Research Of High Reliability Multi-core Processing System Based On Online Self-test

People's requirement of reliability is becoming higher and higher when pursuing high performance of integrated circuits. At the same time, the development of manufacturing process makes integrated circuits be easy to failure under the condition of all kinds of interference and noises, which also brings the problem of reliability. And for some special fields such as aerospace and weapon engineering, reliability is quite important.As a new kind of processor design idea and technology, Application Specific Instruction Set Processor(ASIP) combines the programmability of DSP and the high performance of ASIC, which is also suitable for implementing large-scale parallel structure. Thus in specific areas such as network, automotive, digital signal processing and so on, ASIP is widely used. This article mainly focused on the reliability design method of ASIP parallel structure, complete the following three aspects of work:First, after an in-depth study on the field of digital signal processing, design and complete a single cycle Application Specific Instruction Set Processor, then on the basis of this ASIP, design and complete a Multiple Instruction Stream Multiple Data Stream(MIMD) system on shared memory.Second, after a research and contrast on a variety of classic reliability design methods, come up with a reliability scheme aimed at the MIMD parallel processing system, which combines partial redundancy, built-in self-test(BIST) and error correcting code. The parallel processing system is equipped with two redundant units, one redundant unit replaces the processing unit firstly, then the processing unit tests itself using BIST. If there is a failure, the redundant unit will replace the fault unit, or it will perform similar operation on next unit.Third, test and analyze the reliability of the parallel processing system, first of all, the normal function of the system is tested, then test the reliability using the technology of fault injection, lastly, test results and resource occupation of the system are given and analyzed, they show that this design improves the reliability of the system significantly.