Fuzzy Paralleliser

Frank Hoffmann (fhoffman@cs.berkeley.edu)
Thu, 25 Mar 1999 19:55:41 +0100 (MET)

*********************************************************************
Berkeley Initiative in Soft Computing (BISC)
*********************************************************************
Heres an idea I wish to implement. All suggestions, comments,
critics are eagerly awaited.. Please also let me know if anyone
has come across such an implementation.

Fuzzy Parallelising Compiler.

Aim: To parallelise an existing sequential code using Fuzzy Sets &
Logic
Given:
1: A sequential code
2: Number of nodes/computing systems which can communicate
with each other. One node is configured as Master, and
the rest are Slaves connected in some network configuration.
Proposed Idea:
What is the basic requirement?
I feel that the basic requirement is to break up the code into
small segments of code such that the number of computations within the
segment is higher than the number of communications required
between the segments. The segments can then be assigned to the
various nodes. If the number of segments is lower than the
number of nodes, we will be utilising the computing power to
below its full capacity. And otherways, if the inherent number
of segments is larger than the number of computing nodes available,
then the segments will have to be scheduled according to the
priorities required by the code segments' variables.
Essence:
To find a criteria equivalent to conditions of homogeneity
and heterogeneity, such that homogeneity refers to the ratio
of number of computations that can be carried out within a
cluster of code lines, to the number of inherent communications
required for the end-result to be achieved from that particular
segment of code. And heterogeneity can be refered to the figure
which aggrandises the need for breaking a code segment into
two or more code segments.
We can consider homogeneity to be a figure of merit for
the intra-segment case, and heterogeneity to be the figure of
merit for the inter-segment case.
Condition:
A number of segments such that the intra-segment homogeneity
is larger than the inter-segment heterogeneity.
How to go about it:
One idea is to do it the way normal parallisers do it. By
using directed graphs. What I wish to figure out is a method
using FSS which will look at the segments as small Fuzzy sets
and Fuzzy Logic to decide how to partition these sets so that
the above mentioned criteria is optimised. What will be required
is a form of communication channel or a way to figure out how
the small fuzzy sets of code could interact with each other, so
that they themselves can form optimised clusters of code. This
interaction assumes that lines of code can be passed from one
set to another just like data elements, and these flow around
various "near-by" clusters and settle down into a number of
segments of code such that the total "energy" of the system
coupled with the number of nodes is minimum.
But exactly how....is the main question!

With this rough idea in mind, I'd wish to get some kind
of pointers as to how such an idea can be implemented. All
suggestions welcome.

with warm regards,
Akash Mohabey.

If you ever want to remove yourself from this mailing list,
you can send mail to <Majordomo@EECS.Berkeley.EDU> with the following
command in the body of your email message:
unsubscribe bisc-group
or from another account,
unsubscribe bisc-group <your_email_adress>
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
Do NOT send unsubscribe requests to bisc-group@cs.berkeley.edu
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

############################################################################
This message was posted through the fuzzy mailing list.
(1) To subscribe to this mailing list, send a message body of
"SUB FUZZY-MAIL myFirstName mySurname" to listproc@dbai.tuwien.ac.at
(2) To unsubscribe from this mailing list, send a message body of
"UNSUB FUZZY-MAIL" or "UNSUB FUZZY-MAIL yoursubscription@email.address.com"
to listproc@dbai.tuwien.ac.at
(3) To reach the human who maintains the list, send mail to
fuzzy-owner@dbai.tuwien.ac.at
(4) WWW access and other information on Fuzzy Sets and Logic see
http://www.dbai.tuwien.ac.at/ftp/mlowner/fuzzy-mail.info
(5) WWW archive: http://www.dbai.tuwien.ac.at/marchives/fuzzy-mail/index.html