Embedded Microprocessor Extension Design and Optimization for Real-Time Edge Computing

Abstract

With the development of 5G communication technology, more and more applications could be integrated into one system. Edge computing system and mixed-criticality system may integrate tasks of different criticality levels, which brings better balance in isolation and performance. Such advantages make it gradually become a research hotspot in edge computing and real-time systems with 5G. The important content of designing a mixed-criticality system is how to reduce interference between tasks and how to schedule tasks efficiently to ensure that tasks of different criticality levels can meet time constraints. Instruction extension and hardware software cooperative support may be an effective solution. Based on a fine-grained multithreaded RISC-V processor, this article gives some extensions for real-time operations and proposes a hardware software cooperative real-time scheduling mechanism. Experimental results show that, compared with FlexPRET hardware, the performance of thread scheduling is improved by 22.94% on average. Compared with software scheduling, the performance of scheduling same programs and multiple programs are improved by 15.46% and 26.00%, respectively.