3

Embedded GPU Design

BYEONG-GYU NAM and HOI-JUN YOO

3.1 INTRODUCTION

As the mobile electronics advances, the real-time three-dimensional (3D) graphics has been widely adopted in mobile embedded systems such as smartphones and smart pads. The realization of 3D graphics on these power- and area-limited mobile devices has been a challenging issue because of the highly complex computations inherent in the 3D computer graphics. There have been studies on the hardware acceleration of 3D graphics pipeline to provide the required computing power within a limited power consumption [1, 2]. However, these hardwired accelerators just provide limited graphics effects but cannot afford the programmability to adapt to advanced graphics algorithms. Recently, mobile 3D graphics standards like OpenGL ES [3] introduced the programmable graphics pipeline to accommodate a wide variety of advanced graphics effects on mobile devices. Therefore, high-performance 3D graphics processors consuming only limited amount of power are proposed to address these technology trends [4–7]. However, their design is optimized for fast matrix-vector multiplications for 3D geometry transformations but area efficiency is somewhat neglected.

In this chapter, a high-speed, low-power, and area-efficient 3D graphics processor, that is, a graphics processing unit (GPU), implementing a full 3D graphics pipeline is presented [8]. It demonstrates 141 Mvertices/s peak geometry processing performance with 968 k transistors only ...

Get Embedded Systems: Hardware, Design and Implementation now with the O’Reilly learning platform.

O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.