CRPC ANNOUNCES UNIVERSITY OF MARYLAND AS FIRST AFFILIATED SITE

Ken Kennedy, Director for the Center for Research on Parallel Computation (CRPC), has announced the establishment of a CRPC- affiliated site at the University of Maryland-College Park. The site consists of faculty, graduate students, and postdoctoral researchers lead by Joel Saltz from the university's computer science department. "This affiliation with the University of Maryland marks an important step for the CRPC, which has continually extended its collaborative reach into the high-performance computing community since its creation in 1989," said Kennedy. "The center's inter-institutional nature has always been an important part of the its success because of all the resources that participating institutions bring into the picture," Kennedy added.

"The University of Maryland is delighted to be formally affiliated with the CRPC," said Saltz, who is director of the High Performance Systems Software Laboratory at the University of Maryland. "The CRPC has played a vital role in the high-performance computing community by pooling resources to solve a broad range of challenging real-world problems and by helping to educate students, scientists, and engineers."

In addition to joint work with researchers at the University of Maryland, significant collaborations have occurred between CRPC researchers and colleagues at the University of Illinois, NASA, Oak Ridge National Laboratory, the Courant Institute, Boeing Computer Services, and the University of California at Berkeley, among others. The CRPC plans to add additional affiliated sites in the near future.

The research group at the University of Maryland is developing methods that will make it possible to produce machine-independent compilers that generate efficient multiprocessor code for irregular scientific problems, such as problems that are unstructured, sparse, adaptive, or block structured.

"Affiliation with the CRPC is a very natural step for us due to our strong ties to the CRPC institutions and to our shared commitment to the CRPC's goals," said Saltz. The Maryland group is collaborating with most of the CRPC sites in areas of systems research that focus on the unique challenges posed by irregular problems. This research encompasses development of irregular problem compilation methods for both Fortran and C++ along with the development of compiler optimizations and runtime support needed to support irregular computations on terabyte datasets.

The Maryland group also works closely with applications scientists and engineers whose problem areas include computational fluid dynamics, computational chemistry, computational biology, structural mechanics, and electrical power grid calculations.

A key aspect of the research on irregular scientific problems is the development of portable runtime support libraries that coordinate interprocessor data movement, manage off-processor data, support a shared name space, and couple runtime data and workload partitioners to compilers. This runtime support is then used in distributed memory compilers by CRPC researchers at Rice University and Syracuse University.

The runtime support is also used to port applications codes to a variety of multiprocessor architectures. Additional research involves methods to reduce interprocessor communication costs and overhead costs associated with runtime preprocessing.

"The CRPC has made significant progress in enabling scientists and engineers to use parallel Fortran and C++ for 'regular' problems in computational science," said Kennedy. "I think our partnership with the University of Maryland and other institutions will grow in importance, especially as the CRPC moves more into research on irregular scientific problems."

Table of Contents