论文部分内容阅读
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。
A quasi-double-precision multiply-accumulate operation method is proposed, which can realize the standard precision decoding of ISO / IEC 13818-3 on cheap 16-bit DSP and also can be applied to other high-precision real-time DSP system design. The average number of instructions is about three times the single-precision operation, for the MPEG Layer III audio decoder can achieve 24 to 25 bits of the overall accuracy. The core code and instruction count based on Motorola DSP56824 are given.