Research And Apply The Shared Cache Partition In Real-time Operate System

CNC technology use digital information on the high-performance control of themechanical movement and process. This way greatly improves the quality，accuracy,efficiency of the product. CNC technology are developing rapidly on thehigh-speed and high-precision performance; the open and network architecture; theintegrated, complex, intelligent function. So developing the open architecture CNCsystem has become inevitable. Open architecture CNC systems not only requiremulti-core processors as its hardware platform but also require high-performancereal-time operating system as its software platform. Mainstream multi-core processors,however, still use shared Cache structure.This structure will bring more serious theshared cache conflict phenomenon in multi-core environment. The system will beforced to frequently flush the Cache data set in order to ensure the implementation ofconcurrent tasks on different core needs. Thus extending the execution time of tasks,reducing the overall system performance because of tasks could not schedule.There are a variety of solutions to solve this problem. Shared Cache Partition is aspecial one among those. It can partition the L2Cache to tasks with a non-exclusiveshared Cache space until the end of the task execution. It can help the tasks to avoidsharing Cache conflicts. Therefore, this article used software-based shared cachepartitioning to transform the existing NC system. First, after establishing the HCDMmodel, analysis the shared cache conflict impact on the system and do the prediction ofthe shared cache behavior under task environment. Second, establish static anddynamic partition function according to the performance of the model and calculate thedivision space of tasks in CNC system.Finally, use the "page coloring" technology totransform the existing CNC systems. Experiments show that shared Cache partitioningscheme significantly improves the real-time performance of CNC system.