摘要
人机界面软件不论在何种情况下都要求具有极高的实时性,而软件必须调用的一些复杂算法则会严重影响到软件的实时性。文中介绍了两种解决问题的思路。通过开辟算法专用线程实现前后台线程并行,可以降低数据处理对界面响应过程的影响;GPU编程利用显卡上的流多处理器能提供很高的并行运算能力,但在小规模运算中无法充分体现出其速度优势。两种方法各有优缺点,根据实际情况选择合适的方法,可以有效地降低复杂计算导致的软件响应延迟。
It is important for a man-machine interface (MMI)program to respond to an operation in real time. However,complex arithmetics called by the program tend to delay the response sub- stantially. This paper proposes two methods to resolve the problem:multithreaded programming and GPU programming. By the first method,a dedicated thread is created in the foreground for interface display, parallel with the background main thread for data processing. The second method makes use of the stream processor on the graphic card to provide a high level of parallel processing, but the high speed of the processor does not show itself in processing data of small volume. By presenting both methods' advantages and drawbacks ,the paper points out the neces- sity of making a proper choice between them in different situations to improve the performance of an MMI program substantially.
出处
《指挥信息系统与技术》
2010年第5期55-58,共4页
Command Information System and Technology
关键词
人机界面
实时系统
多线程
图形处理器
通用运算设备架构
man-machine interface (MMI)
real-time system
multithreading
graphic processing unit (GPU)
Compute Unified Device Architecture (CUDA)