mersenneforum.org  

Go Back   mersenneforum.org > Prime Search Projects > No Prime Left Behind

Closed Thread
 
Thread Tools
Old 2009-01-13, 23:30   #177
MyDogBuster
 
MyDogBuster's Avatar
 
May 2008
Wilmington, DE

1011001001002 Posts
Default

Quote:
Perhaps when we get up to 375k or 400k on the 8th drive, we can start the 200k-350k range. I'll put it up to other's opinions as to when we start but I think we definitely need to wait 1-2 months.
I vote for doing anything BUT 200K-350K. I have never seen such a waste of resources. Sure it has to be done, but at the expense of everything else. The only thing I would commit to either drive 9 or 200k-350k would be my slowest three laptops (and that ain't gonna happen). I might get it done by myself in about a year and the answer I get then will be just as useful as the one everyone seems to want to get to by Saturday. Boy, did I waste resources sieving for THAT. Maybe the next time, the people who are so hot to trot on this stuff can do their own sieving.
MyDogBuster is offline  
Old 2009-01-14, 00:16   #178
gd_barnes
 
gd_barnes's Avatar
 
May 2007
Kansas; USA

1024610 Posts
Default

Quote:
Originally Posted by MyDogBuster View Post
I vote for doing anything BUT 200K-350K. I have never seen such a waste of resources. Sure it has to be done, but at the expense of everything else. The only thing I would commit to either drive 9 or 200k-350k would be my slowest three laptops (and that ain't gonna happen). I might get it done by myself in about a year and the answer I get then will be just as useful as the one everyone seems to want to get to by Saturday. Boy, did I waste resources sieving for THAT. Maybe the next time, the people who are so hot to trot on this stuff can do their own sieving.
Keep in mind that we're wanting to find all of the primes, not just the top-5000 primes. On the sieving effort, it's much more efficient to include the entire range in sieving. Had we excluded n=200K-350K from sieving, we would have only seen a 40% increase in total sieving speed because sieving speed varies with the sqaure of the n-range. In other words, had we sieved the ranges separately, it would have taken 60% more total CPU time.

We're only in a bit of a hurry for the n=350K-500K part of the sieve but while doing that, we may as well do the larger range in a more efficient manner. We'll be running n=200K-350K in a while. Some like Carlos will like such a range. Others like you will prefer the higher range...best of both worlds.


Gary
gd_barnes is online now  
Old 2009-01-14, 00:27   #179
gd_barnes
 
gd_barnes's Avatar
 
May 2007
Kansas; USA

2·47·109 Posts
Default

Quote:
Originally Posted by em99010pepe View Post
You can do whatever you want just let IB9000 with work (n=200k-350k)for me and for the others interested in the lower ranges. If you like to please everyone do that.

No problem. We can do that. If you want a range for your server, just let me know. Actually, that is a good idea that I should have considered before. Just like we "seamlessly" went from the 1st drive to the 5th drive by loading the 5th drive range before the 1st drive range was done, we can do the same with port 9000. That way, people can leave their cores running and with the multitude of them on port 9000 right now, that's a good thing!

Even if we do that, it'd be really nice if people took some time off from it to run the rally on the n=352K-500K range with us.

Karsten, I know this means that we'll have a little 'too much' running at once for an extended period for your tastes. Likely we won't promote the n=200K-350K portion of the 9th drive too much until we get a little higher in the 8th drive.

I know some people won't agree with our methods at times but we will get all of the primes found in a systematic manner that is agreeable to as many people as we can while also being highly competitive on the top-5000 site. Trying to balance all the tastes and goals is not an easy task.


Gary
gd_barnes is online now  
Old 2009-01-14, 00:32   #180
MyDogBuster
 
MyDogBuster's Avatar
 
May 2008
Wilmington, DE

22×23×31 Posts
Default

Quote:
Keep in mind that we're wanting to find all of the primes, not just the top-5000 primes.
I understand that, and the sieving was needed, but we sure as heck don't have to commit 20-30 cores to it (as ReiselSearch starts pulling away again). 2000-3000 new primes later, we will still be 290 primes behind them.
MyDogBuster is offline  
Old 2009-01-14, 00:52   #181
gd_barnes
 
gd_barnes's Avatar
 
May 2007
Kansas; USA

101000000001102 Posts
Default

Quote:
Originally Posted by MyDogBuster View Post
I understand that, and the sieving was needed, but we sure as heck don't have to commit 20-30 cores to it (as ReiselSearch starts pulling away again). 2000-3000 new primes later, we will still be 290 primes behind them.

I see your point. All that I'll ask is to try to think a little longer term. Call it an "investment" for the future. Having quickly done fully sieved files (that are not over-sieved!) will put us in a position to make a big move vs. other projects once we get crunching on the range.

One thing is clear: We make our biggest moves during the summer months. We pretty much smoked every project last summer. If people can prepare now to have their machines properly cooled for the upcoming summer months, then that is what I'd suggest that people do. All of my machines run 24x7x365.25 if they don't have a blown mobo.

Many large searchers go into "hiding" during the summer months at other projects due to the heat. If we can maintain our load, we'll make a huge move again this summer. I think we gained nearly 100 primes on RPS last summer. They are difficult to compete with in the winter months.

I appreciate you expressing your concerns, Ian.


Gary
gd_barnes is online now  
Old 2009-01-14, 03:56   #182
gd_barnes
 
gd_barnes's Avatar
 
May 2007
Kansas; USA

280616 Posts
Default

Quote:
Originally Posted by gd_barnes View Post
No problem. We can do that. If you want a range for your server, just let me know. Actually, that is a good idea that I should have considered before. Just like we "seamlessly" went from the 1st drive to the 5th drive by loading the 5th drive range before the 1st drive range was done, we can do the same with port 9000. That way, people can leave their cores running and with the multitude of them on port 9000 right now, that's a good thing!

