[FFmpeg-devel] [PATCH][WIP][RFC] Add support for MPEG-4 Simple Studio Profile

[Kieran Kunhya]

Hi,

I have added support for MPEG-4 Sstp using the available samples on trac.
Yes it doesn't pass fate, yes it's not format-patch, yes it uses printfs.
https://trac.ffmpeg.org/ticket/4447

Being MPEG-4, it depends on mpegvideo.c so has tons of yuv420p assumptions
baked in which are of course undocumented.
Here are my questions (line number refers to attached patch):

Line 35: How do I signal to this idctdsp thing that I want an idct with
32-bit coefficients AND intermediates? A lot of that code has assumptions
that intermediates will be the next size up, i.e 8-bit coeffs, 16-bit
intermediates, or 16-bit coeffs and 32-bit intermediates.

Line 906: Why do RGB samples not decode unless the GBRP format is moved to
the top of PIX_FMT. I get "[mpeg4 @ 0x7f945c029600] format change not
supported" otherwise.

Line 932: What's going on with this branch. Normal mpeg-4 video does
dequant during unpack, why is it not part of this condition?

Line 987: Are there more assumptions baked into mpegvideo.c about "square"
macroblocks, i.e ones where (width == height)?

Line 997: What is all this stuff going on with -1U, unless I remove this I
get a segfault. I do get a stripe on the left though.

Line 1055: How can I make the existing code use 32-bit coefficients
cleanly? I can't reuse block[a][b] because it's allocated in a single
malloc.

Regards,
Kieran Kunhya
-- 

Sent from my mobile device
-------------- next part --------------
A non-text attachment was scrubbed...
Name: studio.diff
Type: text/x-patch
Size: 46708 bytes
Desc: not available
URL: <http://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20170107/a62e124a/attachment.bin>