PRACE Course: HPC Code Optimisation Workshop 2020

Learning Goals

Through a sequence of simple, guided examples of code modernization, the attendees will develop awareness on features of multi and many-core architecture which are crucial for writing modern, portable and efficient applications.

A special focus will be dedicated to scalar and vector optimizations for the latest Intel® Xeon® Scalable processor, code-named Skylake, utilized in the SuperMUC-NG machine at LRZ.

The workshop interleaves frontal and practical sessions. Here is a preliminary outline:

Day 1

  • Introduction to systems and software
  • Introduction: Code modernization approach
  • Basic compiler optimizations
  • Vectorization
  • Skylake optimizations

Day 2

  • Introduction to roofline model
  • Intel® Advisor analysis
  • Intel® Math Kernel Library (MKL) and other libraries

Day 3

  • Introduction to Intel® VTune™ Amplifier
  • Introduction to Intel® Application Performance Snapshot (APS)
  • LIKWID ( “Like I Knew What I’m Doing.”)  Performance Tools
  • Q&A Session


The workshop is a PRACE training event organized by LRZ in cooperation with Intel and RRZE.

Agenda


Main Session

1st day morning
(10:00-12:00)

Intro (Volker/Gerald/Thomas)

Vectorization  (Igor)

1st day afternoon

(13:00-16:00)

Intel Compiler (Igor)

2nd day morning

(10:00-12:00)

Roofline Model and Advisor (Fabio)

2nd day afternoon

(13:00-16:00)

MKL & libraries (Gennady)

3rd day morning

(10:00-12:00)

VTune / APS  (Michael)

3rd day afternoon

(13:00-16:00)

LikWid  (Carla, Thomas) 

3rd day Q&A

All

Presenters

  • Fabio Baruffa (Intel)
  • Gennady Fedorov (Intel)
  • Mathias Gerald (LRZ)
  • Thomas Gruber (RRZE)
  • Carla Guillen (LRZ)
  • Michael Steyer (Intel)
  • Igor Vorobtsov (Intel)

Assistants

  • Momme Allalen (LRZ)
  • Volker Weinberg (LRZ)

Slides and Exercises

Day 1

Interesting external links:

Day 2 Morning Session

Interesting external links:

Day 2 Afternoon Session

Interesting external links:

Day 3 Morning Session

Day 3 Afternoon Session


PRACE Survey

  • Please fill out the PRACE online survey under https://tinyurl.com/cow-survey
  • This helps us and PRACE to
    • increase the quality of the courses,
    • design the future training programme at LRZ and in Europe according to your needs and wishes,
    • get future funding for training events,
    • shape the future system architecture at LRZ.

Accounts

  • Each user gets an own account: m85q00XX during registration on the first day of the workshop.
  • $HOME with 10GB capacity (NFS)
  • Contact the registration desk during the registration via private Zoom chat:
    • Thomas Gruber
    • Gerald Mathias
    • Carla Guillen
  • Please prepare your passport ID number or identification card ID number
  • The accounts will be valid until 11 June 2020, 23:59 CEST

Recommended Access Tools

System

  • Exercises will be done on the RRZE Meggie cluster, see: https://www.anleitungen.rrze.fau.de/hpc/meggie-cluster/
    • 728 compute nodes
    • 2x Intel Xeon E5-2630 v4 @ 2.2GHz (10 cores per socket + SMT)
    • 64 GB RAM
    • Intel OmniPath interconnect: ≈ 100GBit/s bi-directional
    • Lustre-based parallel filesystem: ≈ 1 PB capacity
    • Batch system: Slurm
    • Vendor: Megware
  • Connect to dialog server: ssh m85q00XX@cshpc.rrze.fau.de
  • Connect to frontends of meggie: ssh meggie


  • Select the following options to use NoMachine NX with TDE (see: NoMachine-Howto-v2.pdf on how to use the (German) Windows NoMachine Version in detail)
    • Protocol: SSH
    • Host: cshpc.rrze.fau.de
    • Port: 22
    • Authentication: by Password
    • Proxy: None
    • User name: m85q00XX
    • Password: (you will get your password during registration on Monday)
    • Create new user-defined session
    • Include the following command to be run: starttde
    • This will start a TDE Desktop for you
    • Further information is available under https://www.anleitungen.rrze.fau.de/hpc/dialogserver/

Information on ZOOM

Information on further HPC courses