01:05 airlied[d]: karolherbst[d]: probably should drop sector promotion from gallium driver https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41143
01:05 karolherbst[d]: heh...
01:06 karolherbst[d]: sure that this is what the bit does in nv50?
01:14 karolherbst[d]: anyway, I'm in ffma hell right now, ask me next week 🙃
01:22 airlied[d]: yes the bit is in the headers, just not used in the code, because you know why bother
01:28 karolherbst[d]: heh
02:46 airlied[d]: mohamexiety[d]: can you give https://gitlab.freedesktop.org/nouvelles/kernel/-/commits/nouveau-frl-support a spin?
02:46 airlied[d]: found one bug
03:12 airlied[d]: but I suspect it's DSC related
03:15 airlied[d]: pushed another fix that should disable dsc if it does gets turned on
03:21 asdqueerfromeu[d]: airlied[d]: Will there be more vibe-coded changes in nouveau kernel driver? 🤖
03:22 airlied[d]: depends on what you mean by vibe-coded
03:23 asdqueerfromeu[d]: airlied[d]: `This was claude code assisted.` implies it's partially vibe-coded
03:24 airlied[d]: no it implies I used claude to assist in developing it, vibe-coding isn't that
03:25 asdqueerfromeu[d]: airlied[d]: How about changes with that assistance then?
03:25 airlied[d]: depends if anyone submits changes made using assistance
03:25 airlied[d]: I don't think there is any plans to add new code to nouveau at all
03:27 asdqueerfromeu[d]: I hope that DP audio at least can be fixed
03:56 airlied[d]: I send claude off and it found two inconsistencies between openrm and nouveau around dp audio
04:05 airlied[d]: asdqueerfromeu[d]: if you feel like testing some totally untested stuff https://gitlab.freedesktop.org/nouvelles/kernel/-/commits/nouveau-dp-audio-fixes
10:06 mohamexiety[d]: airlied[d]: still the same behavior 🙁
10:06 mohamexiety[d]: setting it to anything higher than 60Hz fails
10:08 mohamexiety[d]: airlied[d]: It probably wasn’t DSC related but this was still needed probably. 4k120 doesn’t need DSC, but 4k240 does (and anything higher than 120 wasn’t in the menu)
10:12 airlied[d]: mohamexiety[d]: thanks, I'll have to try it on a bit more hw I expect
10:31 chikuwad[d]: time to build full CI
10:31 chikuwad[d]: and look at how I broke everything except NVK
10:33 phomes_[d]: Sottr benchmarks are now in the sheet. I plan to add more games with proper benchmarks. Hitman 3, borderlands 3, and warhammer 2
10:35 phomes_[d]: some of them show results that jump several fps for the same run. So not all are useful to check diff of new mrs
12:01 karolherbst[d]: phomes_[d]: even for borderlands 3? In my testing it was always fairly stable
13:44 karolherbst[d]: or well.. did you run the benchmark or just measured in game?
13:44 phomes_[d]: it is not the worst but still jumps. I just did 3 runs with ultra settings on current mesa main: 36.67, 37.90, and 37.26
13:44 karolherbst[d]: ahh
13:44 phomes_[d]: benchmarks
13:46 phomes_[d]: I will do the benchmarks and add the data to the sheet. Perhaps I will highlight which ones are not stable
13:55 karolherbst[d]: you could display the variance, but yeah... normally you want to automate it, do 3-5 runs and enter the avarage + variance
14:25 klaidonepper: so the release of Estonian science mainly from a man who held the country on his shoulders, you were lacking boolean algebra as last thing. so 12+12+110-42=92 is the reference point 12 and 12 are the low states 16 are high states for the same power X, so 12+16+110-42=96, 16+16+110-42=100 , so the forward states are 0 4 or 8 so final carry applied states are 0 , 4 or 8, where as single operand states in the middle of the algebra are 92-12-110+42=12
14:25 klaidonepper: 96-12-110+42=16 and 100-12-110+42=20 so if that was first power than do the same with next power but out states are now 16 and in states 20, so this will get 12 and 16 etc. I am tired of your scams dudes, I ain't gonna hold Estonian scammers thieves and tyrants on my shoulders anymore either, all the abuse parties get hurt from my lines. Contract with me and military and new world order is that bad people willl be booted out.
