Languages and Compilers for Scientific Computation

High performance computing (HPC) represents the leading edge of computing technology. HPC architectures are characterized by multi-level parallelism, deep memory hierarchies and complex organizational structures, covering a broad spectrum that ranges from high performance workstations to parallel architectures and all the way to heterogeneous systems and the Grid. HPC systems have been applied to many problems in science and engineering; in particular, their use for the simulation and optimization of complex processes occurring in nature, industry, or scientific experiments, has led to important new developments and breakthroughs. However, the development of software for HPC systems has not kept pace with the evolution of hardware systems. This has been the starting point of the research agenda for the AURORA project since its beginning in 1997. AURORA is an interdisciplinary project based upon a highly synergetic cooperation of subprojects ranging across a number of different disciplines. The explicit common goal that unites these projects is to push the state-of-the-art in the field of high-level software for HPC systems. The related research issues range from models, applications, and algorithms to languages, compilers, and programming environments. This research is part of a world-wide effort aiming at making present-day computing systems easier to use; it thus contributes to the growing importance of a global software industry dealing with these issues. As a consequence, the medium and long-term results of the work in AURORA will play an important role in improving the competitivity of the Austrian software industry. Key results of recent research in AURORA include a new release, WIEN2k, of a quantum-mechanical package that is used by more than 500 institutions around the world, the development of new parallel algorithms for large-scale linear programming, new algorithms for eigenvalue problems and FFT, as well as advanced language and compiler work in the context of the Vienna Fortran Compiler and major new work in performance analysis and prediction. AURORA consists of the following 6 Projects: Coordination Project Leader: Hans P. Zima Languages, Compilers, and Runtime Systems for Scientific Computing Leaders: Siegfried Benkner and Hans P. Zima Tools Leader: Thomas Fahringer Numerical Algorithms and Software Leader: Christoph Überhuber High Performance Computing in Finance Leader: Georg Pflug Parallelization of the Program Package WIEN97 Used for Quantum Mechanical Calculations of Solids Leader: Karlheinz Schwarz