<!-- Some styling for better description lists --><style type='text/css'>dt { font-weight: bold;float: left;display:inline;margin-right: 1em} dd { display:block; margin-left: 2em}</style>
***: xiaer1921 has quit IRC (Remote host closed the connection)
<br> xiaer1921 has joined #linux-media
<br> lexano_ has quit IRC (Ping timeout: 480 seconds)
<br> BrianG61UK has joined #linux-media
<br> mrpops2ko has quit IRC ()
<br> mrpops2ko has joined #linux-media
<br> camus has joined #linux-media
<br> jarthur has joined #linux-media
<br> mrpops2ko has quit IRC ()
<br> mrpops2ko has joined #linux-media
<br> darkapex has joined #linux-media
<br> xiaer1922 has joined #linux-media
<br> xiaer1921 has quit IRC (Ping timeout: 480 seconds)
<br> xiaer1922 is now known as xiaer1921
<br> eelstrebor has quit IRC (Quit: Ex-Chat)
<br> nickware has quit IRC (Quit: kthxbye)
<br> nickware has joined #linux-media
<br> mvaittin has joined #linux-media
<br> jarthur has quit IRC (Quit: Textual IRC Client: www.textualapp.com)
<br> djrscally has joined #linux-media
<br> shoragan has quit IRC (Quit: quit)
<br> ao2 has joined #linux-media
<br> mtretter has quit IRC (Remote host closed the connection)
<br> GBenji has joined #linux-media
<br> ao2 has quit IRC (Quit: Leaving)
<br> frieder has joined #linux-media
<br> xroumegue has quit IRC (Ping timeout: 480 seconds)
<br> xroumegue has joined #linux-media
<br> epoll has quit IRC (Ping timeout: 480 seconds)
<br> epoll has joined #linux-media
<br> mripard has joined #linux-media
hverkuil: <u>GBenji</u>: please check my reviews of patch 1+2 of your v16 series. I would prefer to get those sorted done and merged before working on the remainder. When those two (or three, since I think the first patch needs to be split up) are in, then I can work independently on fixing drivers that incorrectly use min_queued_buffers.
***: xiaer1922 has joined #linux-media
<br> xiaer1921 has quit IRC (Remote host closed the connection)
<br> xiaer1922 is now known as xiaer1921
<br> frieder has quit IRC (Remote host closed the connection)
<br> frieder has joined #linux-media
<br> xiaer1922 has joined #linux-media
<br> xiaer1921 has quit IRC (Ping timeout: 480 seconds)
<br> xiaer1922 is now known as xiaer1921
GBenji: <u>hverkuil</u>: Thanks for the review, I'm working on fixing your remarks
***: lexano_ has joined #linux-media
<br> wallacer has quit IRC (Ping timeout: 480 seconds)
<br> wallacer has joined #linux-media
<br> mtretter has joined #linux-media
<br> shoragan has joined #linux-media
<br> wallacer has quit IRC (Ping timeout: 480 seconds)
<br> wallacer has joined #linux-media
<br> mrpops2ko has quit IRC ()
<br> mvaittin has quit IRC (Ping timeout: 480 seconds)
<br> mvaittin has joined #linux-media
<br> ten157237743246 has quit IRC (Remote host closed the connection)
<br> ten157237743246 has joined #linux-media
<br> xiaer1922 has joined #linux-media
<br> xiaer1921 has quit IRC (Ping timeout: 480 seconds)
<br> xiaer1922 is now known as xiaer1921
<br> eelstrebor has joined #linux-media
<br> mvaittin has quit IRC (Ping timeout: 480 seconds)
<br> xiaer1922 has joined #linux-media
<br> xiaer1921 has quit IRC (Remote host closed the connection)
<br> xiaer1922 is now known as xiaer1921
<br> GBenji has quit IRC (Ping timeout: 480 seconds)
<br> frieder has quit IRC (Ping timeout: 480 seconds)
<br> GBenji has joined #linux-media
<br> xiaer1922 has joined #linux-media
<br> xiaer1921 has quit IRC (Ping timeout: 480 seconds)
<br> xiaer1922 is now known as xiaer1921
hverkuil: <u>GBenji</u>: finished the review of your v16 series.