14:42 klaidonepper: I implemented everything to you, and you do not want to fight the devil for the honor of Estonian science, that is already coward type of fraud.
17:21 karolherbst[d]: Pushing my FFMA MR through the CTS: `ERROR - Test dEQP-VK.sparse_resources.shader_intrinsics.2d_array_sparse_fetch.r10x6g10x6_unorm_2pack16.503_137_3: UnexpectedImprovement(Pass): See "/data/git/VK-GL-CTS/out-14660/c13.r1.log"` 🙃 okay....... I hope that's unreated
17:21 karolherbst[d]: huh...
17:21 karolherbst[d]: other sparse tests also improve
17:41 karolherbst[d]: `Pass: 299678, UnexpectedImprovement: 26, ExpectedFail: 3, Warn: 1, Skip: 304292, Duration: 21:26, Remaining: 1:21:44` 🥲
17:42 karolherbst[d]: yeah.. it's all `12x4` or `10x6`
17:50 hanspietrich: You should me awarding science breakthroughs per person indeed cause most of the science you were unaware about was not committed by the Swedish Norwegians nor Finnish anal individuals who are abortion leftovers just like you, so are most of the Estonians, but soon comes the news of the slaughters that hit you one my one one after another, it's because we had enough of such bullies, so technically mark my words that most of your fecalist lineup
17:50 hanspietrich: is soon dead, same happens also to Estonian bullies and my family who had betrayed me. It's ridiculous as to what they do. You have a lot of courage to come to me, saying that i am nothing and no one where i feed all my family out of my blood, notes that a real man is my dad, and real business is done by my sisters and shit like this, where it is clearly known those snakes and monsters are tyrants in addition to being abortion leftovers. And we
17:50 hanspietrich: had enough of such people. I do not know what would anyone expect from anal enthusiast like mwk or other transgenders here, who have never achieved neither worked through their whole lives. Yeah only misleading nobel shows and conspiracy. it does not matter what the country is, russians estonians norwegians swedish they all get their purest steroids from my resources, and same with the code, we can proudly name all of them abortion leftovers,
17:50 hanspietrich: but there is a difference, some are thankful and have some moral and beliefs like Russians, and some traitors like you are moral less. As said it does not last long all your ponies will be executed quite soon.
20:10 karolherbst[d]: okay.. gonna try out HDMI 2.1
20:10 karolherbst[d]: wish me luck
20:11 jja2000[d]: HE'S BACK wow
20:17 karolherbst[d]: airlied[d]: getting `[ 201.835822] nouveau 0000:01:00.0: gsp: cli:0xc1d00001 obj:0x00730000 ctrl cmd:0x0073029a failed: 0x00000065`
20:18 karolherbst[d]: and it also only detects 120 Hz out of 165Hz
20:18 airlied[d]: If it needs DSC then it won't show up hopefully
20:19 karolherbst[d]: well it doesn't even work on 60hz, the display stays black 🥲
20:19 airlied[d]: Did you try patches or the branch?
20:19 airlied[d]: But that error isn't something I've seen
20:20 karolherbst[d]: the branch
20:20 karolherbst[d]: I'm trying my 4K TV now
20:20 karolherbst[d]: that only supports up to 120Hz anyway
20:20 airlied[d]: Cool. Guess I should try more diverse set of GPUs. I've had 120 and 144 work on capture card
20:22 karolherbst[d]: okay.. works on my TV at 4K@120
20:22 karolherbst[d]: but switching to 4K@60 breaks
20:23 karolherbst[d]: mhhhhh
20:23 karolherbst[d]: airlied[d]: okay soo swtiching from 60 to 120 also breaks 🙃 So the initial modeset on boot seems to work and then I see the gsp call failing when trying to change refresh rates
20:24 karolherbst[d]: maybe something about FRL and non FRL transitions?
