Abstract:
The finite deformation elasticity is a theory that describes the capability of the elastic materials undergoing deformations. The finite element method (FEM) is constructed to solve problems based on this theory. The FEM method subdivides the whole problem domain into simpler parts and obtains the approximate results by connecting these simpler parts over subdomains. Solving these problems in real life situations require significantly high computing power, highlighting the need for high performance computational devices in order to accelerate the calculation process. Altera announced industry’s first OpenCL framework for FPGA devices. This tool combines the FPGA with the OpenCL standard to construct powerful system acceleration. In this thesis, an OpenCL solution for finite deformation elasticity is implemented on Altera manufactured Cyclone V SoC development kit. The Cyclone V SoC contains the hard processor system with integrated ARM processor and FPGA, allowing for the host program of the OpenCL application to be executed on the ARM processor and use FPGA’s parallel performance capability to run the OpenCL kernel. The OpenCL kernel is developed to concurrently calculate all the deformation gradient tensors for all elements and a comparison benchmark is conducted to compare the execution time and power consumption between FPGA and GPU setups. The results show that FPGA is 3.5 times faster than GPU and consume significantly lower power.
