Thoughts about upcoming LLRnet rallies - Page 2

gd_barnes · 2008-04-13, 20:51

Quote:

Originally Posted by Anonymous

May 3rd-4th would work fine for me, though I think we might actually be able to do the rally as originally planned anyway. I'm going to send you a PM about that shortly...

After PM discussion with Anon, it's best that we make the rally May 3rd-4th. Hopefully that will come in after most vacations and before any graduations, etc.

Gary

Mini-Geek · 2008-05-02, 12:04

Quote:

Originally Posted by gd_barnes

After PM discussion with Anon, it's best that we make the rally May 3rd-4th. Hopefully that will come in after most vacations and before any graduations, etc.

Gary

Is the rally still on for May 3rd-4th? There hasn't been any discussion about it in a while...what server is it on, and what time is it again?

mdettweiler · 2008-05-02, 14:53

Quote:

Originally Posted by Mini-Geek

Is the rally still on for May 3rd-4th? There hasn't been any discussion about it in a while...what server is it on, and what time is it again?

I think the plan is to run it on port 5000--as for the time, Gary will have to answer that. (He should have just arrived back home yesterday.)

gd_barnes · 2008-05-02, 20:21

Sorry...I just got back late yesterday and will have everything updated today and tomorrow; primes reported to prime search, results verified, web pages updated, etc.

Yes, I would like to run a rally on May 3rd-4th but let's run it on port 300 instead. I think it's going to be more popular to run it on port 300 because a lot of the primes past n=480K have already been found on drive 3; probably > 60-65% of them. Benson did a lot of searching for n=480K-600K for k=300-400 way back and found many of them.

Also, with my, just now, completion of n=473K-480K on the 3rd drive, we're now in good shape there with all ranges filled in past n=480K. And lastly, on drive 1, there isn't a known prime until n>430K so all primes found will most likely be new for a rally.

Let's do the usual...7 PM GMT Sat. to 7 PM GMT Sun. That would be 2 PM-2PM CDT or 3PM-3PM EDT in the U.S. I'll post a separate and new thread for it.

mdettweiler · 2008-05-06, 05:23

Hi all,

Now that we're all done with our May 3-4 rally, we should probably get started on planning the next one.

First of all, when does everybody think it should be? The weekend of the 17th? The 24th? As far as I know I'm OK for those weekends, though I obviously can't speak for Gary, Karsten, or anybody else.

As for which server to run it on, I know that Gary had expressed that he wanted to do our next rally on one of the servers belonging to our sister project, Conjectures 'R Us. For the uninitiated, the goal of Conjectures 'R Us is to prove all the Riesel and Sierpinski conjectures generalized for bases <32 that aren't being worked on by other projects (and actually, for bases that are powers of 2, we have extended our search range to base 256). Currently the project has two LLRnet servers, both graciously hosted by IronBits--one for Sierpinski base 6 and the other for Sierpinski base 4. (Both are on the server crus.ironbits.net, on ports 6 and 4 respectively.) The Sierpinski base 6 numbers, being a non-power-of-2 base, require that users use the SR5 version of LLRnet to connect to the server (it's the same as regular LLRnet, but with some slight modifications that allow it to work with base other than 2); the Sierpinski base 4 server can be connected to with any LLRnet client, since base 4 is a power of 2 and thus we were able to convert the numbers to base 2 before loading them into the server.

The numbers in the Sierp. base 6 server are a little bigger than the ones here (that's in comparison to the port 5000 NPLB server, where the numbers are a little bigger than the ones on the port 300 server), though not too much so. The Sierp. base 4 numbers, though, are ginormous--close to n=2M in base 2 (n=1M in base 4). Consequently, though, any primes found will be absolutely HUGE--and work has been going pretty slowly on the base 4 server so far, so I'm actually somewhat in favor of doing our rally there.

It would be interesting to test such big numbers in a rally for a change, I think. Though of course Sierp. base 6 would be fun too.

Of course, if everybody would rather, we can still do our next rally here at NPLB--we have plenty of work that needs to be done here in order to make our goal of n=600K for all of k=300-1001 by the end of the year.

Opinions, anyone?

