Vectorization of SPHinXsys, a strong-coupling, meshless multi-physics and AI-aware library

Applicant

PD Dr. Xiangyu Hu
Institute of Aerodynamics and Fluid Mechanics
Technical University of Munich

Project Overview

SPHinXsys is an acronym from Smoothed Particle Hydrodynamics (SPH) for industrial complex systems (https://github.com/Xiangyu-Hu/SPHinXsys and https://www.sphinxsys.org). It provides open-source C++ APIs for physically accurate simulation and aims to model and optimize coupled industrial dynamic systems including fluid, solid, multi-body dynamics and beyond with SPH, a meshless computational method using particle discretization, and AI-aware optimization algorithms.

The long term objective that we are perusing within the framework of KONWIHR is to develop a hybrid high-performance single- and multi-node SPHinXsys as an integrated simulation and optimization tool. The specific objective of this 3-month short term project, as a first step working within KONWIHR, is to accelerate SPHinXsys with explicit vectorization which is more scalable and compatible for complex libraries than the implicit approaches using compiling switches and OpenMP SIMD directions.