本文详细分析了Tesla GPU图形与计算架构和CUDA统一计算设备架构，详细描述了如何对计算任务进行并行分解，并通过CUDA的双层并行编程模型映射到Tesla GPU上。 在本文的实现部分，以软件开发流程为主线，描述了如何利用CUDA实现模式识别中的三种常用算法-The paper analyzed the Tesla GPU computing architecture and CUDA graphics and computing devices unified architecture, a detailed descr iption of how parallel computing tasks broken down and through the double CUDA parallel programming model is mapped to the Tesla GPU. In the implementation section of this article, the main line to the software development process, describes how to use CUDA to achieve the three commonly used pattern recognition algorithms