Performance Analysis of Matrix Multiplication Algorithms Using MPI and OpenMP

Tigran M. Galstyan

doi:10.51408/1963-0029

Authors

Tigran M. Galstyan Institute for Informatics and Automation Problems of NAS RA

DOI:

https://doi.org/10.51408/1963-0029

Keywords:

Hybrid, OpenMP, MPI, Matrix Multiplication, Fox

Abstract

The combination of OpenMP and MPI in programming is called hybrid programming. Hybrid programming (through messages and shared memory) has gained an important role since the appearance of cluster architectures. A hybrid programming method combines the MPI and OpenMP libraries to use this hierarchical multi-core architecture. The purpose of this work is to carry out the performance analysis of matrix multiplication algorithms in a cluster system. Each node in the cluster consists of multiple core CPUs, in which memory is distributed among the nodes and shared memory. Algorithms use MPI as a message-passing mechanism and OpenMP as shared memory.

References

J. Ali and R. Zaman, Performance Analysis of Matrix Multiplication Algorithms Using MPI: Khan Department of Computer Science, Aligarh Muslim University, Aligarh. Parallel Programming in C with MPI and OpenMP – by Michael J. Quinn.

Parallel Programming in C with MPI and OpenMP – by Michael J. Quinn.

[Online]. Available: https://www.open-mpi.org/

[Online]. Available: https://www.mpich.org/

[Online]. Available: https://www.openmp.org/

[Online]. Available: https://en.wikipedia.org/wiki/Message_P assing_Interface

[Online]. Available: https://en.wikipedia.org/wiki/OpenMP

Ned Nedialkov, Communicators and Topologies: Matrix Multiplication Example, McMaster University Canada CS/SE 4F03 March 2016.

[Online]. Available: http://pages.tacc.utexas.edu/~eijkhout/pcse/html/mpi-topo.