A Closer Look at GPUs

被引：64

作者：

Fatahalian, Kayvon ^{[1
]}

Houston, Mike ^{[1
]}

机构：

[1] Stanford Univ, Comp Graph Lab, Stanford, CA 94305 USA

来源：

COMMUNICATIONS OF THE ACM | 2008年 / 51卷 / 10期

关键词：

D O I：

10.1145/1400181.1400197

中图分类号：

TP3 [计算技术、计算机技术];

学科分类号：

0812 ;

摘要：

The graphics processing unit (GPU) is a processor that can be used for multicore parallel computing architectures. GPUs has a collection of fixed-function, software-programmable processing resources, high arithmetic logic unit (ALU) count, and peak floating-point rates. The GPU uses hardware components and their software interfaces to keep GPU processing resources busy. The stages of graphics pipeline include vertex generation, vertex processing, primitive generation, primitive processing, fragment generation, fragment processing, and pixel operations. GPU's resources can exist within the programmable processing and it can maintain high efficiency by using milticore designs with hardware multi-threading and SIMD (single instruction, multiple data) processing. GPU provide a high level of performance by maintaining simple programming interface.

引用

页码：50 / 57

页数：8