基于上下文自适应算术编码的可重构配置信息压缩算法

摘要

针对可重构计算系统中配置信息压缩问题,提出了一种基于上下文自适应算术编码的配置信息压缩算法.该算法采用动态调整参数的方法逼近符号概率,采用移位操作消除算术编码的乘除法,在压缩和解压过程中引入上下文模型分配机制,实现对配置信息的压缩,从而可减少所需的配置信息存储空间和重构时配置信息的传输量.该算法用现场可编程门阵列(FPGA)配置信息文件进行验证.同时,提出了一种能够有效提高硬件在线解压速度的改进方法,该方法在Virtex-4 FPGA上得到实现.实验表明,对于器件占用率为90%以上和20%的FPGA配置,该算法的压缩率分别达到25%和9%.与现有的方法相比,该算法在压缩率上具有优势,并可适用于多种器件.%Aiming at the configuration compression in reconfigurable computing systems, the paper puts forward a configuration compression algorithm based on contextually adaptive arithmetic coding. The proposed algorithm obtains the approximate symbol probability by dynamically adjusting the parameters, and eliminates the multiplication and division by using shift operations. It introduces the distribution mechanism based on context into the compression and decompression process to realize configuration compression, so the required configuration storage space and the amount of configuration being transmissed when reconfiguratiny can be reduced. The algorithm was verified by testing field programmable gate anay (FPGA)configuration files. Simultaneously, a decompressing circuit, which accelerates the decompression speed, was designed and implemented on the Virtex-4 FPGA. The experiments show that, for the configuration with the device occupancy rate of 90％ and 20％, the proposed algorithm can achieve the compression ratio reaching to 25％ and 9％ respectively.Compared with the existing methods, the presented gets the higher compression ratio and also can be adopted to various equipments.