Re: [Jack-Devel] JACK2 for Android newly released!

On Saturday 25 January 2014 14:55:11 김정연 wrote:
> > First, the framework only supported I/O via built-in audio devices. No
> > USB. This means audio quality is compromised by their audio hardware.
> > Importantly, they are restricted to 48k/16 bit.
> 
> Yes, it is true, but most of things are platform dependent. Not a jack
> domain.

What do you mean with "platform dependent"? External USB audio devices usually 
support the "USB audio class" standard protocol nowadays. Support for such 
devices is provided by a standard Linux kernel module which is platform 
independent.

> I think these XRUNS related cpufreq on demand governor when a jack client
> runs during low CPU frequency mode in mobile device. Please refer:
> http://bamboopuppy.com/android-cpu-frequency-using-cpufreq-ondemand-governo
> r/ As I know, there were already discussion in here, so I found Robin's
> jackfreqd, How do you think for solution?
> http://rg42.org/oss/jackfreqd/start

Yes, CPU scaling (CPUFreq) is one of the major reasons for xruns. You could 
first try whether this is the case by:

	cpufreq-set -r -g performance

to find out whether that is sufficient to solve the xruns. That would set the 
CPU cores to maximum performance. If that looks well, you could then go ahead 
with a production solution like i.e. jackfreqd or something based on that idea 
to both save power whenever possible and not having xruns in audio 
performances.

> >> > What about low latency/RT stabilty, have you tested running JACK2 with
> >> > low latency period settings (i.e. <= 5ms) on Android? Does it run
> >> > quite stable without xruns?
> >> 
> >> Well, it is hard to say, latency depends on platform limitation or hw
> >> spec. for example, android's fast mixer introduces 10ms of latency.
> >> http://source.android.com/devices/latency_design.html
> >> 
> >> Generally, mobile devices do control power consumption much sensitive
> >> than PC, of course this limitation may possible occur XRUNS in some
> >> scenarios.
> >> We always think how to improve performance better, *** so I need
> >> expert's advice here. ***
> >> But, modern mobile devices increase number of cores(quad-core or
> >> more...), so I guess JACK2 works better performance along with number
> >> of cores. Of course, it works stable if allows reasonable clock-speed
> >> and cores.

I don't know what kind of hardware components you are using exactly, but I 
somehow doubt that latency issues are caused by your hardware. iOS devices 
deliver rock solid low latency from day one (for more than 6 years now). And 
there are various other ARM based hardware on the market which also deliver 
reliable low latency if being equipped with the right OS & co.

Are you running a hypervisor kernel like OKL4 on lowest level and 
Android/Linux on top of it? Or is the Android/Linux kernel running directly on 
the ARM CPU?

Best regards,
Christian Schoenebeck

1390838353.25171_0.ltw:2,a <201401271559.46122.schoenebeck at crudebyte dot com>