Re: [ns] It's impossible !!!!! I don't know !!!! This is correct !!!

[Felix Hernandez Campos <fhernand@cs.unc.edu>]

Vincenzo Mazzotta wrote:
> Hi
> to all ...I've a problem if i sum two or much variable with class operation
> :      
> double a = ((double) (now)) + ((double) (delay_line[line]));
>       double
> b = ((double) (punt->pkt.arriveTime)) + ((double) (delay_line[punt->pkt.linedelay]))
> + ((double) (txtime(punt->pkt.pkt)));
>       fprintf(stderr,"%g,%g,%g,%g,%g
> %g %g\n",now,delay_line[line],punt->pkt.arriveTime,delay_line[punt->pkt.linedelay],txtime(punt->pkt.pkt),a,b);Go
> out on my screen :                               
> 1,1e-05,1,1e-05,8e-07 1.00001 1.00001 IT'S
> IMPOSSIBLE !!!!!Somebody have a reazon and a solution for this ???
It's actually far from impossible. %g default precision is 6 decimal places.
Therefore, for that precision, both sums should give the same result. Increase
your precision with something like %.10g and you will see what you are
expecting.

As a proof of concept, the following program

main() {

        double a = 1, b = 1, c = 1e-05, d = 1e-05, e = 8e-07;

        printf("%g+%g=%g %g+%g+%g=%g\n", a, c, a+c, b, d, e, b+d+e);

        printf("%g+%g=%g %g+%g+%g=%.10g\n", a, c, a+c, b, d, e, b+d+e);

}

prints out:

1+1e-05=1.00001 1+1e-05+8e-07=1.00001
1+1e-05=1.00001 1+1e-05+8e-07=1.0000108

-- Felix Hernandez