摘要
本文为满足G级像素帧的实时性处理需求,针对DCT变换计算量大和常规处理中并行度不足的问题,提出一种基于SIMD PE阵列的DCT数据并行实现方法.该方法因PE阵列本身所具有的可裁减特性,可应用于不同并行度需求的嵌入式系统中.文中提出一种基于PE标识的数据并行操作方式,不但解决了局部计算中的"PE自治"问题,又省去了数据寻址时间开销.该操作方式规则、简洁,满足SIMD操作规则性强的要求,符合并行处理技术的发展方向.
In order to meet the challenge of real-time processing of G-level pixel frame,to solve such problems as high-volume calculation and lack of parallelism in conventional DCT, a parallel computation method implemented on SIMD PE array for DCT transform is presented in this paper. With the inherited customizability of PE array, this method can be applied to various embedded application systems regardless of their requirements on parallelism. A specific data parallel operation method using PE identifier is proposed. It not only solves the problem of "PE autonomy" in local operation,but also eliminates the cost of time in data addressing. This well-structured and compact data operation satisfies the high-regularity required by SIMD operation, and conforms to the trend of parallel processing technology.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2009年第7期1546-1553,共8页
Acta Electronica Sinica
基金
国防微电子预研项目(No.41308010203)