14:29 <dinuxbg> Siegurd: You need a host-side Linux kernel driver to allocate physically contiguous memory. Alternatively, you may reserve memory in DTS, and mmap it from /dev/mem from your userspace application.

14:31 <dinuxbg> Siegurd: I have not tested it, but the following thread offers a viable solution: https://forum.beagleboard.org/t/pru-c-shared-mem/35095/13

16:36 <zmatt> or use uio instead of rproc

16:38 <zmatt> (if you do through any route that involves manually reserving memory in DT, I would recommend creating an uio_pdrv_genirq device for that chunk of memory so that userspace can mmap it without having to use /dev/mem)

16:45 <zmatt> Siegurd: my py-uio project ( https://github.com/mvduin/py-uio ) includes an example that uses shared memory to stream messages from pru to python. that particular example has both an assembly version (stream.py + fw/stream.pasm) and a C++ version (stream-c.py + fw-c/stream.cc)

16:45 <zmatt> (most examples are in assembly)

16:50 <zmatt> if you want to stream adc samples to linux userspace it probably makes more sense to just use a ringbuffer of samples instead of sending large messages, but it depends on the exact use-case I guess

16:54 <zmatt> (note that the py-uio stream examples use a ringbuffer to stream messages, which works the same except it reads one message at a time while with streaming examples you'll want to read in larger chunks for efficiency

16:54 <zmatt> )

17:00 <zmatt> are you just using PRU to stream ADC data to userspace as an initial test, with the plan being to later actually use PRU to process the ADC data itself?

17:09 <Siegurd> Yes, I want to make PID regulator that reads ADC channel and controls PWM channel

17:11 <Siegurd> But first I want to understand what's ADC values de-facto are

17:16 <Siegurd> the frequency of PWM should be constant = 800 Hz, and the duty-cycle should be regulated by PID

17:18 <Siegurd> Of course I can do it in STM32-Arduino within 2 hours with PID auto tuning and so on, but I want go the harder way

18:18 <zmatt> if your PWM is that slow, why do you need the ADC configured at its highest sample rate possible?

18:21 <zmatt> like, I can understand sampling at a higher rate than you need so that you can implement filtering to reduce the noise (in a way better than the averaging the ADC can do for you), but still

18:24 <zmatt> having the adc sampling 250 times faster than the output rate you need seems... generous

18:42 <Siegurd> Yes, I unerstand. But I wanted to confirure ADC in a proper way and know its maximum F. For some others projects in the future.

22:36 <zmatt> ok, just make sure to configure an appropriate sampling time (see the comments in adc.dtsi) if the thing you're measuring has a source impedance higher than 5.5 kOhm

22:36 <zmatt> (for lower source impedance it should be okay to use minimum sampling time)