Research Of RTEMS Embedded Operating System With Support For Muliticore Processor

In recent years, with the explosive growth of the information, coupled with Moore’s Lawlimit, single-core processors can not meet the needs of computer users in various fields.Single-chip multi-core processors solves the bottleneck problems encountered in thedevelopment of the single-core through integrated multiple lower frequencies executablenuclear on-chip. Researching operating system that supports multi-core processors is the basisof the multi-core applications. RTEMS embedded operating system provides a good supportfor multi-processor. But realizing what kind of mechanism in the multicore design processshould be based on the specific hardware architecture. This dissertation studies the RTEMSimplementation mechanism which support SMP architecture.The dissertation first analyze the RTEMS operating systems kernel, then researchRTEMS system which provides support for multi-core processors. The system uses a sharedmemory architecture. Multicore achieve communication through the interrupt mechanism. ForSynchronization mutex mechanism in the case of the single-core can not maintain the originalsemantics in the multicore system, the system uses the task spinlock and the interruptspinocked combination mechanisms, and to prevent processor core has been in busy state itlimits the waiting time for the spinlock. For the deficiencies of the unified allocation strategyimplemented in RTEMS multicore system, proposing two dynamic task scheduling algorithmwhich can automatically adjusted the task according to the dynamic changes of the systemresources on the basis of the study in the existing dynamic scheduling algorithm.This dissertation experiments by use of random task graph to generate test tasks on theM5multicore simulation platform to verify RTEMS multi-core system feasibility andefficiency, and select the speed up and load balancing efficiency as a strategic performanceevaluation. The test results prove the system correctness, and with the calculation of theincreasing scale, the new mechanism shortens the task execution time and improves theparallel efficiency of the system.