Calc pi with long double in C

Question

i have a problem in calc of pi, when i put long double in variable appears a big number, but not the correct value, if you change for calcf function with double it works. My code:

#include <stdio.h>
#include <math.h>

long double calc(int n, double denominador) {
    //printf("%lf e",(pow(-1,n))/denominador);
    //printf("%d\n",(denominador));
    long double var = (long double) ((4.0*(pow(-1.0,n)))/denominador);
    printf("%Lf\n",(long double)var);
    return var;
}

double calcf(int n, int denominador) {
    //printf("%lf e",(pow(-1,n))/denominador);
    //printf("%d\n",(denominador));
    double var = (4.0*(pow(-1.0,n)))/denominador;
    //printf("%lf",var);
    return var;
}

int main() {
    int NUMERO = 100000000;
    long double pi = 4L;
    double pif = 4;
    int i;
    int n=1;
    printf("%d e %d",sizeof(double),sizeof(long double));
    for (i=3;i<NUMERO;i+=2) {
            pi += calc(n,(double) i);
        //pif += calcf(n,i);

            n++;
    }
    printf("PI: %1.50Lf\n",pi);
}

What i'm doing wrong?

Thank you.

i expect output is equal but more precision than double (calcf function in this case) — kavain, Aug 05 '11 at 04:07
I get the same output with both `calc()` and `calcf()` using `gcc`: `8 e 16PI: 3.14159263358979361680130282241663053355296142399311`. What compiler are you using? — Chris Gregg, Aug 05 '11 at 04:11
i didn't found this information, where can i see this? but i'm using codeblocks to compile — kavain, Aug 05 '11 at 04:24

score 3 · Answer 1 · edited Nov 09 '15 at 09:19

3

pow() takes and returns doubles. You need to use powl() which takes and returns long doubles instead. Otherwise you're just casting a double to long double and are not gaining any precision.

edited Nov 09 '15 at 09:19

Tom Zych

13,329
9
36
53

answered Aug 05 '11 at 04:15

Splat

753
3
10

i try this but the problem is other – kavain Aug 05 '11 at 04:24
2

@thamerhatem If you want to get more help from people here, you need to provide more information than "it didn't work", especially with a program that takes many minutes to run. I think there are probably two issues here. First, I don't think this algorithm converges to a point where there will be a significant difference between double and long double in the number of iterations you're using. Second, floating point math is fraught with potential for inadvertently causing errors. See this Wikipedia article for more information about it: http://en.wikipedia.org/wiki/Floating_point – Splat Aug 06 '11 at 18:32

score 3 · Accepted Answer · answered Aug 26 '11 at 22:51

Well, this program has many issues; but the worst one is that you're using Gregory's series, which is just about the slowest algorithm out there. You're not even going to get double precision with that, let alone long double. You would have to let it run through about 10¹⁶ steps just for double.

Mathematicians use −1ⁿ to indicate an alternating sign, which is fine in that context, but using pow() to calculate it is just silly, especially since it's a floating point operation. Instead, use an integer that starts at -1 and keep doing i = -i.

You need to be more careful with your types and constants. 4L is a long integer, not a long double. You want 4.0L. Of course the compiler casts it anyway, but it's bad practice. You're using plain old double constants like 4.0 when you're aiming for long double precision.

It's also silly to keep multiplying everything by 4.0 at every step. You can just do it at the end.

Calc pi with long double in C

2 Answers2