In recent years, high performance computer system have been rapidly moving from homogeneous CPU systems to heterogeneous CPU/GPGPU or CPU/Intel MIC systems. However, the majority of quantum chemistry codes still remain oriented towards homogeneous CPU computer systems, owing to legacy code base and programming models.This work is focused on the development of a novel quantum chemistry code, GPUnCH, which aims at utilizing advanced features of modern CPUs, GPGPUs and Intel MIC co-CPUs. The GPUnCH program implements the Kohn-Sham self-consistent field energy, and is currently being extended to allow for an efficient structure optimization and transition states search with various exchange-correlation functionals, including the popular B3LYP and M06 functionals.In this work we discuss the performance of the GPUnCH code on various homogeneous and heterogeneous computer systems, which contain the recent Intel Xeon CPUs or combination of Intel Xeon CPUs with GPGPUs (NVIDIA or AMD) or with Intel MIC co-CPUs. Our benchmarking set includes various size molecules, and is designed to test the performance of the GPUnCH code for typical computational scenarios encountered in quantum chemical investigations at different levels of basis sets, ranging from def2-SVP to def2-QZVPP.Apart from benchmarking results, we also discuss the implementation and design choices of the two-electron integral evaluation module and the exchange-correlation integrator.

Quantum chemistry on a heterogeneous computer system: Accelerating the Kohn-Sham method for hybrid CPU/GPGPU and CPU/Intel MIC platforms

248th National Meeting (2014)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.