Stereoscopic Imaging Acceleration Based On GPGPU And Embedded Multi-screen 3D Display Technology Research

With the development of embedded graphic systems,there are increasingly urgent re-quirements for embedded graphic systems supporting multi-screen and three-dimensional(3D)display.However,the existing embedded graphic systems are not able to support multi-screen and 3D display perfectly.In recent years,based on existing embedded graphic systems,re-searchers realize support of 3D display by CPU computing with high CPU utilization,therefore low performance can not meet the requirements.The development of embedded graphic chips supporting multiple output,OpenGL and OpenCL,provides the hardware foundation of multi-screen display and 3D display.In the paper,embedded multi-screen display and 3D display graphic systems with a single graphic card are designed and implemented in layered modulars,which also called EM3D,and it supports multiple output display with a single graphic card and 3D hardware accelerated display.On embedded systems,limited resources make the performance a prominent problem,especially for multi-screen 3D applications.With the rapid development of GPGPU,in order to improve system performance,in the paper,to accelerate 3D graphic applications,heterogeneous computing and graphic rendering cooperation model based on OpenGL/OpenCL is proposed,and stereoscopic imaging is accelerated based the model.In the model,OpenCL provides the calculated acceleration,and OpenGL/OpenCL cooperation reduces communication between CPU and GPU.The model is suitable for accelerating graphic applications.It is main works in the follow areas:First,heterogeneous computing and graphic rendering cooperation model based on Open-GL/OpenCL is described in the paper,and stereoscopic imaging acceleration based the model is proposed.Offset calculation of stereoscopic imaging is suitable for SIMD parallel computing,and the results will be displayed.So the stereoscopic imaging can be accelerated by the model.OpenGL and OpenCL provide mechanisms and corresponding API for cooperation,so OpenGL objects can be operated by general-purpose GPU computing.GPU computing of OpenGL objects reduces the amount of communication between CPU and GPU and computation of CPU compared to CPU computing of OpenGL objects,which can improve the performance.Second,the overall structure of graphic systems supporting multi-screen and 3D display with a single card is designed and realized in layered modulars.On the basis of functional and performance requirements of the system,the system designed by layered modular method mainly consists of three layers:the underlying hardware driver layer,the graphic engine layer,the graphic library layer,and five modules:initialization module of graphic device,the graphic engine module,the 3D accelerated rendering module and the 2D,3D graphic library function modules.Furthermore,principles and methods of the multi-screen detection supporting of 3D display,and communication modules are descripted detailly.Final,it builts a platform to test functionality and performance of the system.It testes the heterogeneous computing and graphics rendering cooperation model of graphic system.Experimental results show that the functionality and performance of the graphic system meets the requirement of design,and the model can significantly accelerate graphic applications,and stereoscopic imaging is accelerated to 7 times.