c++ function help

[atown282]

Can someone just help me get started with this. I believe I need to break this into two functions? one called x1 and another called x2? then have the calculations inside the functions rather than being in main? here are my instructions listen below. Then I posted the code as well. Thank you for your time and help.

This version uses smart numerical analysis techniques to find more accurate
solutions in cases where a straightforward use of the quadratic formula
would not work or would not give very accurate answers (because of
"catastrophic cancellation"). There are two techniques used: the coefficients
are first normalized. Then, in cases where there are 2 solutions, only one
solution, call it x1, is calculated using the quadratic formula, namely the
solution in which cancellation cannot occur. Then the second solution, call it
x2, is calculated using x2 = c / (a * x1).

*/

#include <iostream>
#include <cmath>     // Needed for the sqrt function.
using namespace std;


int main(void)
   {
   char Reply;
   int NumSolutions;
   float a, b, c, ANorm, BNorm, CNorm, x1, x2, Discriminant, Root, Largest;
   
   do
   {
      cout << "Quadratic Equation Solver" << endl << endl;
      cout << "Enter the value of coefficient a: ";
      cin >> a;
      cout << "Enter the value of coefficient b: ";
      cin >> b;
      cout << "Enter the value of coefficient c: ";
      cin >> c;

      if (fabs(a) > fabs(b))
         Largest = fabs(a);
      else
         Largest = fabs(b);
      
      if (fabs(c) > Largest)
         Largest = fabs(c);

      ANorm = a / Largest;
      BNorm = b / Largest;
      CNorm = c / Largest;
      Discriminant = BNorm * BNorm - 4.0f * ANorm * CNorm;
   
      if (Discriminant < 0)
         NumSolutions = 0;
      else if (Discriminant == 0)
         {
         NumSolutions= 1;
         x1 = -BNorm / (2.0f * ANorm);
         }
      else   // Discriminant must be positive
         {
         NumSolutions = 2;
         Root = sqrt(Discriminant);
         if (-BNorm > 0)
            x1 = (-BNorm + Root) / (2.0f * ANorm);
         else
            x1 = (-BNorm - Root) / (2.0f * ANorm);
            
         x2 = CNorm / (ANorm * x1);
         }

      if (NumSolutions == 0)
         cout << "No real solutions" << endl;
      else if (NumSolutions == 1)
         cout << "One real solution: " << x1 << endl;
      else
         cout << "Two real solutions: " << x1 << " and " << x2 << endl;  

      cout << "Solve another (y/n)? ";
      cin >> Reply;
   } while (Reply == 'y');

   return 0;
   }

[TheIdeasMan]

Hi atown282,

Be careful in comparing floats - floats are stored in a binary representation and can problems when you compare them. Example 0.28 might be stored as 0.2799999999999999999999... so 0.56 /2.0 may not equal or compare to 0.28 as you might normally expect.

This code can be a problem if a and b are similar values.
if (fabs(a) > fabs(b))

There can be a similar problem with:

if (Discriminant < 0)
and
else if (Discriminant == 0)

Think about this:

if (10.0-(2.0*5.0)) == 0.0 // very likley to be false

The way to specify a 0 float is 0.0 - the compiler or someone might think that 0 is an int.
Storing the variables as doubles might improves the situation, but it s not really a solution as it just means there will be less numbers that cause problems.

The answer is to make use of the built in constants called epsilon. On my system the math library defines a value for DBL_EPSILON.

Google Epsilon and find out how to use for comparisons with floats or doubles.

Now, do you have a problem with your code? You haven't really said whether there is one or not.

Let us know how you are going .

Cheers TheIdeasMan

[atown282]

Well i was given this code and told to break it down into functions and make it run better. so that's what i need to do. i'm currently trying to break it down as i am typing this