| 引用本文: | 万江华,刘胜,周锋,等.具有高效混洗模式存储器的可编程混洗单元.[J].国防科技大学学报,2011,33(6):31-35.[点击复制] |
| WAN Jianghua,LIU Sheng,ZHOU Feng,et al.A Programmable Shuffle Unit with the Efficient Shuffle Pattern Memory[J].Journal of National University of Defense Technology,2011,33(6):31-35[点击复制] |
|
| |
|
|
| 本文已被：浏览 7760次 下载 5893次 |
| 具有高效混洗模式存储器的可编程混洗单元 |
| 万江华, 刘胜, 周锋, 王耀华, 陈书明 |
|
(国防科技大学 计算机学院，湖南 长沙 410073)
|
| 摘要: |
| 为向量DSP提出并设计了一种具有高效混洗模式存储器的可编程混洗单元，该结构能够使应用程序的混洗指令在执行时和通用寄存器或访存带宽等系统的关键资源分离，从而提高混洗单元的执行效率。采用区分不同混洗粒度和元素索引等方法对混洗模式所对应的开关矩阵进行了压缩，我们的存储效率高于当前其他方案。该混洗单元具有可编程的特点，应用程序所需要的混洗模式可以提前由程序员编写并通过DMA等途径加载。对该混洗单元进行了VLSI实现及性能评测，结果显示此方案能够在给系统带来0.6%的额外面积开销基础上使应用程序的性能提升7.4%～17.4%。 |
| 关键词: 向量DSP 混洗单元 混洗模式存储器 混洗指令 存储效率 |
| DOI： |
| 投稿日期：2011-09-20 |
| 基金项目:国家“核高基”重大专项资助（2009ZX01034-001-001-006)；国家自然科学基金资助项目（61070036、61133007）；国家863高技术资助项目（2009AA011704） |
|
| A Programmable Shuffle Unit with the Efficient Shuffle Pattern Memory |
| WAN Jianghua, LIU Sheng, ZHOU Feng, WANG Yaohua, CHEN Shuming |
|
(College of Computer, National Univ. of Defense Technology, Changsha 410073, China)
|
| Abstract: |
| This paper presents a programmable shuffle unit with the efficient shuffle pattern memory for vector DSPs. The shuffle instructions can efficiently execute without occupying the system’s key resource such as the general registers or the memory bandwidth. We compress the switch-matrix by differentiating the shuffle granularity and indexing the elements. The memory efficiency of our scheme is higher than the state-of-art methods. Programmers can design the shuffle patterns ahead of time and load them into the shuffle pattern memory by the DMA or other ways. Experimental results show that our scheme can reduce the execution cycles by 7.4%～17.4% for the applications with the shuffle instruction requirement, at the cost of 0.6% additional chip area. |
| Keywords: vector DSP shuffle unit shuffle pattern memory shuffle instructions memory efficiency |
|
|
|
|
|
