Keyboard on demand

[tomelgato]

Hello Dave, one more thing ;)

I use the 5x4 keypad, 20x4 LCD together with the AccelStepper library. When moving the stepper motor i dont call taskManager.runLoop(), otherwise the performance on the Due is not sufficient. So far so good.

But now i need a function where i need to poll the keyboard (nothing menu related) and drive the stepper at the same time. The plan is to poll the keyboard directly. So the Q is if i just can steal the keyboard pins or if i have to close/open them via iOAbstraction/taskmanager/keyboard.

[davetcc]

I'm always surprised that nobody seems to use the inbuilt ARM SAM hardware PWM or interrupt support. It seems to me that either of these would control stepper motors much more effectively. But that's an entirely different discussion outside of tcMenu.

The short answer is yes you could directly read it, but there may be another way. Reading a matrix keyboard is slow, and involves many steps, the guide on keyboards explains it. However, that aside, you could call exec() on the keyboard manager, and it would service the keyboard in the usual way without doing anything else. You could even try reducing the stabilization delay in the exec method until you get failures. In your case though, if you did that you'd need to replace the yieldForMicros(..) with delayMicroseconds(..). If this worked, I'd make that change behind a permanent build flag.

[tomelgato]

Ok i try and give you feedback. Thanks a lot!

[tomelgato]

I think a second exec call like exec_external() which would not call the loop and use Micros would be better then do do some flags, no?

[davetcc]

What would be best is to refactor it slightly and have a method something like readGrid(int whichLine) that allowed more control for these situations.

[tomelgato]

Yes that would be perfect. Line / Row ... for people using a keypad to control something in X/Y direction. Lol, usual featuritis is coming to effect. ;)

[tomelgato]

Ping ;) See below

[tomelgato]

Hello Dave! Finally i could do some speed tests.

taskManager.runLoop() is somewhere in my main file, its not full speed cause some other stuff is running too.

The 2nd measurement is running in a small sub loop without calling taskManager.runLoop(), instead i poll the keyboard with keyboard.exec();

taskManager.yieldForMicros(500);
//delayMicroseconds(500);

taskManager.runLoop(): 184971 Loops/sec
keyboard.exec(): 237798 Loops/sec
---
// taskManager.yieldForMicros(500); 
delayMicroseconds(500);

taskManager.runLoop(): 196871 Loops/sec
keyboard.exec(): 242423 Loops/sec
---
taskManager.yieldForMicros(250); 
//delayMicroseconds(250);

taskManager.runLoop():	192256 Loops/sec
keyboard.exec(): 243672 Loops/sec
—
//taskManager.yieldForMicros(250); 
delayMicroseconds(250);

taskManager.runLoop(): 201026 Loops/sec
keyboard.exec(): 247453 Loops/sec

[tomelgato]

So altogether its possible to get 10k loops more on the Arduino DUE when not calling the taskmanager indirectly and reducing the delay to 250us