Programs in Physics & Physical Chemistry
|[Licence| Download | E-mail| New Version Template] acbg_v1_0.gz(61 Kbytes)|
|Manuscript Title: Pair potential calculation of molecular associations: a vectorized version.|
|Authors: F. Torrens, E. Orti, J. Sanchez-Marin|
|Program title: AMYRVF|
|Catalogue identifier: ACBG_v1_0|
Distribution format: gz
|Journal reference: Comput. Phys. Commun. 66(1991)341|
|Programming language: Fortran.|
|Computer: IBM 3090 150E VF.|
|Operating system: VM/XA SP(CMS), VM/SP HPO(CMS).|
|RAM: 1600K words|
|Keywords: Molecular associations, Pair potential, 1/r expansion, Geometry optimization, Vectorization, Biology, Structure.|
|Classification: 3, 16.1.|
Nature of problem:
This program determines the optimum separation and relative orientation of two interacting molecular systems through a minimization of the interaction energy.
The interaction energy is evaluated by means of a 1/R expansion (with 1/R, 1/R**4, 1/R**6, and 1/R**12 terms), parameterized on the basis of accurate SCF results. The search for the minimum energy can be carried out by one of the following options: (1) simplex; (2) Steepest-Descent (3) Davidon; (4) Broyden-Flecther-Goldfarb-Shanno optimization method.
Calculations should be restricted to molecules consisting of atoms belonging to the 82 classes for which expansion coefficients are available. The initial position of the second system with respect to the first one is defined by means of a six-fold vector (three trans- lations and three rotations) about the Cartesian axes.
Because the number of atom-atom intermolecular interactions is formally proportional to the square of the number of atoms in the molecules, a great incentive exists for incorporating vectorization into the AMYR algorithm. We have rewritten the program code for an IBM 3090 Vector Facility with an exhaustive use of IBM ESSL library calls. The new program vectorizes 97 percent of the computation. A vector-to-scalar speedup of 6.87 in the vectorizable part of the code gives an overall speedup of 5.83. The size of the vectorized version of the program is 33% larger than the scalar one.
CPU time is 196 s for the test run.
|Disclaimer | ScienceDirect | CPC Journal | CPC | QUB|