Re: [FFmpeg-devel] [PATCH] libpostproc: remove big-endian AltiVec acceleration

[Andreas Rheinhardt <andreas.rheinhardt@outlook.com>]

Sean McGovern:
> Hi,
> 
> On Wed, Apr 9, 2025 at 2:55 PM Michael Niedermayer
> <michael@niedermayer.cc> wrote:
>>
>> Hi Sean
>>
>> On Sun, Apr 06, 2025 at 12:42:30PM -0400, Sean McGovern wrote:
>>> While POWER & PowerPC systems have technically
>>> supported little-endian from inception,
>>> it did not come into proper usage until the
>>> introducion of the POWER8 in 2013.
>>>
>>> The AltiVec acceleration present here has not
>>> had meaningful change since before then.
>>>
>>> refs: Trac/11120
>>> ---
>>>  libpostproc/postprocess.c                  |   13 -
>>>  libpostproc/postprocess.h                  |    1 -
>>>  libpostproc/postprocess_altivec_template.c | 1211 --------------------
>>>  libpostproc/postprocess_template.c         |   31 -
>>>  4 files changed, 1256 deletions(-)
>>>  delete mode 100644 libpostproc/postprocess_altivec_template.c
>>
>> What exactly are you fixing here / what is the problem?
>> You say the code hasnt been changed and you link to a ticket
>> where people cant decide if they talk about big or little endian
>> and that links to seemingly unrelated tickets
>>
>> Please clearly explain what the issue is and why this is
>> the correct fix
>>
> 
> We already discovered a few months back that we have no real
> maintainers for these ppc pieces, so as noted in the ticket I am
> working on removing them.

unmaintained != broken

> Is this what you want in the patch notes?
> 
>> Also you are changing public API without version bump
>>
> 
> Oops, yeah OK I can do that.
> Is a minor bump enough or should this be a major?

This would need a proper deprecation period.
(You can also deprecate the 3DNOW define at the same time.)

- Andreas

_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".