Computing graphs on an HPC cluster




  • Location

    ONLINE

    Date And Time

    Wed, 17 April 2019

    10:00 AM – 11:00 AM PDT

    Register Here

    Description

    Computing graphs on an HPC cluster: Working with distributed unstructured data in Chapel
    Over the past 1.5 years in WestGrid we started using Chapel programming language to teach our HPC users the basic concepts of parallel programming. We use Chapel to show how parallel computations can be done on shared- and distributed-memory systems, via both task and data parallelism on either type of system. In these courses, our sample codes utilize dense, rectangular arrays that are either stored in local memory or are distributed among multiple compute nodes on a cluster.
    Although we don’t include it into our basic training, Chapel allows less structured data sets, and some of these can be distributed and processed in parallel by multiple compute nodes. In this webinar we will cover sparse arrays, associative and opaque domains and arrays, and will implement a graph on top of associative domains, and then decompose this graph over several nodes. We will then write our own mapping to make operations on this graph more efficient.
    We don’t assume any prior knowledge of Chapel. We will start with the basics needed to understand the core material. Attendees of all levels are welcome.

    Speaker:

    • Alex Razoumov, Training & Visualization Coordinator, WestGrid

    Webinar Instructions:

    • Click the green “Register’ button on this page to register for this event. All registrants will be emailed the connection instructions for this webinar.

    If you have questions or would like more information, please contact info@westgrid.ca