Scientific computation through a GPU

Abstract

A personal computer's graphics processing unit, or GPU, has been the seed of a growing interest in the academic and research communities of recent months. This paper investigates current technology that enables a GPU to process and solve linear algebra computations, in particular, matrix operations. Matrix operations of linear algebra are the basis of scientific computation, often used in modeling data and describing the forces of the universe. The author wished to compare the speed of the computation through the CPU and the GPU. Utilizing NVIDIA's CUDA technology, they demonstrated that calculations are preformed considerably faster through the GPU than through the CPU. The authors concluded that all computation in the research community has the potential to run significantly faster than current CPU's allow. ©2008 IEEE.