[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"site-home-live-entry-setting":3,"site-header-contests":8,"site-header-camps":49,"study-material-detail-study-35":104},{"enabled":4,"title":5,"statusText":5,"buttonText":5,"campSlug":5,"stage":6,"classroomId":7,"hasClassroom":4},false,"",0,null,{"total":9,"list":10,"pageNum":44,"pageSize":45,"size":9,"startRow":44,"endRow":9,"pages":44,"prePage":6,"nextPage":6,"isFirstPage":46,"isLastPage":46,"hasPreviousPage":4,"hasNextPage":4,"navigatePages":47,"navigatepageNums":48,"navigateFirstPage":44,"navigateLastPage":44},3,[11,25,35],{"slug":12,"title":13,"logo":7,"liveUrl":14,"prize":15,"sponsor":16,"description":17,"organizer":18,"startTime":19,"endTime":20,"registrationDeadline":21,"status":22,"dashboardVisibility":23,"createdAt":24},"spring2026","2026春季启元人工智能大赛","https:\u002F\u002Fwww.bilibili.com\u002Fblackboard\u002Flive\u002Flive-activity-player.html?cid=1822703816&quality=0","200,000","麒麟软件有限公司","“2026春季启元人工智能大赛”是由启元实验室主办的面向全国人工智能领域开发者、科研人员及高校学生的权威赛事。大赛旨在促进与培养具备工程实践能力和创新能力的优秀人工智能领域开发者、推广国产软硬件技术与生态、助力国家人工智能产业高质量发展。\n各参赛选手需在规定的比赛周期内，依据赛事规则选定赛题并完成相应挑战，优胜者将共同角逐总额20万元的赛事奖金池，并且还将有机会获得实习和知名企业优质岗位内推资格。","启元实验室","2026-04-19 16:00:00","2026-07-13 16:00:00","2026-06-16 16:00:00","ongoing","private","2026-04-16 08:00:00",{"slug":26,"title":27,"logo":7,"liveUrl":14,"prize":28,"sponsor":7,"description":29,"organizer":18,"startTime":30,"endTime":31,"registrationDeadline":32,"status":33,"dashboardVisibility":23,"createdAt":34},"autumn2025","2025秋季启元人工智能大赛","150,000","“2025秋季启元人工智能大赛”是由启元实验室主办的面向全国人工智能领域开发者、科研人员及高校学生的权威赛事。大赛旨在促进与培养具备工程实践能力和创新能力的优秀人工智能领域开发者、推广国产软硬件技术与生态、助力国家人工智能产业高质量发展。","2025-11-08 08:00:00","2026-01-13 08:00:00","2025-12-30 08:00:00","ended","2025-11-03 08:00:00",{"slug":36,"title":37,"logo":7,"liveUrl":14,"prize":38,"sponsor":7,"description":39,"organizer":18,"startTime":40,"endTime":41,"registrationDeadline":42,"status":33,"dashboardVisibility":23,"createdAt":43},"summer2025","启元人工智能大赛(系统方向)","80,000","启元人工智能大赛由启元实验室主办，其中系统赛道基于启元实验室九源软件栈的相关开源项目，通过拉动开源社区力量，一方面促进国产智能计算生态的发展，另一方面也为人工智能系统方向的学生及从业者提供一个实践平台。","2025-07-08 08:00:00","2025-08-25 08:00:00","2025-07-22 08:00:00","2025-06-22 08:00:00",1,5,true,8,[44],{"total":45,"list":50,"pageNum":44,"pageSize":45,"size":45,"startRow":44,"endRow":45,"pages":44,"prePage":6,"nextPage":6,"isFirstPage":46,"isLastPage":46,"hasPreviousPage":4,"hasNextPage":4,"navigatePages":47,"navigatepageNums":103,"navigateFirstPage":44,"navigateLastPage":44},[51,62,73,84,94],{"id":52,"slug":53,"title":54,"logo":55,"publicLiveEmbedUrl":14,"description":56,"introduction":57,"organizer":18,"coOrganizer":5,"sponsor":16,"startTime":58,"endTime":59,"status":22,"dashboardVisibility":60,"createdAt":58,"updatedAt":61},"019eeeff-6b00-7d55-a629-efafded3e232","summer2026","2026年夏季InfiniTensor大模型与人工智能系统训练营","https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fuploads\u002F2026-06\u002F019eef7f-7816-7828-9309-79c2c61dd568.jpg?Expires=1792752590&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=fLKIdTmv5ztTF5lNyW4vFO%2FwiU4%3D","由启元实验室发起的InfiniTensor大模型与人工智能系统训练营，面向开源社区、高校及产业界伙伴开放。