A Tour of NTL: Examples: Floating Point Classes

NTL also supports arbitrary precision floating point with the class RR. Additionally, it supports two specialized classes: quad_float, which gives a form of quadruple precision, but without an extended exponent range, and xdouble, which gives double precision, but with an extended exponent range. The advantage of the latter two classes is efficiency.

Here again is a program that reads a list of numbers from the input, and outputs the sum of their squares, using the class RR.

#include <NTL/RR.h>

int main()
{
   RR acc, val;

   acc = 0;
   while (SkipWhiteSpace(cin)) {
      cin >> val;
      acc += val*val;
   }

   cout << acc << "\n";
}

The precision used for the computation can be set by executing

   RR::SetPrecision(p);

which sets the effective precision to p bits. By default, p=150. All of the basic arithmetic operations compute their results by rounding to the nearest p-bit floating point number. The semantics of this are exactly the same as in the IEEE floating point standard (except that there are no special values, like "infinity" and "not a number").

The number of decimal digits of precision that are used when printing an RR can be set be executing

   RR::SetOutputPrecision(d);

which sets the output precision to d. By default, d=10.

See RR.txt for details.

By replacing the occurences of RR by either quad_float or xdouble, one gets an equivalent program using one of the other floating point classes. The output precision for these two classes can be controlled just as with RR. See quad_float.txt and xdouble.txt for details.