11:14clayreinhard: It's easy to crack somebody, it's hard to train to be someone. You are bragging with people that cracked me. Their achievement was to be outlaws and traitors. It took nine years to develop my motor skills and strenghts to be at half-way of glory, and it was destroyed in 2hours with 5 different attacks prepared at me. So i ain't gonna listen to your excuses, you are causing me to lose, and
11:14clayreinhard: if you ain't gonna fuck off from my territory very high forces will help me to get rid all of you i am tired of greating a value to the world to get attacked rawly by brainless animals. It was my strategy not to learn too much idioticy in school, but it wasn't successful enough to avoid injuries. You wanted to know probably as to how those combinations are calculated, 8 samples from 32
11:14clayreinhard: combinations is round about 1024*1024 so 1 million and five thousand and some hundreds to be exact, now as you map those to between 1024 variants, you likely multiply with 1024, that yields to something like more than 1million, and 4 times overcommit is more than maximum, since some numbers are for state. Same with 64bit, so 16 subfields from 64 as sampled are 6*10 ^14 then multiplied with
11:14clayreinhard: 4096 and with overcommit from 4096 aka 10 or so, is maximum 64bit plus some to state on 64bit. What i talked to you is not my delusion it's how those thins are calculated, your friends from Estonia are deluded, we have real people too. the former people achivements are cracking me up, but real value they never provided and for illborn people that is also not possible.
11:16clayreinhard: more than 1billion was there a typo
13:27karolherbst[d]: it looks like the paste crate is unmaintained
13:29gfxstrand[d]: Is there a replacement? Or is it just good enough and no one needs to fix anything?
13:30asdqueerfromeu[d]: gfxstrand[d]: "If it ain't broke, don't fix it"
13:35karolherbst[d]: gfxstrand[d]: https://rustsec.org/advisories/RUSTSEC-2024-0436
13:35karolherbst[d]: https://crates.io/crates/pastey is mentioned
13:36karolherbst[d]: "pastey is the fork of paste and is aimed to be a drop-in replacement with additional features for paste crate"
13:36gfxstrand[d]: Okay. Easy enough to switch then
13:36gfxstrand[d]: IDK why they couldn't just pass the torch but whatever
13:37karolherbst[d]: who knows
13:37gfxstrand[d]: Ooh! They support changing case now!
13:37karolherbst[d]: the fork?
13:38karolherbst[d]: oh indeed
13:41gfxstrand[d]: I don't see anything about allowing pasting numbers, though. That's annoying. 🫤
14:39snowycoder[d]: New kepler bug, the `dEQP-VK.spirv_assembly.instruction.graphics.image_sampler.imageread.storage_image.optypeimage_mismatch.*` family throws a dmesg warning about unaligned addresses
15:10snowycoder[d]: Is it fine if we leave the unaligned warning?
15:10snowycoder[d]: The specs just say that mismatched formats return undefined texel values, and I guess it's pretty rare to find these cases out in the wild.
15:17gfxstrand[d]: It's fine Just another exception we should disable
15:17gfxstrand[d]: I've actually already triaged that one
15:41x512[m]: Sounds like a hack.
15:45gfxstrand[d]: No. I've analyzed it and it's fine.
15:47gfxstrand[d]: The issue is that if the client uses a 32-bit format in the shader and it's an 8 or 16-bit image then we end up doing unaligned 32-bit reads. It's fine because the bounds checking and offset calculations is all based on the surface info therefore still correct So we're never reading memory that doesn't exist. It's just an alignment problem.
15:52kar1m0[d]: What I am interested about is as to if the 80w restriction that I have on my gpu is an OEM/EC thing or an nvidia driver one.
16:53redsheep[d]: Isn't there something Nvidia did recently on windows laptops where they can see spare power budget from the CPU and hand that to the GPU? I wonder if that 145 you saw was only like that because your cpu was idle
16:58karolherbst[d]: there are power limits declared in the vbios
17:13chikuwad[d]: redsheep[d]: this is also a thing on linux
17:13chikuwad[d]: https://download.nvidia.com/XFree86/Linux-x86_64/580.76.05/README/dynamicboost.html
17:28x512[m]: Still called XFree86...