活动以「普及AI全栈知识技术、培养工程实践与创新型人才、推广国产软硬件技术生态」为核心目标，助力国家人工智能产业高质量发展。","...","2026-06-22 08:00:00","2026-10-19 08:00:00","public","2026-10-12 08:00:00",{"id":63,"slug":64,"title":65,"logo":66,"publicLiveEmbedUrl":7,"description":67,"introduction":57,"organizer":18,"coOrganizer":5,"sponsor":68,"startTime":69,"endTime":70,"status":33,"dashboardVisibility":23,"createdAt":71,"updatedAt":72},"019af3f3-e06a-7245-aa8d-37dc99a69c73","winter2025","2025年冬季InfiniTensor大模型与人工智能系统训练营","https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fuploads\u002F2025-12\u002F019afc69-b7ca-7805-91fb-605429671c01.png?Expires=1792752590&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=IfcKyBkOe6Ug2cb8ODO9I1IUNkA%3D","由启元实验室发起的InfiniTensor大模型与人工智能系统训练营，面向开源社区、高校及产业界伙伴开放。活动以「普及AI全栈知识技术、培养工程实践与创新型人才、推广国产软硬件技术生态」为核心目标，助力国家人工智能产业高质量发展。 \n","新华三技术有限公司","2025-12-25 08:00:00","2026-03-27 08:00:00","2025-12-06 08:00:00","2026-10-22 20:24:30",{"id":74,"slug":36,"title":75,"logo":76,"publicLiveEmbedUrl":7,"description":77,"introduction":57,"organizer":18,"coOrganizer":5,"sponsor":78,"startTime":79,"endTime":80,"status":81,"dashboardVisibility":23,"createdAt":82,"updatedAt":83},"01980342-208c-7803-b3f3-19e1d31960ba","2025年夏季InfiniTensor大模型与人工智能系统训练营","https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fuploads\u002F2025-07\u002F0198045e-28fe-78f0-95fe-23835ce58395.jpg?Expires=1792752590&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=Ev4DWdZSXJ0t1T4u2JL5qybAazs%3D","InfiniTensor大模型与人工智能系统训练营由启元实验室及清华大学 AI 系统方向团队发起，面向开源社区、高校和产业界，普及人工智能各方面知识和技术，培养相关人才。训练营联合开源项目开发者、学术界大牛、工业界资深工程师参与授课与项目指导。课程采用「线上直播授课 + 在线答疑 + 项目实践辅导 」的方式进行，确保每位同学都能获得充分成长。训练营全程免费，通过视频、直播课、习题和项目实践，在人工智能飞速发展的背景下，为学员提供一窥深层技术的机会。","北京积算科技有限公司","2025-07-14 08:00:00","2025-09-30 08:00:00","upcoming","2025-07-13 08:00:00","2026-10-22 20:24:58",{"id":85,"slug":86,"title":87,"logo":88,"publicLiveEmbedUrl":7,"description":89,"introduction":57,"organizer":18,"coOrganizer":5,"sponsor":5,"startTime":90,"endTime":91,"status":81,"dashboardVisibility":23,"createdAt":92,"updatedAt":93},"01980cf9-a7d8-7796-ae86-0bbe05334add","winter2024","2024年冬季InfiniTensor大模型与人工智能系统训练营","https:\u002F\u002Ftesting-cdn.infinitensor.com\u002Fuploads\u002F2025-07\u002F01981269-ce23-7624-a287-9f07dd0d85f6.jpg","InfiniTensor大模型与人工智能系统训练营由清华大学发起，面向开源社区、高校和产业界，普及人工智能各方面知识和技术，培养相关人才。训练营全程免费，通过视频、直播课、习题和项目实践，在人工智能飞速发展的背景下，为学员提供一窥深层技术的机会。