AVS量化和反量化C64x+汇编级优化实现  

AVS QUANTIZATION AND INVERSE QUANTIZATION C64x+ASSEMBLY LEVEL OPTIMIZATION IMPLEMENTATION

作　　者：牛承珍[1] 

机构地区：[1]山西工程职业技术学院计算机工程系,山西太原030009

出　　处：《计算机应用与软件》2011年第7期221-223,共3页Computer Applications and Software

摘　　要：AVS标准中,量化和反量化过程被频繁调用,对其优化可以提高整个编码器的编码速度。超长指令字和打包数据处理是在DSP广泛使用的结构。以TI公司C64x+DSP为例介绍了超长指令字和打包数据处理结构特性,并分析了基于这些特性的AVS量化和反量化的快速实现方法。利用TI公司的仿真工具CCS3.3提供的性能分析工具对汇编级优化的效果和C语言级优化的效果进行比较,量化汇编级优化效果所需时钟周期约为C语言级优化周期的10.24%,反量化约为30.94%。In AVS standard,quantization and inverse quantization are frequently called,so that their optimization helps speed up the encoding of the entire encoder.Very Long Instruction Word（VLIW） and compact instruction is a widely used architecture in DSP.Taking TI＇s C64x＋ DSP as example,the article introduces the characters of VLIW and compact instruction architecture,and analyzes AVS quantization and inverse quantization rapid implementation methods based on these characteristics.Comparison is made by TI＇s emulation tool,CCS 3.3,between results of assembly level and C language level optimizations.The quantization of clock cycle required by assembly optimization effect is about 10.24% that of C language level optimization cycle while the inverse quantization is about 30.94%.

关 键 词：AVS 超长指令字 打包数据处理 量化 反量化 

分 类 号：TP331[自动化与计算机技术—计算机系统结构]