At present, the application of the embedded microkernel operating system in military and civil fields has begun to take shape, but it has not yet formed a unified method and standard. Due to its high performance, low frequency, and high reliability, dual-core embedded processors are getting the attention of many chip manufacturers. Compatibility has been favored by many telecom equipment manufacturers and embedded high-end application integrators, but the dual-core embedded processor needs a new real-time operating system to support it, so that it can give full play to the high performance of the dual-core. It paves the way for the application of its processor in the embedded field, but the design of embedded AI engine is not transparent to it; so, it needs the support of operating system. The user code is used to be in the operating system processor environment; so, it can be used on dual core processor first. In the real-time operating system that supports dual-core processors, the part that needs to be modified is mainly concentrated in the kernel part; so, the core design is the key point to support dual-core processors. This article is to seize this key point to carry out in-depth research. The difference and influence of hardware architecture between dual-core processor and single-core processor are the primary content of the study. Through the research of the dual-core processor architecture, the general abstraction of the dual-core processor architecture is obtained, which is the starting point of the follow-up research. This paper mainly studies the design of embedded AI engine based on the microkernel operating system, extracts the security requirements of the operating system, designs and implements the operating system from the perspective of formal verification, and considers the verification problem under the background of RTOS development, so as to avoid using too many complex data structures and algorithms in the system design and reduce the difficulty of experimental verification. In this paper, we use the spatiotemporal data model, data sharing security in the cloud environment, symmetric encryption scheme, and Paillier homomorphic encryption method to study the design of embedded AI engine based on the microkernel operating system. According to the idea of microkernel architecture, the kernel is divided into four main modules: task processing, semaphore, message queue, interrupt, and exception processing, and the lock mechanism to prevent reentrancy in software is analyzed separately. Three core function modules, initialization, process grinding, and interrupt processing, are extracted from the microkernel operating system to form the formal verification area of the operating system. At the same time, the system syntax and semantics of related rights are separated, and the main rationalization rules are described. The results show that the core of the microkernel operating system in five states adopts the microkernel architecture in the dual core environment. The microkernel architecture is a compact system kernel with good adjustability. Based on the analysis of the microkernel operating system, the internal structure of each module in the kernel is summarized, and the modules are modified according to the machine characteristics of the dual core processor; it corresponds to adding modules to meet the characteristics of dual core architecture. Kernel design is a systematic theoretical research process. This paper only uncovers the tip of the iceberg of real-time operating system kernel design on dual-core embedded processors. It is necessary to understand the kernel more deeply and master the kernel in the future work and study.
