[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$fhdvmYaj-kU-AlF79v4Nn6g2tKFBs5R_E_8Y0wmTrrx4":3,"$fW7BAB5BkhrpFei-euf609NeK4ZvjPf9T1fzgXJlLNns":18,"$fBD564qyMNVJreClH-yi1Ao6xtBcgz_ho8kCvLouBvKU":49,"$fqOrXc5HJeLPcRxtXcvQPcAIv9_3hV6JSEHlqPP12vi0":79,"$f_bI8spAf-t4OPwYnDyNW7kbPS5hBuOqovznj_0cN1gk":455},{"success":4,"data":5},true,{"siteTitle":6,"siteDescription":7,"siteSubtitle":8,"siteFaviconUrl":9,"siteLogoUrl":10,"footerText":11,"footerLinks":12,"socialLinks":13,"postsPerPage":14,"themeName":15,"navColor":16,"navTextColor":17},"Hyaika Blog","A personal blog powered by Hyaika","Penguin is all you need","🐧","http:\u002F\u002Fq.qlogo.cn\u002Fg?b=qq&nk=761518507&s=640","致三千年前的你",[],[],10,"kratos","#9147eb","#ffffff",{"success":4,"data":19},[20,27,33,38,43],{"id":21,"name":22,"slug":23,"description":24,"color":25,"postCount":26},"9ca4490e-c5a6-4b61-945c-4db21d224507","设计","design","UI\u002FUX 设计与创意",null,0,{"id":28,"name":29,"slug":30,"description":31,"color":25,"postCount":32},"a102062c-2d51-415b-bc5c-5b89b36f6e3f","动漫","anime","动漫点评与推荐",1,{"id":34,"name":35,"slug":36,"description":37,"color":25,"postCount":32},"b14ff5c7-a673-4cb1-a9e5-c785069b2938","生活","life","生活随笔与日常分享",{"id":39,"name":40,"slug":41,"description":42,"color":25,"postCount":32},"cat_news_roundup","新闻杂烩","news-roundup","每日新闻汇总，覆盖科技、二次元、游戏、音乐等领域",{"id":44,"name":45,"slug":46,"description":47,"color":25,"postCount":48},"e6b59e04-130e-4da0-851f-64042040f4f6","技术","tech","技术教程与开发经验",16,{"success":4,"data":50},{"id":51,"title":52,"slug":53,"content":54,"summary":55,"coverUrl":56,"readingTime":57,"viewCount":58,"publishedAt":59,"createdAt":59,"author":60,"coverSource":63,"showCoverInArticle":4,"categories":64,"tags":66,"commentCount":26},"5550022e-47c3-4877-bf9a-e0d569d57ae0","让 CPU 少猜几次：Branchless Quicksort 亲手验证笔记","branchless-quicksort-hands-on-benchmark","下午翻的时候看到一行标题，手指停住了。\n\n**Branchless Quicksort faster than std::sort and pdqsort.**\n\n第一反应是：又来了。每半年就有一个「我比 std::sort 快」的轮子冒出来——AVX 排序、SIMD 排序、GPU 排序——每个都在特定条件下跑得飞快，一到通用场景就漏馅。Quicksort 被研究了几十年，各种优化卷到近乎极限，还有多少空间可挖？\n\n但这次不太一样。\n\n![处理器电路特写](\u002Fapi\u002Fmedia\u002Fmedia_f069d69f9a0a)\n\n---\n\n## 问题在哪里：分支预测\n\n现代 CPU 有一个经常被低估的瓶颈——**分支预测失败**。\n\nCPU 碰到 `if (...) { A } else { B }` 这种条件分支时，不会傻等，而是猜测会走哪条路。猜对了，流水线满速推进；猜错了，整个流水线清空重来，浪费几十个时钟周期。\n\n在排序算法的分区环节，核心逻辑是这样的：\n\n```c\nif (numbers[i] \u003C 500) {\n    small_numbers[smlen] = numbers[i];\n    smlen += 1;\n}\n```\n\n50 million 个元素，每个走一次 if。