To meet society’s needs for undergraduate students to have engineering skills and to develop students’ability to operate Linux and engage in network software development,this paper proposes the construction of a new...To meet society’s needs for undergraduate students to have engineering skills and to develop students’ability to operate Linux and engage in network software development,this paper proposes the construction of a new specialized course for network engineering major--Linux system and network programming.This paper analyzes the course’s advantages,presents the contents of this course,designs a series of teaching methods aimed at improving students’engineering ability,proposes a course assessment method that will encourage students to practice,lists the development requirements for an examination software designed for this course,and finally,presents the results of our practice in teaching this course.展开更多
With the development of embedded system, GUI becomes more and more important.As one of the most popular embedded GUI, MiniGUI became to be a key in researching the embedded GUI. This article firstly introduced embedde...With the development of embedded system, GUI becomes more and more important.As one of the most popular embedded GUI, MiniGUI became to be a key in researching the embedded GUI. This article firstly introduced embedded Linux system, processor ARM ,and S3C2410.Then, it analyses structures and primary predominances of MiniGUI. Finally, the transplantation of MiniGUI1.6.x based on $3C2410 is described in this paper.展开更多
大语言模型(large language model,LLM,也称大模型)在软件开发技术问答任务中展现出强大潜力,为代码知识获取和理解提供了新途径.然而,在以Linux内核为代表的复杂系统软件领域,LLM在代码实现、关键机制理解、演化历史追溯及设计决策分...大语言模型(large language model,LLM,也称大模型)在软件开发技术问答任务中展现出强大潜力,为代码知识获取和理解提供了新途径.然而,在以Linux内核为代表的复杂系统软件领域,LLM在代码实现、关键机制理解、演化历史追溯及设计决策分析等方面的真实能力仍缺乏系统验证.现有评测基准多针对通用任务,存在领域深度不足、难度逐渐饱和及评测问题与工程实践存在偏差等局限,难以保障特定领域开发知识问答的客观性、准确性和全面性.为客观评估LLM在复杂系统软件中的知识问答能力,研究提出一种LLM问答能力评测基准数据集构建方法,并构建面向Linux内核的高质量问答评测基准LKQABench,同时设计一种多裁判协同的代码知识问答评测方法MJ-CCE.LKQABench基于开发者社区的真实技术问答数据,经过语义分析和人工审核修订,构建202个标准问答对,覆盖Linux内核主要模块和不同认知维度.MJ-CCE方法定义多个裁判大模型的协同评分与投票机制,从关键知识点覆盖度、事实正确性与表达清晰度等维度对回答进行多维度评估.在LKQABench上对主流大模型的实证研究表明,当前大模型能较好地回答内核实现的单点知识问题,但在涉及跨主题知识整合、深度推理及版本演化关联的问题中,存在知识点遗漏、逻辑链条不完整等不足.研究不仅揭示了大模型在软件开发知识问答中的能力边界,也为其在该领域的持续优化提供了实证数据支撑.展开更多
基金supported by the Teaching Research and Reform Project of Qingdao University of Technology under Grant 2024-10335040。
摘要To meet society’s needs for undergraduate students to have engineering skills and to develop students’ability to operate Linux and engage in network software development,this paper proposes the construction of a new specialized course for network engineering major--Linux system and network programming.This paper analyzes the course’s advantages,presents the contents of this course,designs a series of teaching methods aimed at improving students’engineering ability,proposes a course assessment method that will encourage students to practice,lists the development requirements for an examination software designed for this course,and finally,presents the results of our practice in teaching this course.
摘要With the development of embedded system, GUI becomes more and more important.As one of the most popular embedded GUI, MiniGUI became to be a key in researching the embedded GUI. This article firstly introduced embedded Linux system, processor ARM ,and S3C2410.Then, it analyses structures and primary predominances of MiniGUI. Finally, the transplantation of MiniGUI1.6.x based on $3C2410 is described in this paper.
摘要大语言模型(large language model,LLM,也称大模型)在软件开发技术问答任务中展现出强大潜力,为代码知识获取和理解提供了新途径.然而,在以Linux内核为代表的复杂系统软件领域,LLM在代码实现、关键机制理解、演化历史追溯及设计决策分析等方面的真实能力仍缺乏系统验证.现有评测基准多针对通用任务,存在领域深度不足、难度逐渐饱和及评测问题与工程实践存在偏差等局限,难以保障特定领域开发知识问答的客观性、准确性和全面性.为客观评估LLM在复杂系统软件中的知识问答能力,研究提出一种LLM问答能力评测基准数据集构建方法,并构建面向Linux内核的高质量问答评测基准LKQABench,同时设计一种多裁判协同的代码知识问答评测方法MJ-CCE.LKQABench基于开发者社区的真实技术问答数据,经过语义分析和人工审核修订,构建202个标准问答对,覆盖Linux内核主要模块和不同认知维度.MJ-CCE方法定义多个裁判大模型的协同评分与投票机制,从关键知识点覆盖度、事实正确性与表达清晰度等维度对回答进行多维度评估.在LKQABench上对主流大模型的实证研究表明,当前大模型能较好地回答内核实现的单点知识问题,但在涉及跨主题知识整合、深度推理及版本演化关联的问题中,存在知识点遗漏、逻辑链条不完整等不足.研究不仅揭示了大模型在软件开发知识问答中的能力边界,也为其在该领域的持续优化提供了实证数据支撑.