[FFmpeg-devel] AC3 floating and fixed point decoder work and 32 bit fixed point FFT implementation

[Nedeljko Babic]

Hello,

In first patch implementation of 32 bit fixed point FFT is resubmitted. 
Since this is new implementation of FFT, it is used only in new AC3 fixed 
point decoder submitted in third patch and nowhere else for now. That is 
the reason why it is resubmitted (and there are some minor changes in 
regards to previous patch).

In second patch order of downmix and imdct in floating point ac3 decoder 
is changed as requested in one of reviews for implementation of ac3 fixed 
point decoder. With this change, upmix is eliminated and since upmix 
generated glitch in decoded stream, this bug is also fixed with this patch. 
New referent pcms are generated taking in consideration changes in ac3 
floating point decoder and fate test are adjusted to use them.

In third patch there is implementation of ac3 fixed point decoder with 
changes requested in previous reviews and with changes required for use of 
new 32 bit FFT.

Regards,
-Nedeljko