MicroPython Forum

In your code you call both mp_stack_ctrl_init() and mp_stack_set_top(...). Only one of these should be called, usually it's mp_stack_ctrl_init() which sets the stack top automatically. mp_stack_set_top is only used if setting it automatically doesn't work. Otherwise, the stack limit checking is quit...

The PYBD_SF6 uses double precision float, and only provides the double precision math functions (eg sin, cos, not sinf, cosf). I see in your ulab code that you use "float" exclusively as the FP data type (at the C level). Is this a restriction or can it also work with double? I would suggest changin...

This really sounds like a hardware issue with the external QSPI flash, the memory-mapped one. The only thing I can suggest at this stage is to remove the PYBD_SF2 completely from anything connected to it, and power it up over USB and measure the current draw using a USB power meter. It should draw a...

I think it makes sense to create new sub-forums for ports, but only as they see interest/use. The nrf port is a good candidate to start with.

Some of the unix command-line test do run the unix port through a piped connection to a subprocess with redirected input, so that is certainly possible. In such a case raw REPL would be useful.

When developing the pyboard D-series we did a lot of low-power measurements and optimisations, because it was a key goal to obtain optimal lowspower operation of the MCU and WiFi chips. Not all low-power features are available yet, nor all characterised, but to give you an idea: in lowest power mode...

For enabling help on the unix port, there was already a PR submitted for this, see https://github.com/micropython/micropython/pull/3436 for the discussion. I'd be in favour of enabling it, at least so that help('modules') works. As for the raw repl, I don't think that's as easy to enable on unix, pa...

Very interesting! Note that PYBD runs by default not at the maximum possible frequency. SF2 is at 120MHz and SF6 at 144MHz by default. This can be increased to the maximum via machine.freq(216_000000). Raw performance should scale roughly linear with clock speed. Maybe it is worth increasing the GC ...

Wow, that looks really great, thanks for sharing it! IMO it would be good to match numpy naming and semantics as much as possible, even for things that seem a little awkward. People who already use numpy already know about the subtleties like upcasting and so it'd make sense to just copy the numpy b...

The above fix was merged into master.