徐起超, 杜慧敏, 刘青楠. 基于超长指令字的顶点染色处理器设计[J]. 微电子学与计算机, 2018, 35(10): 13-18.
引用本文: 徐起超, 杜慧敏, 刘青楠. 基于超长指令字的顶点染色处理器设计[J]. 微电子学与计算机, 2018, 35(10): 13-18.
XU Qi-chao, DU Hui-min, LIU Qing-nan. Vertex Shader Processor Design Based on Very Long Instruction Word[J]. Microelectronics & Computer, 2018, 35(10): 13-18.
Citation: XU Qi-chao, DU Hui-min, LIU Qing-nan. Vertex Shader Processor Design Based on Very Long Instruction Word[J]. Microelectronics & Computer, 2018, 35(10): 13-18.

基于超长指令字的顶点染色处理器设计

Vertex Shader Processor Design Based on Very Long Instruction Word

  • 摘要: 为了提高嵌入式图形处理器GPU(Graphic Process Unit)中顶点染色处理器, 设计了一款超长指令字格式的可编程顶点染色处理器, 采用六级流水线实现, 每条指令在同一个周期最多执行7种操作, 软硬件协同设计, 降低了功耗.采用基于FPGA的验证方式, 可编程顶点染色处理器在Xilinx Virtex-7 FPGAs V2000T上最大工作频率达到50 MHz, 顶点的处理速度达到0.16 M/s, 处理一个顶点平均44个周期, 在Synopsys公司Design Compiler工具130 μm工艺综合下, 主频150 MHz, 功耗约为177.742 8 mW.

     

    Abstract: In order to improve the vertex shader processor in the Graphic Process Unit (GPU), a programmable vertex shader processor with a very long instruction word format is designed and implemented in a six-stage pipeline. Each instruction has a maximum of one cycle Implementation of seven kinds of operations, hardware and software co-design, reducing power consumption. Using FPGA-based verification, the programmable vertex shader processor achieves a maximum operating frequency of 50 MHz on the Xilinx Virtex-7 FPGAs V2000T, a vertex processing speed of 0.16 M/s, and an average of 44 cycles for a vertex. At Synopsys Design Compiler tools 130 μm process synthesis, clocked at 150 MHz, power consumption is about 177.7428 mW.

     

/

返回文章
返回