Cube GUI User Guide  (CubeGUI 4.6-rc1, revision cfc8481d)
Introduction in Cube GUI and its usage

OpenMP Serialisation Efficiency

OpenMP Serialisation Efficiency indicates serialisation within OpenMP regions across MPI ranks (e.g. time in barriers, critical sections, atomics, etc.) and can be computed as follows:

OpenMP Serialisation Efficiency = max( runtime in OpenMP without serialization ) / max( runtime in OpenMP without management overhead )

If management overhead is not known due to missing Scalasca analysis, total runtime without management overhead is equal to total runtime. OpenMP Serialisation Efficiency computed per MPI rank and average across MPIs is shown. Moreover, MIN and MAX values are also available by click on metric bar. Statistics shows if there is considerable difference in serialisation across MPI ranks.


Cube Writer Library    Copyright © 1998–2021 Forschungszentrum Jülich GmbH, Jülich Supercomputing Centre
Copyright © 2009–2015 German Research School for Simulation Sciences GmbH, Laboratory for Parallel Programming