GBenji: <u>hverkuil</u>: Thanks I will work on that tomorrow
<br> <u>Hverkuil</u>: I send you 2 patches for 6.8, the one you requested on review + a one I found will working on next version
***: GBenji has left
<br> dbrouwer has left The Lounge - https://thelounge.chat
<br> xiaer1922 has joined #linux-media
<br> xiaer1921 has quit IRC (Ping timeout: 480 seconds)
<br> xiaer1922 is now known as xiaer1921
<br> ao2 has joined #linux-media
<br> BrianG61UK has quit IRC (Read error: Connection reset by peer)
<br> jarthur has joined #linux-media
<br> djrscally has quit IRC (Ping timeout: 480 seconds)
<br> xiaer1922 has joined #linux-media
<br> xiaer1921 has quit IRC (Remote host closed the connection)
<br> xiaer1922 is now known as xiaer1921
<br> gouchi has joined #linux-media
<br> djrscally has joined #linux-media
<br> xiaer1922 has joined #linux-media
<br> xiaer1921 has quit IRC (Ping timeout: 480 seconds)
<br> xiaer1922 is now known as xiaer1921
linkmauve: Hi, we have an issue with JPEG decoding in cedrus: for 4:2:2 and 4:4:4 some old hardware supports outputting only to NV16, not NV12 like we’ve exposed so far, but we can’t know which kind of chroma subsampling the user will give us at S_FMT time.
<br> Plus it could even change from one JPEG to another.
mriesch: hey all... how could a camera sensor with two or three virtual mipi CSI channels be represented?
jernej: Making request api based JPEG decoder would solve more than one problem in Cedrus. Would that be of interest?
<br> <u>hverkuil</u>: ^
mriesch: would it feature one pad per vc?
pinchartl: <u>mriesch</u>: one pad, using the streams API
<br> three vcs on a single pad
linkmauve: I’m not sure the request API would help here, because at the point the request is queued the driver has exactly the same information as currently with the kernel JPEG parser, and still can’t change the capture format.
pinchartl: see the ds90ub960 driver for an example, and not how, in ub960_set_fmt(), the format->stream field is used
mriesch: <u>pinchartl</u>: nice, thanks for the pointer
jernej: <u>linkmauve</u>: proper steps for request api decoding: 1. select source format, 2. set controls (depends on the codec which one), 3. select output format, 4. allocate buffers, 5. start streaming, 6. send jobs
mriesch: are there already camera sensor drivers that use this approach?
<br> it seems there are some sensors that use one vc for short exposure, one vc for long exposure,...
jernej: <u>linkmauve</u>: in step 2, Cedrus would see which JPEG format is selected and thus limit number of formats for step 3
linkmauve: Oh so in that order? For JPEG encoding (with hantro on rockchip) I’ve been setting the controls between 5 and 6; for VP8 decoding I’ve been setting them before QBUF.
<br> The V4L2 API is quite confusing in that it accepts calls in the wrong order. ^^'
<br> I’ve also been doing the steps 5 and 6 in the wrong order lately.
jernej: it's documented in kernel docs :)
<br> API doesn't enforce order
pinchartl: <u>mriesch</u>: not yet. I'm working on support for a camera sensor that sends image data and embedded data in different data types
<br> from a uapi point of view, vc or dt don't make a difference, they're all streams
mriesch: <u>pinchartl</u>: would userspace be responsible for correct linking?
