Sierp base 3 reservations/statuses/primes

[gd_barnes]

Originally Posted by michaf

Gary,

do you have an excel sheet available with all sierp 3 values that are remaining upto the tested limit?

If so, I can easily manoever them into a shape to remove them from the 'first test to 25k' batch.
(These are all numbers that will be remaining in the end (at 25k), so it might save quite some time)

No, I haven't kept one. All that is available for k's remaining is my web page. It is, in effect, in comma-delimited format. Perhaps you can manipulate it for what you need. It also shows how far the various ranges of k's have been tested.

I'll be reviewing and adding your k=120M-200M range to the page either today or tomorrow.


Gary

[michaf]

Originally Posted by gd_barnes

No, I haven't kept one. All that is available for k's remaining is my web page. It is, in effect, in comma-delimited format. Perhaps you can manipulate it for what you need. It also shows how far the various ranges of k's have been tested.

I'll be reviewing and adding your k=120M-200M range to the page either today or tomorrow.


Gary

I can mold into a usable form... just that if you had it, it would be a tad easier :)

Also, I just thought that a list of primes above say n=1k would be handy for the same purpose too. (The lower one's are found quickly enough...)
Anything like that to your avail?

[gd_barnes]

Originally Posted by michaf

I can mold into a usable form... just that if you had it, it would be a tad easier :)

Also, I just thought that a list of primes above say n=1k would be handy for the same purpose too. (The lower one's are found quickly enough...)
Anything like that to your avail?

Yes for k=10M-30M and no for k<10M. Unfortunately for k=10M-30M, they're on my main sieving laptop that just yesterday has started not wanting to turn on. I'm going to have a friend look at it and see if he can figure out what the problem is. The charger is fine and the battery is OK but it's acting like it's not getting any power. The little blue light is on around where the charger plugs in so I know that's not a problem.

I need to rerun k<10M to get all of the primes. They were on a work laptop that I failed to copy off when I got laid off.

For k=100M-120M, you'll have to check with KEP.

This is the one main base where I've been quite disorganized in bringing all of the primes together because it's so time-consuming with the millions of small primes.


Gary

[KEP]

@ Gary: It's just a minor correction, but for Sierp Base 3 I only took k>110M to k<=120M... well no harm done

@ Michaf: I can see that I've all the primes for Sierpinski Base 3 for k>110M to k<=120M, on my g-mail, I don't remember if I've your e-mail but if you would like me to send you the primes that I way back send to Gary, feel free to PM me and give me your e-mail, and then as fast as possible I'll send them to you :)

Regards

KEP

[gd_barnes]

Originally Posted by michaf

Sierpinski base 3, 120M - 200M is now complete to 25k.

The following 381 k's are left after removal of all reduces values
(Gary, I'll mail you the excel file I used for checking at a glance)

For now, I think I shall chime in on the mini-drive, but I will sure return to the hard work to be done in starting up...

I checked your analysis on the spreadsheet that you sent me. Nice work. There are only a few minor glitches:

k=125825886 would not be considered remaining. Although k=125825886/3^5=517802 has a prime at n=1 and n=2 and k=125825886/3^4=1553406 has a prime at n=1, k=125825886/3^3=4660218 is still remaining. So continuing to test k=125825886 would be a duplication of work, hence it is eliminated.

The same issue exists for the following k's that can be eliminated:
k-value : divisibile by : reduced k : comments about reduced k
138570858 : 3^2 : 15396762 : Reduced k remained when spreadsheet sent. Now prime at n=46233.
138881448 : 3 : 46293816 : Has no small primes so must have a larger prime.*
178029648 : 3 : 59343216 : Reduced k still remains.
179527644 : 3^2 : 19947516 : Reduced k remained when spreadsheet sent. Now prime at n=44420.
182389698 : 3^2 : 20265522 : Reduced k still remains.
188512128 : 3 : 62837376 : Reduced k still remains.
189011016 : 3 : 63003672 : Reduced k still remains.
197830008 : 3 : 65943336 : Reduced k still remains.


You have to check k/3^q for ALL q and see if ANY of those k's are remaining (instead of only checking k/3^q for the HIGHEST possible q). If any of the reduced k's are remaining, your k can be eliminated.

Base 3 is BY FAR the most difficult in this regard. Besides having one of the highest conjectures, it is the 2nd lowest base so we're frequently having to check k/3^q for 5-6 q-values or more and possibly as high as 15-20 q-values in the future for any potential k remaining.

From your list of 381 k-values remaining, this eliminates 9 k's leaving 372 k's for k=120M-200M. I'll update the web pages later today.

* - I do not know what the prime is for k=46293816 but it must be n<=25K because it is not remaining, was not found by the mini drive, and is not on the top-5000 site. Since you did k=30M-100M, do you know what the prime is?


Gary

[michaf]

Hmm... I was annoyed in that I couldn't find that k's prime in my logs, so I retested it to 25k,
and to my astonishment, there is NO prime upto 25k!
(for k = 46293816)

It could have been a false positive prime in the pfgw-phase?

I sure hope it is not;

I think it might be time to get all the primes in order, sorted by k, and recheck every single one of them :(

[gd_barnes]

Originally Posted by michaf

Hmm... I was annoyed in that I couldn't find that k's prime in my logs, so I retested it to 25k,
and to my astonishment, there is NO prime upto 25k!
(for k = 46293816)

It could have been a false positive prime in the pfgw-phase?

