第 021 期·2024 年 05 月 14 日

卷首语

RuyiSDK V0.10 版本已于今日发布。 如意软件源仓库发布了 PLCT 实验室提交的支持 T-Head Vector 的 LLVM 17.0.6 版本。RISC-V 设备集成与支持方面，RuyiSDK 完成了对沁恒微电子现有 RISC-V 全产品系列的调研，并在设备系统安装器中添加支持。 此外，全志、芯来RISC-V 处理器的开发板也已经加入集成设备清单，目前相关开发板大部分已经完成调研，接下来也会集成到如意设备安装器中。

为了完善国内注册功能，如意官网在进一步的完善后端服务端和推进 ruyisdk.cn 域名备案。RuyiSDK 推广工作也开始筹备，RuyiSDK 社区首次线下 Meetup 完成筹备，将于 5 月 18 举办。

RuyiSDK 更多进展细节详见下方详情，欢迎大家试用并提供反馈和建议。下一个开发版本 RuyiSDK V0.11 版本将在 5 月 28 日发布。

包管理器

RuyiSDK 0.10 对应的包管理器版本也为 0.10.0，已于今日发布。您可移步 GitHub Releases 或 ISCAS 镜像源下载体验。

本次 RuyiSDK 包管理器的更新主要包含了以下内容：

• 支持了全局选项 --porcelain 以方便外部程序与 ruyi 交互。初期适配了 ruyi list 与 ruyi news list 两种操作。
• ruyi news 在某条新闻存在当前系统语言翻译版本时，会优先展示此语言版本了。

本次 RuyiSDK 软件源的更新主要包含了以下内容：

• 新增打包了 PLCT 维护的 LLVM 工具链 toolchain/llvm-plct，主要特色是 XTHeadVector 支持。
• 新增支持了大量沁恒微电子（WCH）RISC-V MCU 评估板。Ruyi 设备安装器现已新增支持以下设备型号：
  • WCH CH32V103 评估板
  • WCH CH32V203 评估板
  • WCH CH32V208 评估板
  • WCH CH32V303 评估板
  • WCH CH32V305 评估板
  • WCH CH32V307 评估板
  • WCH CH582F 评估板
  • WCH CH592X 评估板

欢迎试用或来上游围观；您的需求是我们迭代开发的目标和动力。

IDE

本期无可见交付进展。

GCC

更新了 RUYISDK GCC14 分支，目前正在 rebase rv64ilp32/SIMD/Profiles 等特性中。

LLVM

• 增加针对 T-Head Vector 的优化 Pass

  • 增加了 RedundantVSETVLIElimination 优化过程，用于消除冗余的 vsetvli 指令

• 继续完善 LLVM intrinsic 函数，自上次更新以来，新支持了这些类别下的函数：

  • 14.1. Vector Single-Width Floating-Point Add/Subtract Operations
  • 14.2. Vector Widening Floating-Point Add/Subtract Operations
  • 14.3. Vector Single-Width Floating-Point Multiply/Divide Operations
  • 14.4. Vector Widening Floating-Point Multiply Operations
  • 14.5. Vector Single-Width Floating-Point Fused Multiply-Add Operations
  • 14.6. Vector Widening Floating-Point Fused Multiply-Add Operations
  • 14.7. Vector Floating-Point Square-Root Operations
  • 14.8. Vector Floating-Point Reciprocal Square-Root Estimate Operations
  • 14.9. Vector Floating-Point Reciprocal Estimate Operations
  • 14.10. Vector Floating-Point MIN/MAX Operations
  • 14.11. Vector Floating-Point Sign-Injection Operations
  • 14.12. Vector Floating-Point Compare Operations
  • 14.13. Vector Floating-Point Classify Operations
  • 14.14. Vector Floating-Point Merge Operations
  • 14.15. Vector Floating-Point Move Operations
  • 14.16. Single-Width Floating-Point/Integer Type-Convert Operations
  • 14.17. Widening Floating-Point/Integer Type-Convert Operations
  • 14.18. Narrowing Floating-Point/Integer Type-Convert Operations
  • 17.2. Integer Scalar Move Operations
  • 17.3. Floating-Point Scalar Move Operations
  • 17.4. Vector Slide Operations
  • 17.5. Vector Register Gather Operations
  • 17.6. Vector Compress Operations