<br> i.e. correct assignment of streams vc/dt between e.g. sensor and isp
pinchartl: userspace is responsible for routing streams, but doesn't see VCs or DTs as such
<br> the VC and DT values are internal to the kernel
<br> sinks can query sources for the VC/DT associated with a stream using the .get_frame_desc() operation
***: BrianG61UK has joined #linux-media
paulk: <u>jernej</u>: well I think it does enforce order for format negotiation
<br> i.e. first set coded format, then relevant controls, then picture format
***: ao2 has quit IRC (Quit: Leaving)
paulk: and setting coded format should always reset picture format
<br> which maybe isn't currently the case in cedrus
jernej: ah, yes, I think I added that
paulk: but that's what I do in my rewrite
<br> ah good!
jernej: any ETA on rewrite? :)
paulk: <u>jernej</u>: so my current situation is that I left Bootlin since January 1st
<br> I'll be working as a self-employed contractant
<br> but before that I'm taking some time off to make a movie
<br> which takes a crazy amount of preparation
jernej: oh, interesting
paulk: so I still have a bit of time for free software stuff
<br> but mostly I'm writing a script, doing camera tryouts, comparing lenses
<br> and preping the whole post-production pipeline
<br> with only free software
<br> to produce something qualitative enough to be displayed in a theater
<br> so 2K 10-bit in DCI-P3 colorspace
jernej: ok, that's a real challenge
paulk: absolutely :)
<br> I have a BMPCC 6K Pro camera that's shooting log footage in ProRes
<br> in a scene-referred colorspace
<br> footage looks like this: https://www.paulk.fr/private/your-very-own-worst-friend/prores-log.png
<br> and I grade using Natron, which relies on OpenColorI/O
<br> to get this: https://www.paulk.fr/private/your-very-own-worst-friend/2024-01-15-olive-ocio-taille.jpeg
<br> hopefully someday we can support a sensor as good as this supported in v4l2 :)
<br> and build our own fully free cinema cameras
mriesch: <u>pinchartl</u>: ok but how would userspace know e.g. which stream corresponds to the long exposure image?
paulk: a bit like https://www.apertus.org/axiom but without FPGA
<br> <u>jernej</u>: so anyways I'll work on the rewrite but probably in a few months
<br> like around april
jernej: that was long explanation. good luck with your journey"
-: mriesch reads pinchartl s presentation slides again
paulk: <u>jernej</u>: yeah sorry I get very verbose when talking about this stuff :)
linkmauve: paulk, woah, that’s great!
paulk: started doing that during COVID lockdown
linkmauve: jernej, so yeah, it seems like adding a new ext_ctrl for JPEG would make sense to support both NV12 and NV16 output here.
<br> paulk, feel free to keep me updated, in private if you want, about the progress on that movie. :)
***: gouchi has quit IRC (Remote host closed the connection)
<br> xiaer1922 has joined #linux-media
<br> xiaer1921 has quit IRC (Ping timeout: 480 seconds)
<br> xiaer1922 is now known as xiaer1921
<br> mrpops2ko has joined #linux-media
mriesch: <u>pinchartl</u>: could the sensor provide frame descriptors that indicate the role of the stream?
<br> e.g. hey isp, this stream has the short exposure images in case you need them?
kbingham: paulk, You didn't want to make the camera yourself then ;-) https://www.diyphotography.net/cinepi-is-a-2k-12-bit-raw-50p-open-source-video-camera-for-200/
<br> And for big sensors in v4l2 - the IMX283 is a 1" 20MP sensor that will be on it's way upstream soon. I think a few other large variants should follow after too.
<br> Even the camera module PCB can be open hardware ;-) https://www.willwhang.dev/OneInchEye/
mriesch: ah... it seems the V4L2_SUBDEV_ROUTE_FL_SOURCE flag was dropped in the upstreaming process...
***: djrscally has quit IRC (Ping timeout: 480 seconds)
<br> ten157237743246 has quit IRC (Quit: The Lounge - https://thelounge.chat)
<br> ten157237743246 has joined #linux-media
paulk: <u>kbingham</u>: awesome!
<br> really cool enclosure too
<br> I definitely want to build one with imx477
<br> but probably with rockchip and/or sunxi ISP
<br> and yeah the sensor is too small
<br> will look out for IMX283!