Speeding your graduation with faster simulations - An introduction to GPU computing

Neraaj Sood

The Edward S. Rogers Sr. Department of Electrical and Computer Engineering (Electromagnetics)

Abstract:

Scientific computing using graphics processing units (GPU) has seen a tremendous rise in recent years due to improvements in programmability and increase in the capability of modern GPUs. The past decade has seen the GPUs exceed CPUs in arithmetic throughput and memory bandwidth. Since 2003 GPUs have found use in non-graphical applications where in the beginning developers used various highlevel shading languages such as DirectX, OpenGL and Cg. These languages required programmers to have a thorough understanding of the graphics APIs and GPU architecture. However with the introduction of the Compute Unified Device Architecture (CUDA) the programming model has been greatly simplified and extensions of high level programming languages can now be used to program GPUs. This talk will provide a brief introduction into computing with a GPU co-processor with the help of a simple example of parallelizing a 1-D FDTD code.

Biography:

Neeraj Sood is a Ph.D. student in the Electromagnetics Group under the  supervision of Prof. Costas Sarris.  He studies how wireless signals  propagate in various environments using Ray-Tracing.