Performance evaluation and visualization of scientific applications using PMlib

Kazunori Mikami, Kenji Ono, Jorji Nonaka

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

    1 Citation (Scopus)

    Abstract

    The computational performance of scientific applications on HPC systems is often much lower than user expectation based on the system's maximum performance specifications. To understand the basis for this performance gap, a multi-perspective evaluation is important. For instance, from the user perspective, correlating the theoretical computation coded as a source program with the actual computation workload produced by the compilers is valuable. From the system perspective, evaluating the characteristics of microarchitecture elements such as processor core and memory is of significance. An open source library called PMlib was developed to address these types of synthetic evaluations. PMlib provides an avenue for reporting the arithmetic/application workload explicitly coded in the source program, as well as the actually executed system workload. It also provides detailed utilization reports of processor-specific hardware including the categorized SIMD instruction statistics, the layered cache hit/miss rate, and the effective memory bandwidth, which are captured via hardware performance counters (HWPC). Using PMlib, users can conduct a synthetic analysis of application performance, and obtain useful feedback for further optimized execution of applications.

    Original languageEnglish
    Title of host publicationProceedings - 2018 6th International Symposium on Computing and Networking Workshops, CANDARW 2018
    PublisherInstitute of Electrical and Electronics Engineers Inc.
    Pages243-249
    Number of pages7
    ISBN (Electronic)9781538691847
    DOIs
    Publication statusPublished - Dec 26 2018
    Event6th International Symposium on Computing and Networking Workshops, CANDARW 2018 - Takayama, Japan
    Duration: Nov 27 2018Nov 30 2018

    Publication series

    NameProceedings - 2018 6th International Symposium on Computing and Networking Workshops, CANDARW 2018

    Conference

    Conference6th International Symposium on Computing and Networking Workshops, CANDARW 2018
    CountryJapan
    CityTakayama
    Period11/27/1811/30/18

    All Science Journal Classification (ASJC) codes

    • Computer Networks and Communications
    • Hardware and Architecture
    • Statistics, Probability and Uncertainty
    • Computer Science Applications

    Fingerprint Dive into the research topics of 'Performance evaluation and visualization of scientific applications using PMlib'. Together they form a unique fingerprint.

    Cite this