Research On RISC-V Processor Supporting Floating-point Extension Instructions

In most general-purpose computer architectures,floating-point units are integrated into processor chips,such as MIPS,ARM,x86,etc.But for embedded processors,especially some traditional designs,floating point arithmetic is not supported.Many functions of floating-point processor can be simulated through software library,which saves the additional hardware cost of floating-point processor,but compared with pure hardware circuit,the speed is obviously slow,not enough to meet the real-time requirements of embedded processor.In order to improve the performance of the system,it is necessary to implement the hardware of floating point arithmetic unit embedded in the system.In this context,in order to meet the requirements of floating point arithmetic in embedded systems and improve the computing power and speed of embedded systems,this paper studies and designs a processor that supports floating point arithmetic and has a three-level pipeline structure based on the open source RISC-V architecture.The"RV32I"integer instruction set and"RV32F"single-precision floating-point extended instruction set under RISC-V architecture are realized,including single-precision floating-point number and integer conversion,single-precision floating-point number comparison,addition,subtraction,multiplication,division,square root,multiplication and accumulation and other floating-point operation units.Passed the instruction set test of RV32I and RV32F,and integrated the RTL code to the Prodigy-13P system for FPGA prototype verification.On the FPGA system,the highest integration frequency can reach 90MHz,and the LUT and FF consumption of the processor core are 16476 and 4363,respectively.In addition,it ran Coremark program written by C language and obtains 2.67Core Mark/MHz running score.Compared with several embedded processors of Cortex-M series,the performance is better than that of Cortex-M23.Finally,the hardware synthesis of the RV32IF processor designed in this paper based on the TSMC 22nm CMOS technology is carried out by DC tool.The total Cell area is 306858)~2,the frequency is up to 350MHZ,and the power consumption is0.6849mw,which indicates that the RV32IF processor designed in this paper meets the requirements of low power consumption and small area in embedded systems.