17:41snowycoder[d]: gfxstrand[d]: What bit should we disable? I cannot find any docs on `gr_gpcs_tpcs_sms_hww_warp_esr_report_mask`
17:48snowycoder[d]: Nvm, found a link scraping the IRC logs
18:09gfxstrand[d]: https://github.com/alliedvision/linux_nvidia_jetson/blob/master/kernel/nvgpu/drivers/gpu/nvgpu/include/nvgpu/hw/ga10b/hw_gr_ga10b.h#L847
18:09gfxstrand[d]: Should be bit 15
18:37snowycoder[d]: Ok, I'll reset that bit too and hope for the best.
18:37snowycoder[d]: I would really love to wire up the FALCON04 method in the open firmware
19:23gfxstrand[d]: It'd be neat but it's probably also the most time consuming way to solve this problem
19:25karolherbst[d]: ~~and probably also the most painful one~~
19:26karolherbst[d]: not because of the assembly code you have to write. That's painful for sure, but also the debugging part where it randomly doesn't work and you never figure out why
19:27kar1m0[d]: redsheep[d]: I am not sure. I got 145w by enabling performance mode in omen hub on windows.
19:27kar1m0[d]: It's all proprietary so I can't know
19:29kar1m0[d]: chikuwad[d]: I have it enabled but it does nothing
19:29kar1m0[d]: Like I haven't noticed any difference
19:29karolherbst[d]: well laptops drawing 145W for a short amount isn't unusual
19:30kar1m0[d]: I have a 280w power adapter
19:30karolherbst[d]: yeah...
19:30kar1m0[d]: So it shouldn't be a problem
19:30karolherbst[d]: well it's 280w for a reason 😄
19:31kar1m0[d]: It probably is a vbios/oem/ec thing that restricts the gpu to 80w
19:31karolherbst[d]: those laptops behave really strangely if you supply them with less
19:31karolherbst[d]: kar1m0[d]: well.. yeah, also clocks and sometimes the GPU can't go higher anyway
19:31karolherbst[d]: run furmark, lol
19:32karolherbst[d]: but yeah, it's capped in the VBIOS
19:32kar1m0[d]: I have asked about this on the Nvidia forum and someone with an Nvidia 5070 ti laptop has 110w
19:32karolherbst[d]: yeah, each OEM can configure it as they please
19:32kar1m0[d]: Hp....
19:33karolherbst[d]: it gets even weirder with MXM cards
19:33karolherbst[d]: also cooling is an issue
19:33karolherbst[d]: like
19:33karolherbst[d]: I doubt 110W is anything any laptop can cool
19:33karolherbst[d]: and then you have the rest of the laptop on top
19:33kar1m0[d]: karolherbst[d]: I don't own the laptop I buy 🙂
19:33karolherbst[d]: it's prolly great for 20 second benchmarks
19:33karolherbst[d]: and that's it
19:34karolherbst[d]: like yes, on paper some might go up to 110W, but in reality their gaming perf will be as bad as yours
19:34kar1m0[d]: karolherbst[d]: True, my cpu wasn't happy about it. It reached 92°C once
19:34kar1m0[d]: Because of the gpu
19:34karolherbst[d]: well that's normal
19:34kar1m0[d]: karolherbst[d]: Not really
19:34karolherbst[d]: it is
19:34kar1m0[d]: It shouldn't go above 85
19:34karolherbst[d]: well that's not how laptop run
19:35karolherbst[d]: laptops go to peak thermal limit, that's how it works
19:35karolherbst[d]: if the therm limit is 105ºC of the CPU, they'll go up to 105ºC
19:35sonicadvance1[d]: If you're not hitting 100c in a laptop then you're leaving performance on the table.
19:35kar1m0[d]: 100°C is critical temp
19:35karolherbst[d]: yeah then it's 100ºC
19:35kar1m0[d]: And isn't normal.
19:35karolherbst[d]: it is normal
19:36karolherbst[d]: claiming anything else just means you don't know how those laptops are designed and expected to run
19:36karolherbst[d]: it's perfectly normal
19:39karolherbst[d]: anyway, claims like "a laptop CPU should never reach 80ºC" is usually just fearmongering
19:40kar1m0[d]: karolherbst[d]: I didn't say that
19:40kar1m0[d]: I said that it isn't normal to reach 90°C+
19:40karolherbst[d]: it is normal
19:40karolherbst[d]: seriously
19:40karolherbst[d]: put some load on the CPU and you'll hit 100ºC
19:40karolherbst[d]: that's just how it works
19:41kar1m0[d]: karolherbst[d]: My cpu never hit 100°C
19:42kar1m0[d]: Does that mean that I never utilized it to 100%?
