Event Detail

Registration

Sorry, the registration period for this event is over.

Practical Performance Analysis of Parallel Applications - 6-7 October 2014

 

Applications need to be effectively parallelised, using techniques such as MPI message-passing, OpenMP multithreading and OpenACC/CUDA for GPU accelerators, to fully exploit the capabilities of modern systems for high-performance computing.  The course introduces the open-source community-developed Score-P instrumentation and measurement infrastructure, and the Scalasca and Vampir tools using it, to provide a practical basis for portable performance analysis of parallel applications.

The course will be delivered as a series of presentations with associated hands-on practical exercises using the Piz Daint Cray XC30 system.  It starts with basic application instrumentation and measurement to generate execution profiles, then improves measurement quality via customization capabilities, and progresses to interactive and automated analyses of execution traces.

While analysis of provided example codes will be used to guide the class through the relevant steps and familiarise with usage of the tools, coaching will also be available to assist participants to analyse their own parallel application codes and may suggest opportunities for improving their execution performance and scalability.

On completion participants should be familiar with common performance analysis techniques and how they can be employed in practice (on a range of HPC systems). Those who prepared their own application test cases will have been coached in the tuning of their measurement and analysis, and provided optimization suggestions.

Deadline for registration September 26th, 2014 at 23:59.

Please contact Jean-Guillaume Piccinali  (jeanguillaume.piccinali(at)cscs.ch) for further information.

Instructors

Felix Schmitt (Uni Dresden), Brian Wylie (Julich)

VenueCSCS, Via Trevano 131, Lugano http://www.cscs.ch/about/visitor_Information/index.html
Time
Day 1: 09:30 - 17:30; Day 2:  09:00 - 17:30
Prerequisites

 

This is an intermediate/advanced course

Familiarity with MPI, OpenMP & CUDA programming in Fortran, C or C++, and running parallel applications on Cray supercomputers. Ideally participants will have their own parallel application code which builds and runs successfully on PizDaint Cray XC30 or similar systems.  Participants need to have their own notebook computer configured for SSH access to CSCS computer systems and running X11-based graphical tools.

Maximum number of participants 

30
Minimal number of participantsIf the minimal number of participants is not reached we reserve the right to cancel the course. You will be informed two weeks in advance.
Accommodation
Participants are kindly requested to make their own arrangements for accommodation.

*** 

Agenda

Monday 06/10/2014

09:00   Introduction to parallel performance engineering

             Instrumentation and measurement of parallel applications with Score-P

10:30   (break)

11:00   Profile analysis report exploration

             Configuration and customisation of Score-P measurements

12:30   (lunch)

13:30   Coaching to profile your own parallel application(s) on PizDaint

15:00   (break)

15:30   Parallel execution performance analysis diagnosis and tuning

17:00   (close)

 

Tue 07/10/2014

09:00   Automated trace analysis with Scalasca

10:30   (break)

11:00   Interactive trace analysis with Vampir

12:30   (lunch)

13:30   Coaching to trace your own parallel application(s) on PizDaint

15:00   (break)

15:30   Parallel execution performance analysis diagnosis and tuning

17:00   (close)


Back to listing