报名学员请阅读“导学”课程页的训练营介绍。","2025-01-05 08:00:00","2025-03-22 08:00:00","2025-07-15 08:00:00","2026-10-22 20:24:39",{"id":95,"slug":96,"title":97,"logo":98,"publicLiveEmbedUrl":7,"description":99,"introduction":57,"organizer":18,"coOrganizer":5,"sponsor":5,"startTime":100,"endTime":101,"status":81,"dashboardVisibility":23,"createdAt":92,"updatedAt":102},"01980c20-363a-7dc3-8f5a-58a68e08dcc4","summer2024","2024年夏季InfiniTensor大模型与人工智能系统训练营","https:\u002F\u002Ftesting-cdn.infinitensor.com\u002Fuploads\u002F2025-07\u002F01981269-fbfa-788a-afcd-17fa9be74ff9.jpg","InfiniTensor大模型与人工智能系统训练营由清华大学发起，面向开源社区和高校，普及人工智能各方面知识和技术，培养相关人才。训练营全程免费，通过课程、习题和项目实践，在人工智能飞速发展的背景下，为学员提供一窥深层技术的机会。","2024-07-20 08:00:00","2024-09-29 08:00:00","2026-10-22 20:24:49",[44],{"currentItem":105,"processedData":125},{"img":106,"keywords":107,"treeType":44,"description":108,"showStatus":44,"updateTime":109,"type":110,"title":111,"content":112,"nextArticle":113,"createTime":109,"sortOrder":117,"canonicalPath":118,"id":119,"contentType":120,"prevArticle":121},"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_35_fm.jpg?Expires=1792752590&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=rPEnvr1HBHLGfNnUC37mvGKh7Zg%3D","InfiniTensor,冬季训练营,AI编译器,后端优化,硬件适配","聚焦 AI 编译器后端优化，解析算子生成、代码优化与硬件适配的核心思路。","2026-10-22 20:38:11",2,"InfiniTensor 冬季训练营 AI 编译器第四课《AI 编译器中的后端优化》","![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_35_zw01.jpg?Expires=1792752590&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=AGmoC6iJaL4xuVr%2B150SlYaX3eU%3D)\r\n\r\n\r\n2026 年 1 月 23 日，2025 冬季 InfiniTensor 大模型与人工智能系统训练营 AI 编译器系列第四课《AI 编译器中的后端优化》开讲。\r\n\r\n本课将重点讲解内存管理优化、算子层面优化、Auto Tuning 和代码生成四个核心内容。\r\n\r\n### 内存管理优化\r\n#### 1. 优化思路\r\n- 调整分配和释放的时间点和次数，减少运行时的内存分配和释放开销。\r\n- 优化分配过程：减少总的内存占用。\r\n#### 2. Naive 版本：\r\n运行到算子时即时申请内存，结束即释放，导致频繁的系统调用。\r\n\r\n![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_35_zw02.jpg?Expires=1792752590&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=aUI6aNmo46myjnsERD5K13%2F29AU%3D)\r\n\r\n#### 3. 优化版本：\r\n- 根据拓扑排序确定 OP 执行顺序。\r\n- 模拟执行过程，计算内存峰值。\r\n- 按需分配和释放内存，利用引用计数机制判断 Tensor 是否可释放。\r\n\r\n![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_35_zw03.jpg?Expires=1792752590&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=n4y%2B%2B%2BBbKjqXwiauvkv7OmNf%2BB0%3D)\r\n\r\n#### 4. 惰性分配（Lazy Allocation）\r\n\r\n![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_35_zw04.jpg?Expires=1792752590&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=N0XrEaGlsv%2F%2FZTKqLO6srcmDXz8%3D)\r\n\r\n### 算子层面优化\r\n#### 1. 优化起点\r\n- 理解瓶颈：区分计算密集型和访存密集型任务。\r\n- 诊断工具：使用 Roofline 模型确定优化方向。\r\n#### 2. 常用优化策略\r\n **（1）循环优化**\r\n- 循环分块：将大循环分解成小块，提升数据局部性。\r\n- 循环展开：增加单次循环迭代内执行的操作次数，减少循环控制开销。\r\n- 循环重排：改变循环顺序，匹配数据存储顺序，提升缓存利用率。\r\n- 循环融合：合并具有相同迭代范围和遍历顺序的独立循环，减少循环控制开销，提升数据局部性。