[FFmpeg-devel] [RFC]Fix aic on PPC

Christophe Gisquet christophe.gisquet at gmail.com
Sat Jun 14 18:55:59 CEST 2014


Hi,

2014-06-14 18:34 GMT+02:00 Carl Eugen Hoyos <cehoyos at ag.or.at>:
> Christophe Gisquet <christophe.gisquet <at> gmail.com> writes:
>> but for fate only:
>> http://git.videolan.org/?p=ffmpeg.git;a=commit;h=d77815ee
>
> See:
> http://thread.gmane.org/gmane.comp.video.ffmpeg.cvs/78589/focus=78591

Not suscribed, so I indeed miss a lot of the context of your patch. It
was not made obvious from the commit message.

>> On the other hand, I'm not sure what things of
>> value could be lost to enforce this at the
>> decoder-level like your patch does.
>
> Do I understand correctly that you believe it is
> better to compile in a completely unusable way
> than in a way that we assumne is not bit-exact?

I can't really reply because I'm missing steps in your thinking here.
I don't even know if you are being agressive, or it's just the
phrasing. Sorry if you think I provoked that.

What I'm saying is that:
1) for an intra codec, a slighly less accurate idct may not cause that
much of an issue
2) Your patch seems to remove the possibility for the user to select the iDCT
3) But your patch seems to provide the best accuracy across archs, and
that's probably preferable to the user getting the idct wrong

Best regards,
Christophe


More information about the ffmpeg-devel mailing list