Parallel Programming Workshop 2017

Speedup Your Code on Next Generation Intel® Xeon Phi™ and Xeon™ Processors at HLRN

Current Intel® Xeon Phi™ processors (Knights Landing - KNL) and next-generation Intel® Xeon™ CPUs feature increasing core counts, integration of high-bandwidth memory (Xeon Phi), and wider SIMD units (AVX-512). Existing codes need to be modernized to take full advantage of the architectural features of these CPUs. Newly developed code should be designed to do so from the beginning.

We invite code developers to the workshop on advanced code optimization techniques on state-of-the-art and tomorrow's multi- and many-core processors.

The Hackathon provides the unique opportunity to start or continue the optimization work with the assistance of experts from ZIB, Intel and Cray. Therefore, we highly recommend that the attendees bring their own codes, kernels or proxy benchmarks to the workshop. For the Hackathon, the HLRN resources at ZIB, i.e. the Cray TDS with 80 Intel KNL nodes and the Cray XC40/XC30 system with Intel Haswell/Ivybridge nodes are open for code evaluation, optimization and benchmarking. Attendees with a valid HLRN account can transfer their codes to the HLRN system in Berlin in advance.

DATE: Monday, March 20, 2017, 1pm - Wednesday, March 22, 2016, 12pm

LOCATION: Berlin, Zuse Institute Berlin (ZIB), Lecture Hall, (Eingang Foyer Rundbau)

REGISTRATION & TRAVEL INFORMATION: Please use the Registration Page which provides information about the location of the Zuse Institute in Berlin.

Prerequisites and Additional Information

  • Do not forget to bring your code to this workshop.
  • Please bring your own laptop for the Hackathon sessions.
    • The laptop needs to have WLAN support.
    • For attendees without a HLRN account we provide temporary accounts. Please have a tool ready for private/public key generation.
    • A ssh client should be installed on you laptop for login into the HLRN system. For Windows we recommend MobaXterm.
    • To use the GUIs of various tools your laptop OS should have a support for X11 compatible client connections. For Windows, MobaXterm comes with a X11 server.

Additional Material


DAY 1: Mon, March 20, 2017
12:30 - 13:00 Registration
13:00 - 13:15 Opening (Th. Steinke, ZIB)
13:15 - 13:45 The Intel Xeon Phi Processor (KNL) Architecture for Programmers (Chris Dahnken, Intel)
  AVX-512 (Chris Dahnken, Intel)
13:45 - 14:30 Using the Intel Compiler: OpenMP 4.x SIMD (Michael Klemm, Intel)
14:30 - 15:00 Using the Cray Compiler and Process Placement with 'aprun' (Stephen Sachs, Cray)
15:00 - 15:15 Break
15:15 - 16:00 Advanced Features of Intel VTune for Code Analysis (Michael Klemm, Intel)
16:00 - 16:30 Preparation for Hackathon: Access to Cray KNL System (Thomas Steinke, ZIB)
16:30 - 18:00 Hackathon - Optimize your code, ask the experts.
18:00 Wrap-up Day 1
DAY 2: Tue, March 21, 2017
08:30 - 09:00 Registration
09:00 - 09:20 Optimization Cycle: Example PALM Hackathon 2016 (Matthias Noack, ZIB)
09:20 - 10:00 Roofline Model (Matthias Noack, ZIB)
10:00 - 10:30 Intel Advisor (Michael Klemm, Intel)
09:00 - 11:00 Hackathon (cont.)
11:00 - 11:15 Break
11:15 - 11:45 Q&A
11:45 - 12:45 Hackathon (cont.)
12:45 - 14:00 Lunch
14:00 - 15:00 Hackathon (cont.)
15:00 - 15:30 Cray Reveal for Integrating OpenMP Threading (Stephen Sachs, Cray)
15:30 - 16:00 Break
16:00 - 16:45 HBM Analysis Assistance in CrayPAT (Heidi Poxon, Cray)
16:45 - 18:00 Hackathon (cont.)
18:00 Wrap-up Day 2
18:30 Abendessen in der "Luise" (, Königin-Luise-Straße 40-42, 14195 Berlin (nahe U-Bahn Dahlem Dorf)
DAY 3: Wed, March 22, 2017
09:00 - 11:30 Hackathon (cont.)
11:30 - 12:00 Q&A, Summary
12:00 Wrap-up Day 3

Created by ThomasSteinke - 18 Mar 2017

Last modification: ThomasSteinke - 31 Mar 2017 13:57 / 3 years, 4 months, 2 weeks ago. (Version: 20)

Topic attachments
I Attachment History Action Size Date Who Comment
Unix shell scriptsh r2 r1 manage 1.7 K 2017-03-22 - 02:40 ThomasSteinke  
Unix shell scriptsh r1 manage 1.9 K 2017-03-21 - 14:12 ThomasSteinke  

No permission to view NewsCenter.WebTopMenu

we are moving ...

Norddeutscher Verbund für Hoch- und Höchstleistungsrechnen

Back to top of page

2001-2020 © HLRN Logo Norddeutscher Verbund für Hoch- und Höchstleistungsrechnen (HLRN). - Ideas, requests, problems regarding this site? Send feedbackPowered by TWiki