mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Conjectures 'R Us (https://www.mersenneforum.org/forumdisplay.php?f=81)
-   -   Riesel base 3 reservations/statuses/primes (https://www.mersenneforum.org/showthread.php?t=11151)

Puzzle-Peter 2011-09-04 06:11

Results for k=1200M to 1320M have been emailed to Gary.

Puzzle-Peter 2011-09-05 19:25

Sent results for k=1320k-1360k to Gary.

Puzzle-Peter 2011-09-07 19:05

Sent results for k=1360M-1400M to Gary.

Puzzle-Peter 2011-09-10 06:49

Reserving k=1500M-1600M to n=25k

Puzzle-Peter 2011-09-11 17:21

Sent results for k=1400M-1500M to Gary.

Puzzle-Peter 2011-09-20 08:13

Sent results for k=1500M-1600M to Gary.

gd_barnes 2011-09-21 07:18

[QUOTE=henryzz;272185]Could I suggest that you post reservations when you start a base? Doing that provides another oportunity for this sort of thing to be caught.
Anyone know where Max's instructions for removing a base from a prpnet server are?[/QUOTE]

I usually do but somehow overlooked that one. Sorry again.

I'd be willing to sieve a couple of bases for you. I'm out of town until Tues. Sept. 27th and could start on them then.

henryzz 2011-09-21 11:07

[QUOTE=gd_barnes;272236]I usually do but somehow overlooked that one. Sorry again.

I'd be willing to sieve a couple of bases for you. I'm out of town until Tues. Sept. 27th and could start on them then.[/QUOTE]
Thanks for the offer. I am currently sieving my next base 3 range k=10M-20M to 500k(aiming for optimal for testing upto 250k currentlly). Could you sieve 1T-2.5T about 150 hours of work on one core of a Q6600? If this is a bit much then less would be fine. I think I used a similar amount of time on S379 although it is hard to judge.

gd_barnes 2011-09-21 16:55

[QUOTE=henryzz;272250]Thanks for the offer. I am currently sieving my next base 3 range k=10M-20M to 500k(aiming for optimal for testing upto 250k currentlly). Could you sieve 1T-2.5T about 150 hours of work on one core of a Q6600? If this is a bit much then less would be fine. I think I used a similar amount of time on S379 although it is hard to judge.[/QUOTE]

No problem. I'll start on it early on the 28th.

I use all 8 cores of a single I7 running at 2.9 Ghz for sieving. It's the equivalent of about 5-6 cores running full speed. If it's 150 CPU hours, it would take me ~25-30 hours. Just send me the file sieved to 1T when it's ready and I'll take care of it.

henryzz 2011-09-25 12:45

1 Attachment(s)
[QUOTE=gd_barnes;272279]No problem. I'll start on it early on the 28th.

I use all 8 cores of a single I7 running at 2.9 Ghz for sieving. It's the equivalent of about 5-6 cores running full speed. If it's 150 CPU hours, it would take me ~25-30 hours. Just send me the file sieved to 1T when it's ready and I'll take care of it.[/QUOTE]
The sieve file is attached.

gd_barnes 2011-09-28 03:05

[QUOTE=henryzz;272690]The sieve file is attached.[/QUOTE]

OK, it is started. Estimated completion is ~24 hours from this posting.

You do not have a sieve depth in the file but I assume that it is sieved to 1T. Is that correct?

henryzz 2011-09-28 09:35

[QUOTE=gd_barnes;272924]OK, it is started. Estimated completion is ~24 hours from this posting.

You do not have a sieve depth in the file but I assume that it is sieved to 1T. Is that correct?[/QUOTE]
Yes that is correct.

henryzz 2011-09-28 09:40

[QUOTE=henryzz;269222]Tested to 400k. No more primes.:no:[/QUOTE]
k=5-10M is complete upto n=480k
reserving k=10M-20M to n=500k Gary is currently finishing the sieving optimal to 250k.

gd_barnes 2011-09-29 04:23

1 Attachment(s)
[QUOTE=henryzz;272936]Yes that is correct.[/QUOTE]

Attached is the file sieved to 2.5T.

henryzz 2011-09-29 21:52

[QUOTE=gd_barnes;272969]Attached is the file sieved to 2.5T.[/QUOTE]
Thanks. I will start testing in a week or 2 when my other base 3 reservation is finished.

Puzzle-Peter 2011-09-30 13:19

Reserving k=1.6G to 1.8G to n=25k

Puzzle-Peter 2011-10-05 14:35

Reserving k=1.8G to 2.0G to n=25k

Puzzle-Peter 2011-10-07 18:43

Results for k=1.6G to 1.8G are being sent to Gary at this moment

Puzzle-Peter 2011-10-12 15:36

Results for k=1.8G to 2.0G are being sent to Gary at this moment

Time to take a break from R3.

henryzz 2011-11-02 23:18

[QUOTE=henryzz;273002]Thanks. I will start testing in a week or 2 when my other base 3 reservation is finished.[/QUOTE]
[code][2011-10-10 16:23:50 GMT] 10886318*3^240181-1: Prime returned by david1@mdds.me.uk using pfgw.exe (Prover: pfgw.exe)!
[2011-10-12 17:47:38 GMT] 11005406*3^234173-1: Prime returned by david1@mdds.me.uk using pfgw.exe (Prover: pfgw.exe)!
[2011-10-12 18:56:53 GMT] 11543008*3^114992-1: Prime returned by david1@mdds.me.uk using pfgw.exe (Prover: pfgw.exe)!
[2011-10-12 20:11:51 GMT] 11926424*3^119299-1: Prime returned by david1@mdds.me.uk using pfgw.exe (Prover: pfgw.exe)!
[2011-10-12 21:25:48 GMT] 12518444*3^111100-1: Prime returned by david1@mdds.me.uk using pfgw.exe (Prover: pfgw.exe)!
[2011-10-27 10:31:52 GMT] 14523304*3^249532-1: Prime returned by david1@mdds.me.uk using pfgw.exe (Prover: pfgw.exe)!
[2011-10-27 14:07:09 GMT] 16571998*3^151721-1: Prime returned by david1@mdds.me.uk using pfgw.exe (Prover: pfgw.exe)!
[2011-10-31 08:45:16 GMT] 18232358*3^155836-1: Prime returned by david1@mdds.me.uk using pfgw.exe (Prover: pfgw.exe)!
[2011-11-01 16:44:05 GMT] 19379356*3^123081-1: Prime returned by david1@mdds.me.uk using pfgw.exe (Prover: pfgw.exe)!
[2011-11-02 06:17:26 GMT] 19885204*3^207385-1: Prime returned by david1@mdds.me.uk using pfgw.exe (Prover: pfgw.exe)![/code]
k=10M-20M is tested to 250k. Commencing sieving optimal for 500k.

gd_barnes 2011-11-02 23:36

[QUOTE=henryzz;276894]
k=10M-20M is tested to 250k. Commencing sieving optimal for 500k.[/QUOTE]

OK, thanks.

You last reported in an Email that k=4M-10M was at n=480K and that you were searching it to n=500K. Did you finish that range?

henryzz 2011-11-03 07:14

[QUOTE=gd_barnes;276898]OK, thanks.

You last reported in an Email that k=4M-10M was at n=480K and that you were searching it to n=500K. Did you finish that range?[/QUOTE]
Yes I did. I thought I had sent the results. I will gather together some results files for you and email them.
edit: Sent.

henryzz 2011-11-27 22:48

12850412*3^287506-1 is prime

Pretty certain this is my biggest non-base 2 prime so far.

paleseptember 2011-11-28 00:09

[QUOTE=henryzz;280145]12850412*3^287506-1 is prime

Pretty certain this is my biggest non-base 2 prime so far.[/QUOTE]

Congratulations! :toot::toot:

henryzz 2011-12-08 10:25

Didn't take me long to beat it.
14028218*3^342326-1 is prime
k=10-15M is tested to 350k
k=15-20M is tested to 300k

henryzz 2012-02-11 14:50

[QUOTE=henryzz;281494]Didn't take me long to beat it.
14028218*3^342326-1 is prime
k=10-15M is tested to 350k
k=15-20M is tested to 300k[/QUOTE]
10766804*3^408240-1 is prime
k=10-15M is tested to 425k
k=16874152 is tested to 412k
k=17-20M are still at 300k
Based on 5-10M I should expect 1.67 more primes from this range. Of course this range is larger than k=5-10M so that doesn't mean much:smile:

henryzz 2012-02-13 15:35

[QUOTE=henryzz;289042]10766804*3^408240-1 is prime
k=10-15M is tested to 425k
k=16874152 is tested to 412k
k=17-20M are still at 300k
Based on 5-10M I should expect 1.67 more primes from this range. Of course this range is larger than k=5-10M so that doesn't mean much:smile:[/QUOTE]
17791294*3^325700-1 is prime
k=16874152 is tested to 425k

henryzz 2012-05-05 23:10

Just been applying some probability to my base 3 work.
From 250k-500k, 12ks remaining at 250k and 48389 candidates sieved to 6T. The columns marked crus include stopping the search after a prime has been found for a k. This calculation relies on the ks all having the same weight but that shouldn't make too much difference.
[CODE]n n primes at least n primes n primes crus at least n primes crus
0 0.212% 100% 0.212% 100%
1 1.307% 99.788% 1.707% 99.788%
2 4.021% 98.481% 6.292% 98.081%
3 8.251% 94.460% 14.056% 91.788%
4 12.696% 86.209% 21.195% 77.732%
5 15.629% 73.512% 22.726% 56.538%
6 16.032% 57.884% 17.768% 33.812%
7 14.096% 41.852% 10.206% 16.044%
8 10.844% 27.756% 4.275% 5.837%
9 7.415% 16.912% 1.273% 1.562%
10 4.564% 9.497% 0.256% 0.289%
11 2.553% 4.933% 0.031% 0.033%
12 1.309% 2.380% 0.002% 0.002%[/CODE]I found it surprising how different the probabilities are when crus style searching is included. Also another thing worth looking at is the cumulative probabilities. For the non-crus values 6 is the most likely outcome but we often overlook that there is 41% chance of there being more that 6 primes. I am pretty certain that if at least 6 primes is reached then there is 73% chance of there being more than 6 primes.
This functionally should reach the odds of prime spreadsheet in the near future.

Returning back to this search I have 4 primes so far and the most likely outcome is 5 primes using crus style searching rather that the 6 primes the standard method suggests so I expect 1 more prime. The expected number of primes is 4.815 with crus style searching and 6.155 with standard searching.

henryzz 2012-05-26 15:08

ks 10M-20M are all at 425k continuing

VBCurtis 2012-05-28 07:29

Taking 2000M to 2050M.

grueny 2012-07-30 17:17

reserving k=250e6-500e6 to n=100e3

VBCurtis 2012-08-27 01:19

[QUOTE=VBCurtis;300487]Taking 2000M to 2050M.[/QUOTE]

complete to n=100k. I'll gather the data together and email primes to Gary this week.
-Curtis

VBCurtis 2012-08-30 03:54

Gary has received the data, and informs me there were 682 k's left after testing to 10k, 73 after testing to 100k.

I am embarking on a sieve of the entire region 500M to 2000M from 25k to 100k. I certainly do not plan to test this range in its entirety, but I expect to have enough sieving complete by 1 Dec to share files with anyone interested in doing some primality testing for base 3.
-Curtis

VBCurtis 2012-09-22 17:53

The sieve is at 10G, roughly sufficient to test the lowest N values.
I am testing 25000 to 26000 now, expecting over 200 primes in that range.
26000 to 27000 should be ready 1 Oct when the sieve reaches 12G, if anyone is interested.
-Curtis

gd_barnes 2012-10-11 04:38

[QUOTE=VBCurtis;312444]The sieve is at 10G, roughly sufficient to test the lowest N values.
I am testing 25000 to 26000 now, expecting over 200 primes in that range.
26000 to 27000 should be ready 1 Oct when the sieve reaches 12G, if anyone is interested.
-Curtis[/QUOTE]

k=500M-2G is complete to n=26K. 276 k's were found prime for n=25K-26K. 7529 k's remain for the k-range.

I am showing k=500M-2G as reserved by you for testing to n=27K. Let me know if you plan differently.

VBCurtis 2012-10-17 02:45

k=500M to 2G complete to 27k. Primes sent to Gary.

Reserving 27k to 28k.

The sieve is complete to 16G, sufficient for tests to at least 30k if anyone wants to run a 1k block. Roughly 7250 k's remain in 500M to 2G.
-Curtis

gd_barnes 2012-10-17 04:09

[QUOTE=VBCurtis;314933]k=500M to 2G complete to 27k. Primes sent to Gary.

Reserving 27k to 28k.

The sieve is complete to 16G, sufficient for tests to at least 30k if anyone wants to run a 1k block. Roughly 7250 k's remain in 500M to 2G.
-Curtis[/QUOTE]

265 k's were found prime for n=26K-27K. 7264 k's remain for the k-range.

I do not recommend splitting up n-ranges on such a huge base. For the primes that you find for n=27K-28K, someone else would end up testing those same k's for n=28K-29K. If the range is to be split up, it's better to split by k's.

VBCurtis 2012-10-18 05:27

I suppose this depends on how much work someone wishes to take on. For a 1k block at a time, roughly 3% of the CPU time is wasted in overlap with redundant k's. On the other hand, CPU time is vastly cheaper than human-administrative time in splitting ranges by k, tracking which blocks have been split or not split, etc.

For a user taking a flier on a single 1k block, simplicity wins. If someone wants to do a CPU-month or more, I can split files by k while maintaining a single sieve. I also don't mind pausing my LLR work while someone else does a 1k block- this removes most inefficiency of the sort you observe.

Anyway, I welcome teamwork on this endeavor.

gd_barnes 2012-10-22 20:35

David (henryzz) has completed k=10M-20M to n=425K; 4 primes were found for n=250K-425K previously reported; the base is released.

grueny 2012-10-24 07:55

r3
 
k = 250e6 - 500e6 complete to n=100e3.

859 primes.

results here: [URL]http://primes.ctrl-x-c.de/[/URL]

base released

VBCurtis 2012-11-03 05:30

Reserving 2050M to 2100M.

27k to 28k complete on 500M-2000M. Reserving 28k to 29k.
-Curtis

VBCurtis 2012-11-29 00:33

2050 to 2100M complete to 15k. Reserving 2100M to 2200M.

On the 500M to 2000M sieve, 28k to 29k is complete. Reserving 29k to 30k.

-Curtis

VBCurtis 2012-12-22 05:32

29k to 30k complete on k=500M to 2G.
Pausing for now in favor of getting PFGW script run to 2.5G.
-Curtis

gd_barnes 2012-12-27 02:35

[QUOTE=VBCurtis;322317]29k to 30k complete on k=500M to 2G.
Pausing for now in favor of getting PFGW script run to 2.5G.
-Curtis[/QUOTE]

Have you sent me the primes for n=29K-30K yet? I haven't received them.

VBCurtis 2012-12-27 04:11

Whoops! Primes have now been sent.

While I have paused my LLR work at 30k on the 500-2000M range, I have sieved to 20G on 30-100k, and welcome anyone to request files to do some testing on this range while I push our explored range up to 2500M.

Reserving 2050-2500M up to 30k.
-Curtis

gd_barnes 2012-12-31 23:23

Lennart has completed k=500M-2G for n=30K-35K. 851 primes were found.

gd_barnes 2013-01-02 21:35

Lennart has completed k=500M-2G for n=35K-40K. 655 primes were found.

gd_barnes 2013-01-05 00:55

Lennart has completed k=500M-2G for n=40K-50K. 944 primes were found.

gd_barnes 2013-01-16 21:33

Lennart has completed k=500M-2G for n=50K-100K. 1833 primes were found; 2345 k's remain for the k-range. A tremendous work by Lennart! :bow:

Here are some stats and extrapolations: For this k-range, 7805 k's were remaining at n=25K and 5460 k's were found prime for n=25K-100K so almost exactly 70% of k's were eliminated.

Based on a reduction of k's remaining of 70% for every 4X increase in the n-value, here are some estimates of k's remaining for the entire base:
n=25K; 325,000
n=100K; 97,500
n=400K; 29,250
n=1.6M; 8,775
n=6.4M; 2,633
n=25.6M; 790
n=102.4M; 237

Of course, the k's become progressively lower average weight as the n-range gets higher so this moderately underestimates the k's remaining. A rough estimate at n=~100M would be ~500 k's remaining.

A bold public prediction: Due to its extremely heavy weight, Riesel base 3 will be proven at a lower n-value than Riesel base 2!

gd_barnes 2013-01-17 07:31

Curtis has completed k=2.15G-2.3G to n=15K. 1283 k's are remaining for the range. He is still testing and continuing all of k=2.05G-2.5G to n=30K.

VBCurtis 2013-01-17 23:50

Gary-
Thanks for making sense of the jumble of files I sent you. I think I better understand now how to start a base and produce simpler files for you to administrate.

I expected I'd have 2050 to 2500 done to 30k or 35k by the time Lennart got the big range to 100k. I had no idea the CPU cycles Lennart would bring to this base!! I am currently at 2350 on the script, running at a pace of 25M per week; so, I should finish 2500 to 15k around 1 March.

I will then test 2050 to 2500 to 30k in March, and then sieve 30k to 100k for this range, sharing the file with anyone who wishes to help the project along.
-Curtis

henryzz 2013-01-18 01:03

Now to extend the tested range from ~2G to ~63G.

VBCurtis 2013-01-21 04:23

[QUOTE=henryzz;325080]Now to extend the tested range from ~2G to ~63G.[/QUOTE]

Well, I am 75% done to 2.5G, and I think Lennart will help me get to 3G this winter. So you should begin at 3G to help get us to 63G.
-Curtis

Lennart 2013-01-21 13:40

Yes I am starting 2.5G-3G now.

Lennart

Puzzle-Peter 2013-01-31 15:44

Reserving 3G - 3.4G to n=25k

Puzzle-Peter 2013-02-19 17:08

3G - 3.4G complete. Results can be downloaded here:
[url]http://www.sendspace.com/file/rnlmil[/url]

Lennart 2013-02-19 19:31

Sorry I missed to report that 2.5G-3G is done to 15k (r3)


I shall start sieve the first G's




Lennart

Puzzle-Peter 2013-02-19 20:39

Reserving 3.4G - 3.8G to n=25k

gd_barnes 2013-02-19 20:57

[QUOTE=Lennart;330089]Sorry I missed to report that 2.5G-3G is done to 15k (r3)


I shall start sieve the first G's




Lennart[/QUOTE]

Have you sent the pl_primes and pl_remain files to me or are you waiting until you have searched the range deeper? I haven't received any files for the range yet.

Lennart 2013-02-19 23:42

[QUOTE=gd_barnes;330115]Have you sent the pl_primes and pl_remain files to me or are you waiting until you have searched the range deeper? I haven't received any files for the range yet.[/QUOTE]


I have not sent them but will do that soon.


Lennart

gd_barnes 2013-02-28 09:39

Lennart has completed k=2.5G-3G to n=25K. 2787 k's remain for the range.

Lennart, will you be continuing with this k-range?

Lennart 2013-02-28 13:18

[QUOTE=gd_barnes;331364]Lennart has completed k=2.5G-3G to n=25K. 2787 k's remain for the range.

Lennart, will you be continuing with this k-range?[/QUOTE]


No I stay here now. I will finish up r7 first.

Lennart

Puzzle-Peter 2013-03-01 19:10

Reserving k=3.8G to 5G to n=25k

Puzzle-Peter 2013-03-04 19:16

Reserving k=5G - 7.4G to n=25k

MyDogBuster 2013-03-04 21:07

[QUOTE]Reserving k=5G - 7.4G to n=25k [/QUOTE]

Looks like you are seriously trying to get to 63G. A noble goal.

Good luck. :cool:

VBCurtis 2013-03-04 22:17

Update:
I had an unexpected windows update reboot my machine during a script run, and then had other priorities for the CPU that put this job on the back burner. I expect another 3-5 weeks to finish 2350-2500 to 15k, followed by getting my range to 25k to match the avalanche of work that has been done on this base in 2013.

Since I like sieving, I will proceed with a sieve from 25k for 2050 to ?? once I get my range done to 25k. I think my machine can handle a sieve from k=2G to 5G, but a larger range may run into memory difficulties due to the number of k's.

-Curtis

Puzzle-Peter 2013-03-05 15:38

[QUOTE=MyDogBuster;331966]Looks like you are seriously trying to get to 63G. A noble goal.

Good luck. :cool:[/QUOTE]

I'll head for 10G first, then re-evaluate. Any help is highly appreciated. Well, maybe not so much by Gary who has to handle a lot of primes and remaing k's. :grin:

MyDogBuster 2013-03-05 17:42

[QUOTE]I'll head for 10G first, then re-evaluate. Any help is highly appreciated. Well, maybe not so much by Gary who has to handle a lot of primes and remaing k's. :grin:[/QUOTE]

He hated it when I was doing S3. Maybe it's time to do some more of that.:devil:

Puzzle-Peter 2013-03-05 18:12

[QUOTE=MyDogBuster;332084]He hated it when I was doing S3. Maybe it's time to do some more of that.:devil:[/QUOTE]
Well, who started the project?

"Spirits that I've cited..."

Puzzle-Peter 2013-03-11 17:14

Reserving k=7.4G - 10.6G to n=25k

Puzzle-Peter 2013-03-13 21:24

Reserving k=10.6G to 11G to n=25k

Uploaded results for 3.4G to 3.8G here: [URL]http://www.sendspace.com/file/lkl7qi[/URL]

KEP 2013-03-16 14:47

Reserving as new to n=25K:

k=11G to
k=12G

Regards

KEP

Puzzle-Peter 2013-04-03 19:53

Results for 10.6G - 11G can be downloaded here:

[url]http://www.sendspace.com/file/ccdz67[/url]

Due to a wide range of machine speeds and uptime per day the subranges are finishing in a kind of randomized order... I'll post them as they come and won't reserve new work until everything under 11G is complete.

Puzzle-Peter 2013-06-17 19:22

I just finished uploading the rest of my reservation to sendspace. As always, a download link has been sent to Gary.

This didn't run as smoothly as I had hoped. I'll have to investigate what caused the huge speed differences on - to the best of my knowledge - identical machines.

No new reservations for now.

gd_barnes 2013-06-17 20:51

Thank you Peter. You did a huge amount of work! It's also a several hour effort to update everything from my perspective. I'll be making some time later today to start downloading and updating everything. Unfortunately sendspace has a 300 MB hourly and 1 GB daily download limit so I'll have to space out the updates over several days.

VBCurtis 2013-06-21 06:29

Gary-
I have one more 25k chunk to finish my part to 2500M; once I complete that in the next few weeks, I'll re-learn how to send you the relevant data. Then, I suppose, I can test out the limits of sr2sieve's ability to handle thousands of k's? I think it will be more efficient to sieve as many k's as possible over a smaller n-range than previous sieves (say, 50k for the first pass), but memory constraints may decide this for me.
With Peter's immense amount of work done, we're over 20% of the way to the conjectured k-value, right?
-Curtis

gd_barnes 2013-06-21 06:52

Oops
 
[QUOTE=Puzzle-Peter;343668]I just finished uploading the rest of my reservation to sendspace. As always, a download link has been sent to Gary.

This didn't run as smoothly as I had hoped. I'll have to investigate what caused the huge speed differences on - to the best of my knowledge - identical machines.

No new reservations for now.[/QUOTE]

Peter,

Unfortunately I did not get the range of k=9.6G-10.2G downloaded from Sendspace in time before the site deleted them. Can you upload them to Sendspace again?

Sorry about that.


Gary

gd_barnes 2013-06-21 06:56

[QUOTE=VBCurtis;343983]Gary-
I have one more 25k chunk to finish my part to 2500M; once I complete that in the next few weeks, I'll re-learn how to send you the relevant data. Then, I suppose, I can test out the limits of sr2sieve's ability to handle thousands of k's? I think it will be more efficient to sieve as many k's as possible over a smaller n-range than previous sieves (say, 50k for the first pass), but memory constraints may decide this for me.
With Peter's immense amount of work done, we're over 20% of the way to the conjectured k-value, right?
-Curtis[/QUOTE]

Will you be finished testing k=2.05G-2.5G to at least n=25K?

When your range is complete to n=25K, we'll not quite be at 20%. We'll be at 11G / 63G = 17.5% of the conjecture complete to n=25K.

Puzzle-Peter 2013-06-21 13:23

[QUOTE=gd_barnes;343985]Peter,

Unfortunately I did not get the range of k=9.6G-10.2G downloaded from Sendspace in time before the site deleted them. Can you upload them to Sendspace again?

Sorry about that.


Gary[/QUOTE]

No problem. I'll re-upload them on them on monday.

Lennart 2013-06-21 19:15

I have started a sieve from 2.5G-9.6G to see if it works.


Lennart

VBCurtis 2013-06-24 22:23

Gary-
I ran PFGW script to 15k instead of 25k; I'll sieve and test to 25k so that everything is consistent to everyone else's work.

Lennart 2013-06-25 20:47

I am taking 2.5G-9.6G up to 100k.

Lennart

VBCurtis 2013-06-26 04:54

[QUOTE=Lennart;344400]I am taking 2.5G-9.6G up to 100k.

Lennart[/QUOTE]

Can you provide some stats on that sieve? Specifically:
How many k's?
How many MB does sr2sieve take to run- or do you have to use srsieve?
How many k/n pairs are in the sieve at whatever depth you're at?

That's one big sieve.

Lennart 2013-06-26 12:49

[QUOTE=VBCurtis;344440]Can you provide some stats on that sieve? Specifically:
How many k's?
How many MB does sr2sieve take to run- or do you have to use srsieve?
How many k/n pairs are in the sieve at whatever depth you're at?

That's one big sieve.[/QUOTE]

I am only using srsieve.

I sieve 2.5G<k<9.6G 25k<n<100k in one file.

Then I cut out 0.5G and run it alone to get a range I can LLR to 100k

You can't use sr2sieve on 0.5G k's maybe later after 100k are done.

I shall see how many k's I can use in sr2sieve but I think that will make very small ranges and you may save some time with a lot more work. :)


The first range is not sieved very high, it's 2.5G-3G 25k<n<35k. 2787 k's They are soon done. It seems to end up in about 700 primes so I expect the range will have about 65% primes (25k-100k).
That will reduce the amount of k's and it will be easier to manage the base.

If you have any better ideas pse. share them :)
'

I use srsieve on a i7 2600k and I am only running 1 sieve using about 6G memory total on that computer. I have only 8GB on that computer. I think I will split it in 3 ranges and use 2 computers (3 ranges) later




Lennart

Puzzle-Peter 2013-07-01 19:57

[QUOTE=Puzzle-Peter;344013]No problem. I'll re-upload them on them on monday.[/QUOTE]

Gary, did you get the email generated by sendspace?

MyDogBuster 2013-07-01 21:29

[QUOTE]Gary, did you get the email generated by sendspace? [/QUOTE]

Gary's out of town and won't be home till Thursday.

gd_barnes 2013-07-03 10:24

[QUOTE=Puzzle-Peter;345000]Gary, did you get the email generated by sendspace?[/QUOTE]

Yes I got it. I probably won't have time to update the pages until this weekend.

Puzzle-Peter 2013-07-03 14:08

[QUOTE=gd_barnes;345148]Yes I got it. I probably won't have time to update the pages until this weekend.[/QUOTE]

Don't hurry! I just wanted to make sure it didn't end up in the spam folder.

KEP 2013-08-13 18:04

R3 11.0G-12.0G is complete. Data link has been sent to Gary.

Take care

KEP

Citrix 2013-08-25 06:12

1 Attachment(s)
I wrote some code to help with the initial sieving.
It finds primes for a range of k(start) to k(end) with n=1 to 2000.
base=3
It writes the riesel primes in primes.txt
It writes the candidates left in left3r.txt


It is much faster than PFGW's script and hence useful.
The remaining candidates can then be tested with PFGW.
:smile:

Lennart 2013-08-25 09:10

[QUOTE=Citrix;350761]I wrote some code to help with the initial sieving.
It finds primes for a range of k(start) to k(end) with n=1 to 2000.
base=3
It writes the riesel primes in primes.txt
It writes the candidates left in left3r.txt


It is much faster than PFGW's script and hence useful.
The remaining candidates can then be tested with PFGW.
:smile:[/QUOTE]


Ok I will start a test. I start at 20G to 21G now.

Lennart

KEP 2013-08-25 14:54

[QUOTE=Citrix;350761]I wrote some code to help with the initial sieving.
It finds primes for a range of k(start) to k(end) with n=1 to 2000.
base=3
It writes the riesel primes in primes.txt
It writes the candidates left in left3r.txt


It is much faster than PFGW's script and hence useful.
The remaining candidates can then be tested with PFGW.
:smile:[/QUOTE]

Does this program account for MOB's or is every single k tested?

Citrix 2013-08-25 17:15

I am assuming you mean K=1 (mod 2)-- yes they are removed.

Note:- The program will overwrite the files if you start the program again. So the files need to be copied to another folder before starting the program again.

KEP 2013-08-25 18:17

[QUOTE=Citrix;350800]I am assuming you mean K=1 (mod 2)-- yes they are removed.

Note:- The program will overwrite the files if you start the program again. So the files need to be copied to another folder before starting the program again.[/QUOTE]

Unfortunantly no, K=1 (mod 2), is only accounting for the general factors. The MOB's is [B]M[/B]ultiple [B]O[/B]f [B]B[/B]ase, hence MOB. A MOB is a k such that K=0 (mod base-value). For these values if k-1 is composit, the testing of the k can be excluded because this k will have the same prime as k/b. If we are using the "starting bases script" and runs PFGW with this script, then if a k meets the previously described conditions, we will in the testing directory, see a file named: pl_mob.txt, while primes will be stored in pl_prime.txt and so forward :)

Regards

KEP

Ps. It may not mean a lot, your program seems to be up to speed, nice work :)

Citrix 2013-08-25 18:46

I will put this in the next release and provide more choices for base etc.

henryzz 2013-08-25 21:16

Is there any chance of source at some point? If it is faster for riesel base 3 it should be for other large conjecture ks.

gd_barnes 2013-08-25 21:58

[QUOTE=Citrix;350810]I will put this in the next release and provide more choices for base etc.[/QUOTE]

Citrix,

To be clear, it appears that the starting bases script is quite a bit more complex than you realize. Have you looked at the existing starting bases script? Even for relatively simple base 3, you can't just tell it to test the even k's for a given k-range to a specific n-value. As KEP says, you need to account for k's that are a multiple of the base (MOB) where k-1 is not prime (Riesel) or k+1 is not prime (Sierp). Such k's should not be tested and need to be written to a separate file.

Also, please get your base 3 script working with MOBs, etc. before attempting to generalize it across all bases. In other words, no enhancements before bug fixes. If you are going to generalize it, it becomes quite a bit more difficult. You need to account for entire sets of k's with trivial factors as well as for k's that are GFNs (Generalized Fermat numbers).

Please everyone, this is a beta test version. Do not use it for new ranges. I will not accept the results of them. It should be parallel tested using ranges already tested with the existing new bases script. In case that's not clear, this new script needs to be run and then the old script needs to be run across the same range of k and n-value across a variety of conditions. The output files need to match up perfectly.

This project must be very exacting to avoid missing primes and causing additional testing.


Gary


All times are UTC. The time now is 03:13.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2022, Jelsoft Enterprises Ltd.