30 snowycoder[d]: gfxstrand[d]: I opened two MRs, one to fix copy-prop for `i2b(b2i(x))` in predicated execution, the other to implement shared-atomics in KeplerB. The code should be fine also for KeplerA, we just need the ops encoding (and testing).
17:46 gfxstrand[d]: Okay, cool
19:14 jja2000[d]: HdkR: *insert markov chain fueled rant about the weather here*
19:14 jja2000[d]: Slightly OT, but I added a disclaimer to https://wiki.postmarketos.org/index.php?title=Steam to add working games for fex-emu to the fex-emu wiki instead/aswell. So if you get more traffic to the wiki, you know why.
19:17 HdkR: Oh fancy. Not sure if I've ever looked at the egress stats of our wiki.
19:18 jja2000[d]: Page hasn't been updated since, but still, thought it was worth to mention to you
19:40 grandefinale: I made a mistake of dealing with you ever. Or #llvm people, who have few achivements on their name and say i am insignifficant and accomplish so little in fixing all of your delusions. I am the biggest achiever you have ever known, cause all what you have are mostly coming from my research labs. You were not thankful i taught you all the execution model and correcting the deviation, and
19:40 grandefinale: fuzzing things and whatnot. without correcting the code is just 112−72−36−140+24=-112 that fast is access, but it requires naturally alignment to 4digits in decimal for the sequences. which is achieved by 224 index 512 , this ended beatifully in my favour showing as to how deluded you have always been. Markov chain was never a thing, the real computation is expressed in pure high
19:40 grandefinale: mathematics with integral computation i.e derivatives, function curve is elliptic expected, and there is no prediction in that sense either, 360 degrees is 270 180 90 and trigonometry functions, so that is an old hack in mathematics, 360 degrees is whole which corresponds to 280 then 280-58 then 280-58-58 and then 280-58-58-58 to reach to the whole circle again in other words you are
19:40 grandefinale: mindill people. currently we have deviation on the algorithm cause it's not aligned to rules of derivatives in the sequence i.e rules of full circle in fixed integer computation for the last methods given, and MINDILL trash like you nor estonians were ever capable of doing any of the testing needed cause you lack sanity from the edge of the beginning of the time count.
19:43 HdkR: 🎉
21:14 HdkR: jja2000[d]: Your linking that wiki page got me to retest the only "No" game on the FEX side. Looks like Bioshock works here at least.
21:19 jja2000[d]: HdkR could be a thing they tried a while ago, could also be a limitation on their SoC
21:19 jja2000[d]: It's a 778G+
21:21 HdkR: Yea, was useful as a test case, since UE games are relatively well behaved.