[FFmpeg-devel] [PATCH] MV-HEVC decoding

[Anton Khirnov]

Hi,
this second round of the MV-HEVC decoding set should be good enough for
inclusion. Changes since the previous set [1] are:
* Rebased against master.
* Addressed all the review comments.
* Now using C23 stdc_trailing_zeros() instead of POSIX-only ffs() (with
  a fallback for systems lacking stdbit.h, thanks to Rémi).
* View ID is now available in get_buffer2() as frame side data rather
  than via a special read-only decoder-private option, thanks to a patch
  by James.
* Dropped output layer set support. HEVC VPS can contain a number of
  layer-set definitions (i.e. layers that should be selected for output
  together). Previous version of this patchset had decoder options for
  querying the sets and selecting one of them for output. However, this
  feature does not seem useful as there is no way to ascribe semantics
  to the sets; I'm also not aware of anyone using it. So in the name of
  simplicity, I dropped the relevant options until and unless someone
  presents an actual use-case for this. The API for layer selection is
  now based only on view IDs.
* Added a test for decoding a real-world file/selecting view by
  position.

As before, you can fetch the code from git://git.khirnov.net/libav,
branch 'mv-hevc'.

My thanks to Vimeo, Meta, and FFlabs for supporting this work.

[1] https://lists.ffmpeg.org/pipermail/ffmpeg-devel/2024-August/332693.html
    Message-Id <20240827154041.13846-2-anton at khirnov.net>

-- 
Anton Khirnov