SC15 Austin, TX

Task-Based Parallel Computation of the Density Matrix in Quantum-Based Molecular Dynamics Using Graph Partitioning


Authors: Sergio Pino (Los Alamos National Laboratory), Matthew Kroonblawd (Los Alamos National Laboratory), Purnima Ghale (Los Alamos National Laboratory), Georg Hahn (Los Alamos National Laboratory), Vivek Sardeshmukh (Los Alamos National Laboratory), Guangjie Shi (Los Alamos National Laboratory), Hristo Djidjev (Los Alamos National Laboratory), Christian Negre (Los Alamos National Laboratory), Robert Pavel (Los Alamos National Laboratory), Benjamin Bergen (Los Alamos National Laboratory), Susan Mniszewski (Los Alamos National Laboratory), Christoph Junghans (Los Alamos National Laboratory)

Abstract: Quantum molecular dynamics(QMD) simulations are highly accurate, but they are computationally expensive due to the calculation of the ground-state electronic density matrix P via an O(N^3) diagonalization. Second-order spectral projection(SP2) is an efficient O(N) alternative to obtain P from a Hamiltonian matrix H. This poster presents a data-parallel version of the SP2 algorithm that uses an undirected graph representation of the matrix P to divide the computation into smaller independent partitions. These partitions can give rise to undesirable load imbalances in standard MPI/OpenMP-based implementations, as they are often of unequal sizes. The load-balancing problem is addressed by using task-based programming models to schedule parallel computations during runtime. We present CnC and Charm++ implementations that can be integrated into existing QMD codes. Our approach is applied to QMD simulations of representative biological protein systems with more than 10,000 atoms, exceeding size limitations of diagonalization by more than an order of magnitude.

Poster: pdf
Two-page extended abstract: pdf


Poster Index