mersenneforum.org  

Go Back   mersenneforum.org > Prime Search Projects > Conjectures 'R Us

Reply
 
Thread Tools
Old 2016-05-20, 19:38   #804
rogue
 
rogue's Avatar
 
"Mark"
Apr 2003
Between here and the

595310 Posts
Default

Quote:
Originally Posted by pepi37 View Post
KEP
I do test on 13000014446*3^n-1 at 50 K and on 4GHz Intel LLR time is 2.68 seconds
So sieve will be fast because it is easy to sieve to the point where exclusion rate is 3 seconds or more.

If I take for example 1000 candidates on 12 cores then it is easy to give each core 84 K and start LLR .

If this can help your effort I will be glad to help.
Of course in mean time I will stop all my activities on reserved CRUS base ( but I assume nobody will blame me for that)
All is effort :)
So the question will be "how long will that take". Your result would have to be multiplied by 3000 (12 * 250) to know how many CPU core days it would take to get that range to n=50000.
rogue is offline   Reply With Quote
Old 2016-05-20, 19:53   #805
pepi37
 
pepi37's Avatar
 
Dec 2011
After milion nines:)

22×337 Posts
Default

I took first 1000 candidates and make sieve up to 10000000
There is nearly 890000 candidates left
Since test at 50 K is about 2.6 second, average test will be lower ( lets say 1.8 second)
First I got problem to make sr2sieve do work with such long K ( first is 13000014446*3^n-1)
If I overcome this problem that I could sieve much deeper and expect to remove at least 100000 candidates.
So final number will be about 790000 candidates to left
Even if I not found any , so if I test all them it will be done in

790000* 1.8 second= 1422000 seconds /3600/24/12 = is about 1.37 days to complete
pepi37 is offline   Reply With Quote
Old 2016-05-20, 20:11   #806
KEP
Quasi Admin Thing
 
KEP's Avatar
 
May 2005

16538 Posts
Default

Quote:
Originally Posted by pepi37 View Post
I took first 1000 candidates and make sieve up to 10000000
There is nearly 890000 candidates left
Since test at 50 K is about 2.6 second, average test will be lower ( lets say 1.8 second)
First I got problem to make sr2sieve do work with such long K ( first is 13000014446*3^n-1)
If I overcome this problem that I could sieve much deeper and expect to remove at least 100000 candidates.
So final number will be about 790000 candidates to left
Even if I not found any , so if I test all them it will be done in

790000* 1.8 second= 1422000 seconds /3600/24/12 = is about 1.37 days to complete
Pepi, I'm glad that you want to help out. But please start working at k>21G, since R3 k=13G to k=21G is already tested to n=32600 and will continue to test on my machine untill n=100K is reached. Currently n=30K-40K is being tested and afterwards I might run n=40K-50K if further sieving isn't needed. Also Pepi, please remember that sr2sieve cannot be used on these high k's as you may already have read in my PM reply to you
KEP is offline   Reply With Quote
Old 2016-05-20, 21:36   #807
VBCurtis
 
VBCurtis's Avatar
 
"Curtis"
Feb 2005
Riverside, CA

2·3·733 Posts
Default

Quote:
Originally Posted by pepi37 View Post
I took first 1000 candidates and make sieve up to 10000000
There is nearly 890000 candidates left
Since test at 50 K is about 2.6 second, average test will be lower ( lets say 1.8 second)
First I got problem to make sr2sieve do work with such long K ( first is 13000014446*3^n-1)
If I overcome this problem that I could sieve much deeper and expect to remove at least 100000 candidates.
So final number will be about 790000 candidates to left
Even if I not found any , so if I test all them it will be done in

790000* 1.8 second= 1422000 seconds /3600/24/12 = is about 1.37 days to complete
On this sieve you made, did you set n from 25k to 100k, or from 25k to 50k? 25 to 100 is the standard we're using on R3, to reduce the variety of regions/levels tested.

Now that you know 1000 k's takes not long at all, how about tackling a 1G range from 25k to 100k? You don't need your entire 12-thread machine for it, just one core to sieve and then a core or two for LLR. Maybe a month per 1G range on a single core?

