Remote CPU Experimental Platform Distributed Architecture Research And System Implementation

Experiments that many universities have drawn much attention in recent years are one of the essential means for students since they deepen the understanding of theoretical knowledge and promote the ability of practice.Remote virtual experiment is an efficient and flexible method,which not only solves the problems of the traditional experimental mode but also speeds up the construction and sharing of laboratories.At present,many remote experiment systems have been proposed.Most remote experiment systems allow users to remotely control the experiment board on the client.However,there exist some drawbacks that mainly manifest as follows.Firstly,the experimental equipment,results and phenomena cannot be controlled or presented in real time.Users need to write all operations as an incentive file and send it to the server while waiting for the results of unified processing.Secondly,the experimental phenomenon is recorded by the camera and sent to the users for observation and debugging.Nevertheless,only the surface phenomenon can be observed while the internal is omitted.Moreover,the construction cost of such systems is increased.Thirdly,the majority of the existing systems with monomer architecture are in poor usability and concurrency.In this paper,based on a long time research and analysis,we design and develop a remote virtual experiment platform that combines with advanced technology for network platform development.This platform adopts distributed architecture so that users can design and carry out rich and varied FPGA experiments.The main work of this paper includes the following aspects:Firstly,we investigate the new technology of remote experiments.Then we complete the technology selection and demand analysis for the remote experiment platform.On this basis,the overall design of the platform is proposed.In order to improve usability,concurrency,and overall performance,we adopt a distributed architecture.Moreover,we do some other works like split the system,separate the reading and writing of the database,and integrate the Redis cache.Secondly,we design the function of CPU experiment for the proposed platform.we research and develop the virtual panel file standard for CPU experiments,which ensures that users design and carry out experiments flexibly.a clear level of experimental data interaction flow is designed to reduce the coupling degree and the complexity of the program.in order to guarantee the reading and writing of the data in master memory efficiently and stably,we design a reliable reading and writing strategy.we design a mechanism to avoid complicated calculation and improve experimental efficiency that microinstruction encoder and assembly instructions can translate to machine code automatically.we use the "regular polling" and queuing system to manage the distribution and recovery of the experiment board,so as to improve the efficiency of it;In addition,the efficient management of the experimental log is achieved by using AOP mechanism.Thirdly,the general real-time rendering algorithm of information flow is designed and implemented to display the execution process within the CPU in real time.First,the problem that only one of the multiplexer data paths is involved in rendering is solved by leveraging classification processing.After that we introduce the concept of binding.The data path on the multiplexer can only participate in rendering if the data path of bounding has already entered,which solves the problem that the data path on the multiplexer works all the time,resulting in impurities in the information flow.Finally,the platform is tested and validated.The test and usage results indicate that the platform runs reliably,meets design requirements,and has high usability and concurrency.