Genetic algorithm based automatic data partitioning scheme for HPF

good data partitioning scheme is the need of the time. However it is very diflcult to arrive at a good solution as the number of possible dutupartitionsfor a given real lifeprogram is exponential in the size of theprogram. Wepresent a heuristic techniquefor automatic data partitioningfor HPF. Our approach is based on Genetic Algorithms and is very simple, yet very eficient to quicklyfind appropriate data partitions evenfor large programs with large number of altemativesfor data distribution.It makes use of both static as well as dynamic data distribution with the main aim of reducing the overall execution timeof the entire program. Thepeqonnance of a parallel program depends largely on its data partitions. So a ProblemDescription In automatic data partitioning, the compiler is responsibIefor deciding the partitioning of data. GeneticAlgorithms at differentlevels The chromosome encoding of the individuals at each level varies. The inter-procedureCA starts with the startuphandom distrib...

• Type: Conference
• Year: 2005
• Where: HPDC
• Authors: Sunil Kumar Anand, Y. N. Srikant
• Keywords: Automatic Data, Data Partitioning Scheme, Data Partitions, Distributed And Parallel Computing, HPDC 2005