\r\n- 循环拆分：将复杂循环拆分成多个简单循环，消除依赖，允许并行执行。\r\n\r\n![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_35_zw05.jpg?Expires=1792752590&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=DLBy8wo5DfhCMJVpXLpC2P%2BhAMY%3D)\r\n\r\n\r\n![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_35_zw06.jpg?Expires=1792752590&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=mA98cN0WA3%2ByLJmVGL0Adq9B%2Bz0%3D)\r\n\r\n**（2）指令优化**\r\n- 向量化指令：\r\n  - SSE：128-bit 寄存器，4×float32\r\n  - AVX \u002F AVX-512：提供 256-bit \u002F 512-bit 向量指令支持，可在合适的硬件与算子条件下显著提升数据并行吞吐率。\r\n  - ARM NEON：128-bit SIMD\r\n- 张量化指令：\r\n  - NVIDIA Tensor Core：专用矩阵乘加单元\r\n  - MMA 指令格式\r\n\r\n**（3）内存优化**\r\n  - 预取：提前加载数据到高速缓存，隐藏内存访问延迟。\r\n  - 双缓冲：通过计算与数据传输的流水线并行，隐藏或降低 IO 等待带来的性能损失。\r\n\r\n### Auto Tuning 技术\r\n\r\n![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_35_zw07.jpg?Expires=1792752590&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=GKs9qREeOLL6nrWNSQX6YlDJfdY%3D)\r\n\r\n### 代码生成\r\n#### 1. Triton 语言介绍\r\n\r\n![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_35_zw08.jpg?Expires=1792752590&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=h68X5XiXaDDhgJkfwfb8HaVAWW8%3D)\r\n\r\n#### 2. 核心抽象概念\r\n**（1）Layout：** 描述张量元素如何映射到计算资源和内存单元上。\r\n**（2）编译器优化：** 自动进行内存访问合并、延迟掩盖、数据重排等优化。\r\n\r\n#### 3. 九齿\r\n- 简化 Triton 编程接口\r\n- 自动处理 PID 等底层细节\r\n- 提升开发效率，降低 CUDA 编程门槛\r\n\r\n![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_35_zw09.jpg?Expires=1792752590&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=H3VP%2FhdKtHpLdDA6k0kKtrTmNMY%3D)\r\n\r\n### 总结\r\n本节课讲述了内存管理优化、算子层面优化、Auto Tuning 和代码生成四个核心领域的关键技术和实践方法。\r\n\r\n\r\n![2025冬季训练营课程表](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_01_zw6.jpg?Expires=1792752590&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=HSupOOOCDz3YCQaxU2p%2B5BXFnLA%3D)\r\n\r\n### 💡 学习贴士\r\n更多细节及完整课程内容，请观看直播或查看课程回放。\r\n\r\n链接：https:\u002F\u002Fwww.infinitensor.com\u002Fcamp\u002Fwinter2025\r\n\r\n### 📚 训练营报名持续进行中\r\n扫码报名与加入群聊，或点击 **「阅读原文」** 直达官网报名\r\n\r\n\r\n![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_01_zw7.jpg?Expires=1792752590&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=supJZt2u6gSL0%2FcpvOzD6u%2BXpJM%3D)",{"publishTime":7,"createTime":109,"canonicalPath":114,"id":115,"title":116},"\u002Fstudy-material\u002Fstudy-34","study-34","InfiniTensor 冬季训练营 Triton & 九齿第三课《九齿二重：渐悟》",74,"\u002Fstudy-material\u002Fstudy-35","study-35","markdown",{"publishTime":7,"createTime":109,"canonicalPath":122,"id":123,"title":124},"\u002Fstudy-material\u002Fstudy-36","study-36","InfiniTensor 冬季训练营 AI 编译器第五课《拓展 Triton 深度学习编译器——DLCompiler》",{"htmlContent":126,"headings":127},"\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_35_zw01.jpg?Expires=1792752590&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=AGmoC6iJaL4xuVr%2B150SlYaX3eU%3D\" alt=\"\">\u003C\u002Fp>\n\u003Cp>2026 年 1 月 23 日，2025 冬季 InfiniTensor 大模型与人工智能系统训练营 AI 编译器系列第四课《AI 编译器中的后端优化》开讲。