三維電腦繪圖之運算模組設計

Abstract

在本論文中，我們最主要的研究是針對在3D繪圖中的一些相關硬體設計，做最佳化或創新之設計，以期能增進現今3D繪圖之效能。而所選擇的研究對象，包含下列三個部分：（1）座標轉換器：我們從基本的座標轉換分析開始，根據一些轉換矩陣特徵，發展一個較為簡單的架構，以省去不必要的乘法運算。（2）除法、開根號與乘法器之綜合硬體設計：設計一以2為底的浮點運算單元，並運用over-redundant 的表示方式，將三種運算以一相容硬體加以實踐，並符合IEEE浮點算數標準規格。（3）著色器的硬體設計：針對3D繪圖上之著色方式中品質最佳的Phong Shading 方法設計硬體，以期減少CPU運算上的負擔，並達到3D繪圖速度上的要求。

In this thesis, our goal is to design and optimize new efficient processing cores for 3D graphics. This these is divided into three parts: (I) The datapath design for geometric transformation: First we analysis basic geometric transformation operations used in 3D graphics. Based on the special features of matrix-vector multiplication for geometric transformation, we design a simple architecture that eliminates redundant multiplication. (II) A fused efficient VLSI implementation of radix 2 floating-point arithmetic unit is presented, which is capable of doing division, square root and multiplication. To make the quotient/ root digit's selection function simple, over-redundant radix 2 representation is used. The arithmetic unit complies with IEEE's binary floating-point standard. (III) The datapath design for 3D graphic shading. We introduce some shading methods that are usually used for 3D graphics. A VLSI architecture is designed for Phong shading, which has the best performance for 3D graphics shading. Our targets for this architecture are to decrease the computation of CPU and can arrive the speed's requires for 3D graphics.