Weekly outline

  • General

    The seminar covers optimization and parallelization techniques for modern multi- and manycore systems. The topics are chosen from interesting contemporary problems in High Performance Computing on modern hardware like multicore processors, accelerators (e.g., GPGPUs or Xeon Phi), and clusters.

    Lecturer: Prof. G. Wellein (gerhard.wellein@fau.de), Martensstr. 1, Room 01.131. Phone -28136
    Location: 2.037 (e-Studio), Martensstr. 1 (RRZE), 2nd floor
    Time: Tuesday 16:00-17:30

    Either 2.5 or 5 ECTS credits will be granted, depending on whether the student gives one or two talks. In either case, a written seminar report is mandatory.

    Possible topics can be found in the intro talk (see below).

    Tutors:

  • 11 April - 17 April

    First Seminar: Tuesday, April 12, 2016, e-Studio (RRZE 2.037), 16:15

    16:15: G. Wellein: Intro, presentation of seminar topics

    17:00: Guest talk by Souley Madougou, University of Amsterdam: Performance modeling on current heterogeneous systems
    Abstract
    Performance predicting parallel computing is notoriously difficult. For the newest generations of CPUs and GPUs, the problem is far from being solved. The few existing generic performance modeling attempts fail the simplicity, usability, and generality requirements for popular adoption. Therefore, there is an urgent need for straightforward, application and architecture agnostic solutions for both performance analysis and performance prediction.
    Our work targets this need, and constitutes a significant step forward in the design of performance analysis tools. Specifically, we define a methodology based on statistical analysis and hardware performance counters. The methodology in built on the predictive power of random forest regression, and can be used for any application and any architecture that allows access to its hardware counters. The goal of the analysis is two-fold: (1) performance bottleneck analysis, and (2) application performance modeling. Based on the performance model, fairly accurate performance predictions can be made.
  • 18 April - 24 April

    16:15 Guest talk

    Tim Werner, University of Bayreuth: Optimierungstechniken für explizite Verfahren zur GPU-beschleunigten Lösung von Anfangswertproblemen gewöhnlicher Differenzialgleichungen (KONWIHR project presentation)

    • 25 April - 1 May

      This week there will be two seminars: Monday, April 25, 16:15, and Tuesday, April 26, 16:15  (both RRZE e-Studio)

      This will be a hands-on session explaining the use of the LIKWID tool suite and the access to the RRZE HPC systems. Please bring your laptop.

      2nd seminar talk by Viktor HaagImplementing ILBDC on GPGPUs.

    • 2 May - 8 May

      Faisal Shahzad (RRZE HPC group): Automatic fault tolerance and checkpoint/restart interface for ESSEX applications.

      • 9 May - 15 May

        Christie Louis Alappat (RRZE HPC group): Optimizing the KACZMARZ Kernel for the GHOST Library.

        • 16 May - 22 May

          No seminar on May 17 (Berg-Dienstag)

          • 30 May - 5 June

            Q&A Session with all advisors

            • 13 June - 19 June

              Christie Louis Alappat: The semi-stencil algorithm (first seminar talk)

            • 20 June - 26 June

              No seminar (ISC conference week)

              • 27 June - 3 July

                The seminar this week will be on Wednesday, June 29, 15:00-18:00.

                Please send the title of your talk to georg.hager@fau.de

                Talks:

                Michael Hußnätter: Performance analysis of a list-based lattice-Boltzmann code (talk 1)

                Raju Ram: Impact of Silent Data Corruption (SDC) on Conjugate Gradient Solver  (talk 1)

                Fabian Hofmann: Analysis of Mantevo's HPCCG benchmark (talk 1)

                Daniel Zint: Performance Analysis of the Molecular Dynamics SPEC Code (talk 1)

                • 4 July - 10 July

                  The seminar this week will be on Wednesday, July 6, 16:00-18:00

                  Talks:

                  Armin BundleSmith Waterman algorithm - Performance Analysis (talk 1)

                  Tobias Klein: Analysis of the 363.swim benchmark

                  Gaurav Chotalia: Analysis of the Mantevo MiniMD benchmark (talk 1)

                  Diaz Arcrossito: Performance Analysis of SPEC's 350.md Benchmark on Xeon Phi (talk 1)

                  Please send the slides to georg.hager@fau.de after your talk.
                  • 11 July - 17 July

                    Chriostopher Bross: Sparse matrix formats and performance (part 2)

                    • 18 July - 24 July