TLM model for GPU geometric primitive start mark management in plane clipping based on SystemC

Abstract

The invention relates to a TLM model for GPU geometric primitive start mark management in SystemC-based plane clipping, and the model comprises a geometric primitive storage unit which is used for storing a primitive vertex package of a first primitive vertex in a geometric primitive; a primitive command parsing unit which is used for setting a first cutting mark according to a first starting markin the primitive vertex packet; a first primitive clipping unit which is used for judging the relationship between each first line segment in the basic primitives and the cutting plane according to the intersection relationship between the basic primitives formed by the vertexes of the plurality of first primitives and the cutting plane during plane cutting, and then updating the first cutting mark as a second cutting mark or reserving the first cutting mark according to the relationship between each first line segment and the cutting plane; and a primitive output control unit which is used for setting initial mark information of a second primitive vertex in the geometric primitive obtained by cutting according to the second cutting mark when judging that all the first line segments in the basic primitive are processed. According to the TLM model, the design period of the GPU is greatly shortened through transaction-level modeling.

Description

technical field [0001] The invention belongs to the technical field of computer hardware modeling, and in particular relates to a TLM model for the management of the start mark of GPU geometric primitives in Plane Clipping based on SystemC. Background technique [0002] Transaction Level Modeling (TLM) is an advanced digital system modeling method that can quickly build a system model in the early stages of a project. It combines the details of communication between models with the details of functional units or communication architectures. separated. The communication mechanism is modeled as a channel, and transaction requests generally occur when calling the interface functions of these channel models, and the interface functions encapsulate the underlying details of information exchange. At the level of things, TLM puts more emphasis on the function of data transmission itself - the content of data and the start and end points of transmission, and involves as little as p...

AI Technical Summary

Problems solved by technology

[0003] At present, my country is independently developing a domestic graphics processor chip (Graphic Processing Unit, GPU), but because the country has been using foreign imported commercial or military GPU chips before this, there is a lack of design experience in the independent development of GPU chips, and it is necessary to develop GPU chips. Exploration and development of core technologies

[0004] The scale of GPU chip hardware logic is huge, and it is a long project from writing the system architecture document to register transfer level circuit (Register Transfer Level, rtl) design and implementation; and even after the rtl design is completed, the verification of functional performance will require a lot of time, greatly increased the entire design cycle and the design process of the auxiliary graphics processor, and reduced the progress and reliability of the self-developed graphics processor project

Images