2007 · 20 citations · 20 references
EngineeringObject-oriented ModelingScientific Software DevelopersComputer ArchitectureSoftware EngineeringComputer-aided DesignHigh Performance ComputingSoftware AnalysisExtensible Computational MechanicsParallel Scientific ApplicationsSocial SciencesParallel ToolParallel SoftwareSystems EngineeringModeling And SimulationParallel ComputingAspect-oriented ProgrammingObject-oriented DesignOpen Source SupercomputingDesignComputer EngineeringComputer ScienceSoftware EngineersSoftware DesignArchitectural DesignProgram AnalysisFormal MethodsObject-oriented ProgrammingParallel ProgrammingParallel Programming ModelProgramming MethodologySystem Software
HPC scientific computational models are notoriously difficult to develop, debug, and maintain. The reasons for this are multifaceted - including difficulty of parallel programming, the lack of standard frameworks, and the lack of software engineering skills in scientific software developers. In this paper we discuss the drivers, design and deployment of StGermain, a software framework that significantly simplifies the development of a spectrum of HPC computational mechanics models. The key distinction between StGermain and conventional approaches to developing computational models is that StGermain decomposes parallel scientific applications into a hierarchical architecture, supporting applications collectively built by a diverse community of scientists, modelers, computational scientists, and software engineers.
20
Toward a common component architecture for high-performance scientific computing
Rob Armstrong, Dennis Gannon, Al Geist et al. · 2003 · 391 citations · Full text
Cluster Computing, Massively-parallel Computing, Engineering +14