Research On Simulation-based Multi-Processor Functional Verification Techniques

With the increasing demands of the embedded applications,the architecture of the multi-processor is becoming more complex and the scale of the design is becoming larger.Functional verification has become the key bottleneck in the multi-processor design development cycle.The efficient functional verification method can help to reduce verification time and time-to-market of the multi-processor.The research work focuses on the functional verification of multi-processor.The dissertation starts from simulation-based verification,explores the related research fields,and proposes some techniques to improve the multi-processor verification efficiency.The proposed methods have been applied to the functional verification of two multi-processors.The research work covers the following three aspects:1)Research for coverage directed test generation based on genetic algorithmIn multi-processor functional verification process,coverage metrics are usually used to identify the functions which have not been covered,and evaluate the progress of functional verification.Genetic algorithm can close the loop between coverage analysis and test generation and direct the test generation to produce tests of good quality.In this paper,we propose a coverage directed test generation based on improved self-adaptation genetic algorithm.The coverage results are evaluated and the constraints for test generation are modified to direct test generation to produce new tests.The new tests can cover the functional points that the old ones fail to cover.The experimental results show that the proposed method can achieve a higher coverage level in a shorter time,which achieves a 70%simulation time saving.It can significantly improve the functional verification efficiency of the multi-processor.2)Test prioritization technique based on k-means algorithmFor the simulation-based multi-processor functional verification,it is necessary to direct the test generation to produce a large number of tests in a short time.However,simulating all the generated tests will take a lot of simulation time and reduce the verification efficiency.Selecting a limited number of tests of good quality which can obtain a high coverage level in a short time can improve the efficiency of the multi-processor functional verification.In this paper,we propose a test prioritization method based on k-means algorithm to select tests based on their estimated coverage.We also propose three improving k-means algorithms to improve the effectiveness of test prioritization.The experimental results show that the proposed improving k-means algorithms are more accurate and efficient than the standard k-means algorithm for test prioritization.In comparison with simulating all the generated test randomly,the more important tests selected with the proposed test prioritization based on k-means algorithm,achieve almost the same coverage metrics in a shorter time,which achieves a 90%simulation time saving.3)Research for the symmetric multi-processor reference modelThe efficiency of simulation-based multi-processor functional verification highly depends on the simulation speed and of the reference model.This paper propose an efficient symmetric multi-processor reference model.A function-accurate model is used to act as the reference model of CPU pipeline to reduce the building time and improve simulation speed.A timing-accurate model is used to verify the cache coherence of the symmetric multi-processor and a dependent timing sequence is applied to establish the relationship between these two models.The proposed reference model has been implemented in a 32-bit symmetric multi-processor verification.Experimental results show that the difficulty of building the proposed model is low and the proposed model can take into account both simulation speed and accuracy.It can help the verification team to locate design errors easily and accurately,and improve verification efficiency.