Multi-GPU cluster wave propagation and OpenGL visualization
Institutionen fur informationsteknologi, Uppsala Universitet
Uppsala Universitet, 2014
The inherent issues of properly deploying finite difference calculations onto GPUs are described and solutions are suggested. A speedup of 60x is achieved over the CPU version. Four visualization methods were implemented using OpenGL and compared in terms of the clarity of their visual result. A combination of hedgehogs and slices was deemed to give the best result.
March 15, 2014 by hgpu
