Project summary

The objectives of this research include:

• Identify issues in optimal integration of interpretation and compilation techniques for parallel language systems.
• Study the impact of the issues on performance for scientific and engineering applications.
• Develop a set of technologies that aim at a high performance language system with interpreted (Java) front end supported by compiled parallel runtime for solving scientific and engineering problems.

To achieve the above objectives, the methods employed will be of experiemental nature. In particular, an environment will be set up from the off-the-shelf components, which allows experiementation on the interplay and trade-off of various techniques such as JIT compiling, dynamic linking, remote loading, garbage collection, stack handling, communication latency hiding, and runtime support, etc. Quantitative analysis will be conducted whenever possible, in addition to qualitative conclusion.

The significance of the proposed activity may be observed in the following facets.

• While the effectiveness of interpreted languages on text processing tasks (or in general, O(n) algorithms) is well accepted, how it can best help (or can not help at all) for scientific and engineering applications, in the context of contemporary computer technologies, is not well understood.

  While the higher performance of compiled programs over interpreted counterparts is well known, the inconvenience (and inefficiency !) of

   edit-compile-link-run-crash-start-it-all-over-again

  process is also commonly recognized.

  This research would be one of the kind that addresses the potentials of bringing the benifit of both technologies together in a contemporary setting.

• While compilation and interpretation technologies are actually merging, such as JIT compiler and graphics runtime support in Java system, but it's in an ad hoc way.

  This study would try to bring up results that describe the perspective of the merging process, which hopefully benifit the merging process in return.

• While issues in parallel processing with compiled code have been studied extensively, not much are known in how large scale parallel processing may benifit from an interpreted environment, given its other advantages.

  This study would try to generate insight in this direction in the context of distributed system simulations.

Table of Contents

Section Number of Pages

A Project Summary 1

B Table of Contents 1

C Project Description 15

D References 3

E Biographical Sketches 7

F Summary Proposal Budget 11

G Current and Pending Support 10

H Facilities, Equipment and Other Resources 1

I Special Information/Supplementary Documentation 0

J Appendix 0