View Event

Optimizing for the Intel Xeon Phi (OSC Workshop)
Date : 09/25/2013 - 9AM - 4:30PM

Register for Event

BALE Theater
Ohio Supercomputer Center
1224 Kinnear Road
Columbus OH 43212

Researchers and developers interesting in learning how to use Intel development tools to develop for Intel Xeon Phi hardware.

Intel will be coming to the Ohio Supercomputer Center in September to provide some in-depth training on using their software development tools to develop code that performs well on Intel Xeon Phi hardware.

In the afternoon, some time is allocated for attendees to interface with Intel and OSC staff to examine attendee code and problems, and explore the tools on OSC's Ruby Transitional Cluster.

9:00 – 11:30 – Exploiting Parallelism for Intel Processors & Coprocessors - JD Patel

  • Parallelism Growth:
    • Rapid growth in parallelism
    • Enabling advances in parallelism
  • Intel Compilers Key Features:
    • Vectorization, interprocedural optimization (IPO), profile-guided optimization (PGO), high level optimization (HLO), Parallel programming models (Cilk Plus)
    • Other optimization considerations
  • Intel Xeon Phi:
    • Phi hardware overview
    • Phi software overview
    • Phi Programming Models:
    • Native mode
    • Offload mode – synchronous and asynchronous offloads
    • Symmetric mode

11:30 - 12:30 – Performance Tuning, or “Am I ready to move to Xeon Phi™ and how to improve performance once there?” - James Tullos

  • Considerations for tuning
  • Overview of Intel® VTune™ Amplifier XE
  • Efficiency metrics
  • Common problem areas for Intel® Xeon Phi™ performance

12:30 - 1:15 LUNCH (provided on site)

1:15 – 1:45 MKL – Sridevi Allam (REMOTE via conference bridge)

  • Latest features in MKL 11.0
  • Current Work
  • Intel Xeon Phi support
  • Bitwise Reproducibility

1:45 - 2:30 – Intel® Trace Analyzer and Collector and hybrid programming (shared memory/distributed memory) - James Tullos

  • Collecting a trace with thread information
  • How to interpret results of hybrid programs in Intel® Trace Analyzer

2:30 - 4:30 - Hands on work through Intel Xeon Phi labs OR work with OSC developers 1:1 on Intel Xeon Phi systems

  • Hands-on demos with Intel code OR work with participant code

Please bring your own laptops to connect to OSC resources with.




Love it? Hate it? Want to suggest new features or report a bug? We would love to hear from you.

Bug Report Content Suggestions
Compliment Other