Real Time data acquisition preventing menu tasks from running

[leelx78]

Hi Dave,

I'm working on a project where my ESP32-S3 CPU is maxed out doing real-time data processing in the loop() function on Core 1, together with the Arduino system. This means:

void setup() {
    DEVICE.begin();
}

void loop() {
    DEVICE.loopFunction();
}

...where DEVICE is my class with a begin() and loopFunction() methods.

Then I wanted to add TCMenu and a SSD1306 display plus a rotary encoder/switch but it doesn't seem to work at all. the function taskManager.runLoop(); seems blocking or just uses too much CPU time. I tried 2 approaches i.e.

void setup() {
    setupMenu();
    DEVICE.begin();
}

void loop() {
    DEVICE.loopFunction();
    taskManager.runLoop();
}

and

void setup() {
    setupMenu();
    DEVICE.begin();
    taskManager.schedule(repeatMicros(100), [] {
      DEVICE.loopFunction();
    });
}

void loop() {
    taskManager.runLoop();
}

With the same result.

Do you have any suggestions? In your video I remember you mentioned running a second taskManager on another core but I didn't see how to do it. Ideally I'd like to keep the UI on core 0 and all the existing real-time processing on core 1

Let me know what you think

thanks!

[davetcc]

Please only raise actual issues using issues Anything else create a discussion. The format of issues must not be changed and they are always to be created using templates. Thanks.

[davetcc]

First I would confirm that everything works properly without the resource consuming task in play. IE in a test project.

Then as you say you could try to move that resource consuming task to another core. Using TaskManagerIO as a queue for example. That should be as simple as submitting work to it (maybe using the execute method). Be aware that any data shared between cores will need to be synchronized or atomically safe. Steps I can think of are:

• globally create a second task manager instance.
• create another native task on the other core
• Run the task manager loop within that.
• Submit work to be done on the other taskManager.

[leelx78]

works now, I had to initialise the encoder with unused pins and then call manually. Seems like a hack but works.

I'm curious why my encoder interface works so much better though. it's based on the CFRotaryEncoder library, maybe it's worth you have a quick look to see if it's something that might be an option for the main code?

[leelx78]

..actually I see that menuMgr.performDirectionMove() is not doing what I need, which is up/down.

@davetcc did I miss something in the MenuManager docs? menuMgr.onMenuSelect works well when connected to my encoder interface and menuMgr.performDirectionMove also "works" but it only as a back/forward, not up/down.

Maybe I'm overcomplicating things but I tried so many combinations of things, including the TC_LEGACY_ENCODER flag and it seems to miss loads of encoder clicks with your interface but it's spot-on with mine for some reasons..

[davetcc]

For TC_LEGACY_ENCODER you must be using the unreleased master branch? This is as yet unfinished, and I have yet to test it with Interrupts enabled - early indications show a bug with interrupt support that will need to be worked through before I can release it.

However, if you really want to handle the encoder yourself, the best way is to write an implementation of the RotaryEncoder class. You could take a look at the HardwareRotaryEncoder and EncoderUpDownButtons classes and create an implementation for your case.

[davetcc]

EG - https://github.com/davetcc/IoAbstraction/blob/master/src/SwitchInput.h#L203

[leelx78]

Thanks @davetcc I'll try to implement it as soon as I have a chance.

A way to do this via MenuManager would be handy for the future if you have a chance though. Something like menuMgr.performDirectionMoveUp, menuMgr.performDirectionMoveDown and so forth.

[leelx78]

Hi @davetcc I close this and open a new issue since I'm narrowing it down to the encoder class thanks