Re: Precision

Re: Precision

[Eljay Love-Jensen]

Hi Feodor,

#include <iostream>
    using std::cout;
    using std::endl;

template <typename FP>
int MantissaBitSize(FP one)
{
    int rv = 1;
    FP fXeno = one / 2;
    FP fLast = one;
    FP f = one + fXeno;

    while(f != fLast)
    {
        fLast = f;
        fXeno = fXeno / 2;
        fLast = f + fXeno;
        ++rv;
    }

    return rv;
}

int main()
{
    cout << "float      : " << MantissaBitSize(1.0f) << endl;
    cout << "double     : " << MantissaBitSize(1.0 ) << endl;
    cout << "long double: " << MantissaBitSize(1.0l) << endl;
}

HTH,
--Eljay

[Non-HTML encoded, this time.  Sorry about that.] 

Precision

[Feodor Pisnitchenko]

Hi all!
How can I find out the precision (mantissa length) of a floating point
number in i386 architecture?  
I don't understand the output the following program (why long double and
double have the same mantissa length?)

#include <iostream>
using namespace std;
int main() {
    float c3;
    double c2;
    long double c1;
c3=1.1234567890123456789012345678901234567890123456789012345678901234567890;
c2=1.1234567890123456789012345678901234567890123456789012345678901234567890;
c1=1.1234567890123456789012345678901234567890123456789012345678901234567890;
    cout.precision(50);
    cout << "long double c1 = " << c1 << "   " << sizeof(c1) << endl;
    cout << "     double c2 = " << c2 << "   " << sizeof(c2) << endl;
    cout << "      float c3 = " << c3 << "   " << sizeof(c3) << endl;
}

output:

long double c1 = 1.1234567890123456912476740399142727255821228027344   12
     double c2 = 1.1234567890123456912476740399142727255821228027344   8
      float c3 = 1.12345683574676513671875   4