First I thought it was precision but now I don't know anymore.
I thought i would never post a thread like this one but...

this is a part of my code, f_x() is P(x):
left1 = left2 = f_x(0);
for(x = 1e-3 ; x < 6000 ; x += 1e-3)
    {
        y1 = f_x(+ x);
        y2 = f_x(- x);
        ans += ((((left1 + y1)/left1)/y1)) + ((((y2 + left2)/y2)/left2));
        left1 = y1;
        left2 = y2;
    }
    ans *= 0.5*(1e-3);

Try to use dx and limits which can be represented in binary form exactly (e.g. 1/1024.0 instead of 1e-3)

appreciate the tip, but still WA11...

Why are you using plus-minus 6000 as limits?
1/P(x) could have its maximum as far as x = 10^6 and a bit further.

Edited by author 06.12.2008 22:36

nvm, Accepted!

Edited by author 02.06.2009 14:16

Thank you a lot, I AC!!!!!!!!!!!!!!