DIMACS 1993-94 Special Year on Massively Parallel Computing

Special Year Organizers:

Organizing Committee:

Activities planned for the special year include a parallel implementation challenge, a lecture series, and several workshops. Also, the 6th Annual ACM Symposium on Parallel Algorithms and Architectures (SPAA) will be held in New Jersey (probably at Cape May) in late June or early August, 1994. Satish Rao (NEC) [satish@research.nj.nec.com] and Yuh-Dauh Lyuu (NEC) [lyuu@research.nj.nec.com] will make the local arrangements.

The goal of this Special Year is to have a positive impact on the design of the next generation of parallel computers and parallel algorithms, and thus to enhance this country's position as the world leader in supercomputing. An ancillary goal of the special year is to develop New Jersey's leadership role in the field of high performance computing (HPC). The sizes of both current machines and problems are unprecedented. Thus, designing the parallel computers and designing the solutions to these problems requires more than good engineering---it requires a scientific approach. One of the successes of the special year has been the ability of DIMACS to attract top researchers from other areas, including computer architects, application researchers, and commercial vendors.

Although it is too early to assess the impact of this Special Year, it is already possible to point out an important conceptual breakthrough that came in part from the DIMACS workshop Models, Architectures, and Technologies for Parallel Computation . This workshop attracted nearly 200 participants, including many students. Several participants remarked that this was the best workshop they had ever attended, noting the usefulness of extended communication and discussion specifically aimed at bridging the various areas represented.

A consensus was established that parallel computer architecture is converging in the short term on machines based on commodity microprocessors, but that in the next five years we would begin to see hardware support for shared memory in these machines.\footnote{ This convergence was discussed in greater detail at a workshop on "Suggesting Computer Science Agenda(s) for High-Performance Computing", which was held in Arlington, Virginia, and was sponsored by the NSF, the the University of Maryland Institute for Advanced Computer Studies (UMIACS), and DIMACS.} The consequences of convergence on a particular form of architecture are far reaching; it gives the architects a target to aim for, it gives the theoreticians a model in which to design algorithms, and it determines which Grand Challenge problems are likely to succumb to parallel computation in the near future.

IndexIndex of Special Year on Massively Parallel Computing
DIMACS Homepage
Contacting the Center
Document last modified on October 19, 1998.