Look at the function f():

#define SGN(X) (((X)>eps)?1:(((X)<-eps)?-1:0))
const double eps=1e-6;
double sol(double k,double p){
____double l,r,m;
____l=0;
____r=1e5;
____while(r-l>eps){
________m=(l+r)/2;
________if(((k+m)/sqrt(m*m+1))<p)
____________r=m;
________else
____________l=m;
____}
____return m;
}
int f(double k,double p){
____int n;
____double x,w;
____if(SGN(p-1)<=0)
________return 0;
____x=sol(k,p);
____w=(1+x*k)/sqrt(x*x+1);
____if(SGN(p-w)>=0){
________n=floor((p-w)/sqrt(x*x+1));
________return(1+(int)n);
____}else
________return 0;
}

My (like any other) AC program outputs "0" for test "99 100", but it's incorrect! Because we can place coffins near main diagonal.
Let's denote AC-program output as INCORRECT(n,k). Then the correct answer will be

max{INCORRECT(n,k), (n+(n%k))*(n/k)+f(k,n%k)}

and for test "99 100" it will be "82", not "0". Examples:
"5 6" --> "1"
"7 8" --> "2"
"11 12" --> "4"
"9999 10000" --> "9855".
You may cut out respective paper rectangles to check it manually :^). But (it's really fun!) there are no tests like this (i.e. in all cases INCORRECT(n,k) >= (n+(n%k))*(n/k)+f(k,n%k) ), and programs which output INCORRECT(n,k) got AC!

Coffins must be parallel to cemetery borders.
Problem statement was updated.

Thank you.