ITR/NGS: A Software Infrastructure for Computational Biology and Physics

Texas A&M Engineering Experiment Station, College Station TX

Investigators

Abstract

STAPL (the Standard Template Adaptive Parallel Library) is a library of ISO Standard C++ components with interfaces similar to the (sequential) ISO C++ standard library, and encouraging results have been obtained with a shared memory prototype. This project will extend this library significantly in the range of systems it can support (notably systems without shared memory), the range of optimization performed, and the degree to which a program is automaticall y adapted to its run-time environment and workload. The project will extend the STAPL library, redesign the STAPL run-time system, and build a compiler to allow high-level (source-level) optimizations and to generate information and code needed by the run-time system for adaptive algorithms and lower-level optimizations. Thus, STAPL will be a framework for highly-optimizable, adaptable, and portable parallel computations. In addition to the usual vector-based operations, it will support parallel use of less regular structures (such as lists and sparse matrices) and ease the evolution of programs from a serial to a parallel environment. Importantly, this will be done in ways that take advantage of the huge investment in mainstream programming environments and tools and make STAPL's parallel computation facilities accessible to mainstream programmers.