2021 Code Performance Series: From analysis to insight
Performance analysis is at the core of the development of exascale software – to understand why software performs (or not) is the basis of any informed improvement of that same code.
N8 CIR is collaborating with VI-HPS, Durham University’s Department of Computer Science and DiRAC on an EPSRC-funded ExCALIBUR Knowledge Integration Activity that will help translate performance data into language that domain specialists can understand and use.
The activity is a set of 7 tutorials and practical workshops that will start in January 2021. There will be one session each month, featuring a morning lecture followed by a practical, hackathon-type session in the afternoon where you can implement the material from the morning session.
21 January – Introduction
Organisers introduce concept; each team introduces their code base and ambitions; some basic profiling methodology; initial basic profiling of user codes.
18 February – Parallel Profiling
Introduction to Scalasca/Score-P/CUBE; primary focus on OpenMP and MPI
11 March – Trace Collection
MPI & OpenMP trace collection and analysis with Scalasca/Score-P/Vampir
15 April – Parallel Correctness
MPI & OpenMP correctness verification with MUST/Archer
20 May – User workshop
Feedback on achievements so far; roundtable on next workshop steps; presentation by workshop participants how they present their performance data traditionally towards their communities (home-made performance data collection and presentation)