@TECHREPORT\{IMM2005-04450, author = "J. Dongarra and K. Madsen and J. Wasniewski", title = "PARA'04 Workshop on State-of-the-art in Scientific Computing, June 20-23, 2004 Complementary Proceedings", year = "2005", number = "", series = "IMM-Technical report-2005-09", institution = "Informatics and Mathematical Modelling, Technical University of Denmark, {DTU}", address = "Richard Petersens Plads, Building 321, {DK-}2800 Kgs. Lyngby", type = "", url = "http://www2.compute.dtu.dk/pubdb/pubs/4450-full.html", abstract = "The {PARA} workshops in the past have been devoted to parallel computing methods in science and technology. There have been seven {PARA} meetings to date: {PARA'}94, {PARA'}95 and {PARA'}96 in Lyngby, Denmark, {PARA'}98 in Ume{\aa}, Sweden, {PARA'}2000 in Bergen, Norway, {PARA'}02 in Espoo, Finland, and {PARA'}04 again in Lyngby, Denmark. The rst six meetings featured lectures in modern numerical algorithms, computer science, engineering, and industrial applications, all in the context of scientific parallel computing. This meeting in the series, the {PARA'}04Workshop with the title \&\#147;State of the Art in Scientific Computing, was held in Lyngby, Denmark, June 20-23, 2004. The {PARA'}04 Workshop was organized by Jack Dongarra from the University of Tennessee and Oak Ridge National Laboratory, and Kaj Madsen and Jerzy Wasniewski from the Technical University of Denmark. The emphasis here was shifted to High-Performance Computing (HPC). The ongoing development of ever more advanced computers provides the potential for solving increasingly difficult computational problems. However, given the complexity of modern computer architectures, the task of realizing this potential needs careful attention. For example, the failure to exploit a computer's memory hierarchy can degrade performance badly. A main concern of {HPC} is the development of software that optimizes the performance of a given computer. The high cost of state-of-the-art computers can be prohibitive for many workplaces, especially if there is only an occasional need for {HPC}. A solution to this problem can be network computing, where remote computing facilities are exploited via the internet. {PARA'}04 featured invited talks, contributed talks, minisymposia, and software and hardware vendors. The rst day, June 20, was devoted to two parallel tutorials. The minisymposia and contributed talks during the main part of the Workshop, June 21-23, were scheduled in parallel sessions. All invited and contributed talks were noncommercial. The Workshop attracted 230 speakers from all over the world. The {PARA'}06 Workshop with the title State-of-the-Art in Scientific and Parallel Computing will be held in Ume{\aa} (Sweden) on June 17-21, 2006." }