mersenneforum.org  

Go Back   mersenneforum.org > Prime Search Projects > And now for something completely different

Reply
 
Thread Tools
Old 2021-09-07, 10:43   #12
paulunderwood
 
paulunderwood's Avatar
 
Sep 2002
Database er0rr

3,929 Posts
Default

Quote:
Originally Posted by Puzzle-Peter View Post
I gave it a try and I found an AP-6 with 10593 decimal digits:


N=(2,738,129,976+n*56,497,325)*24499#+1 for n=0..5
Congrats, Peter, for this AP-6. I am looking forward to your submissions to the UTM top20 and JK Andersen's list.

paulunderwood is offline   Reply With Quote
Old 2021-09-23, 22:14   #13
Batalov
 
Batalov's Avatar
 
"Serge"
Mar 2008
Phi(4,2^7658614+1)/2

23×1,201 Posts
Lightbulb

New record AP-3 with a 807,954-digit max element:
6325241166627*2^1290000-1+d*(1455*2^2683953-6325241166627*2^1290000), d=0,1,2
Batalov is offline   Reply With Quote
Old 2021-09-24, 00:50   #14
paulunderwood
 
paulunderwood's Avatar
 
Sep 2002
Database er0rr

3,929 Posts
Default

Quote:
Originally Posted by Batalov View Post
New record AP-3 with a 807,954-digit max element:
6325241166627*2^1290000-1+d*(1455*2^2683953-6325241166627*2^1290000), d=0,1,2
This is huge! Surely it will top the second table for top20 APs, with almost unassailable points. Nice ingenuity!
paulunderwood is offline   Reply With Quote
Old 2021-09-24, 06:43   #15
Batalov
 
Batalov's Avatar
 
"Serge"
Mar 2008
Phi(4,2^7658614+1)/2

23·1,201 Posts
Default

Shoot! That didn't last long...
884748 digits is now the new top.
Batalov is offline   Reply With Quote
Old 2021-09-24, 16:43   #16
Dr Sardonicus
 
Dr Sardonicus's Avatar
 
Feb 2017
Nowhere

2·3·23·37 Posts
Default

Quote:
Originally Posted by Batalov View Post
Shoot! That didn't last long...
884748 digits is now the new top.
Congratulations, looks like you've found a huge 3-term AP of primes!

I'm a bit confused by your usage "new top." [Chekov voice]The top of what?[/Chekov voice]

I'm inferring the meaning "largest top term of any known 3-term AP of prime numbers." Is that correct?
Dr Sardonicus is offline   Reply With Quote
Old 2021-09-24, 17:33   #17
Batalov
 
Batalov's Avatar
 
"Serge"
Mar 2008
Phi(4,2^7658614+1)/2

258816 Posts
Default

Yes. I was too excited to speak in full sentences.
Batalov is offline   Reply With Quote
Old 2021-09-24, 19:30   #18
Dr Sardonicus
 
Dr Sardonicus's Avatar
 
Feb 2017
Nowhere

510610 Posts
Default

Quote:
Originally Posted by Batalov View Post
Yes. I was too excited to speak in full sentences.
Quite understandable! Perhaps you are in need of sedation...

Dr Sardonicus is offline   Reply With Quote
Old 2021-09-25, 05:59   #19
Puzzle-Peter
 
Puzzle-Peter's Avatar
 
Jun 2009

22·173 Posts
Default

Quote:
Originally Posted by Batalov View Post
Shoot! That didn't last long...
884748 digits is now the new top.

WOW
Congratulations!
Puzzle-Peter is offline   Reply With Quote
Old 2021-09-25, 16:48   #20
Dr Sardonicus
 
Dr Sardonicus's Avatar
 
Feb 2017
Nowhere

10011111100102 Posts
Default

Actually, you may have done better than you realize. If

a = 33*2^2939064 - 5606879602425*2^1290000 - 1 and d = 33*2^2939063 - 5606879602425*2^1290000, then

a - d = 33*2^2939063 - 1. Just out of curiosity, I wrote a Pari-GP script to check this number for factors up to 2^28, and there were none.

So, I looked up a table, and what to my wondering eyes should appear...

