• Login
    View Item 
    •   DSpace@RPI Home
    • Rensselaer Libraries
    • RPI Theses Open Access
    • View Item
    •   DSpace@RPI Home
    • Rensselaer Libraries
    • RPI Theses Open Access
    • View Item
    JavaScript is disabled for your browser. Some features of this site may not work without it.

    Conformal mesh adaptation on heterogeneous supercomputers

    Author
    Ibanez, Daniel Alejandro
    Thumbnail
    View/Open
    177780_Ibanez_rpi_0185E_10983.pdf (5.915Mb)
    Other Contributors
    Shephard, M. S. (Mark S.); Sahni, Onkar; Carothers, Christopher D.; Anshelevich, Elliot;
    Date Issued
    2016-12
    Subject
    Computer science
    Degree
    PhD;
    Terms of Use
    This electronic version is a licensed copy owned by Rensselaer Polytechnic Institute, Troy, NY. Copyright of original work retained by author.;
    Metadata
    Show full item record
    URI
    https://hdl.handle.net/20.500.13015/1817
    Abstract
    Mesh adaptation is a technique which dynamically modifies the mesh being used to approximately solve a Partial Differential Equation (PDE) in order to improve aspects of the approximate solution including the computer time and memory used to compute it as well as its level of accuracy. Even with the use of mesh adaptation, computing ever more accurate PDE solutions requires significant computer time and memory, motivating the use of supercomputers, which are constructed as networks of cooperating computational hardware. Trends in the computer hardware industry at large are introducing heterogeneous designs for current leadership-class supercomputers, which is both an opportunity and a challenge for programs aiming to make use of these machines.; Several improvements to the design of mesh adaptation programs are presented, including solution transfer methods which preserve mass and momentum, methods for the maintenance of high-quality elements, scalable and deterministic methods for hybrid parallelization of mesh modification operations, and a combination of modification operators which reduce implementation complexity without sacrificing effectiveness.; This thesis presents implementations of mesh adaptation which are designed with memory efficient cache-friendly data structures and algorithms which can effectively leverage both distributed memory parallelism and shared memory parallelism (including GPUs). The data structures used in these implementations are widely applicable to other tasks involving meshes, and the programming paradigms introduced are general enough to be of use in most programs targeting leadership-class supercomputers. The implementations presented are being used by several simulation codes in production, and are available as open-source tools so they may continue providing value to the scientific community.;
    Description
    December 2016; School of Science
    Department
    Dept. of Computer Science;
    Publisher
    Rensselaer Polytechnic Institute, Troy, NY
    Relationships
    Rensselaer Theses and Dissertations Online Collection;
    Access
    Users may download and share copies with attribution in accordance with a Creative Commons Attribution-Noncommercial-No Derivative Works 3.0 License. No commercial use or derivatives are permitted without the explicit approval of the author.;
    Collections
    • RPI Theses Online (Complete)
    • RPI Theses Open Access

    Browse

    All of DSpace@RPICommunities & CollectionsBy Issue DateAuthorsTitlesSubjectsThis CollectionBy Issue DateAuthorsTitlesSubjects

    My Account

    Login

    DSpace software copyright © 2002-2022  DuraSpace
    Contact Us | Send Feedback
    DSpace Express is a service operated by 
    Atmire NV