MDBNCH

Abstract

A single run of MDBNCH performs seven independent molecular dynamics calculations, using systems of different sizes and/or using different codes. The purpose of doing this is to sample a variety of different running conditions. Each run is separately timed, and the total time is reported at the end. This last number is what is reported in the table of results. However, the partial times may also be useful for particular purposes.

DateOfInformation

Wed Oct 29 11:46:31 1997

Domain

Benchmark and Example Programs!Miscellaneous

Name

MDBNCH

Restrictions

MDBNCH was originally developed around 1985 as a "real life" application to be used as benchmark for a machine purchase. That version was not suited for a general distribution, and I replaced it with a brand new code incorporating new algorithms in december 1988. Except for a few cosmetic modifications with no effect on timings, MDBNCH did not change since then, and it will never be changed. All the data available in the current table of results have been obtained with this same version, and can be directly compared with each other. MDBNCH is written in 100% ANSI-standard Fortran 77 to ensure maximum portability. This is of course why its name is not, say, MDBench. It consists of about 2100 lines. It was derived from a production code named MASTER used to study metals with many-body `glue' potentials. This code was developed by starting from a program constructed by Michele Parrinello and Aneesur Rahman at Argonne National Laboratory around 1980, featuring a cell with variable volume and shape. This program was in turn evolved from previous Rahman's codes, whose development was started in the 60s.

TargetEnvironment

Fortran 77

Webpage

http://www.sissa.it/~furio/mdbnch.html

ContactIs

Furio Ercolessi