The Big Loop Project


The big loop project provides a craft tool to deal computations over a grappe of computers (cluster). At the origine, bigloop is a pedagogical experience to learn about programming sockets. I used bigloop 0.1 to find the number of bent functions by means of the host of the computer science departements. This to say that bigloop is ready to use in real life ! The version 0.2 below should be operationnal for any C-programmer in a linux environment, but still a lot of improvements remain to do.


The bigloop approach is interesting for people that have a C-line code :

for( i = 0; i < n; i+=s ) task( i )

where n is large, and independent non trivial task(). In my case, I have a lot of problem that lead to a such loop. Of course, one can use MPI like language to parallelize. But, if you do not need to exchange of complex data, bigloop is a very simple way to distribute the computation without assuming more than the presence of a C-compiler on the friend hosts over a net subject to electrical trouble...


The objective of the project is to find the better way to parallize kindly and safely one C-line loop. Nothing more, nothing less.

bigloop 0.2

bigloop 1.0

This future version will take in account the work factor profile of segments in order to optimize the distribution of the tasks with a step adaptative methode.
Philippe Langevin , Last modification on March 2012.