DEVELOPMENT OF SOFTWARE COMPONENTS TO SUPPORT PARALLEL ADAPTIVE MULTISCALE ANAL

这个子项目是利用资源的许多研究子项目之一。 由NIH/NCRR资助的中心拨款提供。对子项目的主要支持 子项目的首席调查员可能是由其他来源提供的， 包括美国国立卫生研究院的其他来源。为子项目列出的总成本可能 表示该子项目使用的中心基础设施的估计数量， 不是由NCRR赠款提供给次级项目或次级项目工作人员的直接资金。 这项工作是科学计算研究中心(SCOREC)研究工作的一部分，该中心专注于开发支持复杂物理和生物系统的并行自适应分析的软件组件。这个项目的两个目标被描述如下：创建一个致力于在超级计算机上解决并行应用开发的研究小组：研究小组由一名高级研究助理、一名CCNI计算机科学家、三名研究生和两名本科生组成。该团队在为RPI研究人员提供的超级计算机(CCNI：拥有32,000个处理器的蓝色基因L体系结构)上开发和运行应用程序方面已经有了一些经验。该项目的第一个目标是支持建立一个超级计算专家小组，他们将负责通过以下方式向其他学生和研究人员传播知识：开发一个网站，提供教程并推广最佳做法，供所有对超级计算感兴趣的RPI学生使用。每月向所有RPI的学生和研究人员开放研讨会。撰写年度报告，总结研究进展。实现可伸缩的多尺度层次化生物工程应用程序：由一套并行软件组件(场、模型、域、误差估计、适应)的开发支持，这些组件结合在一起以支持复杂物理和生物系统的适应性分析[1]。每个支持特定功能集的软件组件都是单独开发的，以支持其在大规模并行计算机上的执行。这样的策略已经成功地使用主/从方法来对生物工程材料的模拟进行建模，使用分层的多尺度方法[2]。由于主/从模型不太适合大规模并行计算机上的规模，我们正在进行的研究工作的一部分是开发一种新的并行多尺度范例，有效地结合：使用扩展PHASTA策略的求解器的多尺度层次分析[3]，该策略通过结合线性代数求解器的本地实例展示了高达300,000个处理器的可扩展性[4]。使用朱-Zienkewicz SPR技术的并行版本的误差估计[5]。Mesh[6]和模型调整[7]，它们已经展示了可扩展到32,000个处理器。多尺度负载平衡，包括估计网格分割器要使用的多尺度权重[8]。[1]F.Delalondre，C.Smith，M.S.Shephard，支持自适应多模型模拟的协作软件基础设施，应用力学和工程中的计算机方法，2010年接受手稿。[2]罗兴杰，T.Stylianopoulos，V.H.Barocas，M.S.Shephard，复杂几何形状生物人工软组织的多尺度计算，计算机工程，第25卷，第1期，87-95，[3]O.Sahni，C.D.Carthers，M.S.Shephard和K.E.Jansen，强标度分析，并行，非结构化，隐式求解器和操作系统干扰的影响，科学编程，17(3)，261-274。[4]可移植、可扩展的工具包？科学计算(PETSC)，http://www.mcs.anl.gov/petsc/petsc-as/index.html[5]O.C.Zienkewicz和J.G.朱，一种用于实际工程分析的简单误差估计器和自适应策略，国际工程数值方法杂志，第24卷，1987，第337-357页。[6]F·Alauzet，X.Li，E.S.Seol，M.S.Shephard，网格修改的并行各向异性三维网格自适应，英文。计算机，21(2006)247258[7]M.A.Nuggehally，C.R.PICU，M.S.Shephard，并行多尺度问题的自适应模型选择过程，多尺度计算工程杂志。2007年，(5)，369-386。[8]K.Devine，E.Boman，R.Hephy，B.Hendrickson，C.Vaughan，？佐尔坦：面向并行动态应用的数据管理服务，科学与工程计算，2002，(4)，2，90-97。

This subproject is one of many research subprojects utilizing the resources provided by a Center grant funded by NIH/NCRR. Primary support for the subproject and the subproject's principal investigator may have been provided by other sources, including other NIH sources. The Total Cost listed for the subproject likely represents the estimated amount of Center infrastructure utilized by the subproject, not direct funding provided by the NCRR grant to the subproject or subproject staff. This work is part of a research effort of the Scientific Computation Research Center (SCOREC) which focuses on developing software components to support the parallel adaptive analysis of complex physical and biological systems. The two aims of this project are described as follows: Creation of a group of research dedicated to the development of parallel application to be solved on supercomputers: The group of researchers is made of a senior research associate, a CCNI computer scientist, three graduate and two undergraduate students. This group already has some experience in developing and running applications on a supercomputer made available to RPI researchers (CCNI: Blue Gene L architecture with 32,000 processors). The first aim of this project is to support the creation of a group of supercomputing experts who will be in charge of disseminating knowledge to other students and researchers through: The