Log in first for optimal use of all functions!

    Scalable In Situ Computation of Lagrangian Representations via Local Flow Maps (Unknown language)

    In: Eurographics Symposium on Parallel Graphics and Visualization ; 19-29  ; 2021

    How to get this title?

    Check access
    Download
    Commercial Copyright fee: €14.50 Basic fee: €4.00 Total price: €18.50
    Academic Copyright fee: €4.50 Basic fee: €2.00 Total price: €6.50

    Export, share and cite

    In situ computation of Lagrangian flow maps to enable post hoc time-varying vector field analysis has recently become an active area of research. However, the current literature is largely limited to theoretical settings and lacks a solution to address scalability of the technique in distributed memory. To improve scalability, we propose and evaluate the benefits and limitations of a simple, yet novel, performance optimization. Our proposed optimization is a communication-free model resulting in local Lagrangian flow maps, requiring no message passing or synchronization between processes, intrinsically improving scalability, and thereby reducing overall execution time and alleviating the encumbrance placed on simulation codes from communication overheads. To evaluate our approach, we computed Lagrangian flow maps for four time-varying simulation vector fields and investigated how execution time and reconstruction accuracy are impacted by the number of GPUs per compute node, the total number of compute nodes, particles per rank, and storage intervals. Our study consisted of experiments computing Lagrangian flow maps with up to 67M particle trajectories over 500 cycles and used as many as 2048 GPUs across 512 compute nodes. In all, our study contributes an evaluation of a communication-free model as well as a scalability study of computing distributed Lagrangian flow maps at scale using in situ infrastructure on a modern supercomputer.

    Table of contents conference proceedings

    The tables of contents are generated automatically and are based on the data records of the individual contributions available in the index of the TIB portal. The display of the Tables of Contents may therefore be incomplete.

    1
    HyLiPoD: Parallel Particle Advection Via a Hybrid of Lifeline Scheduling and Parallelization-Over-Data
    Binyahib, Roba / Pugmire, David / Childs, Hank et al. | 2021
    7
    Machine Learning-Based Autotuning for Parallel Particle Advection
    Schwartz, Samuel D. / Childs, Hank / Pugmire, David et al. | 2021
    19
    Scalable In Situ Computation of Lagrangian Representations via Local Flow Maps
    Sane, Sudhanshu / Yenpure, Abhishek / Bujack, Roxana / Larsen, Matthew / Moreland, Kenneth / Garth, Christoph / Johnson, Chris R. / Childs, Hank et al. | 2021
    31
    Evaluation of PyTorch as a Data-Parallel Programming API for GPU Volume Rendering
    Marshak, Nathan X. / Grosset, A. V. Pascal / Knoll, Aaron / Ahrens, James / Johnson, Chris R. et al. | 2021
    37
    Faster RTX-Accelerated Empty Space Skipping using Triangulated Active Region Boundary Geometry
    Wald, Ingo / Zellmann, Stefan / Morrical, Nate et al. | 2021
    45
    Performance Tradeoffs in Shared-memory Platform Portable Implementations of a Stencil Kernel
    Bethel, E. Wes / Heinemann, Colleen / Perciano, Talita et al. | 2021
    51
    UnityPIC: Unity Point-Cloud Interactive Core
    Wu, Yaocheng / Vo, Huy / Gong, Jie / Zhu, Zhigang et al. | 2021
    63
    Interactive Selection on Calculated Attributes of Large-Scale Particle Data
    Wollet, Benjamin / Reinhardt, Stefan / Weiskopf, Daniel / Eberhardt, Bernhard et al. | 2021

    How to get this title?

    Check access
    Download
    Commercial Copyright fee: €14.50 Basic fee: €4.00 Total price: €18.50
    Academic Copyright fee: €4.50 Basic fee: €2.00 Total price: €6.50

    Export, share and cite

    Feedback