Speaker
Leonardo Dominguez
(Departamento de Computación de Alta Prestación - Comisión Nacional de Energía Atómica)
Description
In this work we introduce different parallelization schemes implemented in the AIRES (AIR-shower Extended Simulations) software, in order to perform simulations, without thinning algorithm, in HPC clusters.
The AIRES's particle stack was modified to define a new structure allowing its parallelization using MPI library.
Adopting this new structure, three different parallelization tactics were implemented according to how particles are transferred between the working nodes: 1) Transfer based on the amount of particles stored in the stacks, 2) Transfer based on the energy of particles stored in the stacks 3) Transfer based on the energy of particles stored in the stacks, with decisions according the characteristics of the particle's nucleus type.
As part of this paper will be present a comparison of the obtained results between the most performant parallelized version of AIRES and original version of AIRES, considering longitudinal and lateral profiles of vertical showers induced by Fe primaries of $ 10^{16.75} eV $.
Towards the end of this work we will include an analysis of performance results of each parallelization tactic evaluated by different simulations of vertical showers whose energies are between $ 10^{15.75} eV $ and $ 10^{18.75} eV $.
Registration number following "ICRC2015-I/" | 673 |
---|---|
Collaboration | -- not specified -- |
Authors
Diego Melo
(Instituto de Tecnologías en Detección y Astropartículas, (CNEA, CONICET, UNSAM))
Gastón Aguilera
(Departamento de Computación de Alta Prestación - Comisión Nacional de Energía Atómica)
Juan Bertinat
(Departamento de Computación de Alta Prestación - Comisión Nacional de Energía Atómica)
Leonardo Dominguez
(Departamento de Computación de Alta Prestación - Comisión Nacional de Energía Atómica)
Luis Nieto
(Departamento de Computación de Alta Prestación - Comisión Nacional de Energía Atómica)
Roberto Bevilacqua
(Departamento de Computación de Alta Prestación - Comisión Nacional de Energía Atómica)
Roque Iozzo
(Departamento de Computación de Alta Prestación - Comisión Nacional de Energía Atómica)
Sebastián Garavano
(Instituto de Tecnologíias en Detección y Astropartículas, (CNEA, CONICET, UNSAM))