Could this be done better ?

I was wondering if there was any one in here that could tell me if this code could be done better or faster. Take a look.

Code:

---

void VectorAngles( const float *forward, float *angles )
{
        float        tmp, yaw, pitch;
       
        if (forward[1] == 0 && forward[0] == 0)
        {
                yaw = 0;
                if (forward[2] > 0)
                        pitch = 90.0;
                else
                        pitch = 270.0;
        }
        else
        {
                yaw = (atan2f(forward[1], forward[0]) * 57.3f);
       
               
                if (yaw < 0) yaw += 360.0;

                tmp = sqrt (forward[0]*forward[0] + forward[1]*forward[1]);
               
                pitch = (atan2f(forward[2], tmp) * 57.3f);
               
                }

                angles[0] = pitch;
                angles[1] = yaw;
                angles[2] = 0;
}

---

You might want to let the functions take the floats independently, or even use structs, so you don't accidentally pass an undersized pointer/array and get GPFs. Also, if you name them appropriately (IE *X, *Y, *Z, P, Y), it will be much easier to see what you're doing in terms of vector maths.

Secondly, you could just write

Code:

---

if (!forward[1] && !forward[0])

---

instead of

Code:

---

if (forward[1] == 0 && forward[0] == 0)

---

This function works with pointers and changes array elements outside its scope. In big programs this is not desired. Perhaps you should use classes and references (C++ style).

Some people need faster sqrt-function than in <cmath>. Just try FastSQRT(...):

Code:

---

#include <iostream>
#include <conio.h>
#include <cmath>
#include <ctime> // clock_t, clock(), ...

using namespace std;

float FastSQRT( float r )
{
    // Null und negative Zahlen abfangen
    //if( r == 0.0f ) return 0.0f;
    if( r <  0.0f ) r = -r    ;
 
    // Startwert festlegen
    float val = r;
    float halfval = 0.5f * r;
    unsigned long *ptr = (unsigned long*)&r;
    *ptr = ( 0xbe6f0000 - *ptr ) >> 1;

    float temp = r;

    // Zwei Schritte reichen
    temp *= 1.5f - temp * temp * halfval;
    temp *= 1.5f - temp * temp * halfval;
    //temp *= 1.5f - temp * temp * halfval; //höhere Präzision 
 
    // Rückgabe
    return val*temp;
}

int main()
{
  double x;
  cout << "Bitte Zahl eingeben fuer Wurzelziehen: ";
  cin  >> x;
  cout << FastSQRT(x) << endl;
  cout << sqrt(x)    << endl << endl;
   
  // Zeitmessung
  const int NMAX = 20;
  int max;
  cout << "Anzahl Schleifendurchlaeufe? ";
  cin >> max;
  clock_t t1,t2;
  double ts, tm=0;

  double a;

  cout << endl;
  for(int n=0; n<NMAX; ++n)
  {
      t1 = clock(); // Start
      for( int i=0; i<max;i++)
      {
        a = FastSQRT(x); //zu messende Aktion 1
      }
      t2 = clock(); // Ende
      ts = (t2-t1)/static_cast<float>(CLOCKS_PER_SEC); // Zeit in Sekunden.
      cout << "Zeit Aktion 1: " << ts << " sec" << endl;
      tm += ts; // Das ist das Gleiche wie tm = tm + ts     
  }
  tm/=NMAX; // Das ist das Gleiche wie tm = tm / NMAX
  cout << "Durchschnitts-Zeit bei Aktion 1: " << tm << " sec" << endl;

  tm=0;
  cout << endl;
  for(int n=0; n<NMAX; ++n)
  {
      t1 = clock();
      for( int i=0; i<max;i++)
      {
        a = sqrt(x); //zu messende Aktion 2
      }
      t2 = clock();
      ts = (t2-t1)/static_cast<float>(CLOCKS_PER_SEC);
      cout << "Zeit bei Aktion 2: " << ts << " sec" << endl;
      tm += ts;   
  }
  tm/=NMAX;
  cout << "Durchschnitts-Zeit bei Aktion 2: " << tm << " sec" << endl;

  getch();
}

---

Thks for the reply's. according to my output running the code u posted erhard, the sqrt functions is faster.

I think modern compilers use special intructions for sqrt(), making such hacks worthless.

What about using valarray for the code above?If yes, could someone show me how to implement it.

You should be able to create lookup tables for the atan2 calls.

You might also want to read this site. I done too much with valarrays. http://cs.calvin.edu/books/c++/intro.../Valarrays.pdf