Tremendous advancement in microprocessor performance and networking technology in past decades has produced the following senario: a personal computer today delivers higher performance than a mainframe could do 20 years ago; a vast sum of accessible computing power exists world wide.
Many consequences result from this senario. Although the debate over compiler vs interpreter was kind of ended in favor of compilation about two decades ago, the fact that an interpreted program runs fast today than a compiled program ran 20 years ago suggests that interpretation technology may be revisited. Besides the popularity of Perl, Tcl, and Python, etc., this evidence is best manifested by the overwhelming success of Java programming language, which is largely due to its underline interpreted nature.
Nevertheless, it remains a fact that interpretation is slower than compilation-execution model within the same hardware technology. Thus, a trend is to merge (integrate) the two technologies, as examplified by JIT compiler in Java virtual machine, built-in computational functions in MATLAB, and as well as an earlier instance -- interpreted front-end supported by data parallel runtime in *LISP on CM2.
However, not much research is seen explicitly towards optimal integration of the two technologies. In particular, while the usefulness of Java for web application has been proved, how it can best help for scientific and engineering applications is not well understood.
To boost the performance of Java, different approaches have been taken, including Java chip, JIT compilation, and creation of connection between Java and message passing system such as PVM, etc.
It's observed that there is a lack of studies on interpreted front-end supported by parallel/distributed runtime system for solving scientific and engineering problems. It is our belief that Java's attractions to scientific and engineering community do not only lie in cool graphics, platform independence, etc., but also include acceptable performance in problem solving, once various technologies integrated properly.
We therefore propose the project: Studies of Integration and Optimization of Interpreted and Compiled Languages -- aiming at interpreted (Java) front-end to high performance runtime in spirit of MATLAB/APL/HPF Interpreter we once built.