In this problem we need to calculate R(S): F(R) = S,
where F(R) = R – T(R) + L%*R – T(L%*R).
But function F(R) is NOT increase monotonically!
In other words, it is possible to find R: F(R) = F(R+1).
So, it is impossible to calculate unique R(S).

Example:
99
1000000000 10
2000000000 40
3000000000 41
0          77
100000000
2000000000
3000000000
4200000000
5000000000
1100000001
2200000000
3300000001
-1

If in your program R(S) = max{R: F(R) = S}
then answer = 11086345330.78
(and you'll get AC)

If in your program R(S) = min{R: F(R) = S}
then answer = 11086345330.81
(and you'll get WA #7)

Also your program can write another answers.
Which of they are correct? And why?

Admins. I think Walrus is right. may be need to change a statement...

It is mathematical situation and not of statement.
max{R: F(R) = S} may be =1/3 and approxed by double with eps. After rounding this eps may convert 0.4999999999~0.0
in 0.5000000000001~1. Thus we must solve the problem above
fractions __int64/_int64.

Got Ac when R is double but after floor(R*100+0.5)/100
This problem is another example that in double
calculations we need in description of the author algorithm.

Edited by author 26.08.2009 12:56

There is no fractions. I just used int64. :)

Suggestion to find R(S) = max{R: F(R) = S}
must be added with information: In what class of numbers
should we seach of the S?
1) Numbers of pattern ########.## - rubles and copeys
2) All real numbers