An implementation is presented of an uncontracted Rys quadrature algorithm for electron repulsion integrals, including up to g functions on graphical processing units (GPUs). The general GPU programming model, the challenges associated with implementing the Rys quadrature on these highly parallel emerging architectures, and a new approach to implementing the quadrature are outlined. The performance of the implementation is evaluated for single and double precision on two different types of GPU devices. The performance obtained is on par with the matrix−vector routine from the CUDA basic linear algebra subroutines (CUBLAS) library. Disciplines Chemistry | Computer Sciences CommentsReprinted (adapted) Abstract: An implementation is presented of an uncontracted Rys quadrature algorithm for electron repulsion integrals, including up to g functions on graphical processing units (GPUs).The general GPU programming model, the challenges associated with implementing the Rys quadrature on these highly parallel emerging architectures, and a new approach to implementing the quadrature are outlined. The performance of the implementation is evaluated for single and double precision on two different types of GPU devices. The performance obtained is on par with the matrix-vector routine from the CUDA basic linear algebra subroutines (CUBLAS) library.

In this article, a new multithreaded Hartree-Fock CPU/GPU method is presented which utilizes automatically generated code and modern C++ techniques to achieve a significant improvement in memory usage and computer time. In particular, the newly implemented Rys Quadrature and Fock Matrix algorithms, implemented as a stand-alone C++ library, with C and Fortran bindings, provides up to 40% improvement over the traditional Fortran Rys Quadrature. The C++ GPU HF code provides approximately a factor of 17.5 improvement over the corresponding C++ CPU code.

A new coupled cluster singles and doubles with triples correction, CCSD(T), algorithm is presented. The new algorithm is implemented in object oriented C++, has a low memory footprint, fast execution time, low I/O overhead, and a flexible storage backend with the ability to use either distributed memory or a file system for storage. The algorithm is demonstrated to work well on single workstations, a small cluster, and a high-end Cray computer. With the new implementation, a CCSD(T) calculation with several hundred basis functions and a few dozen occupied orbitals can run in under a day on a single workstation. The algorithm has also been implemented for graphical processing unit (GPU) architecture, giving a modest improvement. Benchmarks are provided for both CPU and GPU hardware. Disciplines Chemistry CommentsReprinted (adapted) ABSTRACT: A new coupled cluster singles and doubles with triples correction, CCSD(T), algorithm is presented. The new algorithm is implemented in object oriented C++, has a low memory footprint, fast execution time, low I/O overhead, and a flexible storage backend with the ability to use either distributed memory or a file system for storage. The algorithm is demonstrated to work well on single workstations, a small cluster, and a high-end Cray computer. With the new implementation, a CCSD(T) calculation with several hundred basis functions and a few dozen occupied orbitals can run in under a day on a single workstation. The algorithm has also been implemented for graphical processing unit (GPU) architecture, giving a modest improvement. Benchmarks are provided for both CPU and GPU hardware. INTRODUCTIONAs a rule of thumb, the electronic energy obtained with the Hartree−Fock method accounts for ∼99% of the energy. However, many chemical properties of interest are dependent on the remaining 1%, frequently called the electron correlation energy, or simply the correlation energy. The correlation energy is defined as the difference between the reference Hartree−Fock energy and the true energy,

The Massively Parallel Quantum Chemistry (MPQC) program is a 30-year-old project that enables facile development of electronic structure methods for molecules for efficient deployment to massively parallel computing architectures. Here, we describe the historical evolution of MPQC’s design into its latest (fourth) version, the capabilities and modular architecture of today’s MPQC, and how MPQC facilitates rapid composition of new methods as well as its state-of-the-art performance on a variety of commodity and high-end distributed-memory computer platforms.