CODEWITHSUNDEEP
c++structfloating-pointunions
Jack
Jack

Reputation: 13

Creating own float structure in C++

Yet my lectures in C++ at university began yet i got my first problems. Our task was it to implement a self made structure in C++ for floating points via the IEEE 754 standard:

Create a data structure that allows you to store a float, read its raw byte representation and its internal representation as s, e and m. Use a combination of union and bit-field-struct. Write a program where a float number is assigned to the float part of the structure and the raw and s/e/m representation is printed. Use hexadecimal output for raw and m.

What i had so far is the following:

#include <stdio.h>
#include <math.h>

union {
    struct KFloat {
        //Using bit fields for our self made float. s sign, e exponent, m mantissa
        //It should be unsigned because we simply use 0 and 1
        unsigned int s : 1, e : 8, m : 23;
    };
    //One bit will be wasted for our '.'
    char internal[33];
};

float calculateRealFloat(KFloat kfloat) {
    if(kfloat.s == 0) {
        return (1.0+kfloat.m)*pow(2.0, (kfloat.e-127.0));
    } else if (kfloat.s == 1) {
        return (-1.0)*((1.0+kfloat.m)*pow(2.0, (kfloat.e-127.0)));
    }
    //Error case when s is bigger 1
    return 0.0;
}

int main(void) {
    KFloat kf_pos = {0, 128, 1.5707963705062866};//This should be Pi (rounded) aka 3.1415927
    KFloat kf_neg = {1, 128, 1.5707963705062866};//Pi negative

    float f_pos = calculateRealFloat(kf_pos);
    float f_neg = calculateRealFloat(kf_neg);

    printf("The positive float is %f or ",f_pos);
    printf("%e\n", f_pos);

    printf("The negative float is %f or ",f_neg);
    printf("%e", f_neg);
    return 0;
}

The first error with this code is clearly that the mantissa is absolutely wrong but i have no idea how to fix this.

Upvotes: 0

Views: 4012

Answers (2)

masrtis
masrtis

Reputation: 1310

Bruce Dawson has an excellent series of blog posts on floating point representation and arithmetic. The latest in the series, which has a bunch of links to previous posts that discusses this subject matter in detail, is here.

Upvotes: 1

vlad_tepesch
vlad_tepesch

Reputation: 6915

please reread the task:

Create a data structure that allows you to store a float, read its raw byte representation and its internal representation as s, e and m.

this does not mean that you should store a string

I would do this the following way:

union MyFloat
{
  unsigned char rawByteDataRep[4];
  unsigned int  rawDataRep;
  float         floatRep;
  struct{   // not checked this part just copied from you
   unsigned s : 1;
   unsigned e : 8;
   unsigned m : 23;
  }             componentesRep;
}

but be careful! Besides the fact that this union-conversion pattern is widely used, the C-Standard states that the result is undefined behaviour if you read another unionmember than the one that was written.

Edit: added uint32 rep

void testMyfloat()
{
  MyFloat mf;
  mf.floatRep = 3.14;
  printf("The float %f is assembled from sign %i magnitude 0x%08x and exponent %i and looks in memory like that 0x%08x.\n",
        mf.floatRep, 
        (int)mf.componentesRep.s, 
        (unsigned int)mf.componentesRep.m, 
        (int)mf.componentesRep.e, 
        mf.componentesRep.rawDataRep);

}

Upvotes: 2

Related Questions