[FFmpeg-devel] [PATCH] SSE optimization for DCA decoder

[Alexander E. Patrakov]

Benjamin Larsson <banan <at> ludd.ltu.se> writes:

> Yeah, it looked quite funky. But can the code be made faster with the 
> "lowpass" representation ?

Probably not. The idea was to get understanding of what's really going on, and
do optimizations based on this understanding, instead of just obvious trivial
transformations.

Also, if I understand the other posts on this list correctly, other codecs
(MPEG-1 layers 1 and 2?) also use pseudo-QMF cosine modulation filterbanks. It
is highly unlikely that the MPEG group obfuscated their QMF the same way as DCA,
so switching to one common representation will help code reuse even if it has no
speed benefit.

-- 
Alexander E. Patrakov