| + rottensox (~rottensox@unaffiliated/rottensox) | 00:30 | |
| + BArray (~Thunderbi@125-239-58-43-fibre.sparkbb.co.nz) | 00:43 | |
| - B[] (QUIT: Ping timeout: 265 seconds) (~Thunderbi@125-239-58-43-fibre.sparkbb.co.nz) | 00:45 | |
| * BArray -> B[] | 00:45 | |
| + BArray (~Thunderbi@125-239-58-43-fibre.sparkbb.co.nz) | 01:31 | |
| - B[] (QUIT: Ping timeout: 256 seconds) (~Thunderbi@125-239-58-43-fibre.sparkbb.co.nz) | 01:33 | |
| * BArray -> B[] | 01:33 | |
| - erlehmann (QUIT: Ping timeout: 265 seconds) (~erlehmann@46.114.38.105) | 03:07 | |
| + erlehmann (~erlehmann@46.114.39.247) | 03:21 | |
| + addy (~tux@2a00:6020:21d9:6f00:7b75:b602:7464:4929) | 12:04 | |
| + max30 (509730ea@p509730ea.dip0.t-ipconnect.de) | 13:16 | |
| - max30 (QUIT: Remote host closed the connection) (509730ea@p509730ea.dip0.t-ipconnect.de) | 14:13 | |
| - Jookia (QUIT: Remote host closed the connection) (~Jookia@gateway/tor-sasl/jookia) | 15:52 | |
| + Jookia (~Jookia@gateway/tor-sasl/jookia) | 15:53 | |
| ezequielg | mntmn: \o | 15:54 |
|---|---|---|
| ezequielg | mntmn: please pick https://git.linuxtv.org/media_tree.git/commit/?id=88d06362d1d052e4c844ac95a2ca308ed4d90452 and https://git.linuxtv.org/media_tree.git/commit/?id=28a202c55963386b8bc45bcc52029362e9aa0d33 | 15:55 |
| ezequielg | (you can pick all the patches for hantro from media_tree.git, but those two seems to fix the artifacts I'm seeing here) | 15:56 |
| ezequielg | i'll mark them for stable if needed. | 15:56 |
| ezequielg | we also need to merge "arm64: dts: imx8mq: enable Hantro G1/G2 VPU". | 15:57 |
| ezequielg | BTW, you can use the post-processor as well, and get YUY2. ... parsebin ! h264parse ! v4l2slh264dec ! video/x-raw,format=YUY2 .. | 15:59 |
| ezequielg | would be interesting to test that on wayland, dunno if you have waylandsink. | 15:59 |
| - Jookia (QUIT: Ping timeout: 240 seconds) (~Jookia@gateway/tor-sasl/jookia) | 16:17 | |
| + Jookia (~Jookia@gateway/tor-sasl/jookia) | 16:18 | |
| - josch (PART: !!unknown attribute: msg!!) (~josch@fulda099.startdedicated.de) | 19:04 | |
| mntmn | ezequielg: thanks, gonna try it now | 19:28 |
| mntmn | ezequielg: great! these patches fix all the glitches | 19:35 |
| mntmn | ezequielg: (not for 4:4:4 but that was expected) | 19:35 |
| mntmn | ezequielg: but the other videos are ok now! really cool! | 19:35 |
| cphealy | ezequielg: for the 4:4:4, is it as simple as applying a patch to the kernel driver that changes the capabilities advertisement? | 19:46 |
| cphealy | What I mean to say is a patch to make the driver not try 4:4:4 at all that is. | 19:47 |
| ndufresne | cphealy: wehaven't written the patch yet, but in the SPS there is the chroma_idc, the driver should fail if chroma_idc == 3 | 19:48 |
| ndufresne | cphealy: note that otherwise the driver do report it's supported profiles, I need to implement this in gst (well copy it from the existing v4l2 plugin mostly) | 19:49 |
| ndufresne | when that's in place, decodebin will simply skip over to software decoder on 4:4:4, which is better then garbage ;-P | 19:49 |
| ndufresne | ezequielg: thanks a lot, I knew that glitch was familar, but I didn't remember that it was these patches fixing it | 19:50 |
| cphealy | ndufresne: sorry, incorrect words on my part. When I said apply a patch, I meant create and apply a patch. ;-) | 19:51 |
| cphealy | I understand what you are saying now. | 19:51 |
| ndufresne | ah, so then the answer is just yes, we need a patch to prevent invalid use cases | 19:52 |
