NCI ARM Workshop: Tips for developing and optimising scientific HPC applications

Today’s ever-advancing technology platforms offer huge potential for completing faster and more advanced work.

However, missing the vital warning signs of inefficient code behaviour will hinder your performance at any scale.

ARM Allinea tools are designed to help you get to the root causes of application bottlenecks quickly.

In this training, ARM Allinea experts will provide hands-on advice, customised for NCI users, for parallelisation and development of more robust code.

The workshop will also demonstrate the use of the tools for getting the most out of NCI's compute resources by identifying your biggest priorities for optimization across commonly occurring computation, I/O, memory and workload balance issues.

Sessions will be facilitated by ARM Allinea's Staff Applications Engineer, Florent Lebeau.

NOTICE: The following are recommended pre-requisites for this workshop:
- Knowledge of C/C++ or Fortran programming
- Knowledge of parallel programming: MPI, OpenMP or CUDA
- Knowledge of HPC environment: environment modules, job submission scripts, makefiles

Attendees MUST bring their own laptop with WiFi connectivity. Ethernet connectivity is available; please bring your own cable.

Your laptop must have a terminal application such as PuTTY or MobaXterm on Windows or Terminal on Mac.

Attendees will need to install the ARM Allinea Remote Client before commencement of the workshop.

PROGRAM

DAY 1: Monday 11 December 2017

08:30 Registration
09:00 Introduction to NCI (presented by NCI staff)
09:30 Introduction to ARM Allinea tools
10:00 Maximize application efficiency

10:30 Morning tea

11:00 Maximize application efficiency (continued)
11:30 Fix an application crash

12:30 Lunch

13:30 Optimize memory accesses
14:30 Debugging destructive memory bugs

15:00 Afternoon tea

15:30 Debugging destructive memory bugs (continued)
16:00 Resolve workload imbalances and optimise I/O

DAY 2: Tuesday 12 December 2017

09:00 Debugging deadlocks and race conditions
10:00 Detect memory leaks

10:30 Morning tea

11:00 Detect memory leaks (continued)
11:30 Adopting professional best practices

12:30 Lunch

13:30 Debugging GPU problems
14:30 Resolve computation errors

15:00 Afternoon tea

15:30 Resolve computation errors (continued)
16:00 Free hands-on and questions
16:30 Wrap-up and conclusion

Tea, coffee, morning tea, afternoon tea and lunch will be provided on both days.

Participants will be required to arrange their own transport and accommodation in Canberra. University House is a hotel on campus at ANU, and the venue is close to numerous hotels in New Acton and Canberra CBD. For more information about getting around on campus visit https://services.anu.edu.au/campus-environment/transport-parking

If you have any enquiries please email enquiries@nci.org.au

www.nci.org.au