Even if we do that, it'd be really nice if people took some time off from it to run the rally on the n=352K-500K range with us.

Karsten, I know this means that we'll have a little 'too much' running at once for an extended period for your tastes. Likely we won't promote the n=200K-350K portion of the 9th drive too much until we get a little higher in the 8th drive.

I know some people won't agree with our methods at times but we will get all of the primes found in a systematic manner that is agreeable to as many people as we can while also being highly competitive on the top-5000 site. Trying to balance all the tastes and goals is not an easy task.


Gary

Carlos,

On second thought, I think we need to think this through a little more. We do need to be considerate of Karsten and what he does for us as well as potentially spreading our resources too thin. n=200K-350K is far more work than n=50K-200K and as Max said, the former range is something that could wait for a while until we have the 8th drive "stabilized" and running smoothly; perhaps past n=400K.

Let me have some more private discussions with others on work load, commitments, etc.


Gary
gd_barnes is online now  
Old 2009-01-14, 09:15   #183
kar_bon
 
kar_bon's Avatar
 
Mar 2006
Germany

23·3·7·17 Posts
Default

Quote:
Originally Posted by MyDogBuster View Post
I understand that, and the sieving was needed, but we sure as heck don't have to commit 20-30 cores to it (as ReiselSearch starts pulling away again). 2000-3000 new primes later, we will still be 290 primes behind them.
that's what i suggested before:

Drive #9 is for 'small' CPU's, the big ones belongs to the Drives #5-#7!!!
the ranges for many new Top5000 primes!!!

and there's no hassle to complete Drive#9 right now and putting all available cores on it!

but now it's running, fast and cleared up the next few days. so what...
i't everybody's decision where to crunch.


Quote:
Originally Posted by gd_barnes View Post
We do need to be considerate of Karsten and what he does for us as well as potentially spreading our resources too thin. n=200K-350K is far more work than n=50K-200K and as Max said, the former range is something that could wait for a while until we have the 8th drive "stabilized" and running smoothly; perhaps past n=400K.
Drive #9 is quite hard to verify, because of so many results.
yesterday i've checked 3 result-files and displayed all data (now 69 k's out of 498 are complete upto 2009-01-10) and the next days i'll try to complete more.
consider:
- it's my intention to diplay those results for future help, not only hacking in the found primes
- more ranges/servers are quite more work to check
- another result for a new race will be done too
- my ToDo list grows even higher than i can handle: i got data for 2000<k<10000 and 10k<n<20k but no time to include them!

one whish:
keep the amount of drives and servers small as needed, and the checks will be done promptly!
Max is checking all results by his scripts and i do it with mine, so there's another guarantee of completeness and correctness!

Karsten

Last fiddled with by kar_bon on 2009-01-14 at 09:28
kar_bon is offline  
Old 2009-01-14, 10:38   #184
gd_barnes
 
gd_barnes's Avatar
 
May 2007
Kansas; USA

101000000001102 Posts
Default

Kasten, you have a PM. Let me know what you think. Thanks.
gd_barnes is online now  
Old 2009-01-14, 10:44   #185
gd_barnes
 
gd_barnes's Avatar
 
May 2007
Kansas; USA

2·47·109 Posts
Default

Quote:
Originally Posted by MyDogBuster View Post
I understand that, and the sieving was needed, but we sure as heck don't have to commit 20-30 cores to it (as ReiselSearch starts pulling away again). 2000-3000 new primes later, we will still be 290 primes behind them.

BTW, I think I misunderstood your intent here. Personally, I would hope that we would not be committing 20-30 cores towards PRIMALITY TESTING for n=200k-350k at this point. Those cores were only needed for a few weeks to complete sieving of the entire range of n=200k-500k. I would hope that most of the cores would go towards the top-5000 range of n=352k-500k after sieving is done.
gd_barnes is online now  
Old 2009-01-14, 13:51   #186
IronBits
I ♥ BOINC!
 
IronBits's Avatar
 
Oct 2002
Glendale, AZ. (USA)

3·7·53 Posts
Default

I forget which thread is for which drive for which port, so I'll just say it here and ya'll can figure it out.
I have moved all my cores off port 9000 and put them on port 5000. (I think)
IronBits is offline  
Old 2009-01-15, 06:48   #187
gd_barnes
 
gd_barnes's Avatar
 
May 2007
Kansas; USA

2×47×109 Posts
Default

Known primes before this drive started for k=1200-1600 and n=50K-200K have been added to the 1st post of this thread. There are many large k-range gaps to fill!
gd_barnes is online now  
Closed Thread

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Team drive #10 k=1400-2000 n=500K-1M gd_barnes No Prime Left Behind 61 2013-01-30 16:08
Team drive #12 k=2000-3000 n=50K-425K gd_barnes No Prime Left Behind 96 2012-02-19 03:53
Team drive #11 k=2000-3000 n=425K-600K gd_barnes No Prime Left Behind 42 2010-11-19 10:42
Team drive #8 k=1400-2000 n=350K-500K gd_barnes No Prime Left Behind 101 2009-04-08 02:11
Sieving drive for k=1005-2000 n=200K-500K gd_barnes No Prime Left Behind 118 2009-01-17 16:05

All times are UTC. The time now is 01:39.

Thu Dec 3 01:39:02 UTC 2020 up 83 days, 22:50, 1 user, load averages: 1.82, 1.77, 1.75

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.