[FFmpeg-devel] [PATCH] TwinVQ decoder

[Vitor Sessak]

Vitor Sessak wrote:
> Loren Merritt wrote:
>> I'm not entirely convinced it's a good idea to merge two window sizes 
>> into one wtype variable, but I haven't seen the alternative code so 
>> I'm not sure.

[...]

> 
> I've tested and it works fine. New patch attached.
> 
> -Vitor
> 
> PS @Michael: May I just commit it if Loren has no more comments? This 
> code has already been reviewed by quite a few talented developers...

Loren told me in IRC that he had no more comments about my patch, so I 
committed it. Since Michael suggested me to ask someone else to review 
it (he is busy with reviewing SoC code) I suppose it is not breaching 
the policy (I can always be flamed at -cvslog anyway).

Thanks for everyone who helped me getting this code reviewed and up to 
FFmpeg standards!

-Vitor

PS @Mike: This decoder is float-based, so it is probably not bit-exact 
across arches. But when your one-off testing method is ready, I'd 
suggest the samples mono_08_08.vqf, stereo_11_08.vqf, mono_11_10.vqf, 
stereo_16_16.vqf, mono_22_20.vqf, stereo_22_24.vqf, mono_22_32.vqf, 
stereo_44_40.vqf and mono_44_48.vqf (that can be found at 
http://samples.mplayerhq.hu/vqf/luckynight/) for regtesting all the 
possible codepaths...