在本篇論文中，實現了一顆可參數化的數位信號處理器。它是專用於通訊系統應用。除了提供一般十六位元數位處理器所具備的基本指令集外，還為了特別的功能硬體設計，提供特殊指令。這使得這顆數位信號處理器更適於計算密集的應用。 我們所提出的數位信號處理器具有幾項優越的特性：可參數化的架構，高速的效能，和低功率。我們設計了各種模組產生器以產生可變動(configurable)的資料路徑(datapath)和可重複使用的特殊功能硬體，並設計了一個模組產生器用以整合各個功能模組及產生數位信號處理器的硬體描述語言。平行化的架構也加速了效能，在效能測試程式中兩組乘法器串聯加法器減少一半的指令週期。為了減少功率耗損，我們採用許多種低功率設計技巧，如灰碼記憶體定址法和管線分享技巧等。 以模組產生器所產生的十六位元數位信號處理器為例，其最大工作效能可操作在140百萬指令。 This thesis introduces the design and implementation of an embedded and parameterized digital signal processing (DSP) processor. Besides providing a basic instruction set that is similar to general DSP processors, it also contains unique instructions and optional special function blocks that make this DSP processor more efficient for computation-intensive applications. The proposed DSP processor has some advanced features: a parameterized architecture, special functions for communication application, some low power designs and I/O for embedded consideration. By using window GUI and a Verilog code generator, dedicated DSP for specified application can be generated. In addition, we use high degree of parallelism to speed up its performance. The data path contains two Multiply-Accumulate units to reduce half instruction cycles in the operation of FIR filter. For consideration of reducing power consumption, we adopt some low power design such as gray code memory addressing and pipeline sharing techniques. The chip will be implemented in a cell-based design method with a 0.25 1P5M cell library. The maximum operating frequency of a 16?16 DSP is 140MHz.
