Parallel Profile - Charles Koelbel

Research Scientist, Computer Science Department, Rice University

Charles "Chuck" Koelbel's research interests center on mapping algorithms and programs onto distributed memory multiprocessors. Much of this work is directed at giving users a more high-level, portable programming abstraction than first-generation parallel languages provided. "We're always trying to come up with better abstractions that make it easier for humans to program computers," he says. "I'm particularly proud of bringing the techniques for regular and irregular compilation together in my thesis, and with how far this whole field has come in the last 10 years."

Koelbel received his B.A. in mathematics and computer science from Augustana College (1983) and his M.S. and Ph.D. in computer science from Purdue University (1985 and 1990). His doctoral dissertation presented a compiler that translated Kali, a shared-memory imperative language, for execution on the Intel iPSC/2.

In 1990, Koelbel joined the Computer Science Department at Rice University as a CRPC researcher and began working on the FORTRAN D compiler project with CRPC Director Ken Kennedy and other researchers and students. This project has involved designing and implementing a set of FORTRAN extensions for machine-independent parallel programming, and was influenced by the features and compilation strategies of Koelbel's work with Kali. This work led to his involvement in the D System Programming Tools project, which is focused on designing and implementing a set of tools for programming in data-parallel languages. It was also the genesis of High Performance Fortran (HPF), an extension of Fortran that provides support for portable data-parallel programming. Koelbel and Kennedy have been leaders in the effort to standardize HPF through the HPF Users Group and the High Performance Fortran Forum (HPFF), a collaboration of academic, industrial, and government organizations who are working to define extensions to FORTRAN. Koelbel served as executive director of the HPFF from 1992 to 1993.

More recently, Koelbel has worked with the Scalable Input/Output Consortium, a large project led by Paul Messina to attack the parallel I/O problem on a wide front. Rice's role here was to investigate language, compiler, and runtime issues in parallel I/O. By applying techniques from data-parallel compilers, he, Kennedy, and Mike Paleczny were able to develop one of the first compilers for out-of-core computations.

Today, Koelbel is working on the Department of Defense (DoD) High Performance Computing Modernization Program (HPCMP) as part of a nationwide effort to substantially upgrade the computational capabilities of the defense community for research, development, testing, and evaluation. He and CRPC researchers from sites across the country are involved in the Programming Environment Training (PET), which focuses on emerging advances in programming environments, computational tools, algorithms, software, computational solution techniques, and training. He is also actively involved in the National Science Foundation's Partnership for Advanced Computational Infrastructure (PACI) headed by the National Computational Science Alliance (NCSA) at the University of Illinois and the National Partnership for Advanced Computational Infrastructure (NPACI), headquartered at the University of California, San Diego. (See "Research Focus: PACI Projects at the CRPC.") The mission of the Alliance, and the PACI program in general, is to provide high-performance computational services to the scientific community in the United States. Many CRPC researchers are involved in this project, bringing the fruits of their past research to bear on real-world scientific problems.

"We're going through a lot of changes right now, with the advent of distributed shared memory and fast networks," says Koelbel of his research areas and the High Performance Computing Community in general. "There's a lot of opportunity here, and a lot of new research to be done. We are finding that we have to work at a higher level just to keep track of everything going on. Old work could be lost, but later rediscovered, in the process."

Koelbel's professional activities include serving as a reviewer for IEEE Computer, IEEE Parallel and Distributed Computing, IEEE Software, Concurrency: Practice and Experience, and Journal of Parallel Programming. He has been as a committee member on numerous programs and workshops, including the Fourth Workshop on Languages, Compilers, and Run-time Systems for Scalable Computers; the Eighth Department of Defense High Performance Computing Users' Group Conference; and Supercomputing '98. He is co-author of the High Performance Fortran Handbook, and more than 30 papers and technical reports.

Other Issues of PCR Back to PCR CRPC Home Page