[Iaxclient-devel] Re: IAX and ARM RISC

[Steve K. <st...@st...>]

Stephan,


You're not the only one doing this.

Another list member has already told us (off-list) that he has iaxComm 
(not just iaxclient, but all of Michael's UI) working on an iPaq.  He's 
even made a more compact GUI interface for it.  I've just sent you his 
e-mail off-list.  I would recommend that if you guys work on this 
together, keep the list updated.  I don't have any of these gizmos, and 
neither does Michael (a.k.a. Steve V), and I'm not sure about the other 
devs on the list.

As far as the CPU stuff; I'd start by turning off _all_ the input 
post-processing.  Disable echo-cancellation [doesn't work yet anyway], 
noise reduction and VAD.  If they're all disabled, you won't run the 
post-processor at all.

-SteveK


On Mar 2, 2004, at 11:16 AM, S wrote:

> Hi Steve,
>
> thanks for your last mail. Before I comment your last mail I like to 
> explain what I try to do.
>
> I like to write a IAX Telephone Client for the Zaurus PDA. The Zaurus 
> PDA run's Linux and use a Srong ARM RISC 1110 CPU with about 137 
> BogoMips.
> As basis for this project I use the iaxclient lib and parts of 
> simpleclient. Also I use portaudio with some little changes for my 
> Zaurus.
>
> So back to your last mail:
>
> You wrote :
> > If you're using portaudio, then audio.input should be pa_input, 
> which > is a simple function, which just looks to see if the 
> RingBuffer has
> > data in it.  Unless something is causing things to run very slowly 
> for
> > you, then it should only have data occasionally.
>
> This sounds the Zaurus ARM CPU is to slow. For this reason I have to 
> find out where the time goes in iaxclib.
> My Phone client is still a simple program with out any GUI.
> I suspect there is something wrong in the send_encoded_audio || 
> input_postprocess chain, because I never get silent back (I tried it 
> on my power full PC never silent).
> If you have some other idea where the problem is let me know.
>
> So here is my IAX observation :
>
> One of the main problem porting the IAX lib to the ARM CPU is, the IAX 
> protocol is not word align.
> Word align mean every integer, short or pointer need to be on a 
> address which is divisible thru 4 (4*8bit = 32bit, and ARM CPU is a 
> 32bit CPU).
> For this reason I have made some changes in the libiax2 part.
> I will send you the changes later because I like to get the phone 
> client  running first.
> In different mail I have seen notice from people who write they run 
> iaxclib and asterisk on RISC machines without problems. I can't 
> believe this. This code will crash an any RISC machine sone.
>
> For more details I will send you the libiax2 changes later.
>
> Regards
>
>   Stephan
>
>
>