Performances of parallel codes

Performances of parallel codes can be evaluated by means of dedicated applications like profilers. A profiler is a software specifically designed to analyze the behaviour of parallel programs in order to improve and optimize them. There are many profilers around, some of them commercial, others open source. It is higly recommended to use a toolset for analyzing own applications, for such a good practise will make the code faster and more efficient, preventing waste of time.


In this section we present Scalasca, an open source toolset developed as a joint project of the Jülich Supercomputing Centre and theGerman Research School for Simulation Sciences. Scalasca has been installed on CRESCO HPC system for the Linux platform. The current installation supports GNU, Intel and PGI compilers with many MPI flavours. Moreover, users can also use Scalasca on CRESCO for analyzing hybrid (MPI + OpenMP) codes.

How to use Scalasca on CRESCO

 
Cookies Policy