Logo icon

Profiling of energy consumption of the high-performance MPI-applications on clusters

Developed for the Nizhny Novgorod Foundation for Education and Research supported by Intel

Purpose: analysis of the energy consumption of open computational packages during their start on a cluster

The project was aimed at identifying the potential possibility of profiling energy consumption by HPC applications performed on the cluster to assess energy demands for the application functions and the MPI library functions. The project was implemented using the Intel VTune™ Amplifier XE 2013 tool, which allowed collecting energy consumption at the single functions level. The conducted studies resulted in the receipt of the profiles for the HPC applications (LAMMPS, NAMD, GAMESS, GROMACS, OpenFOAM) and identification of the functions with the biggest energy consumption.

Specification

Client: The Nizhny Novgorod Foundation for Education and Research supported by Intel
Area of use: Reducing energy consumption for the high-performance applications
Type (platform): Research project, the applications were running on the cluster under Linux OS
Technologies and algorithms in use: Intel VTune™ Amplifier, Intel MPI library, MPICH, MVAPICH, OpenMPI, Bash, computing cluster based on processors Intel XEON E5, GAMESS, GROMACS, LAMMPS, NAMD, OpenFOAM
More information:

project presentation;
bibliography:

the import of a 3D geometry in STEP format, triangulation, export to an STL file

fast multiplication of the block sparse matrixes