List of primes k · 2^n − 1 for k < 300
Quote:
k = 33 2, 3, 6, 8, 10, 22, 35, 42, 43, 46, 56, 91, 102, 106, 142, 190, 208, 266, 330, 360, 382, 462, 503, 815, 1038, 1651, 1855, 1858, 1992, 2232, 4462, 4726, 5475, 6702, 9710, 10931, 11503, 20552, 22291, 26575, 35931, 39271, 51326, 57695, 63115, 67182, 109848, 128635, 136971, 173110, 174182, 373446, 585400, 598248, 674050, 773030 [1000000] 1115902 L488, 1813526 L621, 2118570, 2212971, 2215291, 2499883, 2513872, 2939063, 3242126 L3345, 6894190 L4965
So, a + n*d, n = -1 to 3 is a four term AP of primes. Although a - d is only about half as big as a, it still might rank high as a first term of 4-term AP of primes.

I don't know whether a - 2*d = 5606879602425*2^1290000 - 1 is prime or composite.

Last fiddled with by Dr Sardonicus on 2021-09-25 at 16:49 Reason: Formatting
Dr Sardonicus is offline   Reply With Quote
Old 2021-09-25, 18:36   #21
paulunderwood
 
paulunderwood's Avatar
 
Sep 2002
Database er0rr

3,929 Posts
Default

Quote:
Originally Posted by Dr Sardonicus View Post
Actually, you may have done better than you realize. If

......

So, a + n*d, n = -1 to 3 is a four term AP of primes. Although a - d is only about half as big as a, it still might rank high as a first term of 4-term AP of primes.


Should that not be n=-1..2? And...

a=5606879602425*2^1290000-1 and d=33*2^2939063-5606879602425*2^1290000

So a-d is negative

a+d = 33*2^2939063 - 1

Now I can see DrS's mistake and Serge/Ryan's ingenuity.

Last fiddled with by paulunderwood on 2021-09-25 at 19:06
paulunderwood is offline   Reply With Quote
Old 2021-09-25, 22:32   #22
Batalov
 
Batalov's Avatar
 
"Serge"
Mar 2008
Phi(4,2^7658614+1)/2

23×1,201 Posts
Lightbulb

I usually wait for a year or two to pass, then use some old scripts and some new- for those AP-3 weightligting approaches.
Clearly I must give kudos to David Broadhurst - because he invented this trickery and found three hits of this kind; I'd thought of how he did it and then re-engineered all of it once I saw his idea.

If X and Y are some primes (and they have a big enough gcd(X+1,Y+1) ), then we can add 2Y-X, for which we want to have properties: a) can prove primality, b) can devise a relatively fast sieve. Because of a) and the particular choice of seeds, we cannot reach higher than 1.29M / 0.275-ish bits. I am ready to run CHG if we get a very high hit near the very end.

Next technical problem is to sieve deep and reasonably fast. And to satisfy this criterion it really helps to have a large array of Xj that have a huge common part that I can remove (in this case k*21290000-1).
I embrassingly-parallelize (simply separate processes) on Yi. There is an interesting restriction on eligible Yi, left as an thought experiment to the reader. The only input for that mini-puzzle is that in array of Xj is that _all_ of them by prior design are -1 (mod 3).

By having a good array of X[5528] and Y[154], I have >850,000 "very" raw candidates. I would rather want > 10^6 (because of probability of a random odd ~1M digit number being prime is ~1/950,000, if I got it right, did I? *) but that's what we had at the moment. I was prepared for testing all sieved pool and getting no hits; the ME was ~1. Two hits was reasonably lucky and we are not finished yet... maybe we'll get a third.

I call all my special-project sieves "EMsieve", before you ask. This proof-code simply means - there was a sieve and I wrote/tinkered it.
After sieving, the value candidate list was ~51,000 to be run on Ryan's cluster.
___________
* footnote: 2Y-X as constructed and are not only odd (for that reason I already doubled the probability), they are also not divisible by 3 (because both X and Y are -1(mod 3) ), so I can triple the probability (these "very raw" candidates are natively pre-sieved for 2 and 3) and then perhaps it is not so lucky that we have found two hits and is not all that unlikely that we might find third hit.
Batalov is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Exponent Progression storm5510 Information & Answers 15 2017-07-26 18:02
Bertrand's Theorem for Arithmetic Progression literka Math 0 2013-06-01 12:42
sieving primes in arithmetic progressions maxal Software 18 2010-10-04 17:11
nth prime number in an arithmetic progression Unregistered Information & Answers 1 2010-04-04 22:06
Arithmetic and Polynomial Progression of Primes? drake2 Math 13 2006-10-10 00:43

All times are UTC. The time now is 06:35.


Sun Nov 28 06:35:25 UTC 2021 up 128 days, 1:04, 0 users, load averages: 0.59, 0.84, 0.93

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.