Accelerating a MATLAB Application with Nvidia GPUs: a Case Study for GPU Library Construction
Nicholas John Moore, Miriam E. Leeser · 2009
This abstract covers a case study in using the Nvidia CUDA environment to accelerate lung tumor tracking through template matching, and will discuss how this lays the groundwork for creating high performance GPU accelerated applications through a modular library-based programming interface. The tracking algorithm uses motion-enhanced templates and Pearson's correlation. Issues that arise when using CUDA from within MATLAB are investigated, as well as how the particular application was mapped from MATLAB code to a GPU. The case study is being used to develop a tool that allows more rapid exploration of the tradespaces for GPU applications.