[Speex-dev] Speex 1.1.10 on ARM926EJ-Sid(wb) rev 3 (v5l)

[Jean-Marc Valin]

Hi Eduardo,

All I can say is that the timings you have are a bit odd. What
optimizations options are you using (I suggest -O3)? Also, perhaps you
can try --enable-arm4-asm just in case. I've had the encoder running in
real-time on a chip that's at least twice slower than yours with the
settings you're using.

	Jean-Marc

Le jeudi 15 septembre 2005 à 16:23 -0400, Eduardo Bezerra a écrit :
> Hi all,
>  
> 
> 
>   I'm trying to use libspeex 1.1.10 on an ARM926EJ-Sid(wb) rev 3
> (v5l). I executed
> the speexenc and speexdec test files and they can encode and decode.
> But I'm getting 95% of cpu utilization on the codification and 44% on
> the decodification.
>   I saw in the post:
> http://lists.xiph.org/pipermail/speex-dev/2005-June/003485.html
>   that this version of speex works fine on ARM platforms. My configuration
> is:
> 
> # cat /proc/cpuinfo
> Processor       : ARM926EJ-Sid(wb) rev 3 (v5l)
> BogoMIPS        : 95.83 (about 192 MHz)
> Features        : swp half thumb fastmult edsp java
> CPU implementer : 0x41
> CPU architecture: 5TEJ
> CPU variant     : 0x0
> CPU part        : 0x926
> CPU revision    : 3
> Cache type      : write-back
> Cache clean     : cp15 c7 ops
> Cache lockdown  : format C
> Cache format    : Harvard
> 
> #  gcc --version
> arm-linux-gcc (GCC) 3.3.2 (Debian)
> 
> Configuration: --enable-fixed-point --enable-arm5e-asm
> 
> Speex specs: Complexity 0, 8Kbps
> Encoding: ~95%, Decoding: ~44%, without Echo Cancellation / Pre-processor
> 
>   Does anyone know how I can reduce this utilization? Is there any
> float point code on speexenc/speexdec that wasn't converted to
> fixed point?
> 
-- 
Jean-Marc Valin <Jean-Marc.Valin at USherbrooke.ca>
Université de Sherbrooke