Code size report:

   bare-arm:    +0 +0.000% 
minimal x86:    +0 +0.000% 
   unix x64:    +0 +0.000% standard
      stm32:    +0 +0.000% PYBV10
     mimxrt:    +0 +0.000% TEENSY40
        rp2:    +0 +0.000% RPI_PICO_W
       samd:    +0 +0.000% ADAFRUIT_ITSYBITSY_M4_EXPRESS
  qemu rv32:    +0 +0.000% VIRT_RV32

Thanks for the contribution. This looks fine to me.

But note that clocks_extra.c uses XOSC_KHZ... so maybe that needs changing as well? Why didn't that lead to a build error in your case?

Also, pico-sdk provides defines of XOSC_MHZ in terms of XOSC_HZ so I'm also uncertain why that define wasn't provided in your case? What value is your XOSC_HZ?

My added definitions are the following:

target_compile_definitions(${MICROPY_TARGET} PRIVATE
        XOSC_HZ=30720000

        PLL_SYS_REFDIV=4
        PLL_SYS_VCO_FREQ_HZ=960000000
        PLL_SYS_POSTDIV1=4
        PLL_SYS_POSTDIV2=2
        SYS_CLK_HZ=120000000

        PLL_USB_REFDIV=4
        PLL_USB_VCO_FREQ_HZ=1344000000
        PLL_USB_POSTDIV1=7
        PLL_USB_POSTDIV2=4
        USB_CLK_HZ=48000000

There is code for backwards compatibility in platform_defs.h:

// For backwards compatibility define XOSC_KHZ if the frequency is indeed an integer number of Khz.
#if defined(XOSC_HZ) && !defined(XOSC_KHZ) && (XOSC_HZ % 1000 == 0)
#define XOSC_KHZ (XOSC_HZ / 1000)
#endif

// For backwards compatibility define XOSC_MHZ if the frequency is indeed an integer number of Mhz.
#if defined(XOSC_KHZ) && !defined(XOSC_MHZ) && (XOSC_KHZ % 1000 == 0)
#define XOSC_MHZ (XOSC_KHZ / 1000)
#endif

Because of the test XOSC_KHZ % 1000 == 0 the XOSC_MHZ is not defined.

The other point about XOSC_KHZ is true. I will later update my MR to also replace this in clocks_extra.c.

In theory we could/should also replace all usages of SYS_CLK_KHZ/USB_CLK_KHZ and the MHZ variants. The thing is, that those are not only used for clock_configure() (where they are converted to a HZ value) but also for set_sys_clock_khz() where we could replace the whole call with set_sys_clock_hz(). But I am not sure if this would have some side effects because of the comment in check_sys_clock_khz():

Note this impl is kept to preserve previous rounding behavior, vs calling check_sys_clock_hz

What is your suggestion?

Because of the test XOSC_KHZ % 1000 == 0 the XOSC_MHZ is not defined.

Ah, right, that makes sense then. Your crystal is not an exact MHz multiple. So you definitely need the change in this PR to get the correct timing.

The other point about XOSC_KHZ is true. I will later update my MR to also replace this in clocks_extra.c.

Yes, please make this change. Looking at pico-sdk, it seems they also updated their code to use XOSC_HZ, which clocks_extra.c is adapted from.

In theory we could/should also replace all usages of SYS_CLK_KHZ/USB_CLK_KHZ and the MHZ variants.

No, let's leave those as they are for now. Too much of a risk to change that just before a release.

I now also removed the usage of XOSC_KHZ. And I improved the change for XOSC_MHZ even more by removing the unneeded definition of the xosc_hz variable in modmachine.c.

I hope this matches your expectations. And thanks again for this great project!

And thanks again for this great project!

😄