Method and device for generating test program for verifying function of microprocessor

Abstract

The invention relates to a method and a device for generating a test program for verifying the function of a microprocessor. The method comprises the following steps of: A, randomly generating mu+lamda individuals, saving the individuals as an original generation population, and taking the original generation population as the current population; B, converting the individuals in the current population into the test program by combining an instruction library; C, performing simulation execution on the test program obtained by converting each individual in the current population to generate a functional coverage report; D, directly performing the step E if the current population is the original generation population, and performing intersection and variation operation on the individuals in the current population for lamda times, saving the generated new individuals in the current population, and performing the step E if the current population is not the original generation population; E, selecting at least mu individuals from the current population as a new-generation population according to the functional coverage report; and F, outputting test program sets corresponding to the individuals in the new-generation population if the stop condition is met, and taking the new-generation population as the current population and returning to the step B if the stop condition is not met.

Description

technical field [0001] The invention relates to instruction-level verification of microprocessors, in particular to a method and device for generating test programs for function verification of microprocessors. Background technique [0002] With the rapid development of integrated circuit manufacturing technology, the scale and complexity of microprocessor design are getting higher and higher, and many new micro-architecture features are introduced, such as deep pipeline, dynamic scheduling, out-of-order execution, etc. Microprocessor is the core component of computer system, and its functional correctness is the key to the normal operation of the final product. Effective verification methods are needed to find hidden design errors in the design. If there is a functional verification vulnerability, causing design errors to flow into the final product, it will bring huge losses to the enterprise. [0003] At present, microprocessor chip-level verification mainly adopts the m...

AI Technical Summary

Problems solved by technology

Not suitable for functional coverage models and design verification requiring large test programs, because microprocessor functional coverage models usually contain a large number of widely varying points of interest, and large test programs are required to achieve functional coverage goals

In the existing test program evolution method, the test program is obtained from the individual mapping in the population, and the validity of the test program needs to be checked during the operation of the evolutionary algorithm, and the calculation cost is greatly affected by the length of the chromosome code.

[0022] In addition, existing evolutionary test program methods explore a unique optimal test program in the solution space, which requires a large search effort for complex coverage models, greatly increases the running time of the algorithm, and a single optimal solution search is not suitable for functional coverage The multi-interest point feature of the model affects the convergence speed of the algorithm

[0023] The complexity and scale of microprocessor design are large, and the complexity of the functional coverage model is high, including a large number of interesting points with great differences. The traditional test program evolution method needs a large-scale test program to achieve the target of functional coverage. Complex coverage models require significant search effort, greatly increasing the running time of the algorithm

[0024] To sum up, the existing test program evolution method can search high-quality test programs in the solution space, but the effect is not good when it is used for the verification of the microprocessor functional coverage model, and there is a lack of actual verification of the microprocessor functional coverage model. experience

Images