Is there a way to test time.ticks_diff() ?

[MrTinkerman]

I wanted to test the rollover behavior of time.ticks_diff() so I tried the following:

import time

t1 = 2**30 - 1000          # 1000ms before rollover
t2 = 1000

print(time.ticks_diff(t2, t1))

-268469032

When I asked perplexity (ai) what was was wrong it said ticks_diff needs to get actual 'timing values' instead of integers.
Does that make any sense?

And if so, how could timing values be created if at all?

Thanks!

[robert-hh]

ticks_diff() takes integers. Only they must be in the range of the ticks_xxx() functions, usually 30 bits. So your result looks odd. Which platform did you use for the test?

Edit: Using your code snippet on a RP2 I get as result 2000.

[MrTinkerman]

I'm running this code on:
MicroPython v1.22.2 on 2024-02-22; XIAO nRF52840 Sense with NRF52840

Well this gives me a diff of zero, but I have no clue what to make of it:

import time

t1 = 2**30
t2 = 268470024

diff = time.ticks_diff(t2, t1)
print(f't1: {t1}  t2:{t2}')
print(diff)

[robert-hh]

On a XIAO nRF52840 Sense with NRF52840 I get the same result of 2000 with the first code snippet and MicroPython 1.24.0. 2**30 is by 1 out of the range of a 30 Bit number, which will then give you unexpected results.
With 2**30 - 1 instead I get a result of 268470025 with your second test script, which is reasonable.
Try updating the firmware. to v1.24.1.
Edit: It may be that the v1.22.2 version you use has a smaller ticks_xxx() range.

[MrTinkerman]

Thanks for the info and testing, I'll consider 1.24 then.
I didn't yet upgrade because of this issue I posted earlier: https://github.com/orgs/micropython/discussions/16378

Is there maybe a way of finding the max ticks_ms for this firmware/board so I can write my own version of tdiff?

EDIT:
I think I'll use a workaround for now. I use it to run code on intervals but it doesn't matter if the interval is shorter once every several days.

import time

t1 = 123456789
t2 = 100

intv = 10000

if abs(t2-t1) > intv:
    print('do something')

[robert-hh]

The respective code for mp_hal_ticks_ms() and mp_hal_ticks_us() is unchanged since July 2020. And it does not change the definition of MICROPY_PY_TIME_TICKS_PERIOD being used in time.ticks_ms() and time.ticks_us(). The implementation seems to be based on a 32768Hz 24bit counter and a roll-over counter, resulting in a 56 Bit counter, of which the lower

[mendenm]

Note that the warning on the documentation page for ticks_diff() and the ticks_ms() about them only being intended to be used together, and never with any other source of numbers is real. I've been bitten by this. ticks_diff() assumes, blindly, that the number you have handed in is a python small (inline) integer. This doesn't just mean it has to be small (<2**30), it means it needs to have the high bits clear, so it is an inline small integer. If you pass in an integer which has been converted to an object, it will read the high bits (which flag that it is an object), and the rest of the pointer to the integer, as the value, and return nonsense.

[robert-hh]

If you stay in the python domain, it's safe. tick_ms() and ticks_us() just return integers. Small integers in this case. And ticks_diff() accepts small integers in the range of 0..MICROPY_PY_TIME_TICKS_PERIOD-1, which is by default 2**30-1. ticks_diff() cannot tell if this integer originates from a tick_xx() call or just another integer.

[mendenm]

My point was that is is really easy to accidentally pass in an integer that has been 'digested' by some other code, which is no longer inline. You cannot necessarily pass into ticks_diff() the result of an arbitrary python calculation, even if it appears small.

[scruss]

My point was that is is really easy to accidentally pass in an integer that has been 'digested' by some other code

That's why the time.ticks_ms() docs say:

"For the most part, you should treat values returned by these functions as opaque. The only operations available for them are ticks_diff() and ticks_add() functions described below."

They also note that time.ticks_add(0, -1) gives you the platform-specific value of TICKS_MAX