Gerador automático de código Python para processamento heterogêneo de imagens em OpenCL por meio da biblioteca VisionGL

Abstract

Heterogeneous computing can be defined as a set of computing devices that process information in different ways to work together. Some applications, like games, use heterogeneous computing (which in this case, consists on using CPU and GPU) to render more complex or realistic scenes. Image processing applications can also take advantage using the amount of cores that GPU offers to execute the processes more quickly. To use this processing power, there are some frameworks like OpenCL, that helps to use parallel computing, hiding some complexities from the developer. This work describes the development of a Python wrapper in a way it integrates the VisionGL library. An automatic code generator is developed to automatically generate the Python wrapper whenever needed and a comparative performance test between the C++ and Python versions of VisionGL is performed, which showed significant advantage to the Python version that was executed.