GPU命令处理器的存储管理单元设计

Design of memory management unit in GPU’s command processor

针对现有图形处理器(graphic processing unit,GPU)中命令处理器对缓冲区对象命令解引用速度低、实时性差等不足,对其加以改进,即在命令处理器中添加存储管理单元,采用硬件方式实现相应存储管理算法。另外创建8个顶点数组缓冲区对象存储区,对其分开管理,以求提高解引用速度。仿真测试结果表明,所提出的存储管理单元设计可行,能够显著地提高命令处理器的工作效率。

Due to the low speed and the poor real-time of the existing command processor in the graphic processing unit, the storage management algorithms in hardware was proposed to add the memory management unit. This algorithm created eight vertex array buffer objects separately, and improved the speed of dereferencing. The simulation test results showed that the proposed design was feasible. The memory management unit can improve the work efficiency of the com- mand processor significantly.