Bad sieve files found for 2020 and 2021

gd_barnes · 2022-09-09, 06:59

Quote:

Originally Posted by VBCurtis

Is this the most unlucky range result yet in this project? A more-than-doubling of N-range tested with just one out of 21 k-values solved seems rather exceptional.

Quote:

Originally Posted by rebirther

R88 tested to n=500k (250-500k)

nothing found, 19 remain

Results emailed - Base released

No. I think this one is now the most unlucky.

Something is beginning to seem very wrong with our results this year. We've hit no primes at all in the n=300k-500k range for ~20-30 bases that have two k's each on our goals list and now there are two very exceptional cases here. That's ~50 k's with no primes in the n=300k-500k range. I feel like some sort of double-check may be in order in the near future. And by double-check, I mean starting from scratch with the sieving process all the way through to testing. Something is off. I'm going to have to think about this.

rebirther · 2022-09-09, 07:31

Quote:

Originally Posted by gd_barnes

No. I think this one is.

Something is beginning to seem very wrong with our results this year. We've hit no primes at all in the n=300k-500k range for ~20-30 bases that have two k's each on our goals list and now there are two very exceptional cases here. That's ~50 k's with no primes in the n=300k-500k range. I feel like some sort of double-check may be in order in the near future. And by double-check, I mean starting from scratch with the sieving process all the way through to testing. Something is off. I'm going to have to think about this.

I hope this does not affect this old bug from srsieve2:

Code:

2.2.2 - October 22, 2021

   srsieve2, srsieve2cl:  version 1.5.3
      Modified to not remove terms that are prime as that defeats the purpose of
      Sierpinski/Riesel searches.

gd_barnes · 2022-09-09, 08:09

Quote:

Originally Posted by rebirther

I hope this does not affect this old bug from srsieve2:

Code:

2.2.2 - October 22, 2021

   srsieve2, srsieve2cl:  version 1.5.3
      Modified to not remove terms that are prime as that defeats the purpose of
      Sierpinski/Riesel searches.

That is extremely worrisome!

The statement in the program docs makes no logical sense. First, why would a sieving program ever remove primes in any circumstance? Second, why would the fact that it is a Riesel-Sierpinski conjecture affect whether they would remove primes from the sieve even if that was sometimes done for who-knows-what reason?

I have stuck with the srsieve/sr1sieve/sr2sieve suite of sieving programs for my sieving needs. If I need to remove n's with algebraic factors, I do so manually. I've never felt comfortable with the newer programs.

rebirther · 2022-09-09, 08:14

Quote:

Originally Posted by gd_barnes

That is extremely worrisome!

The statement in the program docs makes no logical sense. First, why would a sieving program ever remove primes in any circumstance? Second, why would the fact that it is a Riesel-Sierpinski conjecture affect whether they would remove primes from the sieve even if that was sometimes done for who-knows-what reason?

I have stuck with the srsieve/sr1sieve/sr2sieve suite of sieving programs for my sieving needs. If I need to remove n's with algebraic factors, I do so manually. I've never felt comfortable with the newer programs.

Another fact is that the range for R88_160-250k had 3 primes, sievefile was from 2018 and R88_250-500k had no prime with sievefile from 2022, really bad luck or something wrong with the sieve, we could double check 2-3 sievefiles for these 2k 300-500k ranges.

gd_barnes · 2022-09-09, 08:39

Quote:

Originally Posted by rebirther

Another fact is that the range for R88_160-250k had 3 primes, sievefile was from 2018 and R88_250-500k had no prime with sievefile from 2022, really bad luck or something wrong with the sieve, we could double check 2-3 sievefiles for these 2k 300-500k ranges.

It is very unlikely that R88 had no primes for n=250k-500k. So let's do this: I will personally begin sieving it from scratch later on Friday and I'll throw a lot of resources at it to get it done quickly. I will use srsieve followed up by sr2sieve. We won't be using Yoyo for sieving and sr2sieve is not as fast so we won't be sieving as deeply. When I'm done, would you be able to have the SRBase folks run it again with the new file? For the k/n pairs that match up, I also want to do a Residue compare.

Is that something that you could do?

If there are primes in R88, then a sieving program somewhere is likely the issue unless we are getting a large number of residues not matching. Regardless if we find problems, we will likely have to take a second look at almost all work done in 2022 at higher n-ranges.

For the last ~2 months, I have been personally using old reliable srfile to remove factors that Yoyo has provided me with from the original files that I send them. To create the original files, I use old reliable srsieve/sr1sieve/sr2sieve to sieve with. In no circumstance will srfile ever remove k/n pairs from a sieve if the factors that you give it do not divide the associated k/n pair. So I don't feel like it is somehow bad factors from Yoyo or anything like that.

If the double-check finds no primes in R88, then I would also like to do the same thing with R48 for n=200k-500k where there was only 1 prime in 21 k's. 1 prime for the back-to-back testing efforts of R48 and R88 out of 39 k's combined for a doubling of the n-range simply cannot be right. If we only find the one prime on that rerun, then I think it's fair to say that we've just had incredibly bad luck with finding primes in 2022 and I don't feel like we should do any more double-checking.

rebirther · 2022-09-09, 08:57

Quote:

Originally Posted by gd_barnes

It is very unlikely that R88 had no primes for n=250k-500k. So let's do this: I will personally begin sieving it from scratch later on Friday and I'll throw a lot of resources at it to get it done quickly. I will use srsieve followed up by sr2sieve. We won't be using Yoyo for sieving and sr2sieve is not as fast so we won't be sieving as deeply. When I'm done, would you be able to have the SRBase folks run it again with the new file? For the k/n pairs that match up, I also want to do a Residue compare.

