@agatti FYI. Please can you double check this popcount implementation and my fix here, thanks!

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 98.57%. Comparing base (e57aa7e) to head (6fee099).
Report is 1 commits behind head on master.

@@           Coverage Diff           @@
##           master   #17498   +/-   ##
=======================================
  Coverage   98.57%   98.57%           
=======================================
  Files         169      169           
  Lines       21968    21968           
=======================================
  Hits        21654    21654           
  Misses        314      314           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

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

I run the octoprobe tests against this PR: https://reports.octoprobe.org/github_selfhosted_testrun_76/octoprobe_summary_report.html

These are the results: tests_mpy-cross_native.txt

Now all md5 equal: The error has gone!

I don't have access to any of my boards/dev machines until the end of the month to see what's going on there (along with my copy of "Hacker's Delight" to double check things against), but if this PR is reported by multiple sources to fix the issue then please go ahead!

I took that code from an old project of mine running on a ColdFire board that's been working for a few years now, so now I wonder whether this works due to code consuming that value that either does the shift itself or masks it out and fills the other bits with something else - no idea until I take a look.