OpenCL – The Open Standard for Parallel Programming of Heterogeneous Systems
OpenCL – The Open Standard for Parallel Programming of Heterogeneous Systems
复制标题
作者：
J. Xu
当今的计算机正在依赖并联计算的依赖。处理器正在用多个内核运输，并且正在发现新的计算方法，例如GPGPU（通用图形处理单元）编程[1] [3]。这些新处理技术中的每一个都具有自己的优势，从而使这些技术的合并使用经验。例如，GPGPU允许具有改进的浮点精度[2]进行大规模数据并行计算，而CPU则允许任务并行计算。将这两个结果结合在能够大量平行向量计算的应用程序中，而与最终用户保持互动。
Today's computers are becoming increasing reliant on parallel computing. Processors are shipping with multiple cores, and new ways of computing are being discovered such as GPGPU (General Purpose Graphics Processing Unit) programming [1][3]. Each of these new processing techniques present advantages of their own, thus making the combined use of these technologies empirical. For example GPGPU allows for massive data parallel computations with improved floating point precision [2], while CPU allows for task parallel computation. Combining these two results in applications that are capable of massively parallel vector computations, while remain interactive to end users.