Is that something that you could do?

If there are primes in R88, then a sieving program somewhere is likely the issue unless we are getting a large number of residues not matching. Regardless if we find problems, we will likely have to take a second look at almost all work done in 2022 at higher n-ranges.

For the last ~2 months, I have been personally using old reliable srfile to remove factors that Yoyo has provided me with from the original files that I send them. To create the original files, I use old reliable srsieve/sr1sieve/sr2sieve to sieve with. In no circumstance will srfile ever remove k/n pairs from a sieve if the factors that you give it do not divide the associated k/n pair. So I don't feel like it is somehow bad factors from Yoyo or anything like that.

If the double-check finds no primes in R88, then I would also like to do the same thing with R48 for n=200k-500k where there was only 1 prime in 21 k's. 1 prime for the back-to-back testing efforts of R48 and R88 out of 39 k's combined for a doubling of the n-range simply cannot be right. If we only find the one prime on that rerun, then I think it's fair to say that we've just had incredibly bad luck with finding primes in 2022 and I don't feel like we should do any more double-checking.

Sounds possible, I think this year is very bad in luck.

gd_barnes · 2022-09-09, 09:24

I'm starting the re-sieve for R88. After a very fast sieve to 1e9 and running sr2sieve again using that file to have it estimate the number of factors up to 5e14, the sieve depth of our file, it confirms that the number of candidates remaining in the 5e14 file is approximately correct. Of course this could not identify if primes were somehow removed from the file. I will continue.

Luminescence · 2022-09-09, 10:00

Quote:

Originally Posted by gd_barnes

I'm starting the re-sieve for R88. After a very fast sieve to 1e9 and running sr2sieve again using that file to have it estimate the number of factors up to 5e14, the sieve depth of our file, it confirms that the number of candidates remaining in the 5e14 file is approximately correct. Of course this could not identify if primes were somehow removed from the file. I will continue.

I could sieve with the most recent version to compare the output.

Is sr2sieve multithreaded? I could also do some sieving with my big gun.

gd_barnes · 2022-09-09, 10:05

Quote:

Originally Posted by Luminescence

I could sieve with the most recent version to compare the output.

Is sr2sieve multithreaded? I could also do some sieving with my big gun.

No, sr2sieve is not multi-threaded.

Could you sieve R88 for n=250K-500K from scratch using the k's remaining that are shown on the reservations page and using whatever sieving software that you usually use?

gd_barnes · 2022-09-09, 10:10

I just ran R48 for n=200K-500K and R88 for n=250K-500K through my odds of prime spreadsheet for 93877 and 98687 k/n pairs respectively sieved to 5e14. It shows that we should have had 4 primes for R48. We had 1. It shows we should have had 3.5 primes for R88. We had zero.

If it was just these two bases, I would just write it off as bad luck. But on top of finding no primes for ~50 k's for n=300K-500K for our goals bases...well...that's no longer bad luck. Something is wrong.

gd_barnes · 2022-09-09, 10:47

I sieved R88 19 k's for n=250K-500K on a few cores up to P=200e9.

I'm now running 20 threads on my Ryzen 3950X. It will be at 5e12 in ~10 hours.

Florian, can you make a sieve for R88 for the 19 k's for n=250K-500K up to 5e12 ? Optionally you could just sieve it quickly to P=200e9. Either way, we can then compare files.

I'm assuming that you are familiar with sieving. If not, no big deal. This is just an additional check that we can do.

I'll be signing off shortly.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
RSP Sieve Files for k*2^n-1 from PrimeGrid	pinhodecarlos	Riesel Prime Search	103	2022-11-26 15:02
Largest k*2^n-1 Primes Found In 2020	andyhedges	Riesel Prime Search	5	2021-01-02 00:09
I found a sieve to search all pairs of twin primes	Pietro Maiorana	Twin Prime Search	8	2019-09-26 23:07
Algebraic factors in sieve files	pepi37	Conjectures 'R Us	95	2017-07-04 13:37
program to verify factors found by sr(x)sieve?	mdettweiler	Software	16	2009-03-08 02:06

2022-09-09, 09:24	#7
gd_barnes "Gary" May 2007 Overland Park, KS 12046₁₀ Posts	I'm starting the re-sieve for R88. After a very fast sieve to 1e9 and running sr2sieve again using that file to have it estimate the number of factors up to 5e14, the sieve depth of our file, it confirms that the number of candidates remaining in the 5e14 file is approximately correct. Of course this could not identify if primes were somehow removed from the file. I will continue.

2022-09-09, 10:10	#10
gd_barnes "Gary" May 2007 Overland Park, KS 2×19×317 Posts	I just ran R48 for n=200K-500K and R88 for n=250K-500K through my odds of prime spreadsheet for 93877 and 98687 k/n pairs respectively sieved to 5e14. It shows that we should have had 4 primes for R48. We had 1. It shows we should have had 3.5 primes for R88. We had zero. If it was just these two bases, I would just write it off as bad luck. But on top of finding no primes for ~50 k's for n=300K-500K for our goals bases...well...that's no longer bad luck. Something is wrong.

2022-09-09, 10:47	#11
gd_barnes "Gary" May 2007 Overland Park, KS 2·19·317 Posts	I sieved R88 19 k's for n=250K-500K on a few cores up to P=200e9. I'm now running 20 threads on my Ryzen 3950X. It will be at 5e12 in ~10 hours. Florian, can you make a sieve for R88 for the 19 k's for n=250K-500K up to 5e12 ? Optionally you could just sieve it quickly to P=200e9. Either way, we can then compare files. I'm assuming that you are familiar with sieving. If not, no big deal. This is just an additional check that we can do. I'll be signing off shortly.