This repo collects some testing results with the new arch_switch().

Want to test with your favourite Cortex-M platform? Check out the Testing Howto

• FRDM MCX N947
• NUCLEO F411RE
• STM32H750B-DK

The new arch_switch :

• significantly improves TM_COOPERATIVE: 8-12% compared to the current implementation.
• slightly improves TM_PREEMPTIVE: 4-6%
• negatively impacts TM_INTERRUPT: -13%. That's to be expected. Quoting Andy:

  That's the trade off, btw: defer work to interrupt time (very few apps spin hard on interrupts, and when they do everyone considers it a design bug) to benefit cooperative switch performance (something almost all apps can get stuck on unless implemented very carefully)

• The TM_INTERRUPT_PREEMPTION offers an interresting case: +11.7% for the cortex M4, but -16.2% for the M33. I am somewhat baffled by this result
• Other test shows no improvements, which is to be expected I think.