I played around a little bit and got an example with qemu running :-) - it it platform independend (without makefile - just nimble used)
thank you very much - my intention was to tailor a complete running bare-metal example with newlib support (due to the lack of os:standalone :-( . At the moment I am fiddling with the newlib and race conditions (machine is freezing if you get a index-out-of-bounds for instance). But it's impressive that Nim's exception handling is running like a charm :-). Then the next steps would be: get the timer running and multithreading support ( __impure_ptr handling and that stuff).
If you look at the produced binary it's around 28k (without thumb-mode!) - and there is still room for improvements (still some string-bloat in it - strerror function and friends).
If that stuff is working it should be no problem to move to a board of your choice. QEMU has also PI and STM32F4 support for instance (my future plan is to test it with both physical boards but also support very small targets (no multithreading))