Characterizing realistic high-performance computer applications with existing, advanced tools will be a first thrust in this project. An information repository will be created on the Web, documenting properties of application programs, computer architectures, and their interrelations. A second project part will improve the integration of tools and their user interaction. Compilers will be enhanced to make available to the users the information gathered from their source programs. Performance analysis tools will be integrated with compilation tools and developed to the point where they can interpret performance and present it to the user at several abstraction levels. The new tools will be made available on the Web, serving as a "Remote Parallel Programming Lab". Third, advanced capabilities of optimization and performance analysis tools will be developed. Compilers will be enhanced to increase their understanding of subscripted subscript patterns, found in many irregular applications. Performance analysis tools will be enhanced to do "performance forecasting", that is, predict the performance of future computer architectures running future applications. New course material will be developed based on the results of these projects. This material will be brought into the existing course framework of parallel programming, advanced optimizers, and object- oriented programming at Purdue/ECE.***
