#asahi-dev · 2023-12-15 14:32
eilnmillenialhacker: happy to hear :)#asahi-dev · 2023-12-15 09:31
eilnah, thank you#asahi-dev · 2023-12-15 09:12
eilnmpegs have the itu-t draft conformance samples but vp9 I can't seem to find any#asahi-dev · 2023-12-15 09:12
eilnor do I have to tinker with the encoder#asahi-dev · 2023-12-15 09:12
eilnjanneg: do you might know where to find vp9 test vectors / gnarly samples#asahi-dev · 2023-12-09 11:22
eilnVT (VideoToolbox), V4L2#asahi-dev · 2023-12-09 10:11
eilnthat brings us up to High 10 4:2:2. unsure about 265/vp9 yet#asahi-dev · 2023-12-09 10:10
eilnI thought it couldn't because VT rejected my 10-bit sample. turns out that sample was rejected because it was interlaced, not because it was 10-bit#asahi-dev · 2023-12-09 10:10
eilnawesome, avd can do 10-bit#asahi-dev · 2023-12-01 17:08
eilnthat's exactly what I'm not gonna do#asahi-dev · 2023-12-01 17:05
eiln"everything is a gpu" nvidia fault#asahi-dev · 2023-12-01 17:04
eilnwell we can't use it unless we can convince the vulkan people to support actual video decoders#asahi-dev · 2023-12-01 17:01
eilni509vcb: me? this is the video decoder not gpu#asahi-dev · 2023-12-01 14:56
eilnhevc works, and I think it'll be done before vp9#asahi-dev · 2023-12-01 14:56
eilnstatus update#asahi-dev · 2023-11-30 14:20
eilnniiice#asahi-dev · 2023-11-29 18:09
eilnwrite a bitstream parser, sorry#asahi-dev · 2023-11-29 18:06
eilnthis affects V4L2 UAPI#asahi-dev · 2023-11-29 18:05
eilnwriting one isn't hard, I'm wondering if we have to write one at all#asahi-dev · 2023-11-29 18:05
eilnsomeone's gotta parse the headers#asahi-dev · 2023-11-29 18:04
eilnis bitstream parsing handled entirely in hardware?#asahi-dev · 2023-11-29 18:03
eilnyeaah I'm pretty cautious about that. also I don't find disassembly very helpful#asahi-dev · 2023-11-29 18:02
eilnI only see encode in prores.py#asahi-dev · 2023-11-29 18:01
eilnyes#asahi-dev · 2023-11-29 18:01
eilnis decode blocked by a lack of a bitstream parser? if so we could add one to the avd repo#asahi-dev · 2023-11-29 18:01
eilnArcaneNibble: I might be able to port APR except I don't have one heh#asahi-dev · 2023-11-29 18:00
eilnWubbzy04: if you like video you could help fuzz/test coverage for avd#asahi-dev · 2023-11-24 18:58
eilnoh wait no if there are multiple processors we can definitely make use of them as long as their output doesn't get overwritten. but I haven't seen any evidence of macOS threading these. their firmware IPCs are interrupts and the command for executing VP vs. PP (which there's only one of) is handled inside a single IPC decode command. ig that's one thing we can do better#asahi-dev · 2023-11-24 18:49
eilnat least on mine it's not schedulable i.e. vp9 instruction stream has to go into the vp9 fifo, etc#asahi-dev · 2023-11-24 18:47
eilnmy understanding is that each fifo is the entrypoint to a separate processor. actually I don't think there's any other way because you can see it processing each step with each word written#asahi-dev · 2023-11-24 18:38
eilnif the other ones have multiple fifos (why?) it's possible it's hevc fifo#2 but I recall I couldn't get it to work with hevc. also the version number jump was suspicious#asahi-dev · 2023-11-24 18:36
eilnit's not#asahi-dev · 2023-11-24 18:36
eilnyeah it's probably just some prototype design#asahi-dev · 2023-11-24 18:26
eilnaah I'm gonna snooze off, we'll definitely investigate this ;)#asahi-dev · 2023-11-24 18:18
eilnis it 264/5 for sure? I can't imagine parsing macroblocks is a threaded task. maybe vp9 for threadframes but that's the only non-one#asahi-dev · 2023-11-24 18:15
eilnit has four FIFOs? how are they scheduled?#asahi-dev · 2023-11-24 18:04
eilnwe could crowdsource 'log show'#asahi-dev · 2023-11-24 17:53
eilnwhat soc was clover?#asahi-dev · 2023-11-24 17:52
eilnpost executing PP, it's supposed to poll and do two magic writes into the status register. that value slightly differs per codec. https://github.com/eiln/m1n1/blob/avd/proxyclient/m1n1/fw/avd/decoder.py#L92C9-L92C23. these are after PP is completely done so it doesn't affect the output but it "resets" PP for the following runs.#asahi-dev · 2023-11-24 17:44
eilnj293ap 13.5 viola, it's in the tarball#asahi-dev · 2023-11-24 17:43
eilnwe might be talking about different things, but it's not an infinite loop, it's polling the status register to change. it's why I added the w_40104060 hook#asahi-dev · 2023-11-24 17:40
eiln5, interesting.#asahi-dev · 2023-11-24 17:39
eilnif you unpack that into data/* you should be able to run my forked avd_emu.py#asahi-dev · 2023-11-24 17:39
eilncan you get this? https://www.dropbox.com/scl/fi/255zktwxy4z98bd31vlx9/avd-sample.tar.gz?rlkey=qj8mw6h1fqlbsa1lvwvknndd6&dl=0#asahi-dev · 2023-11-24 17:35
eilnlet me upload a tarball of j293 trace samples#asahi-dev · 2023-11-24 17:26
eilnbut I can't imagine a cortex-m3 is any faster, and I don't think that's worth the IPC#asahi-dev · 2023-11-24 17:25
eilnsorry https://github.com/eiln/avd/blob/main/avd_emu.py#L608-L609#asahi-dev · 2023-11-24 17:21
eilnavd has another power domain which is not in the device tree and not PMGR, lovely. it's enabled by write 0xfff I believe. it's gated on that. and I believe there is no "off" write#asahi-dev · 2023-11-24 17:19
eilnline 607 in avd_emu.py (heh) is the non-aic (PP) interrupt I believe. it does two writes at the polled status register which, if it doesn't, PP fails after the fifth-ish run#asahi-dev · 2023-11-24 17:18
eilnexecuting code in a power-gated range is such a nightmare that I'd like to avoid that if we can