Contributions

The primary contributions of this dissertation are to propose the potential of Java as a scientific parallel programming language and to pursue efficient compilation of the HPJava language for high-performance computing. By experimenting with and benchmarking large scientific and engineering HPJava programs on a Linux machine, a shared memory machine, and a distributed machine, we have proved that the HPJava compilation and optimization scheme generates efficient node code for parallel computing.

• The HPJava design team, composed of Bryan Carpenter, Geoffrey Fox, Guansong Zhang, and Xiaoming Li, designed and proposed the HPspmd programming Language Model. The major goal of the system we are building is to provide a programming model that is a flexible hybrid of HPF-like data-parallel language and the popular, library-oriented, SPMD style without the basic assumptions of the HPF model. When the project was proposed, the underlying run-time communication library was available [11], the translator itself was being implemented in a compiler construction framework developed in the PCRC project [13,51]. Soon afterwards, JavaCC and JTB tools were adopted for the HPJava parser.
• The author, Han-Ku Lee, joined the HPJava development team in 1998. Since then, he has developed the HPJava front-end and back-end environments and implementations including the HPJava Translator, the Type-Analyzer for analyzing source classes' hierarchy, and the Type-Checker for safe type-checking as well as implemented the JNI interfaces of the run-time communication library, Adlib. In addition, he has maintained the HPJava Parser, implemented the Abstract Expression Node generator for intermediate codes, the Unparser to unparse translated codes. At the School of Computational Science and Information Technology (CSIT) at Florida State University (FSU), Tallahassee, FL, the most time-consuming parts such as type-checker has been firmly implemented, and the first fully functional translator were developed.
• Currently, the author and the HPJava development team keep researching and improving the HPJava system. The main concentration of the HPJava system is now to benchmark the current HPJava system on different machines such as Linux, SMPs, and IMB SP3. Moreover, the HPJava development team is concentrating on optimization strategies for the translator and pure-Java implementation for the run-time communication libraries to make the system competitive with existing Fortran programming environments.
• The author has significantly contributed to the HPJava system in developing and maintaining the HPJava front-end and back-end environments at NPAC, CSIT, and Pervasive Technology Labs. He has developed the Translator, the Type-Checker, and the Type-Analyzer of HPJava, and has researched the HPJava optimization scheme. Some of his early work at NPAC are the Unparser and the Abstract Expression Node generator, and original implementation of the JNI interfaces of the run-time communication library, Adlib. Recently he focused on the topic ``Efficient Compilation of the HPJava Language for Parallel Computing'' for optimization of the HPJava system.