«上一篇/Previous Article|本期目录/Table of Contents|下一篇/Next Article»

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:

30

期数:

2020年09期

页码:

54-59

栏目:

智能、算法、系统工程

出版日期:

2020-09-10

文章信息/Info

Title:

Modeling of GPU Fragment Processing Based on UML&SystemC

文章编号:

1673-629X(2020)09-0054-06

作者:

姜丽云¹; 田 泽¹; 2; 吴晓成¹; 张 骏¹; 2

1. 中国航空工业集团 西安航空计算技术研究所,陕西 西安 710068; 2. 集成电路与微系统设计航空科技重点实验室,陕西 西安 710068

Author(s):

JIANG Li-yun¹; TIAN Ze¹; 2; WU Xiao-cheng¹; ZHANG Jun¹; 2

1. Aeronautics Computing Technique Research Institute,Aviation Industry Corporation of China,Xi’an 710068,China; 2. Key Laboratory of Aviation Science and Technology on Integrated Circuit and Micro-System Design,Xi’an 710068,China

关键词:

图形处理器; 统一建模语言; SystemC 语言; 片段; 事务级建模

Keywords:

graphic processing unit (GPU); unified modeling language (UML); SystemC; fragment; transaction level modeling (TLM)

分类号:

TP391

DOI:

10. 3969 / j. issn. 1673-629X. 2020. 09. 010

摘要:

GPU 片段处理单元位于图形处理器渲染管线的末尾,执行对将要存储到缓冲区的片段的测试、混合、屏蔽、逻辑等所有操作。为了在 RTL 设计之前快速研究 GPU片段处理单元的硬件结构及功能算法,文中联合使用统一建模语言(unified modeling language,UML)和 SystemC 语言对 GPU 片段处理单元进行事务级建模(transaction-level modeling, TLM)。首先设计了 GPU 片段处理单元的类图、结构图、行为图等视图,对系统进行静态结构和动态行为的描述;然后将 UML 可视化视图转换为基于 SystemC 的事务级模型,详细描述了 GPU 片段处理单元的模块、接口、端口、进程等建模过程,模块之间通过端口和层次化通道通讯,具有建模效率高和仿真速度快的优点;最后基于 Linux 操作系统仿真平台验证了架构和算法的正确性,实现了在设计的前期进行软硬件的协同仿真和算法的验证、优化,对硬件架构进行探索,有利于加速以图形处理器为核心的设计和开发。

Abstract:

The GPU fragment processing unit is located at the end of the graphics processor shading pipeline and performs all the operations of test,blend, mask,logic and so on for the fragment to be stored in the buffer. In order to study the hardware structure and function algorithms of GPU fragment processing unit before RTL design as quickly as possible,we implement the transaction level modeling of GPU fragment processing unit based on unified modeling language (UML) and SystemC language. Firstly,the view of class diagram,structure diagram and behavior diagram of GPU fragment processing unit are designed to describe the static structure and dynamic behavior of the system,the UML visual diagram is then transformed into a SystemC-based transaction-level modeling,which describes the module,interface,port,process of GPU fragment processing unit in detail. The modules communicate through ports and hierarchical channels, which have the advantages of high modeling efficiency and high simulation speed. Finally,the correctness of the architecture and algorithm is verified by Linux operating system simulation platform. The software and hardware collaborative simulation,the algorithm verification and optimization and the structure of hardware architecture exploration are realized in the prophase of the design,which will speed up the design and development of graphics processor as the core.

常用功能

更新日期/Last Update: 2020-09-10