Title: Using GPUs to solve PDEs in Biological Problems

Abstract:
Researchers today have a variety platforms in which to solve mathematical problems associated with their research. These platforms may range from single core CPUs, to clusters, to massively parallel leading class super computers. Unfortunately, not every researcher has access to or money for large computing systems. At the same, CPUs are not getting faster but incorporating multiple cores thus requiring parallel implementations in order improve performance. Parallel to multiple core CPUs are the Graphics Processing Units (GPU) which provide specialized parallel performance.

When tuned, commodity level GPUs have been shown to have similar performance to cluster based systems but at a much lower hardware cost. However, like any parallel implementation the devil in the details. In this talk we will present work done in using the GPU to solve PDEs associated with several biological problems. Our goal in this work has been not to just adapt CPU based algorithms to the GPU but to compare and contrast the implementations and understand the strengths and weaknesses. We draw some of our examples from PDEs used in an advection-reaction-diffusion system.