Re: [Jack-Devel] Implement a driver as a writable client?

Thanks for the info, Paul.
So, are you basically saying I need an extra buffer? A client takes audio
input from the JACK server and writes it to a buffer, and I have to
manually tweak the size of that buffer to avoid over- or underruns, and
another thread writes from that buffer to the hardware I2S FIFO whenever
it's writable? If I understand the implications of clients being unable to
block as you say, the audio generation programs won't ever wait for my
improvised "driver"... but I'm not sure if I get why that's something to
worry about.
-Chuckk



On Sat, May 14, 2016 at 2:41 AM, Paul Davis <[hidden]>
wrote:

> You can't do duplex properly as a client. The main difference between a
> JACK "driver" (aka "backend") and a client is that a driver consists of two
> halves. One is executed at the start of a process cycle, to collect data
> from the hardware; the other is executed at the end of a process cycle, to
> deliver data to the hardware.
>
> If you're not doing duplex, you can do whatever you want in a client, but
> keep in mind that it cannot block, meaning that you basically have to
> shuffle the data to/from another thread that actually interacts with the
> hardware.
>
>
> On Fri, May 13, 2016 at 7:12 PM, Chuckk Hubbard <[hidden]
> > wrote:
>
>> Hi.
>> I've been sinking lots of time in trying to get ALSA to work with a
>> SGTL5000 (Teensy Audio Adapter) in Raspbian on a Raspberry Pi 3. I have the
>> complete documentation for how to manually set up the audio chip, and
>> sufficient documentation on how to manually set up the Pi's ARM I2S
>> peripheral. The ALSA solution is generalized for any kind of hardware you
>> could want, is something I don't really want to know, at all, in my life,
>> and seems just far too complex when I have detailed guides for both the
>> audio chip and the CPU. The one combination of ALSA platform, machine and
>> codec drivers that I've managed to get sound from plays more than 2x too
>> fast and will not open with JACK.
>>
>> So, I'm reading about the JACK API, wondering about simply writing a C
>> program to configure and write to the Pi's built-in I2S FIFOs, without even
>> telling ALSA that I'm doing it; but I see that all JACK clients should use
>> callback functions. Still, audio input from JACK into a program using the
>> API is a regular thing, right? So how complicated might it be to set up
>> such a client that gets audio input from JACK and writes that to the chip's
>> I2S FIFO? My timing will be controlled by the audio chip, the Pi chip will
>> be a slave to it. So as long as my program writes to the transmit FIFO fast
>> enough, and waits if the FIFO is full, I think I don't need to worry about
>> synchronizing JACK exactly to the audio chip's clock... right?
>>
>> I'm just a little fuzzy on this timing part.
>> Thanks for any clarification.
>>
>> -Chuckk
>>
>>
>> --
>> http://www.badmuthahubbard.com
>>
>> 
>> Jack-Devel mailing list
>> [hidden]
>> http://lists.jackaudio.org/listinfo.cgi/jack-devel-jackaudio.org
>>
>>
>


-- 
http://www.badmuthahubbard.com

1463267595.31470_0.ltw:2,a <CAFA-yJm-EgRgXVBD0qK9bg3mAeukw8KDR_9HeozqM2NPE+usYw at mail dot gmail dot com>