Question

in Raspberry-Pi code, there is a s5p-jpeg codec driver.

drivers/media/platform/s5p-jpeg/jpeg-core.c

Can sombody please tell me where I can find an example of how to use it? Or any other v4l2 codec driver?

I have googled for it, but I cannot find any example which uses a v4l2 codec driver.

Était-ce utile?

La solution

(adding another answer, since it's completely different :-))

A "codec" API in the V4L2 specs, refers to hardware codecs. such a codec is a device that has the following features:

  • the hardware codec shows up as a /dev/videoX device

  • it has a video input, where you're userland application can send a video-stream - e.g. containing JPEG-encoded video frames - to, so it has the V4L2_CAP_VIDEO_OUTPUT capability and

  • it has a video output, where you're userland application can read a video-stream - e.g. containing the uncompressed frames - from, so it also has the V4L2_CAP_VIDEO_CAPTURE capability.

There are numerous applications that can write video to a v4l2 OUTPUT device, here are a few that i know of:

afaik, these application don't have any specific code to deal with "v4l2 codec devices", but can write to/read from v4l2 devices, which is all that you should need.

Also check the v4l-utils.git: Look in utils/v4l2-ctl/v4l2-ctl-streaming.cpp

Autres conseils

v4l2 is very liberal when it comes to formats: e.g. capture devices can deliver frames in virtually any format.

So if you are writing a user-land tool (application, library,...; as opposed to a kernel-driver), though shalt not fiddle with codecs (if you can avoid it). Imagine each and every application in the world that wants to read v4l2-streams, having to add code to decode frames in SQ905C or MJPEG or whatnot codec (each application adding their own set of buggy implementation)

Instead, smart people created a library which will decompress the frames delivered by your capture device and provide these frames in a standard way: libv4l2.

Incidentally, if you insist on writing your own code, libv4l2 is a good reference implementation.

Oh, and if you were thinking about simpy loading a module to the encoding/decoding in kernel-space (e.g. you have webcam "foo" which delivers images in format "XYZ" but you want it to deliver images in format "ABC" by means of adding a kernel-module) then you are out of luck.

Linus T. has been quite clear [missing reference] that codec conversion code is not to be run within kernel-space.

The code here accesses a v4l2 codec driver

It indeeds checks the V4L2_CAP_VIDEO_M2M cap

It is based on the kernel capture example but with some extra code

Licencié sous: CC-BY-SA avec attribution
Non affilié à StackOverflow
scroll top