mersenneforum.org  

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

Reply
 
Thread Tools
Old 2021-06-15, 07:55   #639
rebirther
 
rebirther's Avatar
 
Sep 2011
Germany

1011101001102 Posts
Default

S357 sieved to 10G


Results emailed
rebirther is offline   Reply With Quote
Old 2021-08-07, 17:26   #640
MisterBitcoin
 
MisterBitcoin's Avatar
 
"Nuri, the dragon :P"
Jul 2016
Good old Germany

11010001002 Posts
Default

I have sieved the sieve file for S651 deeper, up to P=25e9. It stoped at 6,7 sec/factor using 8 threads. Sieve file can be found here


Now running: S675


Any recommondations / whishes? (n=2,5k-10k)
MisterBitcoin is offline   Reply With Quote
Old 2021-08-07, 18:22   #641
gd_barnes
 
gd_barnes's Avatar
 
May 2007
Kansas; USA

1052510 Posts
Default

Quote:
Originally Posted by MisterBitcoin View Post
I have sieved the sieve file for S651 deeper, up to P=25e9. It stopped at 6.7 sec/factor using 8 threads. Sieve file can be found here

Now running: S675

Any recommendations / wishes? (n=2,5k-10k)
Any base with < 5000 k's remaining currently at n=2.5K and sieve file only at P=1G. Sieving for n=2.5K-10K to P=15G is usually sufficient.

If all bases with < 5000 k's remaining are done you can work your way up from there to < 6000 k's, < 7000 k's, etc.
gd_barnes is online now   Reply With Quote
Old 2021-08-16, 18:52   #642
MisterBitcoin
 
MisterBitcoin's Avatar
 
"Nuri, the dragon :P"
Jul 2016
Good old Germany

83610 Posts
Default

Sieve files for S895, S675 and S717 can be found here


Each of them is far over 6,5 secs/factor and ~16e9.
MisterBitcoin is offline   Reply With Quote
Old 2021-08-17, 14:55   #643
MisterBitcoin
 
MisterBitcoin's Avatar
 
"Nuri, the dragon :P"
Jul 2016
Good old Germany

22·11·19 Posts
Default

I have seen the discussion in the S3 thread, if we require new sieve files for n=25K-100K i can sieve them up to P=1e12 before sending them to yoyo (if needed).
Takes a little time so i can start with checking S66 first; should be much faster with srsieve2 and multithreading now.



Also: Would be n=10K-25K also be well for BOINC, or are the runtimes to low / too much traffic?
MisterBitcoin is offline   Reply With Quote
Old 2021-08-17, 15:48   #644
rogue
 
rogue's Avatar
 
"Mark"
Apr 2003
Between here and the

3·19·113 Posts
Default

Quote:
Originally Posted by MisterBitcoin View Post
I have seen the discussion in the S3 thread, if we require new sieve files for n=25K-100K i can sieve them up to P=1e12 before sending them to yoyo (if needed).
Takes a little time so i can start with checking S66 first; should be much faster with srsieve2 and multithreading now.

Also: Would be n=10K-25K also be well for BOINC, or are the runtimes to low / too much traffic?
It is more efficient to use srsieve2cl than srsieve (if one has a decent GPU), but you are limited by how many k you can sieve at a time. I can do 7000 at a time, but not much more than that. 1e12 is far too deep for n this small.
rogue is offline   Reply With Quote
Old 2021-08-17, 18:51   #645
MisterBitcoin
 
MisterBitcoin's Avatar
 
"Nuri, the dragon :P"
Jul 2016
Good old Germany

22×11×19 Posts
Default

Quote:
Originally Posted by rogue View Post
It is more efficient to use srsieve2cl than srsieve (if one has a decent GPU), but you are limited by how many k you can sieve at a time. I can do 7000 at a time, but not much more than that. 1e12 is far too deep for n this small.

None of my servers have a GPU and my own is already getting quite old. However the speed is okay-ish:


For ~7200 k´s i have a speed of 6K p/sec. Most bases on that range have a lot less, so the speed *should* be higher.

For instance i could start with S270 and sieve it up to 1 factor every x minutes.
MisterBitcoin is offline   Reply With Quote
Old 2021-08-17, 19:48   #646
rogue
 
rogue's Avatar
 
"Mark"
Apr 2003
Between here and the

144518 Posts
Default

Quote:
Originally Posted by MisterBitcoin View Post
None of my servers have a GPU and my own is already getting quite old. However the speed is okay-ish:


For ~7200 k´s i have a speed of 6K p/sec. Most bases on that range have a lot less, so the speed *should* be higher.

For instance i could start with S270 and sieve it up to 1 factor every x minutes.
In that case you might want to compare the speed of srsieve2 to sr2sieve. Sieve to 1e9 first, then compare times of both programs to sieve from 1e9 to 1e10 when using a single thread.
rogue is offline   Reply With Quote
Old 2021-08-17, 20:50   #647
gd_barnes
 
gd_barnes's Avatar
 
May 2007
Kansas; USA

1052510 Posts
Default

Quote:
Originally Posted by MisterBitcoin View Post
I have seen the discussion in the S3 thread, if we require new sieve files for n=25K-100K i can sieve them up to P=1e12 before sending them to yoyo (if needed).
Takes a little time so i can start with checking S66 first; should be much faster with srsieve2 and multi-threading now.

