mersenneforum.org  

Go Back   mersenneforum.org > Prime Search Projects > Prime Sierpinski Project

Reply
 
Thread Tools
Old 2007-10-23, 06:47   #1
hhh
 
hhh's Avatar
 
Jun 2005

17416 Posts
Arrow Which range with which dat (expert discussion)

Thanks Lars. I just recharged my battery to 100% and should have Internet until Friday, but what is going to happen then, I don't know.

As you see, I opened some higher, small ranges for people who can finish something like that in a couple of weeks. I hope that' s OK like this, and I don't need to correct myself.

H.

Last fiddled with by hhh on 2007-10-23 at 06:57
hhh is offline   Reply With Quote
Old 2007-10-23, 23:15   #2
VJS
 
VJS's Avatar
 
Dec 2004

13×23 Posts
Default

hhh,

Not sure if your looking for how far we can sieve with the manual system.

1120000-1125897 should be redone with the combined dat.

for everything else <1500000 we will have to look at whats been done.

There are quite ranges less than 1500T that need work, manual sievers need not worry about running out of work.
VJS is offline   Reply With Quote
Old 2007-10-24, 01:37   #3
hhh
 
hhh's Avatar
 
Jun 2005

22×3×31 Posts
Default

Quote:
Originally Posted by VJS View Post
hhh,

Not sure if your looking for how far we can sieve with the manual system.

1120000-1125897 should be redone with the combined dat.

for everything else <1500000 we will have to look at whats been done.

There are quite ranges less than 1500T that need work, manual sievers need not worry about running out of work.
I'm not sure that I'm getting what you mean.

Brucifer et al. are doing quite some job cleaning ranges that do need only PSP-sieving. From their choice of ranges I deduced that the small gaps inbetween need combined sieving. And because they are small, I wanted to close them fast, so that they do not become confusing when the main manual effort reaches them.

I don't know if it's not even more confusing like this.

H.
hhh is offline   Reply With Quote
Old 2007-10-24, 02:43   #4
Joe O
 
Joe O's Avatar
 
Aug 2002

10158 Posts
Default

It is definitely more confusing this way.
HHH,
You are correct in presuming that the ranges you picked need to be sieved with the combined dat, as should all those currently unreserved below 1000000G. While it may be possible to find ranges above 1000000G and below 1200000G that could be sieved with other than the combined dat, I would recommend the use of the combined dat for the entire range. It would simplify things a lot.
LTD,
This differs from my emails to you. I need to recalibrate my gap analysis program for the range 1000000G-120000G. At a minimum, please use the combined dat for 1000000-1125897.
VJS,
There are some very large gaps in 1000000-1120000, and the factor density seems to lower than expected. While this could be a statistical anomaly, I would like to see this range done with the combined dat.
Joe O is offline   Reply With Quote
Old 2007-10-24, 04:53   #5
Brucifer
 
Brucifer's Avatar
 
Dec 2005

313 Posts
Default

@Joe O, After reading your above, would you rather that the 886000-997000 range was done with the combined dat? I am presently assigning 903000 to my crunchers. If you would rather it is combined, then I can stop the crunching on non-combined at 904000, and then release the reservation, and just go back to start working up the line with the normal combined dat file. ????

Bruce
Brucifer is offline   Reply With Quote
Old 2007-10-24, 07:18   #6
benjackson
 
Aug 2007
Princeton, NJ

2·5 Posts
Default

From how I understand it, it should be fine to go uncombined until at least 1000T in order to catch the PSP sieving up until roughly the point when some people started using the combined 50M dat. The SOB k's were sieved to n=20M up until that point, and they're still under 15M with their LLR testing, so re-sieving those ranges up to 50M might not be worth the extra time considering they could knock some k's off by the time sieving past n=20M is useful, but the extra factors and whatnot are still useful which is why we still generally use the combined dat in those ranges.

The problem is some ranges in the 1000T to 1200T range have all been done by SOB users, but only some used the combined DAT and it's not clear exactly which did. This means that further gap analysis needs to be done to figure out which ranges were sieved for the PSP factors as well so we know which to resieve with the combined dat. But it seems the current goal is to catch our sieving up to 1200T, because once we get there everybody can just use the Combined Automated Reservation System or whatever it's called and get onto the same page.
benjackson is offline   Reply With Quote
Old 2007-10-24, 10:33   #7
Joe O
 
Joe O's Avatar
 
Aug 2002

3×52×7 Posts
Default

Quote:
Originally Posted by Brucifer View Post
@Joe O, After reading your above, would you rather that the 886000-997000 range was done with the combined dat? I am presently assigning 903000 to my crunchers. If you would rather it is combined, then I can stop the crunching on non-combined at 904000, and then release the reservation, and just go back to start working up the line with the normal combined dat file. ????

