Pranav:

now if you wish to activate another task the blocked task frame is swapped off into xram area.so that all tasks use same stack area.

Dan:

But yet, this is the way it's done in at least one commercial preemptive RTOS (CMX RTX51, IIRC).

Hmmm. Really high overhead!

What if I reserve, say 10 bytes (5 nested CALLs), for each task and move SP on each task switching?

I know that I'm asking too many questions, but your comments and links are really helping me a lot...