Parallel molecular dynamics using OPENMP on a shared memory machine (Q1971552)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Parallel molecular dynamics using OPENMP on a shared memory machine |
scientific article |
Statements
Parallel molecular dynamics using OPENMP on a shared memory machine (English)
0 references
11 November 2001
0 references
The parallelization of a program is its presentation in such way which allows to combine in time the execution of branches (sections) of the program. The main result of the paper is the description of experiences of authors in parallelizing molecular dynamics Fortran (mainly Fortran77) programs on shared memory machines by combining standard do-loop splitting approaches with OpenMP [OpenMP: A proposed standard API for shared memory programming, see \url{http://www.openmp.org} (1997)]. From this point of view, three ingredients of programs are investigated: (i) loops; (ii) subroutines that are utilized to evaluate energies and forces i.e., Cosmos, specific sections handle separately solute-solute, water-water, chain-chain, solute-water, solute-chain and chain-water; (iii) subroutines which correspond to operations performed on arrays; The software package Cosmos for simulation of molecular systems are implemented by B. Owenson, M.A. Wilson, A. Pohorille [NASA - Ames Res. Center, Moffertt Field, CA 94035-10000 (1987)]. In the paper under review the authors tested their approach on three benchmarks. The first molecular system consisted of the 58-residue BPTI in an aqueous solution. The system involved 6962 molecules of water, making a total number of atoms equal to 21778. The second system is a bilayer formed by 110 molecules of 2,4,6-trichlorophenol in contact with 2294 water molecules. The third system consisted of a fully hydrated, fluid membrane of DMPC. The bilayer contained 64 phospholipid units, split into 32 molecules per side, in contact with two water lamellae. Each lamella, formed by 1935 water molecules, was in equilibrium with its respective vapor phase. Results for the benchmark applications on a SILICON GRAPHICS ORIGIN 2000 (195 MHz R10000) using 1 to 16 processors (Table 1) are presented.
0 references
molecular dynamics computations
0 references
parallelization
0 references
shared memory machine
0 references