Design And Implementation Of Imaging Pipeline Verification Platform For GPU

With the development of computer graphics technology, people has higher requirement on the performance of Graphic Processing Unit. Accordingly, academia and the business community is very concerned about high-performance Graphic Processing Unit design, which makes the graphics processing hardware design technology has been developed rapidly. 3D engine as a core part of the GPU graphics processor, it is mainly used for rendering three-dimensional graphic image. Image pipeline module as an important part of the 3D engine, is mainly used to complete the image of the Open GL standard drawing, reading, copying, scaling, color replacement, pixel mapping, texture image definition and bitmap processing defined pixel processing operations.Functional verification is a very complex and critical part of the design of the chip. In order to ensure the design of image pipeline module can be in accordance with the design specifications to achieve its function, this paper designed the image pipeline verification platform to verify the module behavior function, namely functional verification.Based on the function of the image pipeline module of GPU, this paper designed and implemented the image pipeline verification platform. First of all, the various functions of the image pipeline, such as bitmap drawing function, the basic image processing functions(image reading, image writing and image copying), pixel transfering, pixel mapping, image scaling, texture image definition and parameter storage functions are analyzed in detail, as a basis for building verification platform. On this basis, this paper has set up a constraint for the pipe with complete image module function random verification platform. The verification platform is designed by System Verilog language. It includes excitation generator, image pipeline module reference model, image pipeline module and other functional modules of the interface module and the comparator. The generator excitation can be in direct loading of the transaction level constraint random excitation; image pipeline module reference model fully realize the functions required by the image pipeline module; component interface realizes the other module and image pipeline related functions; the comparator with automatic output ratio on the results is correct, the realization of the platform to validate the results of the automatic alignment.Finally, we verify the function of the image pipeline module using this verification platform. Firstly, according to the image classification function of pipeline to verify the scene description, verification will divide the scene into data transmission, data processing and storage of three kinds of parameters. In different test scenarios were verified, then random loading multiple validation scenarios for validation. The analysis of the results of the verification and waveform shows that the function of this module is correct. The image pipeline module verification function coverage rate was 100%, the total coverage of more than 93%, the verification platform can meet the verification requirements of image pipeline module.