图形处理器中图像处理单元的设计与实现

Design and Implementation of the Image Processing Unit in GPU

针对图形处理器中关于图像、纹理的传输、偏移、缩放、映射等操作需求,采用具有泡沫挤压功能的流水线结构以及相同功能单元复用的方法,实现了GPU中图像处理流水线。提高了GPU处理图像的速度,同时减少了电路面积。结果表明,在TSMC 0.13μm CMOS工艺下,电路消耗面积为390 726.09,速度可达200 MHz。现该设计已集成到完整GPU系统中并流片,用经典的测试图像对GPU芯片进行测试:绘制的图像质量与Open GL的运行结果具有可比性;像素填充率可达80 M像素/s,满足实时处理要求。

In view of the image and texture processing requirements of graphics in Graphic Processing Unit （GPU）, this paper proposes an image processing pipeline with foam extrusion function and multiplexing same func- tion unit, thus improving the speed of image processing while reducing the circuit area. The design was synthesized under TSMC 0.13 um CMOS technology by Synopsys Design-Compiler. The frequency of this design is up to 200 MHz at a circuit area of 390726.09. The design is integrated into a complete GPU system and the resuhs are compa- rable with the operation results of OpenGL. The pixel fill rate is 80 Mpixels/s, which meets the requirements of real - time processing.