This was just a quick project I cranked out as a proof of concept, mostly, that I could figure out a preemptive multitasking system at a very simple level. Basically, I took Lee Davison's EhBASIC for MC68K (RIP Lee Davison), which I had previously ported to my board, and I ran 4 instances of it task switching 64 times per second. Each instance used it's own serial port. The threads each run in user mode, and the supervisor implements a simple TRAP API for I/O, choosing the serial port based on the process ID. When the 64hz timer interrupt fired, I would save the context, load the next one, and then resume based on the new context. This is the only way a process releases control in this simplified system. There is no way to give up control early, not even when a blocking I/O call is made. Like I said, it's just a proof of concept. Due to the way Lee Davison wrote EhBASIC, the code is position independent, and it also accepts a pointer and a size to tell it the region of memory to use. No addresses are hardcoded, everything is a relative address. This makes it particular easy to do this with, since I can use the same code segment for each instance, and all I have to do is supply a different RAM region for each instance. I believe I divided a 1MB block of RAM into 4 256K blocks, with each instance of the BASIC interpreter using one block. Lots of things in my multitasking system were hardcoded and fixed. It is fixed at 4 processes, has memory for the 4 contexts statically allocated, etc., but it served fine as proof of concept.