Overloaded judge during contests

Alexander Grushetsky · Tue Oct 07, 2003 4:12 pm

Could something be done with the system so that it would process submissions faster? It is very annoying. The overloaded system destroys the pleasure from very good problem sets. For example, during the last contest I submitted the problem A in hour before the end, and received the answer "Memory Limit Exceeded" only after the end of the contest. Of course, if'd received this answer during the contest, I would immediately have changed the solution. Actually, all that I had to do, to change int to short in one place to get AC. In the previous contests I received "Presentation Error". These mistakes are specially annoying, as they would easily corrected in the normal conditions. Of course, everyone is in the same condtions, but these conditions are bad now.

In any case, thanks again for all your efforts and for very interesting problems.

Andrey Mokhov · Post by **Andrey Mokhov** » Wed Oct 08, 2003 7:56 am

Hello, Alexander!

I fully agree with you on the situation on this server. Actually the reply times are so long that I almost decided to search for another contest sites, and there seems to be a lot of them. Although the problems are not so good as here and the participants are less famous, the system works quite fast and it provides the feel of real contest.

I don't think this is a problem with system performance - it's rather some limitation on CPU load (it's idle time according to online server status seems to never reach low values, while it should).

Anyway don't think of me as of traitor.

I really wish this online judge would be the best. And I think it is indeed the best now - just some temporal inconveniences, I hope.

Best regards,
Andrey Mokhov.

Moni · Post by **Moni** » Wed Oct 08, 2003 8:57 am

Hmm... Alexander, your words are very much true! It happens during the last periods of the contests! Besides while there arise a long queue and the judge working on it... then any submisson and it's reply sometimes not found! Sometimes it also not inserted into the queue

How this can be solved I am not sure! As the judge is upgraded now... there hardware is also upgraded, Now I think the socket server or the submit-0-matic needs to be upgraded... fpnc can tell better

And Andrey, congrates for your team in acm.zju.edu.cn.
Hmm... you are finding another contest site...

But with you I also believe this is yet the best! though it needs improvement but as fpnc suggested about our OJ... I know he can do better than that! Looking for his reply

hewei · Post by **hewei** » Wed Oct 08, 2003 3:50 pm

Hi, Alexander. I fully agree with you that the OJ should be improved with its performance. By the way, if I may ask, how do you solve problem A Blocks? Do you make use of any heuristic searching strategy like A*? Thanks.

Alexander Grushetsky · Wed Oct 08, 2003 5:19 pm

2hewei
Actually, I used dynamic programming. The running time is O(N^4) (however the inner loop contains a few operations, so the constant is small and there is no TL), and the used memory is O(N^3) (at first I had array int [210][210][210] and that caused ML - it was quite annoying). If you want more advices, mail me.

shahriar_manzoor · Post by **shahriar_manzoor** » Wed Oct 08, 2003 5:29 pm

Well this is not a regular problem. The waterloo contests and the contest before was quite OK. We will try our best but please visit the link below to know what your duty is as a contestant.
http://online-judge.uva.es/board/viewto ... highlight=

[/url]

hewei · Post by **hewei** » Thu Oct 09, 2003 6:28 am

Alexander Grushetsky wrote:2hewei
Actually, I used dynamic programming. The running time is O(N^4) (however the inner loop contains a few operations, so the constant is small and there is no TL), and the used memory is O(N^3) (at first I had array int [210][210][210] and that caused ML - it was quite annoying). If you want more advices, mail me.

Thanks a lot, Alexander!
Well, at my first sight of the problem during the contest, I thought it needed DP. But I had given it up after a few tries because I had proven the two DP methods I came up with to be both incorrect. Yes, I should have known it was DP from the memory usage of the program of yours and those who got AC (All of them was about 8M, 16M or 32M. Maybe it was the result of char, short or int). And surely, I will mail you if I have further doubts.

kuasha · Post by **kuasha** » Thu Oct 09, 2003 9:31 am

using some volunteer Judge on other sites where there will be some critical inputs (less than final) and sending there people get idea if their solution will not be accepted or has chance to be accepted.... and then submit to the uva judge - i think can decrease some pressure

hewei · Post by **hewei** » Thu Oct 09, 2003 12:38 pm

But, kuasha, how to report the result on the status page for those who fails at the "volunteer judges"? Insufficient judge data may easily lead to improper status info.

Subeen · Post by **Subeen** » Thu Oct 09, 2003 12:46 pm

may volunteer judges can be used for checking only the following things:-

1. there is no submission error
2. the code compiles.
3. no restricted function used.
4. memory used is within the limits
5. and sample input/output matches

chenyue · Post by **chenyue** » Thu Oct 09, 2003 5:54 pm

but the above list of checkings takes no time of cpu ...

actually one suggestion given by the administrator of acm.zju.edu.cn is that your monitoring program might have costed too much and there might be too many procedures running simutaneously. It's only a guess

Dmytro Chernysh · Post by **Dmytro Chernysh** » Sat Oct 11, 2003 6:27 pm

Well, I have no words...

I've submitted author solution for 'E' -- and got "Compile Error" ???

But I've submitted

begin
end.

and got AC

"Cool"

Yes, this contest is NOT one of best ever held...

Lars Hellsten · Post by **Lars Hellsten** » Sat Oct 11, 2003 8:14 pm

The judge appears to be completely dead now.

This is pathetic. There have been problems with the judge in pretty much every single UVA contest I've participated in during the past two years. Why doesn't someone actually FIX the system BEFORE deciding to run these contests...

Dmytro Chernysh · Post by **Dmytro Chernysh** » Sat Oct 11, 2003 8:19 pm

Lars, completely agreed with you.

By the way, it's my third contest and third contest with bugs

Maybe I bring a bad luck?

Alexander Grushetsky · Sun Oct 12, 2003 12:16 am

The summary:

C: It seems it has incorrect constraints. I checked (but too late) n, at it is not in [2,100].
D: A mistake in the problem statement. But there was a clarification
E: An incorrect output checker. Accepts any output.
G: There should be not such problems, while the judge is very slow.

And the judge... I have no good words. I even thought it was dead and did other things.

Once again, very interesting problems and not good feelings about the contest.

OJ Board

Overloaded judge during contests

Overloaded judge during contests

Come on guys

it is possible to use judge helper machine