19:42kar1m0[d]: That's what I understood from your words
19:42karolherbst[d]: well some laptop vendors are more aggressive with cooling or just speed up fans really really early so you might not quite hit 100ºC all the time
19:43karolherbst[d]: but normally CPUs just ramp up to max clock and run for a brief time
19:43karolherbst[d]: and just hit max temp
19:43karolherbst[d]: and then they cool down again
19:43karolherbst[d]: often it's too quick to notice
19:43karolherbst[d]: sustained load _might_ lead to cooling being more aggressive
19:43karolherbst[d]: and the firmware throttling down more aggressively
19:43karolherbst[d]: depends on the model, vendor
19:44karolherbst[d]: but these days laptops generally hit the max temp and it's perfectly normal
19:44sonicadvance1[d]: Also throw ARM laptop results out the window, those are different beasts. Apple doesn't boost like Intel x86 laptops 😄
19:45karolherbst[d]: 😄
19:45snowycoder[d]: karolherbst[d]: Oh my problems are way before that, I'm not even sure what engine that FALCON04 is interrupting.
19:45snowycoder[d]: Nor what kernel assembly is loaded into what engine.
19:45snowycoder[d]: Envy-tools does not even mention the MME in its GPU introduction.
19:45snowycoder[d]: I can only dream
19:45karolherbst[d]: yeah well intel is quite aggressive there
19:45karolherbst[d]: but that makes websites load really fast
19:45karolherbst[d]: meteorlake is funky
19:46karolherbst[d]: https://gist.githubusercontent.com/karolherbst/6844edf77b18b67a810ca3c6ef3fcb0a/raw/74b39bb47b9a256f0d9f05d7ede78a3118d2e9e2/gistfile1.txt 🥲
19:46karolherbst[d]: `Package id 0: +104.0°C (high = +110.0°C, crit = +110.0°C)` toasty
19:46karolherbst[d]: intel bumped the limit to 105ºC kinda impressive
19:47sonicadvance1[d]: Woo! The lower temp cores the small cores?
19:47karolherbst[d]: yeah
19:47karolherbst[d]: there are now three types of cores 😄
19:48sonicadvance1[d]: The LPE performance is pretty rough
19:48jja2000[d]: sonicadvance1[d]: I 'member the "make it hotter" patch
19:48sonicadvance1[d]: jja2000[d]: Hey, I'm still running that patch.
19:48karolherbst[d]: sonicadvance1[d]: yeah...
19:48karolherbst[d]: pipewire has a moment somtimes, not sure what causes it, but under load sometimes audio cracks and...
19:48karolherbst[d]: pain
19:48karolherbst[d]: like apparently 22 cores with a oversubscribed thermal budget is a great idea
19:48karolherbst[d]: the CPU clock graph is just impressive
19:49jja2000[d]: sonicadvance1[d]: I do sometimes wish I was, but upstream good
19:49karolherbst[d]: some cores just ping pong between 4GHz and 400MHz 😄
19:49sonicadvance1[d]: karolherbst[d]: Oh god, my older Razer did that, but a firmware bug meant the cores pinned to 800Mhz after throttling 😄
19:49karolherbst[d]: oh no
19:49karolherbst[d]: I had a macbook where a thermal sensor was uhm.. broken
19:49karolherbst[d]: that was fun
19:50jja2000[d]: I had a 8750H/GTX1060 Dell and it would power throttle so hard, any games would just CTD
19:51jja2000[d]: The MOMENT Forza Horizon 3 switched to exclusive fullscreen it just crashed immediately
19:52jja2000[d]: Noped the fuck out of that thing and bought a T420 for 60€ off Marktplaats/Leboncoin/Craiglist/whatever
19:52jja2000[d]: Had enough money to go to my first CCC that year
19:56jja2000[d]: Can never catch me recommending a dell to anyone anymore
21:26gfxstrand[d]: I should not be contemplating implementing sync file import/export in nouveau GL...
21:29zmike[d]: what are you doing
21:30zmike[d]: FOCUS
21:56snowycoder[d]: My kernel patch works, but the barrier_bug test needs a job timeout of 10 minutes to pass 0_0.
21:57snowycoder[d]: Only 5 minutes with manual reclocking