Premium
Efficient parallelization of the energy, surface, and derivative calculations for internal coordinate mechanics
Author(s) -
Totrov Maxim,
Abagyan Ruben
Publication year - 1994
Publication title -
journal of computational chemistry
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.907
H-Index - 188
eISSN - 1096-987X
pISSN - 0192-8651
DOI - 10.1002/jcc.540151006
Subject(s) - speedup , dihedral angle , computer science , internal energy , computational science , coordinate space , parallel computing , geometry , physics , mathematics , thermodynamics , quantum mechanics , hydrogen bond , molecule
An efficient algorithm for parallelization of a molecular mechanics program operating in the space of internal coordinates such as dihedral angles, bond angles, and bond lengths is described. The iterative procedure to calculate analytical energy derivatives with respect to the internal coordinates was modified to allow parallelization. Computationally intensive modules that calculate energy and its derivatives, solvent‐accessible surface, electrostatic polarization energy and that update lists of interactions were parallelized with nearly 100% efficiency. The proposed strategy for the shared‐memory computer architecture is easily scalable and requires minimum changes in a program code. The overall speedup for a realistic calculation minimizing the energy of a myoglobin reaches a factor of 3 for 4 processors. © 1994 by John Wiley & Sons, Inc.