🚀 go-pugleaf

On Thu, 13 Apr 2000 16:21:26 +0200, dataman@cyberdude.com wrote:

>I am really confused about float precision ..officially it's 3.4 *
>(10**-38) to 3.4 * (10**+38)

That's just the range, not the precision.

>But, that does not help me much even it confuse me more ....
>
>Here is my problem:
>
>I need to define a world of 10 000 000 units .  I need to use float for
>that . I think to use only the positive part (but will depend from the
>answer of my question).
>What precision (after the dot ) can i expect  ?

This is platform dependent. To work it out, compile and run the
following program

#include <limits>
#include <iostream>

int main()
{
	using namespace std;
	cout << "d.p. for float = " 
		 << numeric_limits<float>::digits10 - 8
		 << "\nd.p. for double = "
		 << numeric_limits<double>::digits10 - 8
		 << endl;
	return 0;
}

My system runs out of precision before the decimal point with float (I
got -2). You will probably want to use double instead, which gives me
an impressive 7d.p. of accuracy after the decimal point, with a number
around the size of 10 000 000.

Tom