Toward a processor core for real-time capable autonomic systems
Toward a processor core for real-time capable autonomic systems
复制标题
面向实时自主系统的处理器核心
DOI：
--
复制
发表时间：
2005
期刊：
影响因子：
--
通讯作者：
T. Ungerer
中科院分区：
文献类型：
--
作者：
S. Uhrig;S. Maier;T. Ungerer
本文提出了一种处理器核心，允许支持嵌入式硬实时系统中的自主计算原则。同步多线程CAR核心处理器具有硬件集成调度方案，将硬实时线程与非实时线程隔离开来。它与英飞凌的TriCore处理器二进制兼容，并设计为片上系统的IP核。处理器设计的挑战是实现同时多线程，使得一个线程不能影响另一个线程的计时行为，以便允许可预测的线程执行时间。因此，提出了新的指令发布和数据存储器访问技术。自主计算要求应通过自主管理器实现，自主管理器作为辅助线程在自己的线程槽中与实时应用程序并发运行。自主管理器线程监视应用程序并决定是否必须触发自配置、自修复、自优化或自保护
This paper proposes a processor core that allows to support the autonomic computing principles in embedded hard-real-time systems. The simultaneous multithreaded CAR-core processor features hardware-integrated scheduling schemes that isolate the hard-real-time thread from non-real-time threads. It is binary compatible with Infineon's TriCore processor and designed as IP core for a system-on-chip. The challenge for the processor design is to implement simultaneous multithreading such that a thread cannot influence the timing behavior of another thread in order to allow predictable thread execution times. Therefore new instruction issue and data memory access techniques are proposed. The autonomic computing requirements shall be implemented by autonomic managers running as helper threads in own thread slots concurrent to the real-time application. The autonomic manager threads monitor the application and decide if self-configuration, self-healing, self-optimization, or self-protection must be triggered