HPF parallelization of a molecular dynamics code: Strategies and performances

B. Di Martino, M. Celino, V. Rosato

Research output: Contribution to conferencePaper

Abstract

The High Performance Fortran (HPF) environment has been used to efficiently parallelize a Molecular Dynamics (MD) code. The MD code is a Tight-Binding code, properly specialized for semiconductor materials and characterized by inhomogeneous data distribution. Furthermore, the electronic properties are taken into account during the atomic dynamics: for this reason a large sparse matrix is built and diagonalized to compute the whole body of its eigenvalues and eigenvectors at each time step. For the diagonalization task, a parallel mathematical routine implemented. The strategy of parallelization, the integration of the mathematical routine in the HPF code are described and discussed. All the benchmark are performed on IBM SP architecture.
Original languageEnglish
Publication statusPublished - 1999
Event7th International Conference on High-Performance Computing and Networking, HPCN Europe 1999 - Amsterdam, Netherlands
Duration: 1 Jan 1999 → …

Conference

Conference7th International Conference on High-Performance Computing and Networking, HPCN Europe 1999
CountryNetherlands
CityAmsterdam
Period1/1/99 → …

    Fingerprint

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Di Martino, B., Celino, M., & Rosato, V. (1999). HPF parallelization of a molecular dynamics code: Strategies and performances. Paper presented at 7th International Conference on High-Performance Computing and Networking, HPCN Europe 1999, Amsterdam, Netherlands.