本文提出了一种基于RISC-V架构的CORDIC指令集及实现方法,可直接减少执行的指令数量,通过采用独热码编码来减少译码逻辑资源的消耗,通过复用处理器算术逻辑单元来减少算术逻辑资源的消耗,通过预存特殊角度结果来减少计算时间,通过优化选择电路来提高工作频率。该方法在Zynq 7020平台上进行了实现。结果表明,相较于使用基础处理器指令计算的方法,指令数量压缩了97%,计算时间减少了43%;相较于传统CORDIC电路进行计算的方法,寄存器资源减少了79%,工作频率提升了65%,计算时间减少了39%。
Abstract
In the paper,a proceccor with CORDIC instruction set extension based on RISC-V architecture is proposed,which can directly reduce the number of executed instructions.One-hot encoding is used to reduce the consumption of logic resources used for decoding.By reusing the processor's arithmetic logic unit can reduce the logic resource consumption used for realizing arithmetic logic.The calculation time is reduced by prestoring the special angle results.And the operating frequency is improved by optimizing the selection circuit.The proposed circuit is implemented on the Zynq 7020 platform.Compared with the method of using basic processor instruction to calculate,the number of instructions is reduced by 97% and the calculation time is reduced by 43%.Compared with the traditional CORDIC circuit calculation method,the register resources are reduced by 79%,the operating frequency is increased by 65%,and the calculation time is reduced by 39%.
关键词
RISC-V /
CORDIC指令集 /
处理器 /
FPGA
Key words
RISC-V /
CORDIC instruction set /
processor /
FPGA
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 何康辉,董朝阳.基于正交变换的五阶容积卡尔曼滤波导航算法[J].系统工程与电子技术,2020,42(3):6.
[2] 姚建宇,张祎维,张广婷,等.基于SIMD的三角函数高性能实现与优化[J].计算机科学,2021,48(12):7.
[3] 高嵩,胥剑涛.基于FPGA的相控阵天线波束控制算法优化研究[J].电子技术应用,2021,47(9):5.
[4] 严翔,张屯厚,邓威,等.FPGA嵌入式语音识别控制系统的设计实现[J].计算机与网络,2017,43(19):3.
[5] 李正平,高杨.基于MIPS32架构三角函数指令集扩展的设计与实现[J].合肥工业大学学报(自然科学版),2021,44(5):4.
[6] Volder J E.The CORDIC Trigonometric Computing Technique[J].Electronic Computers Ire Transactions on,1959,EC-8(3):330-334.
[7] Walther J.ELEMENTARY FLOATING POINT CORDIC FUNCTION PROCESSOR AND SHIFTER[J].US,1973.
[8] 仲雅莉,吴俊辉,刘炫,等.一种基于CORDIC算法的高精度反正切求解[J].电子技术应用,2022,48(1):6.
[9] Miyazaki H,Kanamori T,Islam M A,et al.RVCoreP:An optimized RISC-V soft processor of five-stage pipelining[J]. 2020.
[10] Metzgen P.A high performance 32-bit ALU for programmable logic[C]//Proceedings of the ACM/SIGDA 12th International Symposium on Field Programmable Gate Arrays,FPGA 2004,Monterey,California,USA,2004.
基金
* 国家自然科学基金资助(61704173)。