Per Aspera ad Astra: On the Way to Parallel Processing

Computational Science and Engineering is being established as a third category of scientific methodology; this innovative discipline supports and supplements the traditional categories: theory and experiment, in order to solve the problems arising from complex systems challenging science and technology. While the successes of the past two decades in scientific computing have been achieved essentially by the technical breakthrough of the vector-supercomputers, today the discussion about the future of supercomputing is focussed on massively parallel computers. The discrepancy, however, between peak performance and sustained performance achievable with algorithmic kernels, software packages, and real applications is still disappointingly high. An important issue are programming models. While Message Passing on parallel computers with distributed memory is the only efficient programming paradigm available today, from a user's point of view it is hard to imagine that this programming model, rather than Shared Virtual Memory, will be capable to serve as the central basis in order to bring computing on massively parallel systems from a sheer computer science trend to the technological breakthrough needed to deal with the large applications of the future; this is especially true for commercial applications where explicit programming the data communication via Message Passing may turn out to be a huge software-technological barrier which nobody might be willing to surmount.KFA Jülich is one of the largest big-science research centres in Europe; its scientific and engineering activities are ranging from fundamental research to applied science and technology. KFA's Central Institute for Applied Mathematics (ZAM) is running the large-scale computing facilities and network systems at KFA and is providing communication services, general-purpose and supercomputer capacity also to the HLRZ ("Höchstleistungsrechenzentrum") established in 1987 in order to further enhance and promote computational science in Germany. Thus, at KFA - and in particular enforced by ZAM - supercomputing has received high priority since more than ten years. What particle accelerators mean to experimental physics, supercomputers mean to Computational Science and Engineering: Supercomputers are the accelerators of theory