• 继续完善 Clang intrinsic 函数，自上次更新以来，新支持了这些类别下的函数：

  • 14.1. Vector Single-Width Floating-Point Add/Subtract Operations
  • 14.2. Vector Widening Floating-Point Add/Subtract Operations
  • 14.3. Vector Single-Width Floating-Point Multiply/Divide Operations
  • 14.4. Vector Widening Floating-Point Multiply Operations
  • 14.5. Vector Single-Width Floating-Point Fused Multiply-Add Operations
  • 14.6. Vector Widening Floating-Point Fused Multiply-Add Operations
  • 15.1. Vector Single-Width Integer Reduction Operations
  • 15.2. Vector Widening Integer Reduction Operations
  • 15.3. Vector Single-Width Floating-Point Reduction Operations
  • 15.4. Vector Widening Floating-Point Reduction Operations
  • 17.4. Vector Slide Operations
  • 17.5. Vector Register Gather Operations
  • 17.6. Vector Compress Operations

OpenJDK

1. OpenJDK PRs:

• Initial version: https://github.com/RealFYang/loom/commit/56746e7b9b2e20c999427201479b03f97eac805c (RISC-V vthread support for JVM intrinsic monitors)
• https://github.com/openjdk/jdk/pull/18737 (8330095: RISC-V: Remove obsolete vandn_vi instruction)
• https://github.com/openjdk/jdk/pull/18780 (8330242: RISC-V: Simplify and remove CORRECT_COMPILER_ATOMIC_SUPPORT in atomic_linux_riscv.hpp)
• https://github.com/openjdk/jdk22u/pull/146 (8330242: RISC-V: Simplify and remove CORRECT_COMPILER_ATOMIC_SUPPORT in atomic_linux_riscv.hpp)
• https://github.com/openjdk/jdk21u-dev/pull/507 (8326936: RISC-V: Shenandoah GC crashes due to incorrect atomic memory operations)
• https://github.com/openjdk/jdk17u-dev/pull/2385 (8329823: RISC-V: Need to sync CPU features with related JVM flags)
• https://github.com/openjdk/jdk17u-dev/pull/2417 (8326936: RISC-V: Shenandoah GC crashes due to incorrect atomic memory operations)

2. Reviewed riscv-port-jdk11u backport PRs:

• https://github.com/openjdk/riscv-port-jdk11u/pull/11 (8328065: RISC-V: Add isolation for shared code changes)
• https://github.com/openjdk/riscv-port-jdk11u/pull/12 (8328580: Remove trivial shared code changes which are leftover from riscv port)
• https://github.com/openjdk/riscv-port-jdk11u/pull/13 (8283865: riscv: Break down -XX:+UseRVB into seperate options for each bitmanip extension)
• https://github.com/openjdk/riscv-port-jdk11u/pull/16 (8291893: riscv: remove fence.i used in user space)
• https://github.com/openjdk/riscv-port-jdk11u/pull/17 (8284937: riscv: should not allocate special register for temp)
• https://github.com/openjdk/riscv-port-jdk11u/pull/18 (8285303: riscv: Incorrect register mask in call_native_base)
• https://github.com/openjdk/riscv-port-jdk11u/pull/19 (8297697: RISC-V: Add support for SATP mode detection)

3. CFV: New RISC-V Port Committer: Gui Cao

• https://mail.openjdk.org/pipermail/riscv-port-dev/2024-April/001345.html
• https://mail.openjdk.org/pipermail/riscv-port-dev/2024-April/001367.html

V8

1. 新增 fastcall 特性的 JavaScript 可重入特性。
2. 修复内置模拟器 RVV 寄存器内容未初始化、指针压缩开启后，32 位模式下指针判零等 bug。

官网

RuyiSDK 社区首次线下 Meetup 来啦，5 月 18 日软件所见！

为了完善国内注册功能，如意官网在进一步的完善后端服务端和推进 ruyisdk.cn 域名备案。

操作系统支持矩阵

新增部分全志、芯来开发板的系统支持情况调研；基于现有开发板和相关文档展开测试验证，产出测试报告。

• Milk-V Duo S
  • NuttX
    • 新版本重新测试
  • BuildRoot
  • Debian
• CanMV / Kendryte K230
  • NuttX
  • RT-Thread
• StarFive VisionFive 2
  • NuttX
• Maix-I K210
  • NuttX
• Nuclei DDR200T
  • FreeRTOS
  • RT-Thread
• Longan Nano
  • RT-Thread
• RV Star
  • FreeRTOS
  • RT-Thread
• 100ASK-V853-PRO (Allwinner V853)
  • Melis
• CM32M433R-START
  • FreeRTOS
  • RT-Thread
• TinyVision
  • Melis
• Allwinner V853
  • Melis
• Youmu Pi
  • Melis

内容请详见：ruyisdk/support-matrix