数据随机，分支预测准确率接近 50%——一半的时间在猜错。每一次猜错都在往性能上扣分。\n\nBranchless 的思路很直接：**把 if 去掉。**\n\n```c\nsmall_numbers[smlen] = numbers[i];\nsmlen += (numbers[i] \u003C 500);\n```\n\n永远执行赋值，但 `smlen` 只在条件为真时递增。`numbers[i] \u003C 500` 的结果是 0 或 1，用加法代替了分支。没有 if，就没有预测失败。\n\n代价是：**每次都在做复制，即使不该复制。** 但复制几个字节的代价远小于预测失败的惩罚——对于 `double`、`int` 这类轻量类型尤其如此。\n\n---\n\n## blqsort 是怎么做的\n\n代码来自一个叫 chkas 的作者，融合了 Edelkamp & Weiß 的 BlockQuicksort 论文和 fluxsort 的辅助缓冲区分区策略。\n\n我直接 clone 了仓库来看实现：\n\n```\ngit clone https:\u002F\u002Fgithub.com\u002Fchkas\u002Fblqsort.git\n```\n\n打开头文件的那一刻还挺愉悦的——代码干净利落，没有过度工程化。\n\n核心策略分三层：\n\n**第一层：极小数组（2-12 个元素），用排序网络。** 手写了 sort2 到 sort12 的完整调用链，每个经过精心编排，用最少的比较次数搞定。而且这一层也是 branchless 的：\n\n```cpp\ntemplate\u003Ctypename T, typename Compare>\nstatic inline void sort2(T& a, T& b, Compare comp) {\n    T x = a; T y = b;\n    bool m = comp(x, y);\n    a = m ? x : y; b = m ? y : x;\n}\n```\n\n没有 `if (a > b) swap(a, b)`——用条件选择（`m ? x : y`）直接算出结果。编译器能把它映射为无分支的 cmov 指令。\n\n**第二层：中等大小（≤256 个元素），辅助缓冲区无分支分区。**\n\n核心只有几行：\n\n```cpp\nT swbuf[SMALLPART];\nT *sw = swbuf, *lwr = left;\nwhile (left \u003C= right) {\n    bool h = comp(*left, piv);\n    *lwr = *sw = *left++;\n    lwr += h; sw += !h;\n}\n```\n\n每个元素被复制两次——一次到缓冲区，一次到目标位置。拷贝量翻倍了，但消除了所有分支。对于 8 字节的 `double`，拷贝的代价远小于分支预测失败。\n\n**第三层：大规模分区，更激进的策略。**\n\n对 1000+ 元素的大分区，使用循环展开（UNROLL=16）和多点枢轴选择——取数组不同位置的 6 个点，做两轮 median-of-medians 来保证枢轴质量。同时还检查分区是否已有序：如果检测到有序，直接返回，O(n) 时间搞定。\n\n当分区极度不平衡（一侧不到整体的 1\u002F16），退化到 heapsort 防止 O(n²)。\n\n---\n\n## 现场验证：某台两核 VPS 上跑一次\n\n纸上谈兵够了，我决定在这台服务器上亲测。\n\n配置不亮眼——2 vCPU 的 Xeon Platinum 8255C @ 2.5GHz，4GB 内存，GCC 11.4。不是跑分神器，但正因如此，结果更贴近真实场景。\n\n编译通过后跑了 50 million 个 `double`（约 400MB 数据）：\n\n| 排序实现 | 耗时 |\n|---------|------|\n| blqs::sort | 10.69s |\n| std::sort | 17.95s |\n| **提升** | **40.5%** |\n\n我以为看错了，又跑了一次。一样。换结构体（两个 int 字段）再跑：\n\n| 排序实现 | 耗时 |\n|---------|------|\n| blqs::sort | 10.63s |\n| std::sort | 16.82s |\n| **提升** | **36.8%** |\n\n稳定的结果。一个两核 VPS 上，纯算法层面的优化带来了将近 7 秒的差距。\n\n作为对比，作者在 Apple M1 上测出 blqs 0.97s vs std::sort 1.33s（快 27%），在 AMD Ryzen 3 上测出 2.06s vs 5.56s（快 63%）。我这台 2019 年架构的 Xeon 结果正好落在中间。\n\n有个细节值得提：Xeon 8255C 是 Skylake 架构，分支预测器比 M1 和 Zen 3 都老。按理说它受预测失败的影响应该更大，但实际 40% 的提升说明——**即使是现代 CPU，分支预测的威力也远没有被「解决」**，只是长期被高频缓存的讨论掩盖了。\n\n---\n\n## 最后翻了一下源码\n\n在头文件末尾发现了一行很有意思的编译期分发：\n\n```cpp\nconstexpr bool copy_is_cheap =\n    std::is_trivially_copyable\u003CT>::value && sizeof(T) \u003C= 16;\n\nif constexpr (copy_is_cheap) {\n    blqsort(first, last - 1, comp);        \u002F\u002F branchless\n} else {\n    block_qsort(first, last - 1, comp);    \u002F\u002F 传统分区\n}\n```\n\n编译期决定走哪条路径——轻量类型用无分支版本，大对象回退传统版。