Also: Would be n=10K-25K also be well for BOINC, or are the runtimes to low / too much traffic?
Yoyo cannot sieve S3. The k's are too numerous and the files are too big.

Because base 3 is so prime we generally only sieve max-n/min-n = 2. In this case n=25K-50K. ~40-50% of k's are eliminated with each doubling of the n-range.

Reb has indicated that testing anything n<~90K for base 3 is a problem for BOINC due to the size of the files, the number of k's being tested, and the quickness of the tests. It will be up to individuals to get S3 tested to n=~100K. Even then if we turn part of it over to BOINC it will be need to be a few 1000k at a time...perhaps no more than a k=1G range.

Reb also said that n=25K-100K works for most bases as long as the files are not too big. BOINC did a lot of bases with 100-400 k's remaining for n=25K-100K a few years ago. A base like S66 would be too big for them. Also the tests for n=10K-25K would be too fast.

Last fiddled with by gd_barnes on 2021-08-17 at 20:57
gd_barnes is online now   Reply With Quote
Old 2021-08-17, 21:24   #648
MisterBitcoin
 
MisterBitcoin's Avatar
 
"Nuri, the dragon :P"
Jul 2016
Good old Germany

22×11×19 Posts
Default

Quote:
Yoyo cannot sieve S3. The k's are too numerous and the files are too big.

I am aware of that.


Quote:
Because base 3 is so prime we generally only sieve max-n/min-n = 2. In this case n=25K-50K. ~40-50% of k's are eliminated with each doubling of the n-range.

Reb has indicated that testing anything n<~90K for base 3 is a problem for BOINC due to the size of the files, the number of k's being tested, and the quickness of the tests. It will be up to individuals to get S3 tested to n=~100K. Even then if we turn part of it over to BOINC it will be need to be a few 1000k at a time...perhaps no more than a k=1G range.

Splitting it up into 0,5G ranges would be a hassle, maybe its better to get tested by individuals.






Quote:
Reb also said that n=25K-100K works for most bases as long as the files are not too big. BOINC did a lot of bases with 100-400 k's remaining for n=25K-100K a few years ago. A base like S66 would be too big for them. Also the tests for n=10K-25K would be too fast.

What do you think is the maximum amount of k´s reb could handle; would splitting larger bases into smaller segments work; if its not too much work?

I can do a few n=10K-25K, but it will take a few months / weeks depending how many k´s are remain.
MisterBitcoin is offline   Reply With Quote
Old 2021-08-17, 21:51   #649
gd_barnes
 
gd_barnes's Avatar
 
May 2007
Kansas; USA

1052510 Posts
Default

Quote:
Originally Posted by MisterBitcoin View Post
<snip>
Splitting it up into 0,5G ranges would be a hassle, maybe its better to get tested by individuals.

What do you think is the maximum amount of k´s reb could handle; would splitting larger bases into smaller segments work; if its not too much work?

I can do a few n=10K-25K, but it will take a few months / weeks depending how many k´s are remain.
I feel like n=100K-200K for the base 3's would probably work for BOINC as long as we limit the file size to ~2000-3000 k's. For example there are 1852 k's remaining for R3 k=48G-49G at n=100K after the recent work done by Mark. Such k=1G ranges at n=100K may be something that BOINC could work with going forward. But...see below.

I think n=10K-25K would not work for BOINC for any base. It's just too inefficient to put them into most any server including individual PRPnet servers. There's quite a bit of overhead associated with sending each test out.

Reb's servers do not automatically remove k's as primes are found. He has to remove them himself. This makes work on small n with lots of k's remaining inefficient and more labor intensive.

I think this comes down to general project management and direction. I'm not fond of having BOINC work on small tests like this. They are much better suited for medium and huge tests...namely knocking bases down to 1, 2, or 3 k's remaining or proving them. :-)

Reb, can you weigh in on this again?

Last fiddled with by gd_barnes on 2021-08-17 at 22:06
gd_barnes is online now   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Sieving Drive all base 2/4 k's worked by CRUS gd_barnes Conjectures 'R Us 143 2014-10-21 23:55
Some CRUS stats vmod Conjectures 'R Us 213 2014-02-28 21:23
What are your CRUS plans? rogue Conjectures 'R Us 35 2013-11-09 09:03
CRUS-like sieving challenge CRGreathouse Puzzles 24 2011-10-28 18:30
how high will CRUS go Mini-Geek Conjectures 'R Us 1 2010-11-08 20:50

All times are UTC. The time now is 22:45.


Wed Oct 27 22:45:16 UTC 2021 up 96 days, 17:14, 0 users, load averages: 1.33, 1.06, 0.97

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, 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.