Performance Portable Implementation of Octree Algorithm

Institut

Lehrstuhl für Hubschraubertechnologie

Typ

Bachelorarbeit / Semesterarbeit /

Inhalt

theoretisch /  

Beschreibung

 The In-House Aerodynamic Solver of the Institute for Rotorcraft and Vertical flight (VORTEX) is

a Fast Multipole Method accelerated solver. In current implementation it is rely on octree and interaction list

construction on the CPU and GPU offloading for multipole kernels. We would like to implement an octree

algorithm that is able to construct the octree and interaction list also on the GPU. Ideally this octree would be a

distributed octree with MPI support so that the Aerodynamic Solver (VORTEX) can also run on multiple GPUs or

CPUs.

Möglicher Beginn

sofort

Kontakt

Kursat Yurt
Raum: MW 2711
kursat.yurttum.de

Ausschreibung