Cargando…

GPU-friendly boundary representation geometry model for simulation

The SFT simulation R\&D group is working on optimizing the performance of the Geant4 particle transport simulation toolkit. This includes porting the simulation software to accelerator hardware such as graphics processing units (GPUs), which often implies major rewrites of simulation components....

Descripción completa

Detalles Bibliográficos
Autores principales: Stan, Eduard George, Gheata, Andrei
Lenguaje:eng
Publicado: 2023
Materias:
Acceso en línea:http://cds.cern.ch/record/2872662
_version_ 1780978625955233792
author Stan, Eduard George
Gheata, Andrei
author_facet Stan, Eduard George
Gheata, Andrei
author_sort Stan, Eduard George
collection CERN
description The SFT simulation R\&D group is working on optimizing the performance of the Geant4 particle transport simulation toolkit. This includes porting the simulation software to accelerator hardware such as graphics processing units (GPUs), which often implies major rewrites of simulation components. One of the current projects targets developing a surface-based geometry model within the VecGeom library, to mitigate the geometry performance overhead observed in GPU simulations with the current 3D-solid modeling approach. The project is in a phase where the base required functionality is implemented, but several areas are still incomplete, in particular providing surface support for the full set of supported 3D-solid primitives. This project targets extending the set of supported 3D solids to be meshed, including the unit test coverage, and integration with the GPU particle transport framework.
id cern-2872662
institution Organización Europea para la Investigación Nuclear
language eng
publishDate 2023
record_format invenio
spelling cern-28726622023-09-28T19:34:19Zhttp://cds.cern.ch/record/2872662engStan, Eduard GeorgeGheata, AndreiGPU-friendly boundary representation geometry model for simulationComputing and ComputersThe SFT simulation R\&D group is working on optimizing the performance of the Geant4 particle transport simulation toolkit. This includes porting the simulation software to accelerator hardware such as graphics processing units (GPUs), which often implies major rewrites of simulation components. One of the current projects targets developing a surface-based geometry model within the VecGeom library, to mitigate the geometry performance overhead observed in GPU simulations with the current 3D-solid modeling approach. The project is in a phase where the base required functionality is implemented, but several areas are still incomplete, in particular providing surface support for the full set of supported 3D-solid primitives. This project targets extending the set of supported 3D solids to be meshed, including the unit test coverage, and integration with the GPU particle transport framework. CERN-STUDENTS-Note-2023-177oai:cds.cern.ch:28726622023-09-27
spellingShingle Computing and Computers
Stan, Eduard George
Gheata, Andrei
GPU-friendly boundary representation geometry model for simulation
title GPU-friendly boundary representation geometry model for simulation
title_full GPU-friendly boundary representation geometry model for simulation
title_fullStr GPU-friendly boundary representation geometry model for simulation
title_full_unstemmed GPU-friendly boundary representation geometry model for simulation
title_short GPU-friendly boundary representation geometry model for simulation
title_sort gpu-friendly boundary representation geometry model for simulation
topic Computing and Computers
url http://cds.cern.ch/record/2872662
work_keys_str_mv AT staneduardgeorge gpufriendlyboundaryrepresentationgeometrymodelforsimulation
AT gheataandrei gpufriendlyboundaryrepresentationgeometrymodelforsimulation