Re: [iPAQ] mpg123 v madplay comparison

On Sun, 24 Sep 2000, Christopher J Mason wrote:

> This is a comparison of the mpg123 and madplay mp3 players running on the
> Compaq iPAQ (data from an intel PII is included for comparison). This
> test is for "raw" decoding, where the decoder is just asked to decode the
> file, not to send it to the audio device (it is hoped that this will negate
> the effects of possibly nascent audio drivers for the iPAQ, no offense to
> anyone).

[...]

> Subjectively, I can tell only slight differences in the quality of the
> output from the three different decoders. Maybe mpg123 sounds just
> slightly "more cluttered" in the high-tones than madplay; this is more
> noticable on the 128kbps mp3 and it's damn close.

The difference is even more noticeable with classic music, or songs with
low amplitudes.

For a complete quality comparison, have a look at:
http://www.mars.org/home/rob/proj/mpeg/compliance/

> I guess I'm a bit disappointed that even the 128kbits stream takes up 28%
> of the processor.

However if you test performances while actually sending data to the audio
device, you'll note that CPU usage goes down slightly. This might be
explained by the fact that the SA is allocating the cache on reads only
and if you use /dev/null the audio buffer is never read by the CPU. Oddly
enough, the audio driver is warming the cache when it's actually reading
i.e. playing data.

> I also notice that the audio quality on the iPAQ seems
> to lack the bass response of my Sony laptop. It's still quite acceptable,
> though.

With a mixer program, you'll be able to boost the bass a lot.

> The audio amp on the iPAQ can get /much/ louder than that on my
> laptop. Which is good, I guess; I'm waiting for the time when I forget to
> set the volume level (I'm using -g option to mpg123, or the slider in
> gqmpg; I can't find I user land tool like xmixer or aumix) and blow out my
> eardrums.

See:
ftp://ftp.handhelds.org/pub/linux/arm/carleton-winder-rpms/RPMS/aumix-1.18.2-2.armv4l.rpm
(use cpio to extract the executable without having to bother with rpm)

> The only conclusion that I can draw is that the lack of FPU is really going
> to hamper any serious multimedia codecs. Is this true? Can we do much
> better? Am I just naive to think that it'll take less than 30% of the
> processor to decode mp3?

I pushed madplay down to 16% CPU usage on a 200MHz SA1110. However my
patches were too kludged for public release... But it's doable to get
good performances out of a SA when you take care of good fixed point
tricks, cache issues, etc.

> Honestly, it's pretty impressive that the 206mhz strongarm bests the 333mhz
> pentium at integer performance.
>
> I'll tell you one thing for certain though: I'll never listen to this song
> again!

:-)

Nicolas
Received on Sun Sep 24 21:44:09 2000