[Ffmpeg-cvslog] r5898 - in trunk/libavcodec: dsputil.c dsputil.h i386/dsputil_mmx.c vorbis.c vorbis.h

[Benjamin Larsson]

Loren Merritt wrote:
> On Thu, 3 Aug 2006, Benjamin Larsson wrote:
>
>> If you want to optimize more you could look at the mdct pre and post 
>> twiddle steps in mdct.c. Currently they are scalar operations. 
>> Optimizing this would also give a gain to wma and aac.
>
I forgot that ac3 also would gain from this.
> hmm, those are annoying because the data aren't contiguous.
I don't understand can you elaborate?
> CMUL(z[j].re, z[j].im, *in2, *in1, tcos[k], tsin[k]) may be possible 
> if I change the loop order to go from both ends at once, though the 
> revtab[] part still needs to be scalar.
> CMUL(z[k].re, z[k].im, z[k].re, z[k].im, tcos[k], tsin[k]) looks ok.
> And I don't know what to do with the 3rd loop.
The algorithm is based on the A/52 specification. I'm sure there are 
some hints to be gained from the mplayer liba52 asm optimizations.
>
> --Loren Merritt


MvH
Benjamin Larsson