Research And Implementation Of Single And Double Precision Floating Point Operator Based On FPGA

The computer can process various kinds of analog information like numbers,texts and pictures which is expressed in data form.There are two kinds of data formats: fixed-point number and floating-point number.Fixed-point number can only represent integer and pure decimal.When the range of computer number is fixed,the fixed-point number has a small range and low accuracy;but in the same position,the floating-point number could reach larger values and higher accuracy.With the increase of computer implement,the progress of processing of floating-point data is becoming more and more complicate.As the important part of the data processor,the floating-point number compute unit is usually designed as an independent module.This kind of design could make process floating-point data more flexible and reduce the consumption of hardware resources.This paper designs a single and double precision floating point computer which sharing of a set of hardware resource.Firstly,this thesis introduces the IEEE-754 standard,analyzes the storage format of single-precision and double-precision floating-point,refers to some literature about floating-point computer,describes the basic rules and algorithms of addition,subtraction,multiplication and division in floating-point compute,analysis and study the structure of floating-point operator like adder,subtractor,multiplier and divider.Secondly,This paper completed the floating-point operator code preparation and comprehensive simulation on the Quartus II software through the hardware description language Verilog HDL,and download floating-point computer to the DE2-70 development board to test the correction.At last,this paper design floating-point operator that runs a double-precision or two-way,single-precision parallel on the FPGA.The result of DE2-70 development board shows that the floating-point operator this paper designed can effectively and correctly perform single-precision and double-precision floating-point number operations,and the operator could switch easily between single-precision and double-precision compute on a same set of hardware resources.