Parallel stored‐integral and semidirect Hartree–Fock and DFT methods with data compression

Abstract

Recent developments in magnetic disk technology have made stored-integral techniques competitive with the currently more widely used direct methods, which involve the recalculation of the basic two-electron integrals. We present efficient conventional (all integrals stored) and semidirect Hartree–Fock and DFT algorithms with data compression for single-processor and distributed memory parallel computers, and compare them with the corresponding direct algorithms. On inexpensive modern personal computer-based hardware, the stored integral method is up to three times more efficient than the direct method in terms of total elapsed job time. © 2002 Wiley Periodicals, Inc. J Comput Chem 24: 154–160, 2003