I sure hope it is not;

I think it might be time to get all the primes in order, sorted by k, and recheck every single one of them :(

Oh, I'm very bad here. 46293816*3^26776+1 is prime. I found it in my k<50M for n=25K-35K search as shown in the threads. I didn't think to check there. I checked back through the threads and you correctly had it remaining at n=25K. I'm sorry I caused you extra testing time and hassle. Regardless, either way k=15431272 is not remaining.

No need to recheck every one of them. There's no reason to think that you missed any primes. Can you send me all of your primes for k=30M-110M and 120M-200M for n>=1000? It's up to me to keep and organize them all. That shouldn't be too large of a file zipped and I can easily recreate a list of primes for n<1000 for all other k's if and when needed.

KEP, can you send me in an Email all of your primes for n>=1000 for k=110M-120M for Sierp base 3? I need to get an organized list of all primes together sorted by k-value for this base. Not having such a list is creating too much confusion, especially on my part.


Thanks,
Gary

[KEP]

Originally Posted by gd_barnes

Oh, I'm very bad here. 46293816*3^26776+1 is prime. I found it in my k<50M for n=25K-35K search as shown in the threads. I didn't think to check there. I checked back through the threads and you correctly had it remaining at n=25K. I'm sorry I caused you extra testing time and hassle. Regardless, either way k=15431272 is not remaining.

No need to recheck every one of them. There's no reason to think that you missed any primes. Can you send me all of your primes for k=30M-110M and 120M-200M for n>=1000? It's up to me to keep and organize them all. That shouldn't be too large of a file zipped and I can easily recreate a list of primes for n<1000 for all other k's if and when needed.

KEP, can you send me in an Email all of your primes for n>=1000 for k=110M-120M for Sierp base 3? I need to get an organized list of all primes together sorted by k-value for this base. Not having such a list is creating too much confusion, especially on my part.


Thanks,
Gary

Just send you my range Regarding my propersition (suggestion) for a future battleplan for overcoming Sierpinski base 3, would it make more sence to run the initial phase to only n<=1000? I think it will remove a lot more redundant k's and unescesary testing at n<=1000 than it will at n<=2500... also it will speed up the "carpeting" (as i like to call it) a huge deal... sadly it would also, however mean more manual work to be done, but with such a low depth of the carpet, running the entire 125G range can be done in maybe just a few years... and then with the proper sieveing and testing afterwards, n<=25000 could lie just a decade or maybe less into the future

Regards

KEP

[michaf]

I got wise at about 150M...
Since then I tested with pfgw up until n=1k.

It proves to be the 'quickest' way.
It allows for more sieving, which it more efficient then the trial factoring that pfgw does.

But alas, the removal of the 'primed k's ' is a bit more tedious, but easily doable scriptable.

Oh,and I think 10M ranges is more managable, since srfile takes too long to remove a k when there are more then say 200k's remaining.

[gd_barnes]

Originally Posted by KEP

Just send you my range Regarding my propersition (suggestion) for a future battleplan for overcoming Sierpinski base 3, would it make more sence to run the initial phase to only n<=1000? I think it will remove a lot more redundant k's and unescesary testing at n<=1000 than it will at n<=2500... also it will speed up the "carpeting" (as i like to call it) a huge deal... sadly it would also, however mean more manual work to be done, but with such a low depth of the carpet, running the entire 125G range can be done in maybe just a few years... and then with the proper sieveing and testing afterwards, n<=25000 could lie just a decade or maybe less into the future

Regards

KEP

It's not so much the manual effort required as it is that finding billions of small primes would get very boring very fast. That's why we pause to find larger primes at times. It has been my intent to 'pause' every k=50M to do sieving and then searching up to n=100K. But for k>100M, I'm going to suggest doing it every k=100M. This is so that mini-drives (i.e. mini-team efforts) are reasonable in scope and size. It's very difficult to administer if 100's and 1000's of primes are being found. Keep in mind that I have to remove k's from files, update web pages, cross reference k's that are multiples of 3, update a full list of primes for all k's, etc. as the team effort progresses. Starting a team effort at n=10K would take too much time to administer. It's better if one person does it and then sends a list of primes to me. I can then easily sort the list by k-value (if not already done) and we have what we need for future reference.

Reference your search for k<500M on the Riesel side, it looks like you are doing 'pieces' up to n=5K, then n=10K, etc. The problem with that: Can you imagine trying to do a team effort for k<500M starting at n=5K or 10K. We're pretty much stuck waiting for you to finish at least part of your range to n=25K so that we can get a reasonable-sized team effort started.

If I had to make one request of your effort: Please search all k<100M to n=25K and send us the primes and k's remaining, then do k=100M-200M to n=25K, etc. In other words, don't to all k to n=15K, then all k to n=20K, etc. Otherwise, we're going to be waiting 3-4 months for you to finish the entire range. The sieve files become too large for a team effort if we have too large of a k-range at too low of an n-range.


Thanks,
Gary

[gd_barnes]

Originally Posted by michaf

I got wise at about 150M...
Since then I tested with pfgw up until n=1k.

It proves to be the 'quickest' way.
It allows for more sieving, which it more efficient then the trial factoring that pfgw does.

But alas, the removal of the 'primed k's ' is a bit more tedious, but easily doable scriptable.

Oh,and I think 10M ranges is more managable, since srfile takes too long to remove a k when there are more then say 200k's remaining.

As for 10M ranges being more managable, are you referring to searching n<=25K?

If so, I also found 10M ranges to be the most managable. I still prefer to run PFGW to n=10K even though I'm sure running PFGW to n=1K is more efficient on CPU time. Although it takes more total CPU time, I don't have time for a lot of manual intervention. I have one slower dual-core machine in a corner that I just plug in a range and then totally forget it until it is done.


Gary