| ndufresne | cphealy: for 4:2:2, we are not sure what's the pixel format, so the patch require a bit of research, but ezequielg suffered through getting the reference software to run, so it will be easier to find answers from now on | 19:53 |
| ndufresne | I guessing it will be this NV16 | 19:55 |
| cphealy | Hmm, I wasn't aware that the Hantro G1 supported 4:2:2 or NV16, at least from my read of the NXP reference manual. Perhaps it's just not documented there? | 19:58 |
| ndufresne | let me double check, my memory was telling me that G1 had chroma-id up to 2, and G2 was limited to up to 1 (4:2:0) | 20:01 |
| cphealy | Ahh, never mind, I see the reference to 4:2:2 | 20:01 |
| cphealy | Yea, G1 supports 4:2:0 and 4:2:2 while G2 supports 4:2:0 only (at least in the version of i.MX8M reference manual I'm looking at. | 20:02 |
| + bnys (~bnys@c-71-204-184-118.hsd1.ca.comcast.net) | 20:07 | |
| bnys | hey lukas just curious about a couple of things rn: 1) does the kit include rubber feet or will we be picking our own? 2) what's the microUSB on the motherboard for and is it going to be in the final revision? | 20:08 |
| bnys | i've rarely been this excited to see a credit card charge go through lol | 20:09 |
| bnys | additionally will the pinatex sleeve be available through your online store? | 20:18 |
| ndufresne | cphealy: have you notice REG2 SW_DEC_LATENCY ? | 20:31 |
| ndufresne | It let you skip clocks, in step of 8 clock cycles, to slow down decoding, got a hard time figure-out what it's used for | 20:32 |
| ndufresne | maybe that's a trick for better power consumption | 20:32 |
| mntmn | bnys: rubber feet are included :) and the microusb is for flashing the LPC (cortex-a0 microcontroller) | 20:41 |
| mntmn | bnys: the sleeve will be available separately later, yes | 20:41 |
| bnys | fantastic, thanks :) | 20:41 |
| mntmn | is there any video player that i could build that would easily leverage this version of gstreamer ndufresne? | 20:53 |
| ndufresne | mntmn: you mean that would enable zero-copy to let's say GL ? | 20:58 |
| ndufresne | mntmn: I didn't find any polished apps that handle this right, though the simple player in https://gitlab.freedesktop.org/gstreamer/gst-examples | 21:06 |
| ndufresne | the QT/QML one | 21:06 |
| ndufresne | will provide you seeking, subtitles, and zero-copy to GL | 21:06 |
| ndufresne | it would need some hack to try and force YUY2 instead of NV12, despite the higher bitrate this format runs faster on etnaviv | 21:07 |
| ndufresne | but if you first give it a try as-is, that would at least confirm that QML/GL render is indeed the smothest | 21:07 |
| ndufresne | the GTK one is using GTK3, which unfortunatly do GL to exgternal buffer, and then copy that into a SHM that is sent to the compositor | 21:08 |
| ndufresne | mntmn: and to build that qt example, clone that repo, cd playback/player/qt, qmake-qt5, make | 21:10 |
| ndufresne | not sure why it's not in the meson build | 21:10 |
| mntmn | ndufresne: ah, basically what i want to try is a player that can handle audio and resizing, play-pause as well as playback | 21:19 |
| ezequielg | ndufresne: can you check if yuy2 works when you use decodebin? Any reason why I might have seen it's not working? | 21:34 |
| + adjtm_ (~adjtm@80.red-79-150-248.dynamicip.rima-tde.net) | 22:27 | |
| - adjtm (QUIT: Ping timeout: 256 seconds) (~adjtm@80.red-79-150-248.dynamicip.rima-tde.net) | 22:29 | |
| - bnys (QUIT: Quit: Leaving) (~bnys@c-71-204-184-118.hsd1.ca.comcast.net) | 23:29 | |
| - B[] (QUIT: Ping timeout: 246 seconds) (~Thunderbi@125-239-58-43-fibre.sparkbb.co.nz) | 23:35 | |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!