系统级低功耗SoC 裸机测试数据

该数据集主要是对本项目设计的带有智能神经网络加速器的SoC进行帧率和功耗测试。使用ARM DS5 IDE进行裸机测试程序的编写，裸机运行可有效提升能效，内置编译器为GCC，由DS5工具编译生成.axf可执行程序，这种文件格式相比.bin格式的可执行程序有更多的调试相关的信息，通过DSTREAM调试器和JTAG调试接口可将程序烧录到板子上运行，并且我们编写了.ds运行脚本简化了程序的加载和运行过程。

This dataset is primarily used for frame rate and power consumption testing of the SoC integrated with an intelligent neural network accelerator designed in this project. We developed bare-metal test programs using ARM DS5 IDE, as bare-metal execution can effectively enhance energy efficiency. The built-in compiler is GCC, and the DS5 tool uses it to compile the programs into .axf executable files. Compared with .bin-format executables, .axf files contain more debugging-related information. We employed the DSTREAM debugger and JTAG debugging interface to flash the programs onto the development board for execution, and we authored .ds run scripts to simplify the procedures of program loading and execution.

搜集汇总

背景与挑战

背景概述

该数据集针对带有智能神经网络加速器的SoC进行帧率和功耗测试，通过ARM DS5 IDE编写裸机程序并利用调试工具烧录运行，以提升能效。数据集包含13个文件，总数据量为58.15MB，由北京大学崔小欣创建，属于国家重点研发计划项目的一部分。