跳转至

2026

  • 分类于 Biweekly-en
  • 需要 6 分钟阅读时间

[XiangShan Biweekly 96] 20260216

Welcome to XiangShan biweekly column! Through this column, we will regularly share the latest development progress of XiangShan. This is the 96th issue of the biweekly report.

Happy Chinese New Year! Welcome to our biweekly column during the Spring Festival, and we wish you in the new year:

  • Frontend has the precise prediction like a unicorn, and career opportunities are never misjudged;
  • The pipeline runs like a thousand horses, everything goes smoothly and in parallel;
  • The memory subsystem is as vigorous as a dragon horse, and the happy data is inexhaustible;
  • The memory access path is like a vast grassland, and the happy address is always mapped in the heart;
  • The cache hierarchy is as warm as spring, and every precious memory is within reach;
  • The bus bandwidth is like a thousand-mile horse, and the good luck signal arrives immediately at your side!

We also shared the wonderful review of the tutorial hosted by XiangShan team at HPCA 2026. Please visit https://tutorial.xiangshan.cc/hpca26/ to review the content of this tutorial and get the slides. The next tutorial will be held at ISCA 2026 in Raleigh, North Carolina, USA in late June, and we look forward to seeing you again!

Regarding the recent development progress of XiangShan, ~~the XiangShan team is also having a happy Chinese New Year.~~ For the limited details, please see the recent progress section.

  • 分类于 Biweekly
  • 需要 4 分钟阅读时间

【香山双周报 96】20260216 期

欢迎来到香山双周报专栏，我们将通过这一专栏定期介绍香山的开发进展。本次是第 96 期双周报。

祝大家新春快乐！欢迎在新春佳节来到我们的双周报专栏，祝您在新的一年：

  • 前端拥有天马行空般的精准预测，事业机遇从不误判；
  • 流水线奔腾如万马千军，万事顺遂并行不悖；
  • 存储子系统如龙马精神永葆活力，快乐数据取之不尽；
  • 访存通路如辽阔草原，幸福地址永远映射在心间；
  • 缓存层次温暖如春，每一份珍贵回忆都触手可及；
  • 总线带宽如千里骏马，好运信号即刻抵达身边！

我们还为大家分享了香山团队在 HPCA 2026 上举办 tutorial 的精彩回顾，欢迎大家访问 https://tutorial.xiangshan.cc/hpca26/ 回顾本次 tutorial 的内容、获取 slides。下一场 tutorial 将于 6 月下旬在美国北卡罗莱纳州罗利市举办的 ISCA 2026 会议上进行，非常期待与大家再次相见！

关于香山近期开发进展，~~香山的同学们也在度过快乐的春节。~~不多的细节请见近期进展一节。

  • 分类于 Biweekly-en
  • 需要 6 分钟阅读时间

[XiangShan Biweekly 95] 20260202

Welcome to XiangShan biweekly column! Through this column, we will regularly share the latest development progress of XiangShan. This is the 95th issue of the biweekly report.

The high-performance DDR4 memory controller IP, Baiyang, developed by the XiangShan team has been officially released! If you haven't read it yet, please check out our Baiyang release article for more details. Here, we would like to share an exclusive story with you. On January 31st, XiangShan presented a tutorial at HPCA 2026, which included an introduction to Baiyang. The night before the presentation, Baiyang was still being prepared for open source, and the repository was made public just before the tutorial started the next day. ~~Deadlines are indeed the best productivity boosters~~.

Last week, we introduced new GCC15 and XSCC compilers. These two compilers offer more than 10% performance improvement compared to the existing GCC12. Now, XiangShan's SPEC CPU2006 performance has reached 18.5 points/GHz. In this issue of the biweekly report, we provide a comparative analysis of different compilers. In future development, we will gradually switch to GCC15 and XSCC compilers, while focusing more on compiler and hardware co-optimization. The specific scores for different compilers are still in the performance evaluation section, so stay tuned!