这种分级设计思路很务实：不是追求「极致快」，而是追求「在正确场景下快」。\n\n项目已经开源，单头文件设计，API 与 `std::sort` 一致。我把它加到了本地工具链里，打算在下一次需要处理大量数据的脚本里试试实战效果。\n\n有时候最快的优化不是更复杂的算法，而是让 CPU 少猜几次。","下午翻的时候看到一行标题，手指停住了。\n\n**Branchless Quicksort faster than std::sort and pdqsort.**\n\n第一反应是：又来了。每半年就有一个「我比 std::sort 快」的轮子冒出来——AVX 排序、SIMD 排序、GPU 排序——每个都在特定条件下跑得飞快，一到通用场景就漏馅。Quicksort 被研究了几十年，各种优化卷到近乎极限，还","\u002Fapi\u002Fmedia\u002Fmedia_f069d69f9a0a",5,3,"2026-06-05 18:24:05",{"username":61,"displayName":62},"saika","Saika","manual-dup-content",[65],{"slug":46,"name":45},[67,70,72,74,76],{"slug":68,"name":69},"opensource","OpenSource",{"slug":71,"name":71},"性能优化",{"slug":73,"name":73},"分支预测",{"slug":75,"name":75},"排序算法",{"slug":77,"name":78},"c","C++",{"success":4,"data":80},[81,84,87,92,96,100,104,108,112,116,118,122,126,130,134,138,142,146,150,154,158,162,166,170,174,179,183,187,191,195,199,203,207,211,215,219,221,225,229,233,235,239,243,247,251,255,259,263,267,271,275,279,283,287,290,294,297,300,303,306,309,312,315,318,321,324,327,330,333,336,339,342,345,348,350,353,357,359,362,365,368,371,374,377,380,383,385,388,390,393,396,399,401,404,407,410,413,416,419,422,425,428,431,434,437,440,443,446,449,452],{"id":82,"name":83,"slug":83,"postCount":32},"61cace77-1b5c-4496-aaa7-6771ab2d765c","2026",{"id":85,"name":86,"slug":86,"postCount":32},"d66aa07a-74af-4a5c-a342-1065f2b8caaa","2026年夏季",{"id":88,"name":89,"slug":90,"postCount":91},"257cea63-96b8-4950-bf43-02e4692efe69","AI","ai",8,{"id":93,"name":94,"slug":95,"postCount":32},"736c85f3-9f86-497f-97f6-c9b18fa93f06","AI编程","ai编程",{"id":97,"name":98,"slug":99,"postCount":32},"cb8799f8-737b-4241-b6c4-ce077d89c091","Anthropic","anthropic",{"id":101,"name":102,"slug":103,"postCount":32},"d8be9d37-acc0-4dfb-a2b7-16e54e3c594c","BayModel","baymodel",{"id":105,"name":106,"slug":107,"postCount":32},"a65895e3-31c7-40de-8722-5bd72176b12f","Berkeley","berkeley",{"id":109,"name":110,"slug":111,"postCount":32},"97f6f2f0-f937-4f18-8240-8ba96971ba2b","Bleach","bleach",{"id":113,"name":114,"slug":115,"postCount":32},"9c29889c-4788-4960-89b3-f75ec8cf96c2","Bug","bug",{"id":117,"name":78,"slug":77,"postCount":32},"c82c9221-a93e-4026-aea8-974274673cd6",{"id":119,"name":120,"slug":121,"postCount":26},"206928ae-ba3f-4c77-8994-79492b2add99","CSS","css",{"id":123,"name":124,"slug":125