摘要
提出了一种准双精度乘累加运算方法，可以在廉价的16比特DSP上实现ISO／IEC 13818－3的标准精度解码，也可应用于其他高精度实时DSP系统设计。其平均指令数约为单精度运算的3倍，对于MPEG Layer III音频解码器可实现24～25比特的总体精度。给出了基于Motorola DSP56824的核心代码和指令数统计。
In this paper a quasi－double precision MAC multiply－accumulation algorithm is proposed for the implementation of ISO／IEC 13818－3 audio decoder on 16 bit DSP and for other real－time high precision DSP systems． The number of instructions of this algorithm is about 3 times of single－precision MAC and the final accuracy of 24～25 bits can be achieved in the implementation of MPEG Layer III audio decoder． The kernel source code and the instruction count based on Motorola＇s DSP56824 are presented in this paper．
出处
《电声技术》
北大核心
2001年第5期3-6,共4页
Audio Engineering
