Agner`s CPU blog

There are a few features of ARM Cortex-M (the microcontroller profile) that I miss when writing system-level programs in other architectures

Data stack and call stack are preferably separate. The call stack can stay on-chip most of the time. There may be a separate call stack for device drivers and system code. We can keep the system call stack on-chip by banning recursive function calls in device drivers. This would make it possible to make a device driver that doesn't even touch the data cache unless it needs to do a lot of data processing. The device driver has access to a limited block of the application program's data space which it is likely to have good caching. I don't know how an interrupt handler gets its memory. Maybe the application program can call a device driver to enable the interrupt handler and assign some of the application's memory space to it.

I have not decided how interrupts are dispatched, but we may have the interrupt vector table stored on-chip.