Parallel computation of second derivatives of RHF energy on distributed memory computers

Abstract

A parallel implementation of the computation of RHF energy second derivatives with respect to the nuclear coordinates is described. The algorithm and organization of the code are described in detail on the most computationally demanding steps with special emphasis on the integral transformation code. Key features of the proposed algorithm are its large degree of concurrency, limited interprocessor communication, and critical memory needs distributed over the processors. The cpu times and computer and network resources used are reported and discussed for a few examples. © 1997 by John Wiley & Sons, Inc.