\u003C\u002Fp>\n\u003Cp>本课将重点讲解内存管理优化、算子层面优化、Auto Tuning 和代码生成四个核心内容。\u003C\u002Fp>\n\u003Ch3 id=\"内存管理优化\">\u003Ca class=\"anchor\" href=\"#内存管理优化\" aria-label=\"跳转到标题：内存管理优化\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>内存管理优化\u003C\u002Fh3>\u003Ch4 id=\"1-优化思路\">\u003Ca class=\"anchor\" href=\"#1-优化思路\" aria-label=\"跳转到标题：1. 优化思路\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>1. 优化思路\u003C\u002Fh4>\u003Cul>\n\u003Cli>调整分配和释放的时间点和次数，减少运行时的内存分配和释放开销。\u003C\u002Fli>\n\u003Cli>优化分配过程：减少总的内存占用。\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Ch4 id=\"2-naive-版本\">\u003Ca class=\"anchor\" href=\"#2-naive-版本\" aria-label=\"跳转到标题：2. Naive 版本：\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>2. Naive 版本：\u003C\u002Fh4>\u003Cp>运行到算子时即时申请内存，结束即释放，导致频繁的系统调用。\u003C\u002Fp>\n\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_35_zw02.jpg?Expires=1792752590&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=aUI6aNmo46myjnsERD5K13%2F29AU%3D\" alt=\"\">\u003C\u002Fp>\n\u003Ch4 id=\"3-优化版本\">\u003Ca class=\"anchor\" href=\"#3-优化版本\" aria-label=\"跳转到标题：3. 优化版本：\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>3. 优化版本：\u003C\u002Fh4>\u003Cul>\n\u003Cli>根据拓扑排序确定 OP 执行顺序。\u003C\u002Fli>\n\u003Cli>模拟执行过程，计算内存峰值。\u003C\u002Fli>\n\u003Cli>按需分配和释放内存，利用引用计数机制判断 Tensor 是否可释放。\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_35_zw03.jpg?Expires=1792752590&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=n4y%2B%2B%2BBbKjqXwiauvkv7OmNf%2BB0%3D\" alt=\"\">\u003C\u002Fp>\n\u003Ch4 id=\"4-惰性分配-lazy-allocation\">\u003Ca class=\"anchor\" href=\"#4-惰性分配-lazy-allocation\" aria-label=\"跳转到标题：4. 惰性分配（Lazy Allocation）\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>4. 惰性分配（Lazy Allocation）\u003C\u002Fh4>\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_35_zw04.jpg?Expires=1792752590&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=N0XrEaGlsv%2F%2FZTKqLO6srcmDXz8%3D\" alt=\"\">\u003C\u002Fp>\n\u003Ch3 id=\"算子层面优化\">\u003Ca class=\"anchor\" href=\"#算子层面优化\" aria-label=\"跳转到标题：算子层面优化\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>算子层面优化\u003C\u002Fh3>\u003Ch4 id=\"1-优化起点\">\u003Ca class=\"anchor\" href=\"#1-优化起点\" aria-label=\"跳转到标题：1. 优化起点\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>1. 优化起点\u003C\u002Fh4>\u003Cul>\n\u003Cli>理解瓶颈：区分计算密集型和访存密集型任务。\u003C\u002Fli>\n\u003Cli>诊断工具：使用 Roofline 模型确定优化方向。\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Ch4 id=\"2-常用优化策略\">\u003Ca class=\"anchor\" href=\"#2-常用优化策略\" aria-label=\"跳转到标题：2. 常用优化策略\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>2. 