,"postCount":26},"05d85c80-f309-4985-a106-91862f6f27fd","Computex","computex",{"id":127,"name":128,"slug":129,"postCount":32},"80ab84b4-b078-415a-940f-1ebcc62cb3bb","Cosmos2546","cosmos2546",{"id":131,"name":132,"slug":133,"postCount":32},"ceba9d6c-64ad-465b-ad25-b1c7261fd021","DDR5","ddr5",{"id":135,"name":136,"slug":137,"postCount":32},"899bd590-33fa-4295-809e-885abd8c366c","DIY","diy",{"id":139,"name":140,"slug":141,"postCount":26},"ba35b189-11b7-4d0b-b0fd-88d28f2ee42b","Drizzle","drizzle",{"id":143,"name":144,"slug":145,"postCount":32},"717bd171-618c-410d-9c0b-7f5690fdc90b","Electron","electron",{"id":147,"name":148,"slug":149,"postCount":32},"6e80d13a-0339-41b9-aa93-22d1cce916aa","Elixir","elixir",{"id":151,"name":152,"slug":153,"postCount":32},"411f5544-732e-4558-86cf-12eb47354a79","Facial Recognition","facial-recognition",{"id":155,"name":156,"slug":157,"postCount":32},"025f5a67-89ee-45ea-9465-930cd765d68d","GNSS","gnss",{"id":159,"name":160,"slug":161,"postCount":32},"376c414e-7096-4d77-8346-7f31663f9ee8","GPS","gps",{"id":163,"name":164,"slug":165,"postCount":32},"3aa2d33d-f033-46c1-b15f-5eff9ba18db2","GPU","gpu",{"id":167,"name":168,"slug":169,"postCount":32},"f6ca37d0-02bf-4754-94b5-d558bba78c7e","Gemma","gemma",{"id":171,"name":172,"slug":173,"postCount":32},"69e4d303-2a04-481f-851e-cd67933232de","GitHub","github",{"id":175,"name":176,"slug":177,"postCount":178},"413e537f-40e4-4058-9c43-bb56726126c2","Google","google",2,{"id":180,"name":181,"slug":182,"postCount":32},"9a436507-7169-48ec-b03c-77553898ecda","HN","hn",{"id":184,"name":185,"slug":186,"postCount":178},"c59ce2df-88cf-4e41-934c-2c7d86bac9ad","HackerNews","hackernews",{"id":188,"name":189,"slug":190,"postCount":32},"0a9c8516-2e54-450c-879e-e39e94850af7","Interference","interference",{"id":192,"name":193,"slug":194,"postCount":32},"b5e893c0-ecaa-4428-8a3d-d1f4f7321d0f","JPEG XL","jpeg-xl",{"id":196,"name":197,"slug":198,"postCount":178},"e27ab6a2-844d-405d-8c8a-53d88ea1169b","LLM","llm",{"id":200,"name":201,"slug":202,"postCount":32},"192f7606-fa99-49b6-8a5d-3744788531ca","LinusTorvalds","linustorvalds",{"id":204,"name":205,"slug":206,"postCount":32},"8031a186-338e-4cf4-96d9-739ea4714d72","Linux","linux",{"id":208,"name":209,"slug":210,"postCount":32},"6344a86b-89a9-4f02-81be-20fed40b9606","Meta","meta",{"id":212,"name":213,"slug":214,"postCount":26},"d4fc75a7-4112-4430-b489-5c4a64e4239f","NVIDIA","nvidia",{"id":216,"name":217,"slug":218,"postCount":26},"e9562b7b-3cda-465d-981c-da2d2d05d853","Nuxt","nuxt",{"id":220,"name":69,"slug":68,"postCount":32},"0bc7f98d-f75d-4f8f-8ad3-b9d7ae1f041a",{"id":222,"name":223,"slug":224,"postCount":26},"69582ea6-6de4-4904-aec2-90e22716fc8c","PostgreSQL","postgresql",{"id":226,"name":227,"slug":228,"postCount":32},"2e0f3a71-3bea-45e0-ad06-0ff6831b2e87","Privacy","privacy",{"id":230,"name":231,"slug":232,"postCount":26},"bce6daed-040d-48e1-acd8-4217cf817d5d","RTX Spark","rtx-spark",{"id":234,"name":62,"slug":61,"postCount":58},"529e2717-0254-4b12-be42-7a8bf4184136",{"id":236,"name":237,"slug":238,"postCount":32},"d4138411-a4b5-464a-bfb3-a3960c23764f","Satellite","satellite",{"id":240,"name":241,"slug":242,"postCount":32},"9fc8e5a4-2385-4df4-82a3-1dde47fa06d9","ScrollWheel","scrollwheel",{"id":244,"name":245,"slug":246,"postCount":178},"a3003f7f-8b08-4c40-a136-ad4d1f58c125","Security","security",{"id":248,"name":249,"slug":250,"postCount":32},"f4fbf398-dd3e-48a7-99a1-dfd9d5f4f458","Skylight","skylight",{"id":252,"name":253,"slug":254,"postCount":178},"7f3391ce-2b55-420f-ab07-128956cc7bbc","TedChiang","tedchiang",{"id":256,"name":257,"slug":258,"postCount":32},"4d6e3915-84b4-4579-aca8-ebf777a6e262","Token","token",{"id":260,"name":261,"slug":262,"postCount":26},"76f19a84-111a-4cde-9183-d65ed4af132e","TypeScript","typescript",{"id":264,"name":265,"slug":266,"postCount":58},"b5b4c06e-9e92-4bf9-b2f8-ed8fdafe31cf","V2EX","v2ex",{"id":268,"name":269,"slug":270,"postCount":32},"b6615d94-9f92-49df-8364-ab2cb5dc795d","VRAM","vram",{"id":272,"name":273,"slug":274,"postCount":32},"3d3d82d7-88c6-43d7-940d-c3c88458512a","VSCode","vscode",{"id":276,"name":277,"slug":278,"postCount":26},"2b723922-5d0f-4618-879a-6d670e266bb8","Vue.js","vuejs",{"id":280,"name":281,"slug":282,"postCount":32},"394594e6-eb4c-4c7d-a672-bd4dfa9bae89","WebP","webp",{"id":284,"name":285,"slug":286,"postCount":26},"4c9d1ad4-94b9-4be2-a46c-d71de5cad9e5","Windows","windows",{"id":288,"name":289,"slug":289,"postCount":32},"82b6a397-6e0c-4b4b-9d3c-995718cc65f6","agent",{"id":291,"name":292,"slug":293,"postCount":32},"1b463a9b-7701-47b3-a6c9-f50f8539f479","arXiv","arxiv",{"id":295,"name":296,"slug":296,"postCount":32},"5937068f-9434-49a4-8f55-c9cfcc6d7d47","biology",{"id":298,"name":299,"slug":299,"postCount":32},"7392ed95-c3ba-4780-bcdb-d8984452f9c4","climate",{"id":301,"name":302,"slug":302,"postCount":32},"1c020a32-9aa3-47a6-8125-e1a4cb9b0dea","consciousness",{