首页|面向边缘节点的RISC-V处理器的研究与设计

面向边缘节点的RISC-V处理器的研究与设计

扫码查看
边缘节点作为分布式计算体系的基本组成部分,部分工作场景的能源受限,对处理器的成本和功耗较为敏感.针对边缘节点的特殊工况,设计了一款基于RISC-V架构的低功耗处理器SparrowRV.SparrowRV采用指令、数据总线分离的哈佛结构,支持RV32IMZicsr指令集,2级流水线设计.为提高除法计算效率,提出了一种基于恢复余数除法的动态迭代算法,减少了除法的迭代次数.处理器使用iverilog进行功能仿真,通过了RV32IMZicsr指令集功能测试.处理器在XC7K325T FPGA上完成原型验证,Coremark跑分达到2.78 CoreMark/MHz.SparrowRV内核相比于Tinyriscv和蜂鸟E203,同频性能提升了15.8%和29.9%,动态功耗降低了4.9%和32.6%.
Research and Design of RISC-V Processor for Edge Nodes
As a basic component of distributed computing system,edge nodes are sensitive to the cost and power consumption of proces-sors due to limited energy in some work scenarios.A low-power SparrowRV processor based on RISC-V architecture is designed for the special working conditions of edge nodes.SparrowRV adopts Harvard structure,and instruction bus and data bus are separated.It sup-ports RV32IMZicsr instruction set and 2-level pipelined architecture.In order to improve the efficiency of division,a dynamic iterative algorithm based on the division of recovery remainder is proposed,which reduces the number of iterations of division.The processor uses iverilog for function simulation and the function of RV32IMZicsr instruction is verified to be correct.Prototype verification is completed on XC7K325T FPGA for the processor,and the score of Coremark reachs 2.78 CoreMark/MHz.Compared with Tinyriscv and Humming-bird E203,the SparrowRV core improves the same frequency performance by 15.8%and 29.9%,and reduces the dynamic power con-sumption by 4.9%and 32.6%.

RISC-Vprocessorpipelineedge nodedivider

吴言、乔建华、雷光政、栗亚宁

展开 >

太原科技大学电子信息工程学院,山西 太原 030024

RISC-V 处理器 流水线 边缘节点 除法器

2024

电子器件
东南大学

电子器件

CSTPCD
影响因子:0.569
ISSN:1005-9490
年,卷(期):2024.47(6)