Brucifer · 2008-05-06, 06:46

Quote:

Originally Posted by Anonymous

Of course, if everybody would rather, we can still do our next rally here at NPLB--we have plenty of work that needs to be done here in order to make our goal of n=600K for all of k=300-1001 by the end of the year.

Opinions, anyone?

I would prefer to leave my stuff on NPLB where it's at now. I'm for pushing towards your end of year goal.

Flatlander · 2008-05-06, 11:58

Quote:

Originally Posted by Brucifer

I'm for pushing towards your end of year goal.

Me too.

btw I will not have internet access 10th-23rd May.

mdettweiler · 2008-05-06, 14:35

Okay, it looks like we have a pretty strong opinion so far to stick with the NPLB servers for all our rallies until the goals are met. That is probably a wise idea, considering that on 400<k<1001 we might just squeak our our goal by the end of the year unless we do a LOT more on it (I'm not really worried about 300<k<400, though of course we still need to keep up our current level of work on it).

If this is the general consensus of the project, then the next question is: port 5000 or 300? I was thinking 300 since, as I just mentioned above, 400<k<1001 needs some serious attention in order to make our goal comfortably. However we should, of course, collect a consensus and decide based on that.

Anon

P.S.: Flatlander, even though you won't have internet access from May 10-23, will your machines at home be online and crunching? If so, then you could always just leave them on whatever server we choose for the rally before you leave, though you'd probably want to give them each a HUGE workunit cache (maybe ~100 for a fast machine) in case either your internet or the LLRnet server has a connection hiccup.

Flatlander · 2008-05-06, 16:19

Quote:

Originally Posted by Anonymous

P.S.: Flatlander, even though you won't have internet access from May 10-23, will your machines at home be online and crunching? If so, then you could always just leave them on whatever server we choose for the rally before you leave, though you'd probably want to give them each a HUGE workunit cache (maybe ~100 for a fast machine) in case either your internet or the LLRnet server has a connection hiccup.

I did think of that but 3 or 4 times I have found an LLRNET Client stuck on 99% so I would rather do manual work.

EDIT
Or I could have two clients running on each core. Then if a client locks the other one will get 100% of that core.

Brucifer · 2008-05-06, 16:29

My systems are running (knock on wood) nice and happy on port 300 and I'm content to just leave them there. :)

To reach the year end goal, what percentage of work needs to be done on port 300 versus port 5000? Obviously you need more work on 300 but I'm just trying to get an idea of the proportion of needed work where.

mdettweiler · 2008-05-06, 17:13

Quote:

Originally Posted by Brucifer

My systems are running (knock on wood) nice and happy on port 300 and I'm content to just leave them there. :)

To reach the year end goal, what percentage of work needs to be done on port 300 versus port 5000? Obviously you need more work on 300 but I'm just trying to get an idea of the proportion of needed work where.

Well, port 5000 is at ~490K (rounded up a little bit), and port 300 is at ~426K. That means ~110K remaining for port 5000 to reach the goal, and ~174K remaining for port 300. 110 divided by 174 is ~0.63, so that means there is 63% more work on 400<k<1001 (i.e. port 300) to be done than on 300<k<400 (i.e. port 5000).

2008-05-06, 14:35	#19
mdettweiler A Sunny Moo Aug 2007 USA (GMT-5) 3·2,083 Posts	Okay, it looks like we have a pretty strong opinion so far to stick with the NPLB servers for all our rallies until the goals are met. That is probably a wise idea, considering that on 400<k<1001 we might just squeak our our goal by the end of the year unless we do a LOT more on it (I'm not really worried about 300<k<400, though of course we still need to keep up our current level of work on it). If this is the general consensus of the project, then the next question is: port 5000 or 300? I was thinking 300 since, as I just mentioned above, 400<k<1001 needs some serious attention in order to make our goal comfortably. However we should, of course, collect a consensus and decide based on that. Anon P.S.: Flatlander, even though you won't have internet access from May 10-23, will your machines at home be online and crunching? If so, then you could always just leave them on whatever server we choose for the rally before you leave, though you'd probably want to give them each a HUGE workunit cache (maybe ~100 for a fast machine) in case either your internet or the LLRnet server has a connection hiccup. Last fiddled with by mdettweiler on 2008-05-06 at 14:38 Reason: added P.S.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
At least one upcoming paper on polsel	Dubslow	Msieve	0	2016-03-16 06:24
Upcoming features	Xyzzy	Forum Feedback	1	2007-11-26 18:57
Translations of upcoming press releases	tha	Lounge	6	2005-12-23 13:18
Upcoming Work	R.D. Silverman	NFSNET Discussion	22	2005-11-18 20:25
Upcoming INTEL chips?????	georgekh	Hardware	28	2004-11-20 03:53