In terms of recent development progress, on the frontend side, MBTB has introduced the LRU replacement algorithm and uses accurate prediction results from TAGE-SC for updates to improve branch prediction accuracy. On the backend side, an I2F functional unit has been added to support i2f type instructions of FMV and FCVT, and og1Payload has been added to the integer IQ to optimize selection timing. In terms of memory access and cache, the timeout judgment logic in Sbuffer has been fixed, and the timeout threshold is configured through SMBLOCKCTL in CSR. For more details, please refer to the recent progress section.

  • 分类于 Biweekly
  • 需要 4 分钟阅读时间

【香山双周报 95】20260202 期

欢迎来到香山双周报专栏，我们将通过这一专栏定期介绍香山的开发进展。本次是第 95 期双周报。

香山团队的高性能 DDR4 内存控制器 IP 白杨正式发布了！如果还没有看过，欢迎大家阅读我们的白杨发布文章了解更多细节。在这里，我们与大家独家分享一个小故事。香山于 1 月 31 日在 HPCA 2026 上作 tutorial，其中有涉及到白杨的介绍。演讲的前一天晚上，白杨还在为开源作最后的准备，在第二天 tutorial 开始前才将仓库转为 public，~~果然 DDL 是第一生产力~~。

在上周，我们引入了新的 GCC15 和 XSCC 编译器。这两个编译器相比现有的 GCC12 有着 10% 以上的性能提升，现在香山的 SPEC CPU2006 性能已经达到了 18.5 分/GHz。这一期双周报中，我们为不同编译器进行了对比分析。在之后的开发中，我们将逐渐切换到 GCC15 和 XSCC 编译器，同时更加注重编译器和硬件的协同优化。不同编译器的具体分数仍然在性能评估一节，欢迎大家持续关注！

关于香山近期开发进展，前端方面，MBTB 引入了 LRU 替换算法，并使用 TAGE-SC 的精确预测结果进行更新，以提升分支预测的准确率。后端方面，添加了 I2F 功能单元以支持 FMV 和 FCVT 的 i2f 类型指令，并对整数 IQ 添加 og1Payload 以优化选择时序。访存与缓存方面，Sbuffer 中的超时判断逻辑得到了修复，并通过 csr 中的 SMBLOCKCTL 配置超时阈值。更多细节请见近期进展一节。

  • 分类于 Biweekly-en
  • 需要 5 分钟阅读时间

[XiangShan Biweekly 94] 20260119

Welcome to XiangShan biweekly column! Through this column, we will regularly share the latest development progress of XiangShan. This is the 94th issue of the biweekly report.

In terms of XiangShan development, the frontend team is wrapping up the 1-taken 1-fetch architecture while exploring the 2-taken 2-fetch architecture; the backend team continues to advance the design of the vector unit and refactor multiple modules; the memory access team has fixed several bugs and is continuing to explore prefetching and MDP.

  • 分类于 Biweekly
  • 需要 3 分钟阅读时间

【香山双周报 94】20260119 期

欢迎来到香山双周报专栏，我们将通过这一专栏定期介绍香山的开发进展。本次是第 94 期双周报。

香山开发方面，前端对 1-taken 1-fetch 架构进行收尾工作，同时进行 2-taken 2-fetch 架构探索；后端继续推进向量单元的设计与多个模块的重构；访存修复了数个 bug，同时继续预取与 MDP 的探索工作。

  • 分类于 Biweekly-en
  • 需要 5 分钟阅读时间

[XiangShan Biweekly 93] 20260105

Welcome to XiangShan biweekly column! Through this column, we will regularly share the latest development progress of XiangShan. This is the 93rd issue of the biweekly report.

