Virtual Institute — High Productivity Supercomputing

12th VI-HPS Tuning Workshop (JSC, Jülich, Germany)

Date

Monday 7th - Friday 11th October 2013.

Location

The workshop took place west of Cologne at Jülich Supercomputing Centre (JSC), building 16.4, Forschungszentrum Jülich, Jülich, Germany.

  • How to get to JSC (by car or train)
  • Public bus SB11 between Jülich town and JSC: route map + timetable
    Jülich, Neues Rathaus is the closest bus-stop for town centre hotels.
    FZJ Seecasino is the closest bus-stop to Jülich Supercomputing Centre, however, it's also only a short walk from the main entrance.
On arrival at the main entrance, first go to the Visitor reception to collect a visitor badge. Please note that for security reasons, access to Forschungszentrum Jülich is only permitted on production of a valid identity document (passport or identity card).

Organizing Institutions

JSC PRACE

Goals

This workshop organized by VI-HPS for the German PRACE Advanced Training Centre hosted by Jülich Supercomputing Centre will:

  • give an overview of the VI-HPS programming tools suite
  • explain the functionality of individual tools, and how to use them effectively
  • offer hands-on experience and expert assistance using the tools

Programme Overview

Presentations and hands-on sessions are planned on the following topics:

  • Setting up, welcome and introduction
  • Parallel file I/O bottlenecks and solutions
  • Score-P instrumentation and measurement infrastructure
  • Scalasca automated trace analysis toolset
  • Vampir interactive trace analysis toolset
  • Periscope automated performance analysis tool
  • TAU performance system
  • Paraver trace analysis tool
  • MAQAO binary analysis & optimization tool
  • MUST runtime error detection tool for MPI
  • STAT stack trace analysis tool

A brief overview of the capabilities of these and associated tools is provided in the VI-HPS Tools Guide.

The workshop will be held in English and run from 09:00 to not later than 18:00 each day, with breaks for lunch and refreshments. There is no fee for participation, however, participants are responsible for their own travel and accommodation.

Classroom capacity is limited, therefore priority will be given to applicants with parallel codes already running on the workshop computer systems (Juqueen), and those bringing codes from similar BlueGene systems to work on. Participants are therefore encouraged to prepare their own MPI, OpenMP and hybrid OpenMP/MPI parallel application codes for analysis.

Programme in Detail

Day 1: Monday 7 October
09:30 (registration & set-up of course accounts on workshop computers)
[Optional] Individual preparation of participants' own codes.
12:30
(lunch)
13:30 Welcome
  • Introduction to VI-HPS & overview of tools [Brian Wylie]
  • Introduction to parallel performance engineering
  • Parallel file I/O bottlenecks and solutions [Wolfgang Frings]
  • 15:30 (break)
    16:00 Lab setup
  • Computer systems and software environment
  • Building and running NPB-MZ-MPI/BT-MZ on JUQUEEN (JUROPA)
  • 17:30 (adjourn)

    Day 2: Tuesday 8 October
    09:00 Score-P instrumentation & measurement toolset [Ronny Tschüter]
  • Score-P hands-on exercises
  • CUBE hands-on exercises [Markus Geimer]
  • 10:30 (break)
    11:00 Score-P analysis scoring & measurement filtering 
  • Measuring hardware counters and other metrics 
  • TAU ParaProf / PerfExplorer hands-on exercises [Sameer Shende]
  • 12:30 (lunch)
    13:30 Hands-on coaching to apply tools to analyze participants' own code(s).
    17:00 Review of day and schedule for remainder of workshop
    17:30 (adjourn)
    19:30 Social dinner: Trattoria Rossini (Bauhofstr. 28, Jülich)

    Day 3: Wednesday 9 October
    09:00 Scalasca automated trace analysis [David Böhme & Markus Geimer]
  • Scalasca hands-on exercises
  • Vampir interactive trace analysis [Ronny Tschüter & Bert Wesarg]
  • Vampir hands-on exercises
  • 10:30 (break)
    11:00 Periscope automated analysis [Yury Oleynik]
  • Periscope hands-on exercise for JUROPA
  • Score-P advanced configuration
    12:30 (lunch)
    13:30 Hands-on coaching to apply tools to analyze participants' own code(s).
    17:00 Review of day and schedule for remainder of workshop
    17:30 (adjourn)

    Day 4: Thursday 10  October
    09:00 Paraver tracing tools suite [Judit Gimenez & Harald Servat]
  • Paraver hands-on exercises
  • 10:30 (break)
    11:00 TAU performance system [Sameer Shende]
  • TAU hands-on exercises
  • 12:30 (lunch)
    13:30 Hands-on coaching to apply tools to analyze participants' own code(s).
    17:00 Review of day and schedule for remainder of workshop
    17:30 (adjourn)

    Day 5: Friday 11  October
    09:00 Debugging, correctness checking & optimization tools
  • STAT [Alexandre Strube]
  • MUST [Joachim Protze & Felix Münchhalfen]
  • MAQAO [Andres Charif-Rubial & Emmanuel Oseret]
  • 10:30 (break)
    11:00 Hands-on coaching to apply tools to analyze participants' own code(s).
    12:30 (lunch)
    13:30 Hands-on coaching to apply tools to analyze participants' own code(s).
  • Review
  • 15:00 (break)
    15:30 (adjourn or continue with work to 16:30)
     

    Hardware and Software Platforms

    Juqueen: this 28-rack BlueGene/Q system will be used for the hands-on exercises:

    • 28,672 compute nodes with IBM PowerPC A2 processors (1.6GHz, 16 cores per processor, 4-way hardware threads per core) and 16 GB memory per node
    • software: IBM lightweight CNK, BlueGene MPI and XL compilers

    Juropa: 3,288 compute nodes with dual Intel Xeon X5570 (Nehalem) 2.93 GHz quad-core processors with 2-way SMT and 24 GB memory per node, SuSE Linux, Parastation MPI, Intel compilers. Primarily expected to be used for analysis presentation tools.

    The local HPC system Juqueen is the primary platform for the workshop, with the Juropa x86_64 Linux cluster used for analysis presentation tools not available on Juqueen. Course accounts will be provided during the workshop to participants without existing accounts. Other systems where up-to-date versions of the tools are installed can also be used when preferred, though support may be limited and participants are expected to already possess user accounts on non-local systems. Regardless of whichever systems they intend to use, participants should be familiar with the relevant procedures for compiling and running their parallel applications (via batch queues where appropriate).

    Registration

    Registration via the PRACE training portal closed in September: the number of participants is limited and selection made based on the information provided when registering. We had 20 participants from 10 institutions in 8 countries.

    Contact

    Brian Wylie
    Jülich Supercomputing Centre
    Forschungszentrum Jülich
    Phone: +49 2461 61-6589
    Email: b.wylie@fz-juelich.de

    Sponsors

    PRACE This workshop is a PRACE Advanced Training Centre (PATC) course, organised by Jülich Supercomputing Centre for the Gauss Centre for Supercomputing.