Teensy4.0 CPU count reset

[kdharbert]

Teensy 4 goes through it's full CPU count in about 7.15 seconds. I need to reference it in order to fix timing. However, there's a jump when ever the CPU flips. Is there a slower clock I can reference to get around this?

Also is there a way to reset the CPU count without effecting anything? I might be able to solve the problem by resetting before it can flip.

 

[MarkT]

Normally one uses timestamp differences to avoid issues with the wrap-around - perhaps this could fix the problem
in the code you've not posted?

 

[defragster]

micros() lasts longer with 1 instead of 600 million ticks per second. But it takes a bit more time - like 40 cycles to read than the CYCCNT which takes 3 or less to transfer the value.

Indeed what is being timed for how long and what fashion - update start and check end periods only - or continuous checks from start to end of a process?

The wrapping of the 32 bit value can be handled depending on use or need.

Also if micros() is good enough there is a wrapper for with an elapsedMicros variable.

 

[luni]

Here a simple method to extend the cycle counter to 64bit https://github.com/TeensyUser/doc/wiki/Using-the-cycle-counter

 

[kdharbert]

Darn...looks like detecting the overflow and accounting for it is the only way.

 

[jonr]

It's non-intuitive, but by using unsigned arithmetic, the wrap-around issue can be made to go away.

 

[KurtE]

Again don't know how accurate you need. For example is the micros() sufficient?

if so the elapsedMicros object take care of this for you.

you can do things like:

 elapsedMicros em;
 ... 
  em = 0;  // resets to zero.
  <Do your stuff>
  Serial.print((uint32_t)em, DEC);

And the object takes care of the book keeping for you. But note, you can do similar with just using uint32_t...

  uint32_t start_time = micros();
  <do your stuff>
   uint32_t delta_time = micros() - start_time;
   Serial.print(delta_time, DEC);