Virtual Institute – High Productivity Supercomputing

11th VI-HPS Tuning Workshop (MdS, Saclay, France)

Date

Monday April 22 - Thursday April 25, 2013.

Location and How to Get There

The workshop took place near Paris, France, at the Maison de la Simulation (MdS), building 565, CEA Saclay, 91191 Gif-sur-Yvette.

  • How to get to MdS: pdf html
  • Search your itinerary for any public transport in and around Paris: vianavigo (bus stop "CEA porte 306" is close to MdS)

Organizing Institutions

UVSQ PRACE

Goals

This workshop organized by VI-HPS for the French PRACE Advanced Training Centre hosted by Maison de la Simulation was arranged to:

  • 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 were on the following topics:

  • Setting up, welcome and introduction
  • Score-P instrumentation and measurement
  • Scalasca performance analysis toolset
  • Vampir trace analysis toolset
  • TAU performance system
  • MAQAO analysis toolset
  • Intel VTune Amplifier XE profiler
  • Measurement & analysis of heterogeneous HPC systems using accelerators

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

Classroom capacity was limited, therefore priority was given to applicants with parallel codes already running on the workshop computer systems, and those bringing codes from similar systems to work on. Over 25 participants from France, other parts of Europe and Japan prepared their own MPI, OpenMP and hybrid OpenMP/MPI parallel application codes for analysis.

class photo

Programme in Detail

Day 1: Monday 22 April
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 (Brian Wylie, JSC)
14:00
Lab setup
  • Computer systems and software environment
  • Building and running NPB-MZ-MPI/BT-MZ (Brian Wylie, JSC)
  • 15:00 (break)
    15:30 MAQAO optimization tool (Andrés Charif-Rubial & Emmanuel Oseret, UVSQ)
  • MAQAO hands-on exercises
  • 17:30 (adjourn)

    Day 2: Tuesday 23 April
    09:00 Score-P instrumentation & measurement toolset 
  • Score-P hands-on exercises (Ronny Tschüter, TUDresden)
  • CUBE hands-on exercises (Monika Lücke, GRS)
  • 10:30 (break)
    11:00 Score-P analysis scoring & measurement filtering 
  • Hardware counters and other metrics (Ronny Tschüter, TUDresden)
  • ParaProf / PerfExplorer hands-on exercises (Sameer Shende, UOregon)
  • 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 3: Wednesday 24 April
    09:00 Vampir interactive trace analysis (Frank Winkler, TUDresden)
  • Vampir hands-on exercises
  • Scalasca automated trace analysis (Brian Wylie, JSC)
  • Scalasca hands-on exercises
  • 10:30 (break)
    11:00 Score-P advanced configuration
    TAU performance system (Sameer Shende, UOregon)
  • TAU/PDT 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 4: Thursday 24 April
    09:00 VTune Amplifier XE & MPI (Vladimir Tsymbal, Intel)
    10:30 (break)
    11:00 Concluding session (Ronny Tschüter & Brian Wylie)
  • Support for heterogeneous HPC systems using accelerators
  • Performance engineering workflow
  • Review of participant workshop experience
  • 12:30 (lunch)
    13:30 Hands-on coaching to apply tools to analyze participants' own code(s).
    15:00 (break)
    15:30 (adjourn or continue with work to 16:30)
     

    Hardware and Software Platforms

    Poincaré: this cluster, owned by Maison de la Simulation and operated by IDRIS, was used for the hands-on exercises with accounts provided during the workshop:

    • 92 compute nodes "poincare[001-092]" (clallmds) comprised of Sandy Bridge E5-2670 (2.60GHz, dual-socket, 8 cores per socket) and 32 GB shared memory
    • 4 GPU nodes "poincaregpu[01-04]" (clgpu), dual-socket Sandy Bridge E5-2670, 64 GB memory per node, GPU Tesla K10/K20 (CUDA Capability 3.0, 3.5 GB per GPU)
    • 4 interactive frontends "poincareint[01-04]",  dual-socket Sandy Bridge E5-2670, 32 GB memory per node
    • 2 file systems: gpfs1l,  7 TB accessible via HOME; 37 TB gpfsdata
    • software: RedHat Linux, Intel & GNU compilers, Intel MPI,....

    Curie: Intel Xeon X5760 (Nehalem) eight-core processors and Nvidia T20A GPUs, Bullx Linux, Bull MPI, Intel & GCC compilers. Existing account required!

    The local system Poincaré was the primary platform for the hands-on part of the workshop, where Intel compilers (v13) and Intel MPI are supported: . Participants with applications using CUDA (with MPI and/or OpenMP) are encouraged to analyse performance on the GPU nodes of Poincaré. Other systems, e.g. Curie, where up-to-date versions of the tools are installed can also be used when preferred, though support may be limited. Participants are expected to already possess user accounts on non-local systems they intend to use, and should be familiar with the procedures for compiling and running parallel applications on those systems. Please note that during the Tuning Workshop, maintenance is foreseen for Curie (Tue-Wed 23-24 April) and therefore Curie cannot be used on these days.

    Registration

    Registration via the PRACE training portal closed 8 April 2013. Note: the number of participants is limited, and preference was given to those bringing parallel application(s) to analyse and tune as part of the workshop.

    Accommodation

    • MdS list of hotels
    • Trivago meta search engine for hotel bookings

    Contact

    Tuning Workshop Series

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

    Local Arrangements

    Bettina Krammer
    Université de Versailles
    Phone: +33 139 253 615
    Email: bettina.krammer@uvsq.fr

    Sponsors

    Exascale computing research CEA Genci Intel