On Oct 26, 2005, at 8:45 AM, Andrea Pacini wrote:
I ' ve implemented a computational expensive operation on TMote Sky
and the time it takes is
large (about 30 seconds): this value was confirmed by some papers I
read some other papers about other platforms computing the same
things and they takes less than
one seconds to do the same but they wrote all code in Assembler.
My question is: is it possible that writing in assembler my code
reduce time from 30 seconds to
much lower times ? (Lower times OK, but so low (less than one
second ?))
Well, part of the answer depends on how good an assembly hacker you
are. :)
30-fold performance increases are unlikely. As a general rule of
thumb, really good assembly hacking can get you a 2-4x speedup. But
this really depends on the quality of the C optimizer, your assembly
skills, your ingenuity for data layout/transformation, and the
computational problem. I've been in situations where really well
tuned assembly has taken on tenth the time of a C implementation, and
also where it was only about 50% faster than a C implementation. (In
these sorts of cases, the C/nesC distinction is not too relevant.)
Phil
_______________________________________________
Tinyos-help mailing list
[email protected]
https://mail.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help
