CAPS Announces Parallel Programming Week 2 in Paris (France)

Rennes, May 15th - CAPS entreprise announces new Parallel Programming Week 2 (PPW2) that will take place from June, 22nd to July, 1st 2009.

PPW2 gathers four reputed speakers who will open the event with a first day plenary session discussing multicore challenges, followed by 5 intensive developers’ workshops focused on parallel computing.

Monday executive networking day will be co-presented by four ‘gurus’ of parallel computing whose presentations will focus on “How to be successful in the multicore era?” :

• Jack Dongarra

University Distinguished Professor

Director, Innovative Computing Laboratory

University of Tennessee, USA

Presentation abstract:

In this talk we examine how high performance computing has changed over the last 10-year and look toward the future in terms of trends. These changes have had and will continue to have a major impact on our software. Some of the software and algorithm challenges have already been encountered, such as management of communication and memory hierarchies through a combination of compile--time and run--time techniques, but the increased scale of computation, depth of memory hierarchies, range of latencies, and increased run--time environment variability will make these problems much harder.

We will look at four areas of research that will have an importance impact in the development of software.

We will focus on following themes:

· Redesign of software to fit multicore architectures

· Automatically tuned application software

· Exploiting mixed precision for performance

The importance of fault tolerance

• Barbara Chapman

CEO of cOMPunity

University of Houston, USA

Presentation abstract:

Application Programming in the Multicore Era

Applications outlive computers. Although they change and grow along with the demands placed upon them, individual computer programs are typically deployed for a relatively long period of time. It is therefore important to base their development on widely supported standards. In this presentation, we discuss the status of parallel programming interfaces that are suitable for multicore programming, with a particular focus on the OpenMP industry standard. We also consider some of the ways in which current standards are expected to evolve to meet the demands of emerging manycore, hybrid and heterogeneous architectures.

• Guillaume Colin de Verdière

Advisor for Future Visualization Systems, CEA, Centre DAM France

Presentation abstract:

From CEA's needs and available machines, we will try to demonstrate the different ways to handle multi- and many-core. We'll show their advantages and limitations. Then we'll try to picture what could be the future in terms of massive parallelism.

• Bernd Mohr

Senior Scientist

Jülich Supercomputing Center, Germany

Presentation abstract:

Programming parallel computers is much harder than programming conventional sequential ones, as the programmer not only has to find the right data structures and algorithms for the given task but also has to distribute the data and work to the different processor cores in a way that the much more powerful resources of a multi-core chip or a HPC cluster are used efficiently. New developments like using GPUs or Cell B.E. chips for hardware acceleration make this process even harder. This presentation will introduce well-proven methods to analyze the performance of parallel programs and will give an overview of commercial and open-source tools which help the programmer in tuning their parallel programs. Finally, examples how companies could optimize their parallel applications for enhanced performance will be presented.

Following this first opening day, PPW organizers have assembled a rich developers’ dedicated program covering main parallel computing technologies and allowing attendees to select specific theme days of interest.

These high-level developers’ workshops are part of a new inter-company training catalog implemented by CAPS Training Team following the successful feedbacks of its in-house training sessions to large industrial accounts.

Mixing theory and practical labs, CAPS opens to the developers’ community its training program around parallel computing:

· 2 workshops around CUDA programming model (basic and expert)

· 1 workshop introducing HMPP, software developed by CAPS enhancing the performance of manycore systems by using hardware accelerators while reducing development efforts

· 1 openMP workshop

· 1 MPI workshop

« In this new edition of PPW, CAPS Training Team innovates twice…», declares Benoît Raoult, PPW2 organizer at CAPS, « … by proposing an international networking day around multicore challenges and by giving each attendee the desktop they will have used during the labs of workshops for free. PPW2 will enable each participant to serenely enter the multicore era, some thanks to new contacts, some others thanks to quick hands-on practices. »

PPW2 addresses profiles from basic to expert and aims at producing immediate practical results and giving operational capacity to developers.

The richness of this offer associated to educational strategy of its experts-trainers definitely makes CAPS a key reference in parallel computing training.

About CAPS entreprise

CAPS develops and commercializes HMPP (Hybrid Multicore Parallel Programming), the industry leading development tool suite for high performance applications. CAPS is committed to working with hardware vendors and ISVs to provide their customers with solutions to help them efficiently develop applications that leverage the performance of multicore processors. The company is headquartered in Rennes, France.

Further information is available at www.caps-entreprise.com

To contact training department: training@caps-entreprise.com