development of a website providing tutorials and promoting best practices to be used by all RPI students interested in research in supercomputing. Monthly seminars opened to all RPI students and researchers. Writing yearly report summarizing research progress. Implementation of a scalable multiscale hierarchic bioengineered application: Is supported by the development of a suite of parallel software components (Fields, Model, Domain, Error estimation, Adaptation) that are brought together to support the adaptive analysis of complex physical and biological systems [1]. Each software component, which supports a specific set of functionalities, is individually developed to support its execution on massively parallel computers. Such a strategy has been successfully implemented using a master/slave approach to model the simulation of bioengineered material using a hierarchic multiscale approach [2]. As the master/slave model is not well suited to scale on massively parallel computers, part of our on going research effort is to develop a new parallel multiscale paradigm efficiently combining: Multiscale hierarchic analysis using a solver that extends PHASTA strategy [3] which demonstrated scalability up to 300,000 processors by incorporating local instances of linear algebraic solvers [4]. Error estimation using a parallel version of the Zhu-Zienkewicz SPR technique [5]. Mesh [6] and model adaptations [7] that already demonstrated scaling up to 32,000 processors. Multiscale load balancing that consists of estimating the multiscale weights to be used by the mesh partitioner [8]. [1] F. Delalondre, C. Smith, M.S. Shephard, Collaborative software infrastructure to support adaptive multiple model simulation, Computer Methods in Applied Mechanics and Engineering, accepted manuscript, 2010. [2] X.-J. Luo, T. Stylianopoulos, V.H. Barocas, M.S. Shephard, Multiscale computation for bioartificial soft tissues with complex geometries, Engineering with Computers, Volume 25, Number 1, 87-95, [3] O. Sahni, C.D. Carothers, M.S. Shephard and K.E. Jansen, Strong Scaling Analysis of a Parallel, Unstructured, Implicit Solver and the Influence of the Operating System Interference, Scientific Programming, 17 (3), 261-274. [4] Portable, Extensible Toolkit for? Scientific Computation (PETSc), http://www.mcs.anl.gov/petsc/petsc-as/index.html [5] O. C. Zienkewicz and J. G. Zhu, A simple error estimator and adaptive strategy for practical engineering analysis, International Journal for Numerical Methods in Engineering, vol. 24, 1987, pp. 337-357. [6] F. Alauzet, X. Li, E.S. Seol, M.S. Shephard, Parallel anisotropic 3D mesh adaptation by mesh modification, Eng. Comput., 21 (2006) 247258 [7] M.A. Nuggehally, C.R. Picu, M.S. Shephard, Adaptive Model Selectionprocedure for Concurrent Multiscale Problems, Journal of Multiscale Computational Enginnering. 2007, (5), 369-386. [8] K. Devine, E. Boman, R. Heaphy, B. Hendrickson, C. Vaughan, ? Zoltan: Data Management Services for Parallel Dynamic Applications, Computing in Science and Engineering, ?2002, ?(4), ?2, ?90-97.