CERN Accelerating science

Implementation and test of MLFit application using OpenMP and MPI parallel technologies

Date published: 
Monday, 8 August, 2011
Document type: 
Summer student report
Author(s): 
R. Caravita
The work succeeded to implement a data analysis application using MPI techniques for parallelization. The application is used for the likelihood function evaluation in maximum likelihood fits. Implementations are already parallelized for shared memory using OpenMP. Tests were done in different hybrid configuration MPI and OpenMP on a single dual-socket node, achieving good scalability using different models. Also part of the activity was devoted to develop and understand tools for profiling the application and checking the results.
Technical document file: