[Alsa-user] Recording sound on ARM processor causes 30% CPU load

[Rudi v. d. L. <ru...@an...>]

Hi All,

I am writing an application on an ARM based embedded computer.  And have
found a few difficulties with the process.  Any help or explanations on the
following points would be appreciated.

1. The hardware is running at an incorrect clock rate.  This causes a +-46.8
kHz frame rate to be supplied to the CPU.  When I compile the kernel with
this value as the ALSA clock I get "double the error" (i.e. an actual data
rate of [requested data rate]*(46.8/48)*(46.8/48) bps).  But when I compile
it with an ALSA clock rate of 49.1 kHz (i.e. fix it with 48000*(desired
clock rate)/(actual clock rate)) an actual stream of 48kHz is reproduced
(i.e. [requested data rate] * (46.8/48 {hardware fault})*(48/46.8) {49.1kHz
setting in kernel}) My question is,  why does it work "the wrong way round"?
Should one not supply the actual hardware clock rate, instead of a
"corrected" clock rate?

2. I have noticed that recording with the "snd_pcm_readi" function uses 30%
of the CPU (I have tried blocking and non-blocking modes).  As well as
registering a callback, again blocking and non-blocking, as well as using
poll/select (blocking and non-blocking).  The best I can do is 30% CPU load,
which seems a lot.  'arecord' also uses 30% CPU (measured with 'top'). The
strange thing is that 'aplay' hardly uses any CPU power (2-3%).  Any ideas
how I can reduce the CPU loading? Is this an ARM limitation? A known ALSA
driver/library problem?  An implementation problem? Any other ideas?


I am not sure what information would be required, but my configuration is as
follows:
ARM based PC-104 CPU running at 400MHz -  The Viper PC-104 SBC from Arcom
UK, running a Debian based Linux kernel tailored for the CPU, Linux 2.6.13
(patched) to run on the Viper, to correct the ALSA clock rate and some FLASH
issues.
I am not sure how to obtain the version of ALSA itself, amixer, aplay and
arecord all return a version of 1.0.8.

I am a bit reluctant to try and compile a new version of the ALSA drivers
and libraries since I do not know what changes would be required to get it
to work on this embedded CPU.

Regards,
Rudi

Rudi van der Linde
ru...@an...

Note:
This message (and attachments) is subject to restrictions and a disclaimer.
Please refer to http://www.ansys.co.za/index.html or ge...@an... for
full details.