GPU Graphics Acceleration Technology Research And Application Based On FT

Domestic CPU plays a very important role in promoting our country’s information security and the modern process.However, due to domestic CPU starting lately, there is a big gap of the performance between domestic CPU and international advanced CPU.Mostly Domestic CPU deals with a lot of multitasking, which is already overworked. In order to enhance the overall performance of FT platform and graphics acceleration effect, this paper uses domestic FT CPU and GPU heterogeneous technology.In this way, GPU can deal with lots of graphic data which heavily consume CPU resources, and thus not only CPU can focus more efficiently handle other tasks and graphics acceleration performance and display effect has also been greatly improved.This paper deeply studies the system and software framework architecture and hardware accelerated graphics subsystem interactions based on FT. Centering on Vivante GC600 and Radeon R6 XX series application on GPU platform, we researches the working principle and the system of GPU kernel related technologies, including the frame buffer device, mode setting, memory space management, command buffer, the command transport mechanism and related other key technology.Firstly, this paper implements Xorg/EXA 2D and Mesa 3D graphics acceleration system for Vivante GC600 on FT platform.Vivante Corp mainly focuses on mobile devices, but it is not mature in desktop graphics acceleration system especially the Xorg applications. It is the first times to apply Vivante GC600 to fit the desktop graphic system based on FT, which is hard to avoid appearing problems. After studying FT1000 A, FT desktop graphics acceleration,Vivante GC600 and acceleration graphics system structure and characteristics, we design and implement Vivante GC600 graphics memory space management, mode setting and frame buffer device,then successfully apply Vivante GC600 to FT desktop graphics acceleration system. In the process of implementation, we mainly resolve the FT1000 A and Vivante GC600 kernel memory space in the address space of the alignment, byte order and memory allocation and management mechanism, multi-resolution and asynchronous chip clock synchronization of mode setting and other technical difficulty issues.According to IOMMU mechanism, we break through the key technology of combining the static reserve and dynamic allocation of memory management.Secondly, this paper increases Radeon R6 XX kernel/application interactive mode, then applying R600 to Direct FB, thereby making it more suitable to the development of graphics acceleration technology. Our study find that R6 XX had great progress in technology, such as adopting more efficient command transmission mechanism and PM4 package, etc. Therefore, this paper increases the kernel/application of Radeon R6 XX series driver structure and successfully implements that Radeon R600 is applied to the Direct FB accelerated graphics system. In the process of implementation the R600, we researches and resolves main memory address space allocation and management, the GPU command from the command processor ring buffer gain command flow PM4 command of packages, such command transmission mechanism, the simple frame buffer device and other hardware and software interaction of a series of technical difficulties.And we break through host computer(CPU) system main memory mapping into the GPU address space and kernel PM4 command package implementing the pointing to line, drawing the rectangle, triangle and Blit basic operations such as related technical problems.At last, on the basis of the research and implementation of graphics acceleration, this paper evaluates and tests graphics acceleration system based on GC600 and Direct FB accelerated graphics system.Evaluation chooses Vivante company providing test kits to test GC600 functional completeness. On the basis of the confirmation,we evaluates Xorg/EXA and the Mesa basic graphics acceleration performance.Meanwhile, for the Direct FB system implementation, in order to compare the performance of Direct FB graphics acceleration, we firstly realizes Blit copy screen operation to test the limits of the physical properties of the nearest R600 in the kernel, and then evaluates the Direct FB graphics acceleration system functionality and performance.In a word, this paper, on the basis of the principle of GPU accelerated graphics, deeply studies GPU memory address space,command buffer and command transmission mechanism, frame buffer device and other related the GPU accelerated graphics core technology. What is more meaningful, the relevant research combines with application development needs, it is successfully applied ralated researches to the FT platform, so as to achieving the combination of the theory research and concrete practice, and fulfils the needs of the applicating development goals.