arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2302.03732cs.AR
收藏

向RISC-V ISA添加显式加载获取和存储释放指令

Adding Explicit Load-Acquire and Store-Release Instructions to the RISC-V ISA

Bryce Arden, Zachary Susskind, Brendan Sweeney

AI总结:

本项目向RISC-V ISA添加显式加载获取和存储释放指令,支持herd、gem5及LLVM/Clang,旨在防止ABI碎片化,并验证语义编码正确。

摘要:

弱内存模型允许简化硬件并提高内存层次结构中的性能,但代价是增加了软件复杂性。在弱内存模型中,需要显式同步来强制不同处理器之间的顺序。获取和释放语义提供了一种强大的原语,用于仅表达正确性所需的排序。在本项目中,我们探索向RISC-V ISA添加加载获取和存储释放指令。我们向herd形式内存模型、gem5周期近似模拟器和LLVM/Clang工具链添加了支持。由于这些指令在RISC-V标准中不存在,因此固有地迫切需要批准显式加载获取/存储释放指令,以防止多种ABI实现和生态系统碎片化。我们发现,对于具有高共享度和重度争用的工作负载,较少内存排序的影响被削弱,但我们的更改成功编码了我们期望的语义。

英文摘要:

Weak memory models allow for simplified hardware and increased performance in the memory hierarchy at the cost of increased software complexity. In weak memory models, explicit synchronization is needed to enforce ordering between different processors. Acquire and release semantics provide a powerful primitive for expressing only the ordering required for correctness. In this project, we explore adding load-acquire and store-release instructions to the RISC-V ISA. We add support to the herd formal memory model, the gem5 cycle-approximate simulator, and the LLVM/Clang toolchain. Because these instructions do not exist in the RISC-V standard, there is an inherent urgency to ratify explicit load-acquire/store-release instructions in order to prevent multiple ABI implementations and ecosystem fragmentation. We found that for workloads with a high degree of sharing and heavy contention, the impact of less memory ordering is muted, but our changes successfully encode the semantics we desire.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