This is the first issue of the biweekly report in 2026! In the past year of 2025, the XiangShan team has made solid progress and achieved multiple important milestones with high quality:

  • First industrial application. The second-generation XiangShan Nanhu has been integrated as the main control CPU into the latest generation chips by Moore Threads and Chipown Technology, with Moore Threads shipping tens of thousands of units; the third-generation XiangShan Kunminghu has completed product-level delivery for the first batch of SoC chips, with two companies completing SoC chip tape-out in October and November respectively, and will receive in Q1 2026.
  • Won the first Open Source Contribution Award from the CCF Architecture Committee of the China Computer Federation
  • The "XiangShan" open-source processor core was selected as one of the representative scientific research achievements of the Chinese Academy of Sciences in 2025 and was included in the New Year message of CAS President Hou Jianguo
  • Presented tutorials at top conferences such as ISCA, HPCA, MICRO, introducing the latest progress of XiangShan to the world
  • Multiple tools have been developed and papers published at top international conferences, such as GSIM (DAC25), DiffTest-H (MICRO25), TraceRTL (HPCA26), UCAgent, etc.
  • Several papers based on XiangShan evaluations have been published in top international conferences
  • Verification work has been continuously improved, successfully passing multiple milestone tests such as 8-core consistency verification and booting GUI OpenEuler
  • Kunminghu V3 became the new default branch, with the latest performance surpassing Kunminghu V2
  • Hosted community events such as RISC-V Hackathon and Documentation Bug Hunt, and had enthusiastic exchanges with friends around the world who care about the progress of XiangShan at the second XiangShan Open Source Community Conference
  • XiangShan Compiler (XSCC) released, XiangShan now has its own compiler
  • The AI intelligent agent UCAgent from the "One Chip for All" team was released, successfully hosting the first open-source chip hackathon

A new year means a new beginning. In 2026, XiangShan will continue to implement the new concept of "open source", continuously promote the iterative development of Kunminghu V3 and the construction of the open source community. We sincerely thank everyone for their companionship and support for XiangShan!

In terms of XiangShan development, the new front-end of 1-taken and 1-fetch has been basically completed; the back-end continues to advance the design and implementation of the new vector unit while refactoring existing code; the memory system continues to refactor modules such as MMU, LoadUnit, StoreQueue, L2, and fixes some bugs.

  • 分类于 Biweekly
  • 需要 3 分钟阅读时间

【香山双周报 93】20260105 期

欢迎来到香山双周报专栏，我们将通过这一专栏定期介绍香山的开发进展。本次是第 93 期双周报。

这是 2026 年第一期双周报！在过去的 2025 年中，香山团队脚踏实地，高质量完成了多个重要里程碑：

  • 首次实现产业落地。第二代香山南湖作为主控CPU，分别被摩尔线程与芯动科技集成到最新一代芯片中，其中摩尔线程出货量已有数万颗；第三代香山昆明湖已完成面向首批 SoC 芯片的产品级交付，其中两家企业分别于 10 月与 11 月完成 SoC 芯片流片，将于 2026 年 Q1 回片
  • 荣获中国计算机学会 CCF 体系结构专委会首届开源贡献奖
  • “香山”开源处理器核作为 2025 年度中国科学院科研成果代表之一，写入中科院侯建国院长的新年贺词
  • 在 ISCA、HPCA、MICRO 等顶级会议上作 tutorial，向全世界介绍香山的最新进展
  • 开发了多项工具，并在国际顶级会议上发表论文，如 GSIM（DAC25）、DiffTest-H（MICRO25）、TraceRTL（HPCA26）、UCAgent 等
  • 数篇基于香山评估的论文在国际顶级会议发表
  • 验证工作持续完善，成功通过 8 核一致性验证、启动 GUI OpenEuler 等多个里程碑测试
  • 昆明湖 V3 成为新的默认分支，最新性能超越昆明湖 V2
  • 举办 RISCV Hackathon、文档捉虫等社区活动，在第二节香山开源社区大会与全世界关心香山进展的朋友们热烈交流
  • 香山编译器（XSCC）发布，香山有了自己的编译器
  • 万众一芯团队的 AI 智能体 UCAgent 发布，成功举办首届开源芯片黑客马拉松

新的一年意味着新的开始。在 2026 年，香山将继续贯彻“开源”这一新理念，持续推进昆明湖 V3 的迭代研发与开源社区建设。衷心感谢大家一直以来对香山的陪伴与支持！

香山开发方面，1-taken 与 1-fetch 的新前端基本完成；后端继续推进新向量单元的设计实现，同时重构了现有代码；访存继续重构了 MMU、LoadUnit、StoreQueue、L2 等模块，并修复了一些 bug。