Fw: Number of computers

From: S?bastien Paquet (spaquet@iad.ift.ulaval.ca)
Date: Mon 21 Jun 2004 - 12:37:53 GMT


Hi,

Your only one computer short of having three sets with two clients. I think that this would be the best compromise. It would enable us to do more simulation than last year, but with a less important decreasing of the cpu power.

For my part, our agents should work on one computer, but it is really border line. I know that we will lost some cycles on the maps with many agents. We have work all year long to develop algorithms that uses as efficiently as possible the cpu power we have, but we tought that we would have at least two computers. We will participate either way, we have invested too much in this competition, but I found it really strange that the comitee wants to change everything just one or two weeks before the competition. Those discussions about the cpu power and the competition itself (learning competition, reducing messages, reducing visibily, etc.) should have been discussed in december, not in june.

That's said, I hope that we will have a good competition and that every team will be able to perform well with those last minute changes.
See you in Lisbon.

S?bastien Paquet
-----------------------------------------------------
Leader of DAMAS-Rescue
http://www.damas.ift.ulaval.ca/projets/RobocupRescue/index.php
-----------------------------------------------------
  ----- Original Message -----
  From: jteu004@cs.auckland.ac.nz
  To: r-resc-l@usc.edu
  Sent: Sunday, June 20, 2004 4:34 PM
  Subject: Re: Number of computers

  To all concerned,

  While I agree that there is a significant decrease in computing power since
  last year's competition, I find it unlikely that any teams should be affected.

  Our agents perform some quite complex analysis at each timestep, using
  unoptimized code. What's worse is that it is written in Java. Running all
  agents under Windows on a 1.8GHz machine never uses more than 20% of the CPU
  time available. Infact, the only problems experienced are network ones.

  It seems to me that the time we have available should be sufficient to allow
  teams to implement any ideas they have. Particularly since costly processes
  can now be pre-processed.

  Cheers,
  Jonathan Teutenberg.
  The Black Sheep

  ÍéÒ§ÍÔ§ Mohsen Izadi <roboresc_mi@yahoo.com>:

> Dear Dr. Takahashi,
>
> > Yes, the number of PC is less that at 2002 and 2003.
> > However, the power of CPU and memory are greater
> > than 2002 and 2003,
> > if I am correct. (There is no descriprion on my HP
> > 2003 memo, sorry:)
> >
>
> In 2003 there were 3 Pentium 4 2.4GHZ with 256MB
> Memory.
>
> > As a team leader of Hinomiyagura, I prefer less
> > number of PCs with high
> > specification to more number of PCs with low
> > spec.!!Becaue the former!!
> > configuration provides more stable network traffic
> > to the latter one.
>
> I agree with you! But I think that the change in
> computing power should logically match the decrease in
> the number of PCs. In our case, I think
> one P4 2.8GHZ cannot reach the computing power of
> 3 P4 2.4GHZ PCs.
>
> > I think this configuration is one for preliminary
> > games and after
> > semifinals, 2sets: 1 server and 3 clients. Is it
> > right ?
> > Then, I prefer more games at preliminary with less
> > PCs to
> > less games with more PCs.
>
> I did not find anything in Dr. Akin's e-mail
> indicating that in semifinal there would be 2 sets
> with 1 server and 3 clients.
>
> However, Although it's right that in semifinals we
> have fewer simulations to perform, I do not think that
> having preliminaries with 1 client and then adding 2
> more ones for semifinal would be a fair and logical
> decision.
>
> If having 1 client for preliminaries is because there
> is a limitation on computing resources for each agent,
> then this limitation should be applied to other rounds
> too. But if it is just because of not having enough
> PCs, I think it's not fair to reduce the quality of
> simulations and competitions for this reason. A better
> decision would be to have 2 clusters and dedicating
> more time to run the simulations on
> each day.(e.g. From 9:00 AM to 7:00 PM, 3 situations
> each day)
>
>
> Regards,
> Mohsen Izadi,
> S.O.S. Member
>
>
>
>
>
> __________________________________
> Do you Yahoo!?
> New and Improved Yahoo! Mail - 100MB free storage!
> http://promotions.yahoo.com/new_mail
>



This archive was generated by hypermail 2.1.3 : Mon 21 Jun 2004 - 12:39:03 GMT