Eigensolver performance comparison on Cray XC systems

摘要

Hermitian (symmetric) eigenvalue solvers are the core constituents of electronic structure,quantum-chemistry, and other HPC applications such as Quantum ESPRESSO, VASP, CP2K,and NWChem to name a few. Our understanding of the performance of symmetric eigenvaluealgorithms on various hardware is clearly important to the quantum chemistry or condensed matterphysics community but in fact goes beyond that community. For instance, big data analyticsis increasingly utilizing eigenvalues solvers, in the study of randomized singular value decomposition(SVD) or principal component analysis (PCA). Noise, vibration, and harshness (NVH) isanother field where fast and efficient eigenvalue solvers are required. Most eigenvalue solverpackages feature numerous different parameters which can be tuned for performance, eg, thenumber of nodes, number of total ranks, the decomposition of the matrix, etc. In this paper, weinvestigate the performance of different packages as well as the influence of these knobs on thesolver performance.