20:25 airlied: that's link training failing
20:26 airlied[d]: okay yes that msg is link training timedout
20:28 airlied[d]: probably need to juice up link training fallbacks/retries
20:28 karolherbst[d]: `[drm:drm_parse_cea_ext] [CONNECTOR:46:HDMI-A-2] HF-VSDB: max TMDS clock: 600000 KHz, HDMI 2.1 support: yes, DSC 1.2 support: yes` heh
20:29 karolherbst[d]: but yeah.. seems like it doesn't like this display..
20:29 karolherbst[d]: maybe HDR in the way?
20:30 karolherbst[d]: let me try with HDR disabled, just in case
20:32 karolherbst[d]: yeah... somehow nouveau doesn't like this display
20:32 karolherbst[d]: I'll try stock nouveau next in case it's already broken
20:32 karolherbst[d]: _though_ with my TU102 card that display always worked...
20:33 karolherbst[d]: ohh that was displayport
20:33 airlied[d]: there is nouveau-hdmi-frl-hacks with some untested try link train harder
20:34 airlied[d]: what gpu are you testing with?
20:34 karolherbst[d]: on FHD@60 I get `drm: wndw-0: timeout` 🥲
20:34 karolherbst[d]: airlied[d]: `GA103M [GeForce RTX 3080 Ti Mobile]`
20:34 airlied[d]: the handover from bios to nouveau isn't well done in the nouveau code, so many holes to fall down
20:35 airlied[d]: the main blocker on the spark is the uefi setup is very different so nouveau confuses the hardware
20:35 karolherbst[d]: heh...
20:35 karolherbst[d]: not surprising
20:36 airlied[d]: Gotta go walk bbl
20:36 karolherbst[d]: I'll try your new branch but it seems like nouveau just chokes on that display on that GPU anyway...
20:37 karolherbst[d]: okay.. 6.19 nouveau seems to be doing alright at 4K@60
20:38 karolherbst[d]: will try the hacks branch
20:44 mhenning[d]: karolherbst[d]: I haven't seen that test fail but some sparse residency stuff was fixed for me by cts 1.4.5.3
20:44 karolherbst[d]: airlied[d]: so it looks like it just doesn't work, like no error on the nouveau side, but display stays black: https://gist.githubusercontent.com/karolherbst/6fbf38bc35e40f8e868be0f3df16bbd4/raw/fa9d73e9290aef4ae87474863b68ffd63474b547/gistfile1.txt
20:44 karolherbst[d]: mhenning[d]: heh.. I havne't updated the CTS yet
20:45 karolherbst[d]: still on 1.4.5.0
20:49 airlied[d]: Link training times out 5 times, okay probably need to look into that and reducing frl rate
20:51 karolherbst[d]: ohh didn't see those fails...
20:52 karolherbst[d]: mhhh
20:52 karolherbst[d]: let me check something
20:53 karolherbst[d]: okay so the display can do FRL6
20:53 karolherbst[d]: in case that's important to know
20:54 karolherbst[d]: I think that's the 48G one
20:55 airlied[d]: I should test with my super long shitty HDMI cable
20:55 karolherbst[d]: heh....
20:55 karolherbst[d]: okay but my cable isn't shitty 😄 it's one of those 16k HDMI 2.2 ones 😛
20:56 matt_schwartz[d]: but is it gold plated
20:57 karolherbst[d]: `Military Durability】 Triple Shielding & High Quality Workmanship - Triple braided nylon fabric, aluminum housing and gold-plated connectors resist EMI interference and survive 45,000+ bends. 28AWG tinned copper cables guarantee stable 8K HDR signals without dropouts. ` 😄
20:58 karolherbst[d]: I think I paid like 22€ 🙃
20:59 karolherbst[d]: anyway.. I kinda doubt it's anything to do with the connection, just something not correctly set up
21:01 karolherbst[d]: used the same cable on the 4K@120 TV anyway