This project aims at implementing a GPU accelerated non-linear solver for hyperelastic neo- Hookian models with contact. The project will build on the PolyFEM [11], a C++ and Python finite element library (https://polyfem.github.io/), coupled with the Incremental potential contact method which allows accurate simulation of surface-to-surface contact (https://ipc- sim.github.io/). Currently, PolyFEM relies on multithreading libraries like TBB, the strategy is to reengineer the non-linear solver section of the code to make it suitable for CUDA [10] device functions taking advantage of libraries such as cuBLAS, cuSPARSE, Thrust and external sparse linear solvers with GPU support. Our goal is to obtain a performance improvement of stress tests in comparison with multithreading implementation from low to high end CPUs. The project will be conducted in collaboration with Teseo Schneider (University of Victoria).
GPU accelerated contact simulations(2022 Dec 20).
GPU accelerated contact simulations
-
2022-12-20
Abstract
This project aims at implementing a GPU accelerated non-linear solver for hyperelastic neo- Hookian models with contact. The project will build on the PolyFEM [11], a C++ and Python finite element library (https://polyfem.github.io/), coupled with the Incremental potential contact method which allows accurate simulation of surface-to-surface contact (https://ipc- sim.github.io/). Currently, PolyFEM relies on multithreading libraries like TBB, the strategy is to reengineer the non-linear solver section of the code to make it suitable for CUDA [10] device functions taking advantage of libraries such as cuBLAS, cuSPARSE, Thrust and external sparse linear solvers with GPU support. Our goal is to obtain a performance improvement of stress tests in comparison with multithreading implementation from low to high end CPUs. The project will be conducted in collaboration with Teseo Schneider (University of Victoria).File | Dimensione | Formato | |
---|---|---|---|
Alexander Cristhoper Trujillo Ochoa_Thesis.pdf
accesso aperto
Descrizione: MHPC thesis
Tipologia:
Tesi
Licenza:
Non specificato
Dimensione
2.52 MB
Formato
Adobe PDF
|
2.52 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.