[Tinyos-host-mote-wg] Re: [Tinyos-2.0wg] proposed Scheduler interface change

[David Gay]

On 10/15/05, Philip Levis <pal at cs.stanford.edu> wrote:
> On Oct 15, 2005, at 11:53 AM, Cory Sharp wrote:
> > I think the interface change sounds good, since it's more directly
> > expresses in the intended use of the interface -- that sleep(TRUE) was
> > only intended for use inside the taskLoop anyway.
> >
> > But, is the primary savings coming from having the "while no task"
> > loop inside the atomic section?

That's definitely part of it, but see below too.

> > That seems like the only significant
> > practical change from the current method of while(1)
> > runNextTask(TRUE).
>
> I think there are also issues of inlining/component boundaries.

I'd say it's easier for the Scheduler implementer to ensure that
things happen the right way if they get to write the whole loop (e.g.,
in the mica case, I couldn't convince avr-gcc to inline runNextTask
even with an inline directive, so performance sucked). Desperate
implementers now have the option of writing the task loop in assembly
if necessary...

David

_______________________________________________
Tinyos-2.0wg mailing list
Tinyos-2.0wg at Mail.Millennium.Berkeley.EDU
https://mail.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-2.0wg