Bruce
Bruce,
I'm fairly confident that your range can be done with the non-combined dat, especially from 903000-997000. There is one 500G area that you have already passed that I need to look at more closely, but that is all. It is from 1000000G on that I would like to see some caution. For example, 1060500G - 1062000G needs to be done with the combined dat, and maybe 1037000G - 1039000G. The attached graph may be of interest.
Attached Thumbnails
Click image for larger version

Name:	GP_PecanPie_1P-1.1P.png
Views:	116
Size:	15.8 KB
ID:	1986  
Joe O is offline   Reply With Quote
Old 2007-10-24, 12:55   #8
hhh
 
hhh's Avatar
 
Jun 2005

22·3·31 Posts
Default

OK I undid everything and just wait until we reach there.
H.
hhh is offline   Reply With Quote
Old 2007-10-24, 13:43   #9
VJS
 
VJS's Avatar
 
Dec 2004

4538 Posts
Default

Guys,

Sorry for adding to the confusion... I know there are some areas between 1000T-1260T that need to be done with the combined dat for two reasons.

First its possible that SoB factors between 20M<n<50M are missing.
Second I know for certain that some are not done with the combined dat.

We are working those holes.

I think Joe's explaination is perfectly simple, use the combined dat n<50M only for all ranges unless otherwise specified.

Ben,

I understand what you are getting at and we have been on and off this conversation before. The facts always come down to;

The PSP dat is quite a bit larger than the SoB dat, and the PSP-only doesn't have a significant speed advantage over the combined.

There are quite a few people like myself that will use the combined dat even though they know that the PSP only dat is all that is needed.

The worst case senario for using the combined dat is a few % reduced speed.

The consequences of not using the combined dat is having to redo the range with the combined dat anyways.

I personally vote for only using the combined dat from now on, who is with me?

Inorder to push the combined dat I will volunteer myself to do all future ranges that we don't have 100% confidence in with the combined dat.

---------------

On a side note, we may actually eliminate another SoB k from the dat soon. This will mean even a lesser difference between the PSP-only and the combined.
VJS is offline   Reply With Quote
Old 2007-10-24, 15:28   #10
Brucifer
 
Brucifer's Avatar
 
Dec 2005

31310 Posts
Default

Quote:
Originally Posted by Joe O View Post
Bruce,
I'm fairly confident that your range can be done with the non-combined dat, especially from 903000-997000. There is one 500G area that you have already passed that I need to look at more closely, but that is all. It is from 1000000G on that I would like to see some caution. For example, 1060500G - 1062000G needs to be done with the combined dat, and maybe 1037000G - 1039000G. The attached graph may be of interest.
Okay Joe, I will go ahead and finish up to 997000 with the con-comb. Please give the the 500G area that you are questioning, and I will run that for you with the combined dat. If you notice other questionable areas please let me know and I'll re-do those with the combined. :)

@VJS -- There is enough of a speed difference between the non-comb and the combined that it would add probably at least three weeks to the sieve time for me to reach the 997000. So it's easier for me to continue and then just re-do any areas Joe is concerned about if he sees any more other than the 500G one. If he was concerned about more stuff in my reserved range then I'd do the combined. As for everything above the 997000 I intend to just use the combined dat as I agree with your thinking on it.

There were other ranges on the list Joe sent to Lars which Lars in turn gave me (above 997000) that at the time that the reasoning at the time was to run with the non-combined. However in view of Joe's new concerns I'll just do with the combined if I get to them before anyone else. There's a few months work on those. :) I'm estimating a couple months on this one up to 997000 before I can start on another of those ranges.
Brucifer is offline   Reply With Quote
Old 2007-10-24, 15:45   #11
hhh
 
hhh's Avatar
 
Jun 2005

37210 Posts
Default

Quote:
Originally Posted by Brucifer View Post
Okay Joe, I will go ahead and finish up to 997000 with the con-comb. Please give the the 500G area that you are questioning, and I will run that for you with the combined dat. If you notice other questionable areas please let me know and I'll re-do those with the combined. :)
hhh is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Any TI/AM35x/Sitara/WinCE6 expert around? LaurV Lounge 0 2012-12-27 10:00
OK, how can we get a range now? thechickenman Lone Mersenne Hunters 4 2008-12-01 10:45
please help me, cal expert tinhnho Homework Help 11 2007-09-25 12:24
Available range for TF to 2^60? edorajh Lone Mersenne Hunters 2 2003-12-31 16:04
NOVICE INSTRUCTIONS & EXPERT TTn 15k Search 0 2003-06-05 19:13

All times are UTC. The time now is 11:40.

Thu Oct 29 11:40:59 UTC 2020 up 49 days, 8:51, 1 user, load averages: 1.88, 1.76, 1.88

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.