CyberGIS Toolkit Home Page


PGAP is a PGA application that uses MPI non-blocking functions for migrating solutions among PGA demes. The implementation of PGA is based on the island model. Although this solver is written for solving the Generalized Assignment Problem (GAP), it can be easily modified to solve other combinatorial optimization problems. The scalability of this application has been tested on BlueWaters supercomputer at the National Center for Supercomputing Applications (NCSA), Stampede supercomputer at the Texas Advanced Computing Center (TACC), and Trestles supercomputer at the San Diego Supercomputer Center (SDSC). This code can scale to 262K processor cores on BlueWaters with marginal communcation cost.



If you publish your work in which PGAP is used, please cite the following publication(s):

  • Liu, Y.Y. and Wang, S. 2014. “A Scalable Parallel Genetic Algorithm for the Generalized Assignment Problem.” Parallel Computing,
  • Liu, Y. Y., Guo M., Wang, S. 2013. “Large-scale Land Use Optimization by Enhancing a Scalable Parallel Genetic Algorithm Library.” In: Proceedings of XSEDE 2013: Extreme Science and Engineering Discovery Environment: Gateway to Discovery, Jul 22-25 2013, San Diego, CA, USA.


PGAP depends on MPI and SPRNG to compile.

make clean;make

NMI Build and Test Status