High performance computing tools in science and engineering

Abstract

The combination of multicore architectures with the potential acceleration of Graphics Processing Units allows to design, low cost, desktop machines with hundreds of processing cores and at the same time to build large supercomputers with hundreds of thousands of computational units. Heterogeneity in the architecture and programming model appears naturally, so that the solution of many problems in science and engineering poses a different performance depending on the architecture and how they were programmed. Since current compilers are no longer able of automatically moving the computing power of the new generation of processors to applications, an interdisciplinary effort among researchers in science and engineering, and computing disciplines is demanded. The purpose of the Minisymposium on High Performance Computing in Science and Engineering is to bring together applied mathematicians, computer scientists, and in general researchers with costly applications or a fair knowledge/interest in high performance computing, to present, share, and discuss their techniques, tools, and ideas in the area of high performance computing applied to complex large-scale computational problems. This Minisymposium use is to be held as part of the International Conference on Computational and Mathematical Methods in Science and Engineering (CMMSE). This special issue collects a selection of the papers presented at the edition celebrated in La Manga del Mar Menor (Murcia–Spain) in July 2012 during the CMMSE—2012. Although several groupings of the contributions appearing in this issue can be performed, we decided to classify them into the four following groups.