2008-05-02, 20:21	#15
gd_barnes May 2007 Kansas; USA 10395₁₀ Posts	Sorry...I just got back late yesterday and will have everything updated today and tomorrow; primes reported to prime search, results verified, web pages updated, etc. Yes, I would like to run a rally on May 3rd-4th but let's run it on port 300 instead. I think it's going to be more popular to run it on port 300 because a lot of the primes past n=480K have already been found on drive 3; probably > 60-65% of them. Benson did a lot of searching for n=480K-600K for k=300-400 way back and found many of them. Also, with my, just now, completion of n=473K-480K on the 3rd drive, we're now in good shape there with all ranges filled in past n=480K. And lastly, on drive 1, there isn't a known prime until n>430K so all primes found will most likely be new for a rally. Let's do the usual...7 PM GMT Sat. to 7 PM GMT Sun. That would be 2 PM-2PM CDT or 3PM-3PM EDT in the U.S. I'll post a separate and new thread for it.

2008-05-06, 05:23	#16
mdettweiler A Sunny Moo Aug 2007 USA (GMT-5) 3×2,083 Posts	Hi all, Now that we're all done with our May 3-4 rally, we should probably get started on planning the next one. First of all, when does everybody think it should be? The weekend of the 17th? The 24th? As far as I know I'm OK for those weekends, though I obviously can't speak for Gary, Karsten, or anybody else. As for which server to run it on, I know that Gary had expressed that he wanted to do our next rally on one of the servers belonging to our sister project, Conjectures 'R Us. For the uninitiated, the goal of Conjectures 'R Us is to prove all the Riesel and Sierpinski conjectures generalized for bases <32 that aren't being worked on by other projects (and actually, for bases that are powers of 2, we have extended our search range to base 256). Currently the project has two LLRnet servers, both graciously hosted by IronBits--one for Sierpinski base 6 and the other for Sierpinski base 4. (Both are on the server crus.ironbits.net, on ports 6 and 4 respectively.) The Sierpinski base 6 numbers, being a non-power-of-2 base, require that users use the SR5 version of LLRnet to connect to the server (it's the same as regular LLRnet, but with some slight modifications that allow it to work with base other than 2); the Sierpinski base 4 server can be connected to with any LLRnet client, since base 4 is a power of 2 and thus we were able to convert the numbers to base 2 before loading them into the server. The numbers in the Sierp. base 6 server are a little bigger than the ones here (that's in comparison to the port 5000 NPLB server, where the numbers are a little bigger than the ones on the port 300 server), though not too much so. The Sierp. base 4 numbers, though, are ginormous--close to n=2M in base 2 (n=1M in base 4). Consequently, though, any primes found will be absolutely HUGE--and work has been going pretty slowly on the base 4 server so far, so I'm actually somewhat in favor of doing our rally there. It would be interesting to test such big numbers in a rally for a change, I think. Though of course Sierp. base 6 would be fun too. Of course, if everybody would rather, we can still do our next rally here at NPLB--we have plenty of work that needs to be done here in order to make our goal of n=600K for all of k=300-1001 by the end of the year. Opinions, anyone?

2008-05-06, 16:29	#21
Brucifer Dec 2005 313₁₀ Posts	My systems are running (knock on wood) nice and happy on port 300 and I'm content to just leave them there. :) To reach the year end goal, what percentage of work needs to be done on port 300 versus port 5000? Obviously you need more work on 300 but I'm just trying to get an idea of the proportion of needed work where.