USING GPU COMPUTING FOR THE SOLUTION OF A CDEM PROBLEM

Abstract eng:
In this paper we introduce an approach for the solution of a coupled discrete and finite element method (CDEM) designed for effective implementation on a graphics processing unit (GPU). A static problem is solved using dynamic relaxation while an optimum load-time history keeps the inertial forces small, thus eliminating the need for damping. Explicit time integration using central difference method is used. Each element is integrated independently and potential interelement contacts are enforced using penalty forces. This allows to avoid the need for forming global characteristic matrices and leads to a formulation well suited to GPU processing. To test the stability and accuracy of the proposed method, we start with a serial CPU implementation and examine a three-point bending test with the beam composed of dicrete deformable blocks (represented by traditional finite elements) connected by linear springs. The CPU code will serve as the starting point of the GPU implementation.

• Export as: BibTeX | MARC | MARCXML | DC | EndNote | NLM | RefWorks
• Add to your basket