TDDFT in massively parallel computer architectures: the octopus project
Xavier Andrade, Joseba Alberdi-Rodríguez, David A. Strubbe, Micael J. T. Oliveira, Fernando Nogueira, Alberto Castro, Javier Muguerza, Agustin Arruabarrena, Steven G. Louie, Alán Aspuru-Guzik, Angel Rubio, and Miguel A. L. Marque. TDDFT in massively parallel computer architectures: the octopus project. Psi-k Newsletter, April 2012. 2012, Vol. 110, p. 60-2012.
OCTOPUS is a general-purpose density-functional theory (DFT) code, with a particular emphasis on the time-dependent version of DFT (TDDFT). In this article we present the ongoing efforts for the parallelisation of {\sc octopus}. We focus on the real-time variant of TDDFT, where the time-dependent Kohn-Sham equations are directlypropagated in time. This approach has a great potential for execution in massively parallel systems such as modern supercomputers with thousands of processors and graphics processing units (GPUs). For harvesting the potential of conventional supercomputers, the main strategy is a multi-level parallelisation scheme that combines the inherent scalability of real-time TDDFT with a real-space grid domain-partitioning approach. A scalable Poisson solver is critical for the efficiency of this scheme. For GPUs, we show how using blocks of Kohn-Sham states provides the required level of data-parallelism and that this strategy is also applicable for code-optimisation on standard processors. Our results show that real-time TDDFT, as implemented in {\sc octopus}, can be the method of choice to study the excited states of large molecular systems in modern parallel architectures.
OCTOPUS is a general-purpose density-functional theory (DFT) code, with a particular emphasis on the time-dependent version of DFT (TDDFT). In this article we present the ongoing efforts for the parallelisation of {\sc octopus}. We focus on the real-time variant of TDDFT, where the time-dependent Kohn-Sham equations are directlypropagated in time. This approach has a great potential for execution in massively parallel systems such as modern supercomputers with thousands of processors and graphics processing units (GPUs). For harvesting the potential of conventional supercomputers, the main strategy is a multi-level parallelisation scheme that combines the inherent scalability of real-time TDDFT with a real-space grid domain-partitioning approach. A scalable Poisson solver is critical for the efficiency of this scheme. For GPUs, we show how using blocks of Kohn-Sham states provides the required level of data-parallelism and that this strategy is also applicable for code-optimisation on standard processors. Our results show that real-time TDDFT, as implemented in {\sc octopus}, can be the method of choice to study the excited states of large molecular systems in modern parallel architectures.