Profiling workshop

Considering the growing need for computational power in scientific applications, the optimization of computation codes has become a key point in numerical simulation. The step prior to any optimization action is to analyze the performance of the code (profiling) in order to identify the critical points of the code in its execution environment. To do so, there are many analytical tools that should be choosen and used according to the needs.

Sequential code profiling

The Mesocentre in collaboration the CNRS groupe Calcul hosted Laurent Gatineau (Application Support, NEC HPC Europe) for a first workshop on the profiling of sequential computation codes. This training took place on Wednesday, June 11th 2014 at Ecole Centrale Paris. The goal of this workshop was to give the participants good bases to start profiling.

This workshop included :

an introduction to the profiling of sequential codes; a presentation of the gprof and PAPI tools; practical work on a machine using these tools. The workshop supports are available below (in french) :

OpenMP code Profiling

The Mesocentre in collaboration the CNRS groupe Calcul welcomed Brian Wylie (Jülich Supercomputing Center) for a second workshop on the profiling of computer codes dedicated to OpenMP applications. This training took place on Wednesday 10th and Thursday 11th December 2014 at Ecole Centrale Paris.

The goal of this workshop was to give participants good bases for profiling OpenMP applications.

This workshop included :

an introduction to the profiling of OpenMP codes; a presentation of tools like Scalasca; practical work on a machine using these tools. The workshop supports are available below: