New data page - Page 6

Cruelty · 2007-08-15, 08:15

Quote:

Originally Posted by gd_barnes

Do you plan to list ALL primes in this forum of any kind for the range that you mentioned like I did for k < 10K and n < 10K?

I'm afraid it will not be possible as right now @ n~1300 the result file is already 10MB big, so I guess I will have to upload it somewhere rather than list here

Quote:

If so, I'd suggest starting at k = 10K since that's where I left off at on my 'regular' prime search that Karsten has now posted.

Now you're telling me, I thought that you are working on it... when I finnish my current range I will take a look at 10001<k<99999.

Quote:

Also, be sure and let Karsten know well ahead of time since it is a huge effort for him to get them all into the correct format for posting on the web pages.

I estimate that it will take me a week to finnish this range

Initially I was planning some preliminary data-mining, but since it already is so big, I don't think excell can handle it and I'm not that good with access or any other db

gd_barnes · 2007-08-15, 19:48

Cruelty,

Sorry about the confusion. We got the ranges tested by me for 'all primes' and 'twin primes' mixed up.

Here is a final clarification about what I have done in the past and what am currently working on:

1. All primes: Not working on currently. I previously completed 1K < k < 10K for n < 10K. (k < 1K was already complete for this range of n.) Eventually I want to take this range of k up a little higher; perhaps to n = 25K, but that will probably be sometime next year.

2. Twin primes: k < 100K complete for n < 10K. Currently working on k < 1M for 10K < n < 15K. That should be done in 3-4 days.

If your effort was to find twin primes over a large range of k by initially finding all of the Riesel's in the range that you want, I should mention that twins are a big challenge to search over multiple k and you found out how big the effort becomes by first finding Riesel's in order to find matching Proth's. Unfortunately, there is no software that I am aware of specifically set up for this. I decided to use NewPGen's twin-prime search and use the increment counter to sieve each n to P=5G, 1000 n at a time. This takes just over 1 min. per n. Note that P=5G, although it sounds low, is probably a little more sieving then is necessary for the range of n I'm working on (1 sec. removal rate vs. about 1/3 sec. LLR rate), but should be about right for 15K < n < 20K.

But the problem with the above is that it sieves each n individually and creates one file for each instead of doing them all at once like srsieve or sr2sieve do for all-prime searches. This makes the optimum sieve depth much lower and so it takes longer to LLR. Regardless, what I end up doing is taking the 1000 files that it has created and copy them into one big file for LLRing. (Fortunately LLR can handle multiple occurrences of the T:0:2:3 line embedded within it.) The sieved file of 1000 n ends up around 10-12 MB so not too big.

Here is what I might suggest to make your effort manageable in size and scope:
1. If you want to do an 'all prime' search, try one of the following:
a. Search 1K < k < 10K for 10K < n < 25K for all primes using srsieve or something similar like I am currently planning on doing sometime next year. (I think that Kosmaj would prefer that we not add a lot more k's for the time being.)
b. Search the range of 10K < k < 25K for n < 10K, which would give us a plenty more k's to work with but wouldn't make your files overwhelmingly large.

2. If you want to do a twin prime search, do a search only in the manner that I suggested above for the range of k < 1M and 15K < n < 20K. This would effectively extend my current effort. If you do that, I'll continue the effort by starting in on 20K < n < 25K after I'm done with 10K < n < 15K. Having a list of all twins for k < 1M and 10K < n < 25K added to my current list for k < 100K and n < 10K would be excellent!

What would really be cool

is to get a list together of ALL twin primes for k < 1M for 10K < n < 75K. (Note that there are too many twins to list for that range of k for n < 10K and that much information wouldn't really be useful now.) They will start getting pretty rare past about n = 25K but there should still be enough to keep the scope of the project reasonable. I suggest that range as a goal for the effort because any twin with n > 66K would be a reportable twin in the top 20 all time (even if 1-2 additional large twins are found between now and then)!! That is my ultimate objective with the effort.

(Note that if we turn this into a twin-only search and discussion, we might want to move our discussion to the Twin Prime Search forum here.)

Gary

Cruelty · 2007-08-15, 20:21

For n<20000 I am using either PFGW or Proth to find such small primes in given ranges of "k" and / or "n". Proth has an added benefit of automatic testing for some special forms, e.g. twins and SGs, however it is slower than PFGW.

em99010pepe · 2007-08-15, 20:31

Look at the tittle of this thread: "Nueva pagina de los datos"

What about in Portuguese?

Carlos

gd_barnes · 2007-08-15, 22:43

Quote:

Originally Posted by em99010pepe

Look at the tittle of this thread: "Nueva pagina de los datos"

What about in Portuguese?

Carlos

A whole bunch of them in other forums here are in Spanish. I guess it's time for me to learn Spanish! Does anyone know what caused that? The next thing we know, they'll show up in French or Portuguese!

Gary

Cruelty · 2007-08-15, 22:47

Looks like someone is really bored here

VBCurtis · 2007-08-16, 01:44

Since Gary suggested extending his k<10k work, I should mention that I am currently testing in the 1500-1600 range up to 100k, and plan to have 1400-1600 complete to at least 50k in the next 30-60 days. Testing to n=50k is fast enough that I may extend this to k=2000. If either of you decide to tackle the search to 25k, please PM me or use this thread to see where I am.

I'll send my primes to karsten when he returns.
-Curtis

gd_barnes · 2007-08-16, 16:56

Quote:

Originally Posted by VBCurtis

Since Gary suggested extending his k<10k work, I should mention that I am currently testing in the 1500-1600 range up to 100k, and plan to have 1400-1600 complete to at least 50k in the next 30-60 days. Testing to n=50k is fast enough that I may extend this to k=2000. If either of you decide to tackle the search to 25k, please PM me or use this thread to see where I am.

I'll send my primes to karsten when he returns.
-Curtis

Excellent!

It's probably best to put everything that we're working on in one place to keep from duping efforts now and in the future so I'll do that below. Note that APS = 'all prime' search and TPS = twin prime search.

Me:
Currently: TPS for k < 1M for 10K < n < 15K. (should be done by Saturday)
3-30 days from now: TPS for k < 1M for 15K < n < 20K.
Mid-next year: APS for 1K < k < 10K for 10K < n < 25K minus whatever has been completed by Curtis, Cruelty, and others at that point.

Curtis:
Currently: APS for 1500 < k < 1600 for 10K < n < 100K.
30-60 days from now: APS for 1400 < k < 1500 for 10K < n < 50K.
Possibly after 60 days from now: APS for 1600 < k < 2000 for 10K < n < 50K.

Cruelty:
Currently: APS for 100K < k < 200K for all n < 20K.
After the above: APS for 10K < k < 100K for all n < 20K.

Already complete:
APS for k < 10K for n < 10K.
TPS for k < 100K for n < 10K.

Cruelty, can you confirm that I have your planned work correct? After you realized the size of the effort, I wasn't quite clear if you were still planning on doing the entire thing. Also, note that my TPS for k < 1M will probably be up to n = 20K within the next 30 days so that technically is an overlap of any twins that you find for 10K < k < 200K up to n = 20K.

One thing more...As you probably know, Kosmaj has requested that we put any found primes in the format of X, XX, XX, XX, XXX, etc. for large #'s of submitted primes for easy posting on the web sites. Without writing a program to do so, this is tricky at best. Notepad does not have a character for end-of-line (that I'm aware of) so the 'change all' command doesn't work. I end up using the 'concatenate' command in Excel after cutting-and-pasting them into a spreadsheet. This is fast and works fine for a few hundred primes on 2-3 k's but would not work at all for for several thousand primes on multiple k's. Because it takes some time to get it formatted in that manner for large #'s of primes, we might want to ask Karsten if he still has the automated program that I'm sure he created to format the huge # of primes that I submitted for 1K < k < 10K. If he does, it may be OK to just forward the prime files as they come from LLR or whatever. Cruelty, since you would be adding many new k's, in order to help Karsten out, I might also suggest forwarding a file of Nash weights to him for the new k's.

Gary

gd_barnes · 2007-08-16, 19:00

I found the following corrections that are needed on the latest update to the www.15k.org sites:

1. k=19437 and 115029915 should show reserved by me. They were previously shown as reserved by me so I'm not sure what happened. I haven't unreserved any of my high-weight k's yet. They are all in my large sieve for testing up to n = 400K.

2. k=3428677395 should show as tested thru n=200K (vs. n=175K). The correct message is linked there so I think the update of the testing range just got missed.

3. k=151023 should show as tested thru n=160K. It is currently blank.

4. k=65581096558109 should show as k=6558109. Note that it is in the correct sequence for k=6558109. (I thought this repeated value was kind of funny. It'd be interesting to actually test k=65581096558109.)

5. The range tested on k=735 is incorrect. I'm not sure what it should be. But there are several primes shown that are outside of the testing range and then there's a large part of the testing range that has no primes. So something seems off there.

6. On the main www.15k.org < 300 page, the quick link for 10^5 * 1 doesn't work. (i.e. where you have links with '#' signs that allow you to quickly go to various values of k) I tested all of the rest of the links and they all worked correctly.

Gary

Cruelty · 2007-08-16, 19:50

So far I confirm 100001<k<199999 @ n<20000. Afterwards I'll look into 10001<k<99999 - I'm not sure if I will do it immediately after finnishing current work...

em99010pepe · 2007-08-20, 10:40

"Gatos" is cats in portuguese and in spanish......nothing to do with "new data page". Keep trying....

2007-08-15, 19:48	#57
gd_barnes May 2007 Kansas; USA 2²·19·137 Posts	My testing clarification and a suggestion Cruelty, Sorry about the confusion. We got the ranges tested by me for 'all primes' and 'twin primes' mixed up. Here is a final clarification about what I have done in the past and what am currently working on: 1. All primes: Not working on currently. I previously completed 1K < k < 10K for n < 10K. (k < 1K was already complete for this range of n.) Eventually I want to take this range of k up a little higher; perhaps to n = 25K, but that will probably be sometime next year. 2. Twin primes: k < 100K complete for n < 10K. Currently working on k < 1M for 10K < n < 15K. That should be done in 3-4 days. If your effort was to find twin primes over a large range of k by initially finding all of the Riesel's in the range that you want, I should mention that twins are a big challenge to search over multiple k and you found out how big the effort becomes by first finding Riesel's in order to find matching Proth's. Unfortunately, there is no software that I am aware of specifically set up for this. I decided to use NewPGen's twin-prime search and use the increment counter to sieve each n to P=5G, 1000 n at a time. This takes just over 1 min. per n. Note that P=5G, although it sounds low, is probably a little more sieving then is necessary for the range of n I'm working on (1 sec. removal rate vs. about 1/3 sec. LLR rate), but should be about right for 15K < n < 20K. But the problem with the above is that it sieves each n individually and creates one file for each instead of doing them all at once like srsieve or sr2sieve do for all-prime searches. This makes the optimum sieve depth much lower and so it takes longer to LLR. Regardless, what I end up doing is taking the 1000 files that it has created and copy them into one big file for LLRing. (Fortunately LLR can handle multiple occurrences of the T:0:2:3 line embedded within it.) The sieved file of 1000 n ends up around 10-12 MB so not too big. Here is what I might suggest to make your effort manageable in size and scope: 1. If you want to do an 'all prime' search, try one of the following: a. Search 1K < k < 10K for 10K < n < 25K for all primes using srsieve or something similar like I am currently planning on doing sometime next year. (I think that Kosmaj would prefer that we not add a lot more k's for the time being.) b. Search the range of 10K < k < 25K for n < 10K, which would give us a plenty more k's to work with but wouldn't make your files overwhelmingly large. 2. If you want to do a twin prime search, do a search only in the manner that I suggested above for the range of k < 1M and 15K < n < 20K. This would effectively extend my current effort. If you do that, I'll continue the effort by starting in on 20K < n < 25K after I'm done with 10K < n < 15K. Having a list of all twins for k < 1M and 10K < n < 25K added to my current list for k < 100K and n < 10K would be excellent! What would really be cool is to get a list together of ALL twin primes for k < 1M for 10K < n < 75K. (Note that there are too many twins to list for that range of k for n < 10K and that much information wouldn't really be useful now.) They will start getting pretty rare past about n = 25K but there should still be enough to keep the scope of the project reasonable. I suggest that range as a goal for the effort because any twin with n > 66K would be a reportable twin in the top 20 all time (even if 1-2 additional large twins are found between now and then)!! That is my ultimate objective with the effort. (Note that if we turn this into a twin-only search and discussion, we might want to move our discussion to the Twin Prime Search forum here.) Gary Last fiddled with by gd_barnes on 2007-08-15 at 19:50

2007-08-16, 19:00	#64
gd_barnes May 2007 Kansas; USA 2²·19·137 Posts	Corrections on latest update I found the following corrections that are needed on the latest update to the www.15k.org sites: 1. k=19437 and 115029915 should show reserved by me. They were previously shown as reserved by me so I'm not sure what happened. I haven't unreserved any of my high-weight k's yet. They are all in my large sieve for testing up to n = 400K. 2. k=3428677395 should show as tested thru n=200K (vs. n=175K). The correct message is linked there so I think the update of the testing range just got missed. 3. k=151023 should show as tested thru n=160K. It is currently blank. 4. k=65581096558109 should show as k=6558109. Note that it is in the correct sequence for k=6558109. (I thought this repeated value was kind of funny. It'd be interesting to actually test k=65581096558109.) 5. The range tested on k=735 is incorrect. I'm not sure what it should be. But there are several primes shown that are outside of the testing range and then there's a large part of the testing range that has no primes. So something seems off there. 6. On the main www.15k.org < 300 page, the quick link for 10^5 * 1 doesn't work. (i.e. where you have links with '#' signs that allow you to quickly go to various values of k) I tested all of the rest of the links and they all worked correctly. Gary

2007-08-20, 10:40	#66
em99010pepe Sep 2004 101100001110₂ Posts	"Gatos" is cats in portuguese and in spanish......nothing to do with "new data page". Keep trying.... Last fiddled with by em99010pepe on 2007-08-20 at 10:41

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Erroneous data	ATH	Data	8	2013-11-13 19:21
Corrupted data?	Oops!	Information & Answers	2	2013-10-22 03:48
GPU TF vs DC/LL data	bcp19	GPU to 72	0	2011-12-02 16:41
Data available?	Prime95	LMH > 100M	10	2007-06-22 23:55
Conflicting data?	ATH	Data	4	2006-02-27 13:53

2007-08-15, 20:21	#58
Cruelty May 2005 11001011000₂ Posts	For n<20000 I am using either PFGW or Proth to find such small primes in given ranges of "k" and / or "n". Proth has an added benefit of automatic testing for some special forms, e.g. twins and SGs, however it is slower than PFGW.

2007-08-15, 20:31	#59
em99010pepe Sep 2004 2·5·283 Posts	Look at the tittle of this thread: "Nueva pagina de los datos" What about in Portuguese? Carlos

2007-08-15, 22:47	#61
Cruelty May 2005 2³·7·29 Posts	Looks like someone is really bored here

2007-08-16, 01:44	#62
VBCurtis "Curtis" Feb 2005 Riverside, CA 4868₁₀ Posts	Since Gary suggested extending his k<10k work, I should mention that I am currently testing in the 1500-1600 range up to 100k, and plan to have 1400-1600 complete to at least 50k in the next 30-60 days. Testing to n=50k is fast enough that I may extend this to k=2000. If either of you decide to tackle the search to 25k, please PM me or use this thread to see where I am. I'll send my primes to karsten when he returns. -Curtis

2007-08-16, 19:50	#65
Cruelty May 2005 2³·7·29 Posts	So far I confirm 100001<k<199999 @ n<20000. Afterwards I'll look into 10001<k<99999 - I'm not sure if I will do it immediately after finnishing current work...