Re: [iPAQ] mpg123 v madplay comparison

On Sun, 24 Sep 2000 22:56:09 -0400, Christopher J Mason wrote:

[snip]

> To summarize: (cpu time in min:sec) (4:25 mp3, same track, same encoder)
> ARM SA1100 (206mhz) i686 (333mhz)
> madplay mp-speed mpg123 madplay mp-speed mpg123
> VBR (182kpbs avg) 1:08 1:08 1:13 1:24 1:19 0:36
> 128kbs 1:02 1:02 1:11 1:25 1:07 0:35
>
> (mp-speed is madplay compiled with the --enable-speed option; this doesn't
> seem to have much effect on arm.)
>
> In general this translates to about 28% of the processor all the time on
> 206mhz strong-arm; versus less than 5% of the processor on 333mhz intel.
> The difference between the VBR and 128kbs files seems to be minimal (this
> is quite surprising to me; the VBR file sounds noticeably better, and is
> significantly larger).

Did you also try Nicolas Pitre's ARM optimized version of splay? You can
get it at:

  ftp://ftp.netwinder.org/users/n/nico/

FYI: Rob Leslie (the author of madplay) also did a comparison (in April
this year), you can find his results in the LART mailing list archive:

  http://www.lart.tudelft.nl/list/list.php3#body?sd=30&sm=3&sy=2000&ed=4&em=4&ey=2000&sort=t&inl=&mid=200004020128.UAA27092%40deimos.mars.org

You can gain a lot of performance by proper use of the StrongARM CPU. The
StrongARM has a very fast multiply-accumulate instruction that's ideal for
FFT and DCT operations. Another thing to keep in mind is that an uncached
memory operation on a 220MHz SA1100 costs about 60 instructions. By using
the prefetch instruction, the load/store unit already loads the next data
for the IDCT in the read buffer (and D-cache) while the CPU still
processes the current data.

Erik

-- 
J.A.K. (Erik) Mouw, Information and Communication Theory Group, Department
of Electrical Engineering, Faculty of Information Technology and Systems,
Delft University of Technology, PO BOX 5031,  2600 GA Delft, The Netherlands
Phone: +31-15-2783635  Fax: +31-15-2781843  Email: J.A.K.Mouw@its.tudelft.nl
WWW: http://www-ict.its.tudelft.nl/~erik/