"All members who have not previously advanced – limited to the first 2,500 Competitors who register in the Arena" — from the rules linked in the post.

Its scheduled for tomorrow 12th april, still 26 hrs

EDIT: My solution is wrong.

My code failed in 1000. My logic is as follows. I find the max protection needed among all the positions not covered by any local shield and set the global shield to that value. After that I assign local shields greedily from left to right. When, at a given position, the current total protection offered is less than that needed, I find the local shield whose ending point is the rightmost among those active currently, and increment its value as needed. Can anyone please tell me whether this logic is wrong (if so, I made an implementation mistake)?

I did ternary search on the amount the special shield is powered, and it looks like everyone did this. After that, you can sweep from left to right and use greedy with the addition of the simple shields, by using the one that extends furthest to the right.

I unfortunately didn't think much about how it works — it was more based on intuition and the fact that the problem is supposed to be solvable quickly. Can anyone prove why this approach works (or give a counterexample)?

I did ternary search on the value of the special shield. But need to set left boundary to the first position where there is a solution.

First, need to sort intervals and remove redundant ones (e.g. which are fully covered by some other interval).

When checking given value of the special shield, simply sweep through cells from left to right, checking for each interval start/end. Assume interval starts, then before the next interval you have to cover everything with the previous one. But you need to remember the intervals which were opened previously and how much money you have put in them. I used queue for that.

EDIT: Rewrote comment.

The constraint "if len>=n" stop and return -1 applies to the original n, not the last number in the sequence (they never call that n). Not saying it was a misunderstanding; I just know I had a bit of confusion over this and had to reread a couple times.

Others definitely also made this mistake, and 1e9 catches it as hellman_ said below, but even if you add a catch for "if (n==1) return -1;" you can still fail for example on 1162.

1162->42->20->4 (now len>=n and you break)->16->17 (answer).

I think this is where most hacks were from, but also some people in my room just used a bool prime[] array and failed on big primes.

Yes previous round was rated

Насколько я понял, если прошел с раунда 1A или же 1B, то не нужно и не можешь.

Да, правильно понимаешь. Ты во 2-ом раунде.

It's updated now, I can see my latest rankings and stats on my profile!

Maybe they are manually rewriting them :P

leave it to karma

Why downvotes? This guy solved 250 in 21 seconds and 500 in 47 seconds. See here too. Such a gifted coder!

I just saw this post so i need to clear things a bit , First of all apologies , both arrogantidiot and zoomswk are correct but partially , as u can see this was my 3-4 serious contest on topcoder , i was there in my friends room when i saw the questions on his screen and thus coded them there only without knowing the fact( again my ignorance ) that topcoder judges on relative time unlike codeforces ,and that is when things went a bit haywire ,and unknowingly and unintentionally the following scenario happened , I would be more than glad , if u can provide some help on how to unjudge my problems .