基于保护缓存的x86系统仿真优化

Accelerate x86 system emulation with protection cache

为降低x86保护机制的仿真开销,提出了一种基于保护缓存的系统仿真优化方法。分析了x86保护机制的仿真开销来源,确定了与保护相关的具有较大仿真开销的指令和操作;给出了保护缓存的概念和工作机制,通过缓存最近访问过的描述符以及对使用该描述符的指令或操作的保护检查结果,来减少虚拟机内存访问和保护检查操作的次数。基于QEMU对保护缓存机制进行了实现,测试结果表明,引入保护缓存后涉及保护检查的x86指令或操作的仿真效率提升了11%～24.5%,且系统调用、页面故障处理和I/O密集型应用的性能均得到了稳定的提升。

To decrease the emulation overheads of x86 protection mechanism, a protection cache based optimizing method is presented. Firstly, the sources of x86 protection mechanism emulation overhead are analyzed, and the protection related instructions and operations that have great emulation overheads are confirmed. Then, the concept and working mechanism of protection cache are given, by caching the descriptors recently accessed and the protection checking results against those instructions and operations that use these descriptors, times of virtual machine memory access and protection checking operation can be reduced. At last, QEMU based implementation of protection cache is described, and evaluation demonstrates that, when protection cache is enabled, emulationperformance of x86 instruction or operation that related to protection checking gets a speedup d 11%-24. 5 %, and the performances of system call, page fault handling and I/O intensive applications are all steadily improved.