Cargando…

Optimising and Extending A Single-Particle Tracking Library For High Parallel Performance

SixTrackLib is a library for performing tracking simulations on highly parallel systems such as shared memory multi-core processors or graphical processing units (GPUs). Its single-particle approach fits very well to parallel implementations with reasonable base-line performance, making such a libra...

Descripción completa

Detalles Bibliográficos
Autores principales: Schwinzerl, Martin, De Maria, Riccardo, Paraschou, Konstantinos, Bartosik, Hannes, Iadarola, Giovanni, Oeftiger, Adrian
Lenguaje:eng
Publicado: 2021
Materias:
Acceso en línea:https://dx.doi.org/10.18429/JACoW-IPAC2021-THPAB190
http://cds.cern.ch/record/2781835
Descripción
Sumario:SixTrackLib is a library for performing tracking simulations on highly parallel systems such as shared memory multi-core processors or graphical processing units (GPUs). Its single-particle approach fits very well to parallel implementations with reasonable base-line performance, making such a library an interesting building block for various use cases, including simulations covering collective effects. We describe the optimisations applied to SixTrackLib to improve its performance on its main target platforms and the associated performance gain. Furthermore we outline the technical interfaces and extensions implemented to allow its use in a wider range of applications and studies.