常用优化策略\u003C\u002Fh4>\u003Cp> \u003Cstrong>（1）循环优化\u003C\u002Fstrong>\u003C\u002Fp>\n\u003Cul>\n\u003Cli>循环分块：将大循环分解成小块，提升数据局部性。\u003C\u002Fli>\n\u003Cli>循环展开：增加单次循环迭代内执行的操作次数，减少循环控制开销。\u003C\u002Fli>\n\u003Cli>循环重排：改变循环顺序，匹配数据存储顺序，提升缓存利用率。\u003C\u002Fli>\n\u003Cli>循环融合：合并具有相同迭代范围和遍历顺序的独立循环，减少循环控制开销，提升数据局部性。\u003C\u002Fli>\n\u003Cli>循环拆分：将复杂循环拆分成多个简单循环，消除依赖，允许并行执行。\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_35_zw05.jpg?Expires=1792752590&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=DLBy8wo5DfhCMJVpXLpC2P%2BhAMY%3D\" alt=\"\">\u003C\u002Fp>\n\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_35_zw06.jpg?Expires=1792752590&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=mA98cN0WA3%2ByLJmVGL0Adq9B%2Bz0%3D\" alt=\"\">\u003C\u002Fp>\n\u003Cp>\u003Cstrong>（2）指令优化\u003C\u002Fstrong>\u003C\u002Fp>\n\u003Cul>\n\u003Cli>向量化指令：\u003Cul>\n\u003Cli>SSE：128-bit 寄存器，4×float32\u003C\u002Fli>\n\u003Cli>AVX \u002F AVX-512：提供 256-bit \u002F 512-bit 向量指令支持，可在合适的硬件与算子条件下显著提升数据并行吞吐率。\u003C\u002Fli>\n\u003Cli>ARM NEON：128-bit SIMD\u003C\u002Fli>\n\u003C\u002Ful>\n\u003C\u002Fli>\n\u003Cli>张量化指令：\u003Cul>\n\u003Cli>NVIDIA Tensor Core：专用矩阵乘加单元\u003C\u002Fli>\n\u003Cli>MMA 指令格式\u003C\u002Fli>\n\u003C\u002Ful>\n\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>\u003Cstrong>（3）内存优化\u003C\u002Fstrong>\u003C\u002Fp>\n\u003Cul>\n\u003Cli>预取：提前加载数据到高速缓存，隐藏内存访问延迟。\u003C\u002Fli>\n\u003Cli>双缓冲：通过计算与数据传输的流水线并行，隐藏或降低 IO 等待带来的性能损失。\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Ch3 id=\"auto-tuning-技术\">\u003Ca class=\"anchor\" href=\"#auto-tuning-技术\" aria-label=\"跳转到标题：Auto Tuning 技术\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>Auto Tuning 技术\u003C\u002Fh3>\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_35_zw07.jpg?Expires=1792752590&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=GKs9qREeOLL6nrWNSQX6YlDJfdY%3D\" alt=\"\">\u003C\u002Fp>\n\u003Ch3 id=\"代码生成\">\u003Ca class=\"anchor\" href=\"#代码生成\" aria-label=\"跳转到标题：代码生成\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>代码生成\u003C\u002Fh3>\u003Ch4 id=\"1-triton-语言介绍\">\u003Ca class=\"anchor\" href=\"#1-triton-语言介绍\" aria-label=\"跳转到标题：1. Triton 语言介绍\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>1. Triton 语言介绍\u003C\u002Fh4>\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_35_zw08.jpg?Expires=1792752590&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=h68X5XiXaDDhgJkfwfb8HaVAWW8%3D\" alt=\"\">\u003C\u002Fp>\n\u003Ch4 id=\"2-核心抽象概念\">\u003Ca class=\"anchor\" href=\"#2-核心抽象概念\" aria-label=\"跳转到标题：2. 核心抽象概念\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>2. 核心抽象概念\u003C\u002Fh4>\u003Cp>\u003Cstrong>（1）Layout：\u003C\u002Fstrong> 描述张量元素如何映射到计算资源和内存单元上。\u003Cbr>\u003Cstrong>（2）编译器优化：\u003C\u002Fstrong> 自动进行内存访问合并、延迟掩盖、数据重排等优化。