I really do question the need for a fancy task scheduler and computed calls for what is essentially a 'small system'. With most embedded systems, the tasks are known up front so therefore the address can be hard coded. similarly with scheduling the priorities are normally fixed - in most of my apps I use a co-operative tasker with a 10mS tick. I have up to 8 tasks of fixed priority. Each of these tasks has a timer that activates the task when it expires, the task can also be activated by calling "set_task". At each dispatch, the task run flags are evaluated for the highest priority to run, then that task is run for that tick. Whilst inter-task comms and semaphores are nice, they have a bit of overhead and a probably best left to a pre-emptive type RTOS. By definition - a co-operative task system need no interlocks since only one task can run at one time, and the task runs to completion. Sorry to pour water on your idea!
Nowadays, if I want a pre-emptive type RTSO with semaphores and inter-task communication (pipes etc) - I wouldn't be choosing to use a 8051 type micro, most probsbly it would be ARM.

I did see a product doing pre-emptive task switching by saving the whole cpu state (including stack) to xram - and with a 12 clock cpu! Needless to say a task switch was expensive for cpu time. I couldn't help but think at the time a co-operative tasker would've been more suitable.