Design Of Graphics Processing Unit For Embedded Systems

With the continuing growing demands of graphics processing ability in Embedded Systems, a CPU will be overloaded if it needs to process all the graphics tasks as well as other control tasks for the entire system. And sometimes the graphics program is so complex that it becomes impossible for CPU to execute in real time for the interactive requirements. To deal with this difficult situation, GPU has been introduced to enhance rendering speed and free CPUs from complex graphics computation task. So researching and designing GPUs suitable for embedded systems makes great sense under this trend.The thesis firstly reviews the history of GPUs and discusses the components for graphics systems. A typical graphics system is composed by graphics APIs, the driver for GPU and the hardware. This thesis adopts OpenGL as its graphics API, selects 23 most commonly used commands from OpenGL, and designs the rendering lists for these selected commands. These rendering lists will be used as the GPU's design specification. Based on this, the thesis also designs the driver for the GPU which will be used to convert the selected OpenGL APIs into the rendering lists. The thesis spends lots of efforts on graphics pipeline——the core of GPUs and according to the contents of graphics pipeline, designs the architecture for the GPU. After that, a transaction level SystemC model has been developed for the GPU. This Model implements the geometry phase and rasterization phase of graphics processing flow. The RTL model can be acquired by refining the SystemC model.Constructing and debugging the GPU concept model is under the Visual C++6.0 environment. According to the experiment results, this model can execute all the implemented OpenGL APIs, and it performs translate and rotate transforming, lighting, culling and clipping, backface culing, projection as well as rasterization for objects in 3D word properly.