Weekly outline

  • General

    Collapse all Expand all

    Lecturer: Prof. G. Wellein (gerhard.wellein@fau.de), Martensstr. 1, Room 01.131. Phone -28136
    Location: The seminar will be conducted online for the time being. We may migrate to face-to-face over time.

    Zoom Link: https://fau.zoom.us/j/64340968729?pwd=VnAwVERWV251aGJDNStTN3JEa3Bmdz09

    (Passcode: 099278)

    Time: Monday 16:15-17:45

    Credits: 5 ECTS credits. This requires two talks and a written seminar report.

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

    Tutors:


  • 18 October - 24 October

    First seminar (online): Monday, October 18, 4:00 p.m.

    Zoom Link: https://fau.zoom.us/j/64340968729?pwd=VnAwVERWV251aGJDNStTN3JEa3Bmdz09

    (Passcode: 099278)

  • 25 October - 31 October

    Seminar (online): Monday, October 25, 4:15 p.m.

    Topic: Final assignment of topics

    Zoom Link: https://fau.zoom.us/j/64340968729?pwd=VnAwVERWV251aGJDNStTN3JEa3Bmdz09

    (Passcode: 099278)


  • 1 November - 7 November

    No seminar on November 1 (public holiday)

    Important notes:

    • Please register in the Moodle system (click on "Login" in the upper right corner and register as a new user if you haven't done so already) and enroll in the course. This way we can reach you via Moodle.
    • HPC accounts have been set up for all students. Please take a look into your IdM; your account should appear under "Services." You have to set a password to use it. Give it a couple of hours after a PW change to propagate to all your systems.

  • 8 November - 14 November

    Monday, November 8, 4:15 p.m.: 

    Thomas Gruber: Introduction to using the clusters and the LIKWID tool suite (part 1)

  • 15 November - 21 November

    Monday, November 15, 4:15 p.m.: 

    Thomas Gruber: Introduction to using the clusters and the LIKWID tool suite (part 2)

  • 22 November - 28 November

    Monday, November 22, 4:15 p.m.: 

    Gerhard Wellein: Introduction to the Roofline model (slides)

    Video: Introduction to GPU Performance Analysis by Dominik Ernst (IdM login required)  


  • 29 November - 5 December

    November 29, 4:15 p.m.:

    Alexander Lieret: Analysis and Optimization of the Matrix Transpose operation (2nd seminar talk)

  • 6 December - 12 December

    Monday, December 6, 4:15 p.m.: 

    Gonzalo Pinzon: SpMV Performance on modern CPUs (2nd seminar talk)

  • 13 December - 19 December

    Monday, December 13, 4:15 p.m.:

    Maximilian Gaul: MD-Bench CUDA port (first seminar talk)


    All students are also warmly invited to a "Christmas Lecture" on Monday, December 13, at 11:00 a.m.:

    Title: Thirteen modern ways to fool the masses with performance results on parallel computers

    Speaker: Georg Hager, NHR@FAU

    Abstract: In 1991, David H. Bailey published his insightful paper "Twelve Ways to Fool the Masses When Giving Performance Results on Parallel Computers." In that humorous article, Bailey pinpointed typical "evade and disguise" techniques that were used in many papers for presenting mediocre performance results in the best possible light. At that time, the supercomputing landscape was governed by the "chicken vs. oxen" debate: The famous question "If you were plowing a field, which would you rather use? Two strong oxen or 1024 chickens?" is attributed to Seymour Cray, who couldn't have said it better. Cray's machines were certainly dominating in the oxen department, but competition from massively parallel systems like the Connection Machine was building up. In the past two decades, hybrid, hierarchical systems, multi-core processors, accelerator technology, and the dominating presence of commodity hardware have reshaped the landscape of High Performance Computing. It's also not so much oxen vs. chickens anymore; ants have received more than their share of hype. However, some things never change, including the tendency to sugarcoat performance results that would never stand to scientific scrutiny if presented in a sound way. My points (which I prefer to call "stunts") are derived from Bailey's original collection; some are identical or merely  reformulated. Others are new, reflecting today's boundary conditions.

    Zoom: https://fau.zoom.us/j/65060789347

  • 20 December - 26 December

    Michael Panzlaff: HPCCG (1st Seminar talk)

    Biruk Amare: CUDA CoMD (1st Seminar talk)


  • 10 January - 16 January

    No seminar this week!

  • 17 January - 23 January

    No seminar this week!

    Do not forget to schedule you second seminar talk!

  • 24 January - 30 January

    Maximilian Gaul: MD-Bench CUDA port (second seminar talk)

  • 7 February - 13 February

    Biruk Amare: CUDA CoMD (2nd Seminar talk)

    Michael Panzlaff: HPCCG (2nd Seminar talk)