Feb. 28 — 2013 is the year that large scale heterogeneous computing becomes a reality. Systems such as Blue Waters, Stampede, and Titan take a major step from modest scale, heterogeneous test beds and prototypes to world-class, extreme-scale heterogeneous computing systems. Through the use of these systems, the science and engineering community has access to excellent resources and services to experience multiple ways to harness manycore processes for their applications.  

The NSF funded Blue Waters and eXtreme Science and Engineering Discovery Environment (XSEDE) projects are hosting the seventh in a series of Extreme Scaling workshops.  The workshop will address large scale heterogeneous computing through a series of talks and discussions for sharing experiences and successes making use of the new types resources of GPUs and manycore processors.  This workshop is set up to provide a sufficient period of time for science and engineering teams to gain understanding on utilizing these systems to advance their applications while still having time to exchange and learn best practices.   

The workshop goal is to inform research and education communities about how to effectively utilize large numbers of accelerated nodes.  The dissemination of information will occur through presentations, discussions, and workshop proceedings that will include recommendations from the presenters and participants who have had experience with these technologies.  The workshop organizers plan to seek ACM SIG affiliation to include the proceedings in the ACM digital library.

The workshop committee seeks submissions of excellent quality addressing your experiences with the following challenges:

• Application porting and tuning to NVIDIA Kepler and Intel Phi systems
• Using GPU/manycore nodes at large scale, including load balancing at scale
• Using heterogeneous and homogeneous cores in a single application (e.g. XE6 and XK7 nodes)
• Performance evaluation of heterogeneous systems
• Experiences and performance comparison with OpenACC, CUDA, and Intel Phi
• Use of heterogeneous computing for solving Big Data problems and challenges

Two types of submissions are encouraged from scientists, engineers, scholars, and high-performance technologists from colleges, universities, laboratories, industry, HPC centers, and other organizations conducting related work.

• Abstracts for papers that include title, abstract, and names and institutions of authors/presenters.  Abstracts should be no more than two pages in length and provide sufficient detail for the workshop committee to make informed judgments about the work. Abstracts should include the names and affiliations of all co-authors and indicate which author will be presenting.  Authors whose abstracts are selected will be asked to submit a full paper by July 1, 2013, and provide a 30-minute presentation during the workshop.
• White papers that identify needs for improving the computing environment for advancing scientific discovery that can be implemented in three to six months.  Please provide a description (at least 2 paragraphs but at most 2 pages) of the desired improvement(s), including your name, institution, and how the improvement will enhance scientific discovery.  The white papers should address the top three improvements that will benefit the community. The workshop committee will select a few white paper authors to be members of a lightning round panel to each speak about the proposed improvements, and to foster discussion among the workshop participants.

There will be discussion time after the paper presentations and after the panels on needed improvements.  Workshop proceedings will be produced that will include the accepted papers and white papers and a summary of recommendations from the discussions among the presenters and participants.

Send submissions to Scott Lathrop at lathrop@illinois.edu by May 15, 2013. Notice of acceptances will be issued by June 1, 2013.

Key Dates

May 15: Paper abstracts and white paper submissions due.
June 1: Acceptances issued.
June 10: Workshop registration opens.
July 1: Final papers due.
Aug. 15-16: Workshop held at University of Colorado, Boulder

Registration, Travel and Accommodations

A registration fee of $150 covers the cost of food and beverages provided at the workshop: breakfast, lunch and dinner on Monday, and breakfast and lunch on Tuesday.

A block of rooms will be reserved at a hotel that is within walking distance of the meeting location in Boulder. The room rates in our block are $124/night plus taxes.

Participants are responsible for the cost of their own travel and accommodations and making their own travel arrangements.  There is a shuttle available to transport people between the Denver airport and Boulder for about $33/person.  Additional details about housing and travel will be made available on the web site.

WORKSHOP WEBSITE: https://www.xsede.org/web/xscale/xscale13 

-----

Source: EGI and XSEDE