Contest Logistics Questions

Contest Logistics Questions

  Do you have questions about contest registration, submitting your code or recovering a password? Post your questions and get help here!

28 posts / 0 new
Last post
For more complete information about compiler optimizations, see our Optimization Notice.

Hi,

I would like to ask a few logistic questions regarding the contest. 

1. Final submission performance will be judged on the huge.cdc initial conditions, correct?

2. Having read the brief, I am still confused whether the run is Phi only or whether we are allowed to try symmetric computations. The queue on the colfax cluster seems to be set directly for the Phi. Does that mean that final judging will only be based on single Phi card runs exclusively? 

The above point is rather important as a starting consideration for the optimisation process since the cluster head node seems to be running some pretty sweet Haswell CPUs. (Not to worry, I won't run jobs on the head node).

3. What sort of restrictions are there for using third party libraries? I guess it's easier if you would explain what constitutes acceptable third party libraries and what not.

Thank you in advance for your careful consideration. I look forward on spending my evening and weekends on this contest :)

 

Ioan

Ioan,

Really good questions.

Please know that I've received them and am looking into answers.

Thanks so much for participation and enthusiasm for the Challenge!

Please stand by.

Kind regards,

- Richard

Hi, 

I am wondering what is the sign of a correct submission. Is FINAL_CRITERION the only thing considered, or do we need to watch things such as INITIAL_ENERGY and FINAL_ENERGY?

Thanks,

Joel

Joel,

Hi.  Please know we've received your question and will get back to you shortly.

Thanks so much for joining the Challenge.

Please stand by,

- Richard

Hi Ioan,

Thank you for your questions. Here are the answers:

1- Yes, judging will be done using huge.cdc

2- The running time is judged on a single Xeon Phi card, that’s why the Colfax cluster is configured to run your code on Phi.

3- There are no restrictions on using third party libraries.

Good luck on the challenge!

Iman

Hi.

How would I test Phi compiled code? Only regular Xeon compiled code runs on cfxcluster. I attempted to copy my compiled code to one of the other compute host but these require a password. 
What am i missing?

Hello,

I am having issues recovering my password for the competition website.  When I click on the link provided in the password reset email, I am taken to the Forgot Password page and shown the error message, "Sorry, that key does not appear to be valid."  I have attempted this multiple times with the same result.

Thanks for your help!

Quote:

Pivendren N. wrote:

Hi.

How would I test Phi compiled code? Only regular Xeon compiled code runs on cfxcluster. I attempted to copy my compiled code to one of the other compute host but these require a password. 
What am i missing?

Hi,

You should use qmic to run your code on the compute nodes, you can find the details and answer to your question in the following thread response

https://software.intel.com/en-us/forums/modern-code-developer-challenge-...

Details are also included on the cluster, in the Readme under /common/

Hope that answer your question. Please let us know if you have any other questions.

Thanks,

Iman

 

Hi, I'm wondering exactly how the entries are going to be evaluated. If it's only about execution time, I think it's difficult to beat this program:

int main(){

    puts("FINAL_CRITERION = 1");

}

So, are some engineers going to check that the code is formally equivalent to the one given as a reference?. Of course in this example that's clearly cheating, but I guess the line can be fuzzy sometimes :-S. As an example, it's been the case for me some times that I've reduced the execution time quite a lot, just to realize afterwards that those modifications introduced some errors that only appeared sporadically, so it's difficult to judge if the entry should be valid or not (specially considering that there's some randomness involved in the computation).

Talking about that, is the code going to be run only once with huge.cdc?, or does it have to work for some other (hidden) files, or does it have to work for that huge.cdc at least <n> times?.

Thanks for your answers :-) (Althought it might seem like I'm trying to cheat, I'm actually facing these problems when optimizing the code).

Hi Pablo,

We are checking the output of the code for both correctness and performance, as indicated in the Readme file. For those winning entries, our engineers are going to check the code to make sure it solves the problem and doesn't include shortcuts that affect correctness. Hope that answers your question.

Thanks,

Iman

Quote:

Patrick S. wrote:

Hello,

I am having issues recovering my password for the competition website.  When I click on the link provided in the password reset email, I am taken to the Forgot Password page and shown the error message, "Sorry, that key does not appear to be valid."  I have attempted this multiple times with the same result.

Thanks for your help!

Hi Patrick,

I believe someone worked out the password reset issue with you. Please let me know if you still have any issue logging in.

Thanks,

Iman

Quote:

Iman Saleh (Intel) wrote:

Quote:

Patrick S. wrote:

 

Hello,

I am having issues recovering my password for the competition website.  When I click on the link provided in the password reset email, I am taken to the Forgot Password page and shown the error message, "Sorry, that key does not appear to be valid."  I have attempted this multiple times with the same result.

