Re: [Jack-Devel] iio and jack - 1MHz sample rate success

Le 12 févr. 2014 à 08:14, Matt Flax <[hidden]> a écrit :

> I have implemented the jack2 iio driver and have done a couple of tests.
> 
> In a preliminary finding, waiting for more time to test and re-test, I have had success.
> I don't see any overruns reported (when setup properly), and I haven't put a signal through the system to highlight any un-reported overruns. Next week I will look into this more.
> 
> Firstly I would like to say that writing a driver for jack2 is a little difficult because there is not much documentation ... at least I couldn't find simple information. Some inline src doc would be good ... mind you, I could have put more into my driver as well :)
> 
> On the system with IIO devices I ran :
> jackd -iio
> 
> I then setup and ran netjack on both computers.
> 
> On the other computer I used :
> jack_capture -c 2 -p system:capture*
> 
> ... wahlah ... it worked ! with an MTU of 1500. Thats 2 channels @ 1 MHz
> 
> Everything reporting 1MHz sample rates and so on a so forth, no xruns and the wav file looked like what I expected ...
> 
> When I tried to record 4 channels @ 1 MHz, I had to increase the MTU to 6000 to get rid of xruns on the ARM core. But still ... 4 channels @ 1MHz over a network ... pretty good start !
> 
> I would like to integrate my both of my iio branchs into the jack1 and jack2 repos, but for now I have forked the jack2 repo here :
> https://github.com/flatmax/jack2/tree/iio

When ready, do a "merge request".
> 
> I have some questions for other developers of jack2.
> is it necessary to use :
>     JackDriver::CycleTakeBeginTime();
> and
>     JackDriver::CycleTakeEndTime();

Sure: this code is used to take "timestamps" at the beginning and end of driver cycle, so that DSP CPU consumption can be computed later on.

Stéphane 

1392191623.3845_0.ltw:2,a <97EA9EB4-7C3F-4E18-AFF7-C54564823DF6 at grame dot fr>