Texture transmission optimization and prefetching in GPU hard accelerated rendering

In GPU 3D hard acceleration process, there will inevitably be the exchange of data between main memory and video memory, such as texture mapping process. Texture data needs to be transferred between the main memory and video memory occupying the system bus, it causes memory can not to be accessed by CPU and idling in this transfer process, affects system real-time performance. Large texture transmission granularity will result in intermittent rendering pause. To improve the situation, this paper presents a novel approach to execute prefetching algorithm of instruction (data) before transferring texture data, even if the system bus to be occupied in texture during transmission, CPU can also be running within relatively longer time depending the instruction (data) cache, while solving the problem of screen rendering intermittent pause.