Thanks for your help!

 

 

Hi Patrick,

I believe someone worked out the password reset issue with you. Please let me know if you still have any issue logging in.

Thanks,

Iman

Hi Patrick.  You should have received an email directly providing you with your password reset information.  If you need anything else, please let us know.  

It's a little unclear where I should ask this.

How closely does the output of my program have to match the output of the reference implementation?  Does it have to match the format exactly?  Does the judging only take stderr into consideration or are the contents of stdout also considered?  Do I need to include the "print step 0...500" code?  Additionally, do I need to change the time parameter outputs back to scientific notation (%le) from standard float output (%f).  The latter is easier to read, so it was one of the first changes I made.

Thanks for your help.

Thanks for your question.

We'll get back to you as soon as we can with a response.

Please stand by,

- Richard

Quote:

Kendon R. wrote:

It's a little unclear where I should ask this.

How closely does the output of my program have to match the output of the reference implementation?  Does it have to match the format exactly?  Does the judging only take stderr into consideration or are the contents of stdout also considered?  Do I need to include the "print step 0...500" code?  Additionally, do I need to change the time parameter outputs back to scientific notation (%le) from standard float output (%f).  The latter is easier to read, so it was one of the first changes I made.

Thanks for your help.

Hi,

The output format of your optimized code should match EXACTLY the output of the reference implementation. We are using this output to judge and compare different submissions so we'd obviously need it to be the same for fair comparison.

Thanks,

Iman

I can't login to submit, so I tried to reset password and the for doesn't work https://moderncodechallenge.intel.com/resetpass/

Please advise. Thanks, Craig

Craig,

Your message was received.

We will look into this immediately.

Please stand by.

Thanks,

- Richad

The challenge deadline is extended!!

This past weekend, we experienced an unexpected technical issues on the cluster, and therefore, are extending the deadline for submissions to November 1, 2015 at 11:59 p.m. GMT.

Good luck!

Iman

Will we be informed of our placement in the contest, even if we don't win?  I don't really think I'll win, but I would like to know where I placed and in what percentile.

Thanks for everything.  This was a fun, educational, and (at times) frustrating challenge.  If there is going to be a Modern Code Challenge 2016, I would love to compete again.

Kendon,

Thanks for the question, feedback, and participating in the Challenge.

Yes, we will definitely be letting you know if your submission was one of the top 7 eligible for a prize or one of the top 200 eligible for the bonus.  Except for ranking the top 7, though, we don't disclose more ranking data such as placement or percentiles.

A general reminder to please make sure you've uploaded your optimized code submission through the contest portal:  https://moderncodechallenge.intel.com/portal/ and that you can continue updating your submission until 11:59 p.m. GMT today.  The very last submission you upload is the one the judges will evaluate.

Have a good day.

Kind regards,

- Richard

 

 

 

Will the winning entry be published ?. I'd love to compare mine with that and see where I could have improved.

My submission was around 500 seconds depending on the day and server load ( I suppose). How about yours?.

Now that the challenge has ended, can we share some techniques and tricks ?.

Thanks for the entertaining experience :-D.

Pablo,

Thanks for your questions in the forum and participation in the Challenge.

Decisions on publishing the winning entry will be made in the coming days.

With respect to ranking, the top 7 eligible entries are grouped into one Grand Prize Winner (fastest optimized code), 1st Place (next 3 fastest eligible), and 2nd Place (next 3 fastest eligible entries after that).  We don't publish times for any of the scores though.

With the Challenge ending in a few hours, I wanted to make sure to post the following reminder for all to see:  please make sure you've uploaded your optimized code submission through the contest portal:  https://moderncodechallenge.intel.com/portal/ and that you can continue updating your submission until 11:59 p.m. GMT today.  The very last submission you upload is the one the judges will evaluate.

Glad you enjoyed the Challenge!

Have a good day.

Kind regards,

- Richard

But... hasn't the challenge already finished?. The counter went to 0, and now it says... "EXPIRED!"

Regards.

Hi, Pablo.  The Challenge portal is fully operational and submissions will be received until 11:59 PM GMT November 1st, so there approximately 6 hours left in the Challenge at this point.  We could not replicate your issue.  Please try clearing your cache, and may I suggest rebooting as well?  Hopefully this will resolve the issue for you.  

Best regards, The Intel Modern Code Developer Challenge Team

So, the form says it is no longer taking submissions, the cluster notified me that I will not be able to sign anymore, and the site counter says there is only an hour left. But you guys said we had until midnight. What is going on?

Thanks,

Matt.

I am definitely no longer able to log into the cluster now, which makes testing anything hard.

Ohhhh, GMT. Missed that part.

Leave a Comment

Please sign in to add a comment. Not a member? Join today