examples/natmod/features2 on rp2350 not working?

[jkorte-dev]

Hi,
I had an issue using floats in a native module on rp2350 and therefore tried the feature2 natmodule example.
Feature2 from the examples compile for arch armv7emsp behaves different on rp2350 and stm32f411:
I tested the same features2.mpy build on a backpill and a pico 2w:

Correct calculation (stm32f411):

stm32f4 % mpremote 
Connected to MicroPython at /dev/cu.usbmodem365D345235351
Use Ctrl-] to exit this shell

>>> import features2
[True, True, True]
>>> 
>>> import os
>>> os.uname()
(sysname='pyboard', nodename='pyboard', release='1.25.0-preview', version='v1.25.0-preview.172.g5d12df51f.dirty on 2025-01-14', machine='WEACT_F411_BLACKPILL with STM32F411CE')

wrong/failure on rp2350:

rp2350 % mpremote                   
Connected to MicroPython at /dev/cu.usbmodem101
Use Ctrl-] to exit this shell

>>> import features2
[False, False, True] 
MPY: soft reboot
MicroPython v1.25.0-preview.49.g0625f07ad.dirty on 2024-11-21; Raspberry Pi Pico 2 W with RP2350
Type "help()" for more information.

on import
https://github.com/micropython/micropython/blob/master/examples/natmod/features2/test.py
is executed

Is arch armv7emsp wrong for rp2350, which one is the right one?

[jonnor]

The arch should be checked when the natmod is loaded, so that seems correct. Can you check what actual values you get on RP2350 (the incorrect ones vs expected)?

[jkorte-dev]

the pico2 does a soft reboot after import.

I had to remove the SytemExit() and recompile.

Here the result fom the pico2:

from features2 import add , add_f , productf

def isclose(a, b):
    return abs(a - b) < 1e-3

v1 = 0.1
v2 = 0.2
expected = 0.3

r = add(v1, v2)
print(f'add(): {v1} + {v2} returned {r} , isclose to {expected} => {isclose(r, expected)}')
r = add_f(v1, v2)
print(f'add_f(): {v1} + {v2} returned {r} , isclose to {expected} => {isclose(r, expected)}')

import array
ar = array.array("f", [1, 3, 0.3])
productf(ar)
print(f'productf(): 3 * 0.3 returned {ar[0]}')

[False, False, True]
add(): 0.1 + 0.2 returned 0.2 , isclose to 0.3 => False
add_f(): 0.1 + 0.2 returned 0.2 , isclose to 0.3 => False
productf(): 3 * 0.3 returned 0.9