Fun with Beagle – Exploring the GPU 《与BeagleBone GPU 愉快玩耍》简体中文版
Version 1.0: 21st January 2021
Simplified Chinese Version 1.0: 26st March 2021
The purpose of this document is to explain how Imagination Technologies PowerVR SGXTM 530 Graphics Processing Unit (GPU) in the processor on the BeagleBone® Black (BBB) board can be used to enhance applications running on the main Arm® Cortex® A8 core.
This is both in the traditional role as a graphics renderer using OpenGL® ES and as an offload engine for tasks that can parallelised using OpenCLTM with a newly published library. The target audience are Hobbyists and Universities who wish to experiment and see if they can improve their embedded systems.
The focus in the document will be on the how to use the new OpenCLTM library and very importantly understand its strengths and weaknesses. This is done using a matrix multiplication example. With this understanding of the type of application that can benefit from an OpenCL offload to the SGX530, it then explains how to integrate OpenCLTM into an ALSA audio Sample Rate Convertor. This implementation can see the CPU load reduced from 55% to 20% by using OpenCL as long as the application can trade off latency.
本文档旨在介绍如何使用BeagleBone®Black(BBB)板上处理器中的Imagination Technologies PowerVR SGXTM 530图形处理单元(Graphics Processing Unit，GPU)增强Arm® Cortex® A8核心上运行的应用程序的能力。这款GPU不仅是使用OpenGL® ES的传统图形渲染器，也 可作为使用OpenCLTM和新发布的库来并行处理任务的减负引擎。本文档的目标受众是希望借 助实验来改进嵌入式系统的业余爱好者和高校。 本文档重点关注如何使用新的OpenCLTM库，并帮助读者了解其优势和劣势。我们通过一个矩 阵乘法的示例来介绍这些内容。在了解可从SGX530的OpenCL减负功能受益的应用程序类型 之后，本文档将继续介绍如何将OpenCLTM集成到ALSA音频采样率转换器中。只要应用程序 可以对延迟采用折衷方案，就可以通过使用OpenCL将CPU负载从55%降低到20%。