Swan
收藏资源简介:
Swan是由密歇根大学开发的一个面向移动应用的向量处理基准套件,包含59个多样化的数据并行工作负载,源自操作系统、网页浏览器、音视频消息应用及PDF渲染引擎等四个常用移动应用。该数据集通过深入分析性能、功耗和能耗,揭示了向量处理在移动设备上的优势与局限。创建过程中,研究团队详细研究了四个广泛使用的移动应用的源代码,并从12个常用库中精心挑选了代表性的工作负载。Swan的应用领域主要集中在探索CPU向量处理在移动平台上的性能,旨在解决移动设备资源受限环境下的性能优化问题。
Swan is a mobile application-oriented vector processing benchmark suite developed by the University of Michigan. It comprises 59 diverse data-parallel workloads derived from four widely used mobile application categories: operating systems, web browsers, audio/video messaging apps, and PDF rendering engines. This benchmark suite reveals the advantages and limitations of vector processing on mobile devices through in-depth analysis of performance, power consumption, and energy usage. During its development, the research team conducted a detailed study of the source code of four prevalent mobile applications, and carefully selected representative workloads from 12 commonly used libraries. The primary application domain of Swan focuses on exploring the performance of CPU vector processing on mobile platforms, with the goal of addressing performance optimization challenges in resource-constrained mobile device environments.

- 1Vector-Processing for Mobile Devices: Benchmark and Analysis密歇根大学 · 2023年



