drahoslavzan / mkp-pso

A Distributed Multidimensional Knapsack Problem Solver using Island Based Particle Swarm Optimization

A Distributed Multidimensional Knapsack Problem Solver using Island Based Particle Swarm Optimization

This package contains an efficient CPU and GPU implementations of the platform for solving various problems using the Particle Swarm Optimization (PSO) technique. The Multidimensional Knapsack Problem (MKP) was chosen as a benchmark for demonstration purposes, but it is one of the most widely used combinatorial optimization problems. This solver can also be used in parallel environment such as HPC cluster, where one or more multi-core CPUs or GPU are allocated for each swarm (process). Running in a such configuration can lead to a better solution and/or reduced time spent in searching for a solution.

External libraries:

• OpenMPI
• PAPI

Build:

$ cd cpu  
$ mkdir build  
$ cd build  
$ cmake ..  
$ make

Example (4 swarms):

$ ./mkp --help  
$ mpirun -np 4 ./mkp --exch 1 --stat 0 --stop 300 --epochs 3000 --particles 128 --vel "5 10 15 20" --verbose ../../data/weing8.dat