Extension Of Real-time Operating System Based On Multi-core Processors

From the born of microprocessor, the promotion of performance is mainly by increasing the frequency of the processor. But the upgrade of frequency demands a huge number of transistors, which requires more power consumptions and can cause significant heat dissipation. All of those problems are critical to the embedded system. With this background, multi-core processors are being paid more and more attentions. A multi-core microprocessor is one which combines two or more independent processors into a single chip; this means it can be got a better performance with a fairly lower frequency. In the embedded system, along with the application of multi-core processor, the system software such as Real Time Operation System (RTOS) and parallel compilers are becoming a research hotspot.First of all, this dissertation discusses the realization technique of multi-core processor with a hardware view, especially the interface cooperating with the operation system. After that, it analyzes some mainly implementations of the RTOSs which support multi-core processor. And those key points, including the bootloader of the multi-core processor, the task scheduler and the mutex and communications between processors, are described in detail.From the version 2.0, Linux begin to provide the support for Symmetric Multi-Processor Architecture (SMP). We analyze the implementations of SMP in the Linux. By using the SMP architecture, we extend a RTOS MicroC/OS-II which only supports single processor to a RTOS named"MicroC/OS-SMP"which can support multi-core processor platform. As the RTOS running on the multi-core processor, MicroC/OS-SMP introduces a more complex multi-task scheduler and a more smart inter-processor communication mechanism. This dissertation discusses those key problems about the implementation, and especially discusses the algorithm of the multi-task scheduling on the multi-core platform. We also set the two conditions for a good task scheduling algorithm, and designed a simple and effective multi-task scheduling algorithm for the multi-core processor platform.