Doublecheck efforts; S66/S79 to start with

[gd_barnes]

We are starting this thread for all doublecheck efforts. To begin with, as stated in the news thread, S66 and S79 have problems and will need some k's retested. Later on, we will add other bases when potential problems are identified or just to simply confirm past work done.

Attached here are two files that each contain several thousand k's. The S66 k's need to be tested for n=1000 to n=25K and the S79 k's need to be tested for n=1000 to n=50K in order to catch them up with all other k's search depth on the conjectures.

Note that these k's have not been searched for n>1K so this is not a true doublecheck effort. They were newly identified as k's that were erroneously removed as a result of incorrect small composite "primes" that were found by a bad version of PFGW.

Initially I was going to start these myself but won't be able to get to them for 2-3 weeks so I'm opening them up to anyone who has a few spare cores available. Any help would be greatly appreciated to get these large-conjectured bases corrected.

S66

k-range   n-range  reservation  last status
1-700K    1K-25K   firejuggler  (complete)
700K-21M  1K-25K   Lennart      (complete)
21M-end   1K-25K   Mathew       (complete)

S79

k-range   n-range  reservation  last status
1-2.2M    1K-50K   Lennart      (complete)
2.2M-end  1K-50K   Mathew       (complete)

[gd_barnes]

Mathew,

I added n-range to the reservation list in case people want to take a wider range of k to a smaller search depth. I'm assuming that you're searching to the max search depth of the conjectures.


Gary

[Lennart]

I am taking s66
1-21M 1K-25K Lennart

s79
1-2.2M 1k-50k Lennart

[firejuggler]

Arg! Oh well, since Lennart take everything left, I guess i'll stop what I was testing.
I took all k upto 700 000 from N=1k to N=2k:
126 primes found , 90 different sequences.
PFGW Version 3.7.7.64BIT.20130722.Win_Dev [GWNUM 27.11]

[Lennart]

Originally Posted by firejuggler

Arg! Oh well, since Lennart take everything left, I guess i'll stop what I was testing.
I took all k upto 700 000 from N=1k to N=2k:
126 primes found , 90 different sequences.
PFGW Version 3.7.7.64BIT.20130722.Win_Dev [GWNUM 27.11]

Sorry...

I have not started any tests I am only sieving at this point.
I sieve all reserved range on s66 1k-25k in one file.

If you like to continue the search up tp k = 700k I can cut that out to you. Just one question . Are you going to do them to 25K ?

Lennart

[firejuggler]

yup, i'll do them up to n=25k.
there was 224 sequence between 470 and 686081 (last k before 700 000).
Continuing with the 134 sequences left.

[Lennart]

Changing this


Releasing 1-21M 1K-25K Lennart

and taking s66
700k-21M 1K-25K

Lennart

[gd_barnes]

Originally Posted by firejuggler

Arg! Oh well, since Lennart take everything left, I guess i'll stop what I was testing.
I took all k upto 700 000 from N=1k to N=2k:
126 primes found , 90 different sequences.
PFGW Version 3.7.7.64BIT.20130722.Win_Dev [GWNUM 27.11]

Firejuggler,

Please reserve a range before working on it. It creates extra work for other searchers and possibly for yourself.

Second, I appreciate that you removed the higher primes for k's with multiple primes. It saves me admin time. But I would strongly suggest simply stopping the search of a k with a prime to save you CPU time and the task of creating and sending an extra file to me. Here is how to do it: Replace the header in your sieve file with this:

ABC $a*66^$b+1 // {number_primes,$a,1}

With this header, the program will stop searching a k when a prime is found for it. This info. can be found in the file "abcfileformats.txt" in the PFGW download directory. Note that the more recent versions of PFGW can be stopped and restarted and it will still "know" which k's have primes. (It reads the pfgw.log and/or pfgw-prime.log files so those will need to be left there if the program is stopped.) Earlier versions would "forget" which k's had primes if stopped.


Gary

[gd_barnes]

Thank you everone for stepping in to get these bases corrected quickly.

[henryzz]

I suspect a reasonable way to detect whether a base has had a faulty version used on it would be to doublecheck the small ns for the current remaining ks. If any primes are found then that base/range needs redoing.

[gd_barnes]

Originally Posted by henryzz

I suspect a reasonable way to detect whether a base has had a faulty version used on it would be to doublecheck the small ns for the current remaining ks. If any primes are found then that base/range needs redoing.

That's how Phillip found the issue with S66/S79. For the most part, when the issue appears, that would work because the problem is prevalent enough. But if it is a sparse problem on some bases, I would still like to know if there are any differences, even if erroneously removed k's eventually had a larger prime causing there to be no issues found when doing your suggested method. On large conjectured bases, what I'm doing is just testing k<100K to n=100 and doing the compare. That is sufficient enough for my comfort level and it doesn't take too long.