What is this new format used for and where? Someone with access to the (I assume it's hardware?) would need to implement that for proper testing.

Is this basically in 8 values / 10 bytes group, or... ?

Yes, hardware video decoder and video render use this format. The size of the 10 bits NV12 video frame buffer is 1.25 times of 8 bits NV12 video frame buffer.

That's not enough information to implement this though.

Please, provide a complete layout iformation.

Our video frame buffer need 16 pixel alignment. It is 16 values / 20 bytes group. All 16 values need reverse if convert from little endian to big endian.

Is it common video format? Or I need add it as our private format?

It does not seem common in the sense I haven't seen it documented anywhere. The "reversing" isn't clear to me yet though, so this still need information.

Maybe you could provide a sample yuv image, reference in well known format, and one packed the way you described ?

8 bit NV12:

Y plane:

pixel:   1     2     3     4     5
bytes:   1     2     3     4     5

10 bit NV12:

Y plane:

P010:

pixel:        1           2          3         4          5
bytes:   1    2     3     4     5    6    7    8    9    10

Our 10 bit format:

pixel:     1     2     3       4
bytes:   1    2    3     4     5

Again, can you attach couple of RAW samples for that format, so we can test with. In general you can attach a ref in a known format (like P010) and one in your new format. Otherwise, propose a patch that implements what you need.

If the helps, here's one of my submission for similar format, which is packed over 32bit (with 2 bit padding / lost).

  https://bugzilla.gnome.org/show_bug.cgi?id=789876

-- GitLab Migration Automatic Message --

This bug has been migrated to freedesktop.org's GitLab instance and has been closed from further activity.

You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/issues/368.