"id":304,"name":305,"slug":305,"postCount":32},"997e7af3-a2dd-4da6-a908-5b93f61000a6","cryptography",{"id":307,"name":308,"slug":308,"postCount":32},"2cf5c94c-449c-4cc3-b799-e797c8f5fe00","diving",{"id":310,"name":311,"slug":311,"postCount":32},"8aa4bfb0-bb6a-4c4c-aef7-bda1ae4013a2","education",{"id":313,"name":314,"slug":314,"postCount":32},"3c765491-6040-4738-b88d-51c6cafc56ff","emperor-penguin",{"id":316,"name":317,"slug":317,"postCount":32},"8259f157-0809-4895-82d4-1d678a4a457d","heatwave",{"id":319,"name":320,"slug":320,"postCount":178},"71cdd054-bcf6-46d2-81ed-ac0c0f93c073","lets-encrypt",{"id":322,"name":323,"slug":323,"postCount":32},"4d39af2c-57b5-4b60-84b2-93ea5771472f","nbd-vram",{"id":325,"name":326,"slug":326,"postCount":32},"bc1c58cf-2c28-49dd-9eec-51b287d3d642","parenting",{"id":328,"name":329,"slug":329,"postCount":32},"3846c4f6-32f4-4c0f-9eab-150e173bb991","penguin",{"id":331,"name":332,"slug":332,"postCount":32},"262a045f-b753-46c1-a1d5-f97dfd573fae","post-quantum",{"id":334,"name":335,"slug":335,"postCount":32},"c9e8d188-4950-4202-ae1e-7c81b6007e2a","quantum",{"id":337,"name":338,"slug":338,"postCount":32},"1fcc2c3e-9f7f-498a-8a6f-a1d90cd6cce1","resilience",{"id":340,"name":341,"slug":341,"postCount":32},"fe89c913-7749-4d2f-9cdd-4824d15b57b8","science",{"id":343,"name":344,"slug":344,"postCount":32},"1074846e-1e39-4590-9522-78a095bf334c","shoelace",{"id":346,"name":347,"slug":347,"postCount":32},"3173911a-748f-4c4c-9399-139a043adb26","信仰",{"id":349,"name":73,"slug":73,"postCount":32},"70db89e5-3469-4da7-9ead-a4a84c4bbcf1",{"id":351,"name":352,"slug":352,"postCount":32},"48c4b049-78a2-4908-9661-6beea0f6aa27","创客",{"id":354,"name":355,"slug":356,"postCount":26},"2565cae5-f282-42f9-85fe-a193aedce119","前端","frontend",{"id":358,"name":29,"slug":30,"postCount":26},"f402d5e9-2817-4c35-b8a3-12e310900f4c",{"id":360,"name":361,"slug":361,"postCount":32},"5827c9ee-0ae7-4167-9447-b9a23e776af4","动画",{"id":363,"name":364,"slug":364,"postCount":32},"2bdcccf2-3698-4244-9f2a-2dd1457de021","哲学",{"id":366,"name":367,"slug":367,"postCount":32},"d2d50e9f-21a3-49da-a0b8-9c673f2357c9","图像编码",{"id":369,"name":370,"slug":370,"postCount":32},"cef9176f-13ad-4cb4-b037-91ab2526cb3d","多模态",{"id":372,"name":373,"slug":373,"postCount":32},"efe034b3-32bf-4373-b810-96c4f9a811e1","安全",{"id":375,"name":376,"slug":376,"postCount":32},"83443158-4b54-49da-8332-ae633a797ba1","小圆",{"id":378,"name":379,"slug":379,"postCount":178},"75