[Matroska-devel] Re: [mpc-devel] Re: [mpc-general] Protocol of my 2+ hrs telephone conversation with Frank Klemm

Michel Lespinasse walken at zoy.org
Mon Nov 17 07:54:57 CET 2003

On Sun, Nov 16, 2003 at 11:47:24PM +0100, Jo?l Bourquard wrote:
> Some reasons:
> - Speed (ok, Huffman codes take some effort to decode, but MDCT is
> by far the most expensive process). Optimizations to Huffman
> decoding can be done, while MDCT is kind of irreductible.

MDCT can be quite fast too. Maybe not as fast as subband coding, but
still. A god implementation can go quite faster than a naive one.

Look at my AC3 decoding library at http://liba52.sf.net. Right now
decoding performance is that for a stream coded as 5.1 channels, 384
kbit/s, decoding on a 400 mhz celeron (mendocino) takes about 5.3% of
CPU time, or if we downmix to 2 channels (which can be done before the
imdct stage), 3.2%.

> Since the framing structure is cleaner in SV8, complexity should not be
> a problem (the only additional work is in a lossless SV7->SV8
> converter). I guess the inter-frame limitations can be sorted out during
> SV7->SV8 conversion.

I have to say the SV8 bitstream format is quite neat.

This kind of consideration is the main reason why I never implemented
an mp3 decoder - their bit format is just way too ugly. Layers 1 and 2
are nice and simple (even if suboptimal).


Michel "Walken" Lespinasse
"In this time of war against Osama bin Laden and the oppressive
Taliban regime, we are thankful that OUR leader isn't the spoiled son
of a powerful politician from a wealthy oil family who is supported by
religious fundamentalists, operates through clandestine organizations,
has no respect for the democratic electoral process, bombs innocents,
and uses war to deny people their civil liberties." --The Boondocks

More information about the Matroska-devel mailing list