Although parallel computer systems have been commercially available for a decade, their impact outside the computer industry has fallen short of expectations. The two principal impediments to widespread use of parallel computers in industry are the difficulty in developing parallel applications and the ever-changing variety of incompatible architectures and programming models supported by vendors.
Under ARPA contract DABT63-92-C-0038 (and formerly under contract DABT63-91-C-0028) researchers at Rice University are investigating techniques to support machine-independent data-parallel programming. Data parallelism is the key to solving many important problems in science and engineering quickly. To simplify the task of parallel programming, we defined the Fortran D programming language and investigated compiler and runtime implementation techniques for it. Fortran D validated the data-parallel approach but also highlighted the difficulty of understanding program behavior in high-level languages. This led us to expand our research efforts to design and build prototype tools, collectively called the D System, to support effective data-parallel programming. A key goal of the D System is to retain the abstract, machine-independent view of the program, shielding the programmer from the complex, machine-dependent implementation details. The D System includes tools for static and dynamic performance feedback, automatic data layout selection, and an underlying repository for interprocedural analysis results.
During 1992 and early 1993, project researchers Ken Kennedy and Charles Koelbel led a coalition of academics, computer vendors, and industrial and government Fortran users in the definition of High Performance Fortran. The HPF language is based on many of the concepts developed as part of our ARPA-supported research on Fortran D. In 1994 and 1995, Kennedy continues to lead the Forum in the exploration of extensions that broaden the scope of HPF. A document describing the scope of activities and motivating applications, and the HPF Journal of Development are available.
Several members of the project team are taking leadership roles in the CRPC Women's Project, which aims to encourage more young women to pursue careers in mathematics and science. This program is a collaboration with the Houston Independent School District, the Rice University School Mathematics Project, and the American Association of University Women.