\u003C\u002Fp>\n\u003Ch4 id=\"3-九齿\">\u003Ca class=\"anchor\" href=\"#3-九齿\" aria-label=\"跳转到标题：3. 九齿\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>3. 九齿\u003C\u002Fh4>\u003Cul>\n\u003Cli>简化 Triton 编程接口\u003C\u002Fli>\n\u003Cli>自动处理 PID 等底层细节\u003C\u002Fli>\n\u003Cli>提升开发效率，降低 CUDA 编程门槛\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_35_zw09.jpg?Expires=1792752590&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=H3VP%2FhdKtHpLdDA6k0kKtrTmNMY%3D\" alt=\"\">\u003C\u002Fp>\n\u003Ch3 id=\"总结\">\u003Ca class=\"anchor\" href=\"#总结\" aria-label=\"跳转到标题：总结\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>总结\u003C\u002Fh3>\u003Cp>本节课讲述了内存管理优化、算子层面优化、Auto Tuning 和代码生成四个核心领域的关键技术和实践方法。\u003C\u002Fp>\n\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_01_zw6.jpg?Expires=1792752590&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=HSupOOOCDz3YCQaxU2p%2B5BXFnLA%3D\" alt=\"2025冬季训练营课程表\">\u003C\u002Fp>\n\u003Ch3 id=\"学习贴士\">\u003Ca class=\"anchor\" href=\"#学习贴士\" aria-label=\"跳转到标题：💡 学习贴士\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>💡 学习贴士\u003C\u002Fh3>\u003Cp>更多细节及完整课程内容，请观看直播或查看课程回放。\u003C\u002Fp>\n\u003Cp>链接：\u003Ca href=\"https:\u002F\u002Fwww.infinitensor.com\u002Fcamp\u002Fwinter2025\">https:\u002F\u002Fwww.infinitensor.com\u002Fcamp\u002Fwinter2025\u003C\u002Fa>\u003C\u002Fp>\n\u003Ch3 id=\"训练营报名持续进行中\">\u003Ca class=\"anchor\" href=\"#训练营报名持续进行中\" aria-label=\"跳转到标题：📚 训练营报名持续进行中\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>📚 训练营报名持续进行中\u003C\u002Fh3>\u003Cp>扫码报名与加入群聊，或点击 \u003Cstrong>「阅读原文」\u003C\u002Fstrong> 直达官网报名\u003C\u002Fp>\n\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_01_zw7.jpg?Expires=1792752590&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=supJZt2u6gSL0%2FcpvOzD6u%2BXpJM%3D\" alt=\"\">\u003C\u002Fp>\n",[128,130,134,137,140,143,145,148,151,154,156,159,162,165,167,170],{"id":129,"text":129,"level":9},"内存管理优化",{"id":131,"text":132,"level":133},"1-优化思路","1. 优化思路",4,{"id":135,"text":136,"level":133},"2-naive-版本","2. Naive 版本：",{"id":138,"text":139,"level":133},"3-优化版本","3. 优化版本：",{"id":141,"text":142,"level":133},"4-惰性分配-lazy-allocation","4. 惰性分配（Lazy Allocation）",{"id":144,"text":144,"level":9},"算子层面优化",{"id":146,"text":147,"level":133},"1-优化起点","1. 优化起点",{"id":149,"text":150,"level":133},"2-常用优化策略","2. 常用优化策略",{"id":152,"text":153,"level":9},"auto-tuning-技术","Auto Tuning 技术",{"id":155,"text":155,"level":9},"代码生成",{"id":157,"text":158,"level":133},"1-triton-语言介绍","1. Triton 语言介绍",{"id":160,"text":161,"level":133},"2-核心抽象概念","2. 核心抽象概念",{"id":163,"text":164,"level":133},"3-九齿","3. 九齿",{"id":166,"text":166,"level":9},"总结",{"id":168,"text":169,"level":9},"学习贴士","💡 学习贴士",{"id":171,"text":172,"level":9},"训练营报名持续进行中","📚 训练营报名持续进行中"]