[Tinyos-devel] message_t thoughts: encapsulations and interfaces

[Eric Decker]

On Tue, May 26, 2009 at 12:52 PM, Miklos Maroti <mmaroti at math.u-szeged.hu>wrote:

> Hi Eric,
>
> > So I would propose we split things up into two parts.
> > 1) Writing encapsulations.
> > needed for AM: dest(16), src(16), len (8), group(8), and type(8)
> > ie.  simple AMencap(dest,  type) which would default to compiled in
> static
> > definitions for the other values or more complete AMencap_full(dest, src,
> > len, group, type).
> > 2) Handing the packet off to a different layer...
> > send?   send(msg).
>
> I agree that we need to split things up, and it is not practical do
> all processing by adding/removing header fields without subsequent
> access to those by some interface.
>
> For example, the AMPacket.setXXX commands are used in the following
> components:


I would propose something of the form:

AMencap.write(part1, part2, ..., part n) that covers the majority of cases
for writing the encapsulation
and we break out with other calls for other parts.  The idea is to avoid the
overhead of figuring out
repeatedly (which I think will foil the optimizer) where the offset is and
the subroutine call and the
inherent hit on the stack.

Once we agree on a reasonable direction we can try some things out and see
how the code looks.
I'm getting good at dealing with the code generated and also looking at how
the system is behaving
using a jtag based development environment.  I'm looking at low level
behaviour of the code I'm
developing.


> apps\BaseStation\BaseStationP.nc
> apps\tutorials\RssiDemo\InterceptBase\BaseStationP.nc
> tos\lib\net\6lowpan\IPP.nc
> tos\lib\net\lqi\LqiRoutingEngineP.nc
> tos\lib\net\tymo\LoopBackM.nc
> tos\lib\net\tymo\mh\MHEngineM.nc
> tos\system\AMQueueEntryP.nc
>
> I have to point out that PacketAcknowledgement also have to access the
> IEEE 802.15.4 header fields, so that should also be either passed down
> with send() or set separately with an interface. I am not even trying
> to search for all users of the PacketAcknowledgement interface.


I'll take a look soon once we start getting closer to a direction.
Sounds like messing with the 802.15.4 code would examine what we are
talking about really well too.

Are you intimately familar with that code?  I haven't looked at it yet.
What about the lowpan6/IP code?

eric


>
>
> Miklos
>



-- 
Eric B. Decker
Senior (over 50 :-) Researcher
Autonomous Systems Lab
Jack Baskin School of Engineering
UCSC
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://www.millennium.berkeley.edu/pipermail/tinyos-devel/attachments/20090526/e7906d1a/attachment.htm