DelftBlue course 2 days (SuperComputer)
DelftBlue course 2 days (SuperComputer)

Program

27 June 2022

09:30 - 09:45

Walk in, grab a coffee

09:45 - 10:00

Welcome to DHPC (Kees Vuik)

What is a SuperComputer

10:00 - 10:30
  • Login nodes, CPU vs. GPU nodes, network
  • Global vs. local vs. remote storage
  • Batch processing/scheduling
  • What is a Supercomputer not?
10:30 - 11:00

What does DHPC offer, and how?

  • Organization and user interaction
  • Accounting and usage policies
  • Using DelftBlue for education purposes
  • HPC Training
11:00 - 12:30

Practical use of DelftBlue

  • remote access, queuing, accounting, data transfer
  • OpenOnDemand vs. ssh access
  • practical exercise
12:30 - 13:30

Lunch

13:30 - 14:15

Programming techniques for SuperComputers

  • Using optimized libraries
  • Introduction to OpenMP
  • Introduction to MPI
  • Native multi-threading and GPU support in C++20 and Fortran 2008
  • Introduction to CUDA
14:15 - 15:00

Hands-on session (bring your own work or try one of our examples)
15:00

Wrap-up and open discussion

28 June 2022

09:30 - 09:45

Walk in, grab a coffee

09:45 - 10:15

Quick recap of Day 1: DHPC, policies, login to DelftBlue
10:15 - 10:30

Overview of DelftBlue

  • Node types, storage, network…
  • Slurm on DelftBlue
  • Special features
10:30 - 12:30

Workflow demos

12:30 - 13:30

Lunch

13:30 - 14:30

Performance analysis

  • performance modelling (roofline)
  • measuring: gprof and using performance counters
  • top-down performance analysis
14:30 - 14:45

Coffee break

14:45 - 15:30

Hands-on session (bring your own code or try out one of our examples)
Open discussion round
15:30
  • what works well/should be improved?
  • what kind of trainings should DHPC offer?
  • using DelftBlue for education