|
|
back to boardWhy? I pass the given tests correctly.
BTW, the problem can have several answers, :S I had WA#1, then WA#22. Can somebody help me? I think there are problems with precision. Wow, I also have WA#22 now.
BTW, my WA#1 was because I didn't #include <math.h> LOL AC now)))
Using epsilon in some more places helped me a lot! Yes!Agree.Impossible get Ac without eps. Please, who can explain me why epsilon is needed in
problems which contain real numbers, and where should
I use it in this problem? I got WA#22, and can't
understand my mistake. There are many trick with epsilon.
For example. Let eps=1e-10 and some geometrical
value d calculated as -1.e-11. We reject this
calculation because d must be >=0 and make mistake
lossing d=0. With eps we do it as
if (fabs(d)<=eps) d=0
thus accepting negative -1e-11. When I used eps=1e-13, I got WA #22, and when I chaged it to 1e-10, I got AC.
Here are fragments, where I use eps:
...
double d=p*p-2*a*b;
if(fabs(d)<eps)
...
(d is discriminant of square equation)
...
if(t1>a+eps || t2>b+eps)
return false;
...
(check that t1<=a and t2<=b).
I can't understand why I had this situation. Maybe the numbers are big, so adding 1e-13 is equal to adding 0. eps = 1e-10
thanks!
Edited by author 30.07.2011 02:06
Edited by author 30.07.2011 02:06 According formule S=sqrt(p*(p-a)*(p-b)*(p-c)) => the problem has 3 different answers. No! Sample #2 has only one answer. |
|
|
