From: Kai Becker (kmb@kai-m-becker.de)
Date: Sat Feb 08 2003 - 05:23:06 MST
Am Samstag, 8. Februar 2003 05:27 schrieb avatar:
> In reference to the below: could people on the Internet allow their
> computers to host a version of one of the games of life type
> simulations, as they do the SETI programme, utilizing their computers'
> idle sections?
Conway's game of life? That is not as easy as SETI. SETI data packets can
be calculated independently, but patches of the GOL board interact with
each other on their borders in each step. You can't compute step n+1 of a
patch until all of its neighbors have reached step n. This is a cascading
effect, where one slow computer can delay the evolution of large parts of
the board.
The other interesting question is the relation between patch size and
efficiency. The computational effort for the state transisition is rather
small compared to the overhead for communicating the patches over the
internet[2, p.11]. SETI packets have a far better ratio. That's why
normally Connection Machine like systems are used for CA. [1] has some
ideas for tuning. And [2] uses an interesting approach to reduce the
communication overhead for GOL and a massive parallel neural network.
Kai
[1] http://cafaq.com/soft/index.shtml#parallel
[2] http://www.ccrl-nece.de/~greg/Papers/icmp.pdf
-- == Kai M. Becker == kmb@kai-m-becker.de == Bremen, Germany == "Any technology distinguishable from magic is insufficiently advanced"
This archive was generated by hypermail 2.1.5 : Sat Feb 08 2003 - 05:25:37 MST