面向CPU+GPU异构计算的SIFT

Parallel Algorithm of CPU and GPU oriented Heterogeneous Computation in SIFT Feature Matching

摘要:

依据图形处理器(GPU)计算特点和任务划分的特点，提出主从模型的CPU+GPU异构计算的处理模式.通过分析和定义问题中的并行化数据结构，描述计算任务到统一计算设备架构(CUDA)的映射机制，把问题或算法划分成多个子任务，并对划分的子任务给出合理的调度算法.结果表明，在GeForce GTX 285上实现的尺度不变特征变换(SIFT)并行算法相比CPU上的串行算法速度提升了近30倍.

Abstract:

According to the basis of features about graphic processing unit(GPU) computation and tasks division, the study tries to bring forward a method of Master/Slave CPU+GPU heterogeneous computation. This paper presents an analysis and definition of the parallel data structures, and a description of the mapping mechanism for computing tasks on compute unified device architecture(CUDA). A logical scheduling algorithm is proposed to divide an issue or algorithm into many sub tasks. The result shows that the speed of SIFT parallel algorithm in the Geforce GTX 285 is about 30 time of the serial algorithm running in the CPU.