FixedPointNumberLibrary

Simple C++ header-only library for Fixed-Point Number operations

C version here, designed to be compact

Main files

• include/FixedPointNumber.hpp
  • the header-only library of class FixedPointNumber
• src/FixedPointNumberTest.cpp
  • the test of class FixedPointNumber
• src/FixedPointNumberDemo.cpp
  • the demo of 1D-convolution using class FixedPointNumber
  • more examples will be listed below
• makefile

Usage

Include library

#include "FixedPointNumber.hpp"

MUST be compiled with C++17 or above, e.g.,

g++ YourCPP.cpp --std=c++17

Create a FixedPointNumber object

• sign_bit is always of width 1
• int_bits is the width of integer part bits
• frac_bits is the width of fraction part bits

The below example instantiates a FixedPointNumber which has

• 1bit sign-bit
• Xbits int-bit
• Ybits frac-bit

FixedPointNumber<X, Y> fp(...);

And the constructor arguments can be of type

• uint32_t
• int32_t
• double
• FixedPointNumber<V, W>
  • This will construct this FixedPointNumber<X, Y> from the value of another object which is a FixedPointNumber<V, W>

e.g.

FixedPointNumber<3, 16> fp1(0xFFF9EU);
FixedPointNumber<7, 8>  fp2(0x0067);
FixedPointNumber<7, 8>  fp3(1.2345678);
FixedPointNumber<3, 16> fp4(fp3);
// support operator=
FixedPointNumber<3, 16> fp5 = 1.2345678;
FixedPointNumber<3, 16> fp6 = fp1;

Functions

• to_double

FixedPointNumber<3, 16> fp(3.14159265359);
cout << setprecision(10) << fp.to_double() << endl;
// output:
// 3.141586304

• get_value

FixedPointNumber<7, 8> fp(0x1234);
cout << "0x" << hex << fp.get_value() << endl;
// output:
// 0x1234

• operator- (negative)

FixedPointNumber<3, 16> fp(3.14159265359);
cout << setprecision(10) << (-fp).to_double() << endl;
// output:
// -3.141586304

• operator+

FixedPointNumber<3, 16> fp1(3.14159265359);
FixedPointNumber<3, 16> fp2(2.71828182846);
cout << setprecision(10) << (fp1+fp2).to_double() << endl;
FixedPointNumber<3, 16> fp3(3.14159265359);
FixedPointNumber<7, 8>  fp4(2.71828182846);
cout << setprecision(10) << (fp3+fp4).to_double() << endl; // result will be FixedPointNumber<3, 16>
cout << setprecision(10) << (fp4+fp3).to_double() << endl; // result will be FixedPointNumber<7, 8>
// output:
// 5.859863281
// 5.856430054
// 5.85546875

• operator*

FixedPointNumber<7, 8>  fp1(3.14159265359);
FixedPointNumber<7, 8>  fp2(2.71828182846);
cout << setprecision(10) << (fp1*fp2).to_double() << endl;
FixedPointNumber<7, 8>  fp3(3.14159265359);
FixedPointNumber<5, 10> fp4(2.71828182846);
cout << setprecision(10) << (fp3*fp4).to_double() << endl; // result will be FixedPointNumber<7, 8>
cout << setprecision(10) << (fp4*fp3).to_double() << endl; // result will be FixedPointNumber<5, 10>
// output:
// 8.5234375
// 8.5234375
// 8.53515625

• operator<<

FixedPointNumber<7, 8> fp1(0x1234);
cout << fp1 << endl;
cout << fp1.to_double() << endl;
FixedPointNumber<7, 8> fp2(18.203125);
cout << fp2 << endl;
cout << fp2.to_double() << endl;
// output:
// 0x1234
// 18.203125
// 0x1234
// 18.203125

Examples

Convolution
FIR Filter
FFT