As KEP observed, sr2sieve is limited to k's less than 2^32 (~4G), with square-free part of k less than 2^31. This is why I have a region testing on BOINC from 1G to 2.147G; that's 2^31. 0-2.147G will sieve on sr2, higher regions must use srsieve (unless someone is willing to fish out the k's from 2^31 to 2^32 with square-free parts below 2^31!!).
VBCurtis is offline   Reply With Quote
Old 2016-05-21, 04:57   #808
Puzzle-Peter
 
Puzzle-Peter's Avatar
 
Jun 2009

67010 Posts
Default

Quote:
Originally Posted by KEP View Post

I really wish that more people would join the effort and start working the range n=25K-100K on their own computers using LLR.
I'll be joining in about a week or so, I'm pretty clogged up at the moment.
Puzzle-Peter is offline   Reply With Quote
Old 2016-05-21, 11:04   #809
KEP
Quasi Admin Thing
 
KEP's Avatar
 
May 2005

3×313 Posts
Default

Quote:
Originally Posted by VBCurtis View Post
Now that you know 1000 k's takes not long at all, how about tackling a 1G range from 25k to 100k? You don't need your entire 12-thread machine for it, just one core to sieve and then a core or two for LLR. Maybe a month per 1G range on a single core?
I actually also thinks that Gary would prefer a 1G range reservation. A conjecture, big as R3, where all k's remaining cannot be run in a single sieve and reservation, it is preferable to run such a conjecture in a 1G range per core or computer (dependant on the choice of the user).
KEP is offline   Reply With Quote
Old 2016-05-21, 11:05   #810
KEP
Quasi Admin Thing
 
KEP's Avatar
 
May 2005

16538 Posts
Default

Quote:
Originally Posted by Puzzle-Peter View Post
I'll be joining in about a week or so, I'm pretty clogged up at the moment.
Now see that is very cool news
KEP is offline   Reply With Quote
Old 2016-05-29, 06:39   #811
Puzzle-Peter
 
Puzzle-Peter's Avatar
 
Jun 2009

10100111102 Posts
Default

I'm not used to the small stuff any more so let's start with a trial range.

Reserving k=21G to 22G, n=25K to 100K.
Puzzle-Peter is offline   Reply With Quote
Old 2016-06-02, 16:41   #812
Siemelink
 
Siemelink's Avatar
 
Jan 2006
Hungary

1000011002 Posts
Default

As a tribute to this massive effort, I'll take a reservation too.

Reserving k=22G to 23G, n=25K to 100K.

Willem.
--
PS
40000093766*3^36414-1
40000132604*3^92956-1
40000249162*3^49500-1
40000263902*3^67723-1
40000612492*3^58970-1
40000802024*3^26071-1
40000857326*3^63562-1
40000857326*3^87319-1
40001001914*3^195180-1
40001016038*3^118209-1
40001123828*3^83824-1
Siemelink is offline   Reply With Quote
Old 2016-06-08, 15:07   #813
henryzz
Just call me Henry
 
henryzz's Avatar
 
"David"
Sep 2007
Cambridge (GMT/BST)

32·72·13 Posts
Default

Approximately what percent of ks are primed between 25k and 100k?
henryzz is offline   Reply With Quote
Old 2016-06-08, 18:50   #814
KEP
Quasi Admin Thing
 
KEP's Avatar
 
May 2005

3·313 Posts
Default

Quote:
Originally Posted by henryzz View Post
Approximately what percent of ks are primed between 25k and 100k?
~71% (about 46% per double of n)
KEP is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Bases 251-500 reservations/statuses/primes gd_barnes Conjectures 'R Us 2184 2020-10-27 14:30
Bases 101-250 reservations/statuses/primes gd_barnes Conjectures 'R Us 853 2020-10-26 17:25
Bases 33-100 reservations/statuses/primes Siemelink Conjectures 'R Us 1672 2020-10-25 07:39
Bases 4-32 reservations/statuses/primes gd_barnes Conjectures 'R Us 1415 2020-10-24 09:18
Sierp base 3 reservations/statuses/primes gd_barnes Conjectures 'R Us 388 2020-10-21 19:42

All times are UTC. The time now is 11:13.

Wed Oct 28 11:13:38 UTC 2020 up 48 days, 8:24, 2 users, load averages: 2.19, 1.96, 1.83

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.