基于移动多核GPU的并行二维DCT变换实现方法  

Parallelization of 2D-DCT based on Mobile Multicore GPU

作　　者：龚若皓 杨斌[1] 

机构地区：[1]西南交通大学信息科学与技术学院,四川成都610031

出　　处：《成都信息工程学院学报》2015年第1期22-26,共5页Journal of Chengdu University of Information Technology

摘　　要：传统的基于CPU的串行程序所实现的二维DCT变换算法时间复杂度高变换效率低,难以满足许多应用的实时要求。特别是在当代以嵌入式处理器为核心的移动端信息处理终端,有限的CPU性能更加难以实现快速的DCT变换。值得欣慰的是新一代嵌入式处理器提供了支持GPGPU技术的GPU,为解决复杂的移动计算问题提供了高效的并行化解决途径。基于最新的ARM Cortex-A15内嵌GPU Mali-T604及Open CL框架设计实现了一种针对二维DCT变换的并行化加速方案并实测了优化效果,实验结果表明文中的并行方案能够提高二维DCT变换的效率,在输入数据量足够大的条件下能够达到近20倍的加速比。The traditional algorithm of 2D-DCT is based on CPU computing and it has high time complexity and its computational efficiency is low especially on mobile information terminal based on embedded processor whose comput- ing power is limited, so it＇s hard to meet requirement of real-time system. The new generation of embedded proces- sor integrates mobile GPU that supports GPGPU technology which can provide efficient parallel computing solution for complex mobile computing problem. This paper researches the parallelization solution of 2D-DCT based on the lat- est ARM Cortex-A15 processor and Mall-T604 GPU using OpenCL framework. The result of experiment shows that when the scale of input data is large enough, the computational efficiency of 2D-DCT on mobile devices can be accel- erated 20 times approximatively using the solution of this paper.

关 键 词：计算机应用技术 异构计算 二维DCT变换 移动多核GPU GPGPU OPENCL 并行化 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]