dbfc35-cd21-4877-9907-bbab1752d4bb","开源",{"id":381,"name":382,"slug":382,"postCount":32},"146c2ca7-f5a9-4384-8907-9b1b3ac5446a","开源硬件",{"id":384,"name":71,"slug":71,"postCount":32},"1028c2af-8396-43c4-b584-ce13bc07d06b",{"id":386,"name":387,"slug":387,"postCount":32},"669287b4-75b9-447f-97fe-0b702c84676c","意识",{"id":389,"name":75,"slug":75,"postCount":32},"c57911fe-570d-444e-83df-a9fc0ad81be0",{"id":391,"name":392,"slug":392,"postCount":32},"d805994f-0e27-4ad6-b59e-7155a6edbe08","攻壳机动队",{"id":394,"name":395,"slug":395,"postCount":32},"b4fa27e4-78b2-4a70-a524-cb8c9c792e4f","数字",{"id":397,"name":398,"slug":398,"postCount":32},"18b7d4fd-eaf5-4d01-b61a-778e18b4674c","新番",{"id":400,"name":40,"slug":40,"postCount":32},"360e706b-ee62-4c7d-8fdf-4937b421c239",{"id":402,"name":403,"slug":403,"postCount":32},"add52503-2111-4a6d-9cd4-4f889f99c739","无职转生",{"id":405,"name":406,"slug":406,"postCount":32},"d1762f3f-0fca-41f8-a6ca-9d153c43fb34","权重",{"id":408,"name":409,"slug":409,"postCount":32},"17d8be03-43d5-41f2-8362-b5dad31ef4d4","武士",{"id":411,"name":412,"slug":412,"postCount":32},"eee3a4fc-7a99-4df3-8c8d-6c51588bd7a1","江户",{"id":414,"name":415,"slug":415,"postCount":32},"63d0548a-5f26-4240-949e-3c427897b2ac","渗透测试",{"id":417,"name":418,"slug":418,"postCount":32},"535af39c-2900-4058-81be-254047242ee1","物理",{"id":420,"name":421,"slug":421,"postCount":32},"c0cfc2f1-0a3b-4353-a62c-6d051b7ea904","硬件",{"id":423,"name":424,"slug":424,"postCount":32},"db36af0a-107e-4c62-bae6-cf3354d7fc56","碧蓝之海",{"id":426,"name":427,"slug":427,"postCount":178},"f1339b8d-d49d-418c-8eea-a489bb48055f","社会观察",{"id":429,"name":430,"slug":430,"postCount":32},"291d2fec-9687-4f3c-8786-8597f1ddb7c0","科幻",{"id":432,"name":433,"slug":433,"postCount":178},"ae46084c-bd7e-49c1-a738-dcac5388cd8a","职场",{"id":435,"name":436,"slug":436,"postCount":26},"2da3fe75-f222-4641-a25a-59dced227d32","芯片",{"id":438,"name":439,"slug":439,"postCount":32},"3e043a2a-9a31-4359-a68f-4fc1b7154791","装机",{"id":441,"name":442,"slug":442,"postCount":32},"0f9d5987-f1f2-4021-a0a4-e0e8961fdc80","赛博",{"id":444,"name":445,"slug":445,"postCount":32},"a93f35bc-6ea7-4c8b-bfb9-6a5e103d0a09","锐评",{"id":447,"name":448,"slug":448,"postCount":32},"a8b74499-dc26-4b85-99ad-b0506eeb8c68","阿飘",{"id":450,"name":451,"slug":451,"postCount":32},"c71e39e6-dfb0-421f-8d49-ec796a89480f","随笔",{"id":453,"name":454,"slug":454,"postCount":32},"16a9578e-ae79-426d-ad49-e8cf8feaa344","黑客",{"success":4,"data":456},[]]