|Name:||Tutorial 10Hands-on with Intel Xeon Phi|
|Time:||Sunday, June 16, 2013
2:00 PM - 6:00 PM
|Room:||Lecture Room 9 (LR 9)
CCL - Congress Center Leipzig
|Breaks:||4:00 PM - 4:30 PM Coffee Break|
|Presenter(s):||William L. Barth, University of Texas at Austin|
|Kent Milfeld, University of Texas at Austin|
|Daniel C. Stanzione, University of Texas at Austin|
|Abstract:||Many of the top supercomputers deployed recently feature heterogeneous architectures with many-core CPU and accelerator technologies. One such system is the Stampede system at the Texas Advanced Computing Center (TACC), the first system with a large scale deployment of the Intel Xeon Phi Many Integrated Core (MIC) technology. Stampede features more than 2PF and 100,000 cores of the traditional Intel Xeon E5, and an additional 7+ PF in more than 6,400 Xeon Phi processors with hundreds of thousands of heterogeneous cores. This tutorial will introduce experienced C/C++ and Fortran programmers to techniques essential for preparing their scientific applications for Stampede and other future systems. Participants in this tutorial will develop an understanding of how to extract optimal performance of their programs on a heterogeneous system with Xeon E5 hosts and MIC co-processors. The tutorial will cover all elements that are crucial for learning techniques for efficient and parallel execution on these new platforms, i.e. SIMD vectorization, high thread count OpenMP shared memory parallelization, and hybrid MPI programming through hands-on use of Stampede.
10% Beginner, 30% Intermediate, 60% Advanced
This tutorial is intended for application developers (or application modifiers) who will wish to port their applications to supercomputers consisting of both multi-core processors and many-core co-processors on a single node; it will also help users take advantage of the extended vector lengths in Sandy Bridge systems (AVX) and MIC coprocessors.
The tutorial will assume that the audience has a background in programming for scientific computing, including knowledge of some variant of Fortran or C/C++ and MPI. Basic experience with MPI and a working knowledge of OpenMP is required.