This work presents the parallelization of an algorithm for finding facility locations for an entering firm which has to make decisions on the locations of its facilities as well as on its price setting in order to maximize profit. Customers buy from the facility that offers the lowest price in the area the consumers belong to. This combinatorial location problem is solved by GASUB, a new multimodal genetic algorithm with subpopulation support that uses a cooling technique similar to simulated annealing. The high computational requirements of the location problem demands the parallelization of the method. In this work two common strategies have been implemented and compared. The first one follows a master-slave model, where the master processor takes charge of making global decisions and distributing the populations of points that must be evaluated. On the other hand the slave processors evaluate the objective function in those populations of points. The second strategy is the coarse grain parallelization, where the different processors execute the same algorithm of optimization on different subpopulations in an independent way. Though the executions are independent, intermediate results are sometimes exchanged.
Current 2016 2015 2014 2013 2012 2011 2010 2009 2008 2007 2006 2005 2004 2003 2002 2001 2000 1999 1998 1997 1996