My understanding, with no empirical basis, is that TOS supports
4 byte floats on the micas in software using a standard math lib.
The controller itself has a set of multiply instructions that
work on 8 bit ints, but no divide(??!). Given that and a
very limited RAM area, on-board FP matrix operations are quite
a luxury. A single 8x8x4 matrix is 1/16th of your total memory.
If your app allows it, using 16bit fixed point ints might work.
I just proved to myself that floats work in T1 with "make pc".
So if all you want to do is simulate reality, you should be set.
MS
Prem Krishnan wrote:
Hi,
I'm implementing a data aggregation scheme that requires each mote to
perform computation involving floating point data types (Gaussian
Elimination - Matrix size of 8x8). However, I'm being told by technical
support at xbow that mica/micaz do not support floating point operations
because of a 8 bit MCU. I was wondering if I could instead use TOSSIM to
simulate my network. I'd really appreciate if someone could answer my
following queries:-
Q1) Is it possible to simulate the floating point operations like
Gaussian Elimination using TOSSIM ?
Q2) I gather that it currently supports only Micaz. Does this mean that
TOSSIM emulates the Atmel Atmega 128L microcontroller? Since the Atmega
128L does not support floating point operations, is it the case with
TOSSIM as well?
Q3) I need to find the computational / packet delay using our
aggregation scheme. Does TOSSIM let us find such data using network
simulation?
thanks,
Prem
------------------------------------------------------------------------
_______________________________________________
Tinyos-help mailing list
[email protected]
https://mail.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help
_______________________________________________
Tinyos-help mailing list
[email protected]
https://mail.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help
