Overflow for Intermediate Results

[zytryt]

I am wondering why the following code works:

double x,y,z,w,v;
x = 2.0e300;
y = 1.0e300;
z = 1.0e300;
w = 1.0e300;
v = (x*y)/(z*w);
cout << &quot;v = &quot; << v << endl << endl;

The correct answer of 2 prints out. However, both numerator and denominator exceed the range of double, which is approximately 10 to the 308 power.

Why isn't there an overflow error on the intermediate results? Either a runtime error should occur or the final result should be nonsense instead of 2.

Thanks.

 

[Salem]

While calculations are being performed within the processor itself, floating point numbers are stored in 'long double' format.

long double has a max exponent of 4932 compared to the 308 of a double, so the intermediate results of 2.0e600 is well within range. When the calculation is complete, the long double is converted back to a double, which is no problem since the result is back in range of a double.