00:05Lynne: no, I had no idea nvidia use off the shelf decoders
00:28dwlsalmeida[d]: avhe[d]: Btw I think these mv0 to mv2 buffers are associated with “use_temporal_mvs” which also happens to have three entries
00:32dwlsalmeida[d]: Lynne: We also have two parameters which control how much of the bitstream to skip, but for key frames they’re zero in the tracer
00:33dwlsalmeida[d]: I wonder if there should be something in them, the v4l2 driver has this “start_bit” value that’s used to offset within the tile data apparently
00:34Lynne: I think it should be 0 for all frames
00:35Lynne: though for av1, maybe you need to skip the uncompressed header
00:37dwlsalmeida[d]: What is this frameHeaderOffset thing? I see you always ser it to 0
01:37Lynne: just something to avoid needing to manually cut the OBUs out, probably
22:12gunsfrompast: So he was my only ally but he gows down with me, you asked about putin, sure I love him and many other russians he is biggest brain and soul of 21century to me, the best effort. china is second and turning point of all what is going to happen after putins and my death. they inserted two chips to me , one is lethal to heart and another to neck. so putin tries his best against jew animals in
22:13gunsfrompast: this world.
22:13gunsfrompast: they had inserted first ones at youth
22:26Jasper[m]: I hate to be awake at the time the spambots start
22:29redsheep[d]: That one may be spam but the repeated themes seem to indicate it's still the same person that some others here have mentioned has been doing it for like a decade
22:33Jasper[m]: oh
22:33Jasper[m]: In more on-topic news
22:35Jasper[m]: https://paste.centos.org/view/25d53068 can anyone tell what's happening here?
22:35Jasper[m]: the nouveau kernel driver seems to be crapping out for some reason, mesa falls back to llvmpipe afterwards
22:36karolherbst[d]: well.. nouveau in tegra isn't really supported, so it breaking is kinda part of "nobody tests it"
22:37karolherbst[d]: the firmware loading is also all different than on the normal desktop gpus
22:37magic_rb[d]: Was bout to ask what gpu this is, i saw "pmOS" and became intrigued
22:37Jasper[m]: In my experience it's been kinda wishy washy annoyingly
22:38Jasper[m]: I've even gotten other people with slight config changes have it work, I don't really understand anymore
22:38Jasper[m]: hell, over kernel releases it just starts working... or the opposite
22:39Jasper[m]: karolherbst[d]: It _has_ worked, that's why I don't understand anymore hahaha
22:39Jasper[m]: Also interesting that it seems to hit a fermi function it seems?
22:43Jasper[m]:uploaded an image: (1172KiB) < https://matrix.org/oftc/media/v1/media/download/AV6t9yGGxQkBSTf3qOCNJDDUqhY7SUCAj3gm9Q3SdKJszauCLxOROJh5lIiEcj-6eCqAmTAZcYD2q69ERQhIMj5CeU2Eh57AAG1hdHJpeC5vcmcvbXVxSnd2YXlVY0tTV3lYR0VrVUFsRHV6 >
22:43Jasper[m]: (same hardware for example)
22:52karolherbst[d]: Jasper[m]: well.. yeah, because there is no testing
22:52karolherbst[d]: and nobody works on it
22:52karolherbst[d]: might as well just remove the code
22:53karolherbst[d]: except somebody volunteers and does the work
22:56Jasper[m]: Bummer, I'm not the right person for the maintenance
22:56Jasper[m]: I do have the hardware though
22:56Jasper[m]: and UART
22:57ermine1716[d]: iirc tegra had its own kernel driver?
23:04mhenning[d]: skeggsb9778[d]: Hey! I'm using your 565 kernel branch right now to try and debug an MMU fault. Is fault_addr in dmesg meant to be 1) the instruction pointer of the instruction that caused the fault, or 2) the data address that can't be accessed?
23:05mhenning[d]: also, is there any key for what `fault_type:00000002` means?
23:06Jasper[m]: @_oftc_ermine1716[d]:matrix.org yes, I had that built as module in that dmesg, I now tried it built in and the display freezes again with no errors in dmesg
23:07Jasper[m]: though something still goes wrong it seems
23:12Jasper[m]: Nor Xorg really, it thinks it's running fine
23:14Jasper[m]: https://paste.centos.org/view/da645cd9 for reference, here's the dmesg
23:30skeggsb9778[d]: mhenning[d]: the fault addr can be from anywhere (you'll have to match it up to your gpu address space mappings)
23:30skeggsb9778[d]: mhenning[d]: take a look at dev_fault.ref.txt in open-gpu-doc, NV_PFAULT_ACCESS_TYPE is what you want
23:31skeggsb9778[d]: err, perhaps NV_PFAULT_FAULT_TYPE actually
23:34mhenning[d]: Thanks!
23:34gfxstrand[d]: NVK folks: How would people feel about having a weekly stand-up to touch base on NVK development so that we all know what's going on? We have enough people that it might be useful. Not the next couple of weeks, probably. We'll wait until I'm 100% back.
23:35mhenning[d]: I'd be up for that
23:36gfxstrand[d]: Collabora has a video chat thing or we can just use discord.
23:41marysaka[d]: sure 👍
23:45gfxstrand[d]: How do folks feel any times? Based on discord usage, it seems most people work on roughly US time even if they live in Europe. 🤭
23:47karolherbst[d]: if you want to potentially chime in folks from Australia there aren't many options
23:48karolherbst[d]: like... this hour and maybe in the slot in ~8 hours might work
23:49gfxstrand[d]: I'm fine with late US afternoon if the Europish folks are.
23:50gfxstrand[d]: But I'm also happy to fill Dave in later as needed.
23:50karolherbst[d]: yeah.. just bringing it up, because I know it's a pain to schedule 😄
23:50gfxstrand[d]: Yeah, I know
23:51mhenning[d]: US afternoon tends to work well for me
23:51karolherbst[d]: around this time is probably best, because Dave already plans for having meetings around this time
23:51gfxstrand[d]: I've thought about trying to live in .au or .nz for funzies and then I remember the time zone...
23:54marysaka[d]: That should be fine for me too :aki_thonk:
23:55karolherbst[d]: yeah... time zones are "fun"
23:58gfxstrand[d]: I want to make sure it's okay for mohamexiety[d] and phomes_[d] as well