[MPlayer-dev-eng] [PATCH] remove mp3lib

[Ivan Kalvachev]

On 9/23/11, Thomas Orgis <thomas-forum at orgis.org> wrote:
> Am Fri, 23 Sep 2011 03:45:49 +0300
> schrieb Ivan Kalvachev <ikalvachev at gmail.com>:
>
>> Even if we assume the extra 6 seconds are because of the ugly float
>> conversion and clipping i've done in af_format, even then mp3lib is
>> still around 16% faster than the float only case.
>
> Would you care to give the ac mpg123 a spin, too? Regardless about the
> performance of ffmpeg's mp3 (and mp2, and mp1) decoding, the comparison
> to the official mpg123 code should work out on removing the internal
> fork.
>
> Because of improved SSE code, mpg123 should be faster on platforms
> supporting that. The game is not that clear for MMX and 3DNow*, where
> subtle effects of the libmpg123 binding could rule.
>
> Make sure to use mpg123 version 1.13.x, please (there was some
> optimization regarding the data transfer from MPlayer).
>
>
> Alrighty then,
>
> Thomas.

Here they are:
athlon:$ for i in 1 2 3 ; do time mpg123 -w /dev/null stream.dump ; done
High Performance MPEG 1.0/2.0/2.5 Audio Player for Layers 1, 2 and 3
        version 1.13.3; written and copyright by Michael Hipp and others
real 0m48.938s 		user 0m47.331s		sys 0m1.591s
real 0m48.790s		user 0m47.013s		sys 0m1.764s
real 0m48.736s		user 0m46.996s		sys 0m1.726s

athlon:$ ./mplayer -loop 3 -ao pcm:fast:file=/dev/null -benchmark
-quiet -ac mpg123 stream.dump
MPlayer SVN-r34123-4.5.3 (C) 2000-2011 MPlayer Team
Selected audio codec: [mpg123] afm: mpg123 (MPEG 1.0/2.0/2.5 layers I, II, III)
AO: [pcm] 48000Hz 2ch s16le (2 bytes per sample)
BENCHMARKs: VC:   0.000s VO:   0.000s A:  55.760s Sys:   0.862s =   56.621s
BENCHMARKs: VC:   0.000s VO:   0.000s A:  53.390s Sys:   0.515s =   53.906s
BENCHMARKs: VC:   0.000s VO:   0.000s A:  52.770s Sys:   0.536s =   53.306s

=====
While the standalone run looked quite promising, the mplayer run
wasn't as good. Even if mpg123 is the fastest among the alternatives,
mp3lib is still 8% faster.

I used the version that comes with my distribution, as that would be
the typical user case. Fortunately it comes with 1.13.3 . It is
entirely possible that the package is not compiled with the best
options for my cpu. I may try with a custom build later.