The High Performance Fortran Applications project came about from a realization that although a huge amount of work had gone into the development of the HPF definition, and of various compilers for HPF and for practical subsets of it, less work had been done in writing and collecting worked example codes in HPF languages.
The main reason for this was a lack of full HPF compiler systems and the inherent difficulties in developing medium to large application codes without being able to validate correct compilation. At the beginning of the HPFA project, NPAC obtained early access to a number of proprietary compilation systems for Fortran 90 and HPF, as well as the subset languages available on several high performance computing systems.
The aim of the HPFA project was to write and collect worked example programs in HPF and other Fortran dialects. Example programs and other educational materials were also developed to help new HPF users. It was also expected that the example programs would help compiler builders and language designers. From the information gained in writing these application programs, more understanding can be gained about which applications are suitable for the language, and what language extensions may be required.
In addition to the example codes that were written for this project, some results were obtained from early compilers. Benchmarking performed in 1996 with the DEC, IBM, and PGI compilers was reported on in "Evaluation of High Performance Fortran Through Applications Kernels", Yau, H.W., Fox, G.C. and Hawick, K.A., in the proceedings of High-Performance Computing and Networking 1997, Vienna, Austria, April 1997. There is also a presentation from this conference.
In response to this paper, Jonathon Harris did more work on the benchmarking of these kernels in 1997/98, producing notes and updated kernels.
The HPFA project has been developed and maintained by a group of researchers at the Northeast Parallel Architectures Center (NPAC) at Syracuse University, under the auspices of the Center for Research in Parallel Computation (CRPC). The HPFA resources are also integrated with the National HPCC Software Exchange, another CRPC project. These resources will also help CRPC projects studying language extensions and run-time support for HPF and other parallel languages, and projects to develop tools and compilers for parallel languages, such as the D System project.
An overview of the technical motivation for this project is also available.