[En-Nut-Discussion] RFC: Nut/OS Tasks

[Nathan Moore]

On Tue, Jul 14, 2009 at 9:37 AM, Timothy M. De Baillie <
debaillie at ciholas.com> wrote:

> It seems like this conversation is trying to make a cooperative
> multitasking system act like a time-sliced system.
>

Not at all.  It's mainly about making timers that can do more work that
current timers are
able to do (current timers are unable to sleep do anything that might
yield).  This keeps
the current timer system from being able to be used for many of the things
timers are
needed for -- notably TCP retransmits.  TCP (and PPP) have state machine
threads that
in large part are just there to act as timers for timeout processing.

If such a substantial change to the OS is done it only makes sense to try to
anticipate
other uses of whatever new functionality is added -- beyond the original
problem.  Trying
to think about if there are other situations where something similar could
simplify some
other code or even make something new and useful possible.

We have not discussed anything resembling time slicing, and the only thing
resembling
preemptive multitasking was when I said that second halves of interrupts (if
used) could
be run right after the first half of the interrupt (what is currently the
interrupt routine).
This was only a high performance suggestion of how to deal with that
particular situation,
but it comes with it's own problems (more likely to overflow stacks).

Nathan