Shared‐memory parallelization of the TURBOMOLE programs AOFORCE, ESCF, and EGRAD: How to quickly parallelize legacy code

Abstract

Abstract:The programs ESCF, EGRAD, and AOFORCE are parts of the TURBOMOLE program package and compute excited-state properties and ground-state geometric hessians, respectively, for Hartree-Fock and density functional methods. The range of applicability of these programs has been extended by allowing them to use all CPU cores on a given node in parallel. The parallelization strategy is not new and duplicates what is standard today in the calculation of ground-state energies and gradients. The focus is on how this can be achieved without needing extensive modifications of the existing serial code. The key ingredient is to fork off worker processes with separated address spaces as they are needed. Test calculations on a molecule with about 80 atoms and 1000 basis functions show good parallel speedup up to 32 CPU cores.

Contrasting the optical properties of the different isomers of oligophenylene

“…37,38 The CAM-B3LYP calculations used NWChem 6.0 39 except in the case of the TD-DFT S1 relaxations, which were performed using GAMESS-US 40 (version 1 October 2010 R1). 37,38 The CAM-B3LYP calculations used NWChem 6.0 39 except in the case of the TD-DFT S1 relaxations, which were performed using GAMESS-US 40 (version 1 October 2010 R1).…”

Deuterium-labelled N-acyl-l-homoserine lactones (AHLs)—inter-kingdom signalling molecules—synthesis, structural studies, and interactions with model lipid membranes

“…SCF convergence criteria were tightened to 10 −8 and the DFT integration grid was of m 4 size. Subsequent frequencies and polarizability derivatives calculations were carried out with the recently parallelised [43] aoforce and egrad modules of the TURBOMOLE package, respectively. The obtained frequencies were scaled down by a factor of 0.965 [44] and the spectrum was plotted with Lorentzian line shape as implemented in Molden [45].…”

Efficient simulation of large materials clusters using the jaguar quantum chemistry program: Parallelization and wavefunction initialization

“…The critical section (line [8][9][10][11][12][13][14] is responsible for serializing the caching procedure of temporary data on disk in order, as those file I/O operations are implemented in sequential manner, as opposed to random assessable manner. As the worst case, this critical section in JK routine, which fetches the PS grid points as well as the strips of Q matrix, only consumes 2 to 4% of the calculation time for one loop iteration.…”

“…An alternative strategy [8][9][10][11] from an implementation perspective is to design and to implement a hardware-efficient parallel implementation for DFT algorithm. The major considerations supporting this approach are (1) the semiconductor industry has been shifted from frequency-driving paradigm to the multicore paradigm for many years and parallel implementation becomes crucial for scalability on large machines; (2) the continuous introduction of new instruction-level parallelism (i.e., SSE, AVX, etc.)…”