Hands-On GPU Programming with Python and CUDA

OpenCL and PyOpenCL

CUDA isn't the only language that can be used to program a GPU. CUDA's most major competitor is called Open Computing Language, or OpenCL. Where CUDA is a closed and proprietary system that will work exclusively on only NVIDIA hardware, OpenCL is an open standard that's developed and supported by the nonprofit Khronos Group. OpenCL can be used to program not only an NVIDIA GPU, but also AMD Radeon GPUs and even Intel HD GPUs—most major technology companies have committed to supporting OpenCL in their products. Additionally, the author of PyCUDA, Professor Andreas Kloeckner of UIUC, has written another excellent (and free) Python library called PyOpenCL, which provides an equally user-friendly interface to OpenCL, with ...