BEGIN:VCALENDAR
PRODID:-//Microsoft Corporation//Outlook MIMEDIR//EN
VERSION:1.0
BEGIN:VEVENT
DTSTART:20151117T170000Z
DTEND:20151117T173000Z
LOCATION:18AB
DESCRIPTION;ENCODING=QUOTED-PRINTABLE:ABSTRACT: The configuration-interaction (CI) method, long a popular approach =0Ato describe quantum many-body systems, is often cast as a very large sparse matrix eigenpair problem with matrices whose dimension can exceed one billion.=0ASuch formulations place high demands on memory capacity and memory bandwidth.=0AIn this paper, we describe an efficient, scalable implementation, BIGSTICK, which, by factorizing both the basis and the interaction into two levels, can reconstruct the =0Anonzero matrix elements on the fly and reduce the memory requirements by one =0Aor two orders of magnitude allowing researchers to trade reduced resources for increased computational time.=0AWe optimize BIGSTICK on two leading HPC platforms.=0ASpecifically, we not only develop an empirically-driven load balancing strategy that =0Acan evenly distribute the matrix-vector multiplication across 256K threads, we also=0Adeveloped techniques that improve the performance of the  =0ALanczos reorthogonalization.=0ACombined, these optimizations improved performance by 1.3-8x depending on platform and configuration.
SUMMARY:Parallel Implementation and Performance Optimization of the Configuration-Interaction Method
PRIORITY:3
END:VEVENT
END:VCALENDAR