mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Riesel Prime Search (https://www.mersenneforum.org/forumdisplay.php?f=59)
-   -   Low weight stats page. (https://www.mersenneforum.org/showthread.php?t=2934)

Thomas11 2006-04-27 07:58

[QUOTE=Kosmaj]Low weight (3rd Drive) sounds good and I also vote for option 2. I'm not sure do you need extra sieving or not. If you do, we have a small "sieving squad" that might be able to help.[/QUOTE]

The extra sieving effort needed is really low. We could already start with -- let's say 1.1-1.5M -- and I'll finish the sieve for the larger candidates easily by myself (I'm expecting only one or two weeks to be necessary for that purpose). You should note that I'm relying on the good old RieselSieve 1.35 by Paul Jobling. On the low-weight candidates I found it to be up to twice as fast as it's successor ProthSieve...

In total we are talking about less than 12000 candidates. This should keep us busy for a few months - depending on the number of coworkers.

-- Thomas

Thomas11 2006-04-27 11:33

This is just a compilation of FFT lengths used by LLR for testing k of the size we will look at in the 3rd drive (e.g., k of about 9 decimal digits):

[CODE] fftlen nmax
-----------------------
114688 1160703
131072 1326659
163840 1650074
196608 1966990
229376 2290405
262144 2618320
327680 3259650
393216 3884481
458752 4522811
524288 5168642
655360 6423302
786432 7652963
917504 8912624
1048576 10182285
-----------------------[/CODE]

("nmax" is the largest n for the given "fftlen")

Note, that we are talking about "Zero Padded IBDWT", which is slightly slower (about 5-10%) than the non-zero padded one of the same FFT length that would be used on the smaller k.

P.S.: These FFT lengths are for SSE2 machines (e.g. P4) and are valid for LLR 3.6/3.7. On Athlons the FFT turnover should appear at slightly larger n.

lsoule 2006-05-01 21:21

I've worked on k=144337 and finished it up to 1M (it's not currently listed
on the low-weight page but if you can add it, that would be great). The
only primes found were:

144337 129
144337 237
144337 609
144337 1077


A sieve to 173G leaves 4.23 n's/1000.

kar_bon 2006-05-10 06:51

status overview
 
i've reserved per email k=20141 and 27253 for me but not posted here.
i'm working on these and k=31511 too.

k=20141 up to n=716000 -> no prime found yet

k=27253 up to n=660000 -> n=272347 refound (Ballinger, 1998) and a new
prime at n=370867 (L150, Nov.2005), mailed and listed on status page.

k=31511 up to n=335000 -> no prime found yet

Cruelty 2006-05-10 12:47

I would like to reserve k=864316301 (it's not listed on 15k.org page).
There are only two primes for n<20000:
864316301*2^30-1
864316301*2^402-1

Right now I am sieving this k for 20000<n<1000000, and at p=13.3*10^9 there are ~11400 candidates left. I plan to take it till p=100*10^9.

amphoria 2006-05-10 18:03

When the page is next updated can you record that I am working on the following low weight k's.

k=48703 [at n=840000] Confirmed prime at n=109415 (Wilfred Keller, May 1998).

k=49151 [at n=600000] Prime at n=97058.

k=50171 [at n=1500000] Prime at n=138.

VBCurtis 2006-05-11 20:04

[QUOTE=Cruelty]I would like to reserve k=864316301 (it's not listed on 15k.org page).
There are only two primes for n<20000:
864316301*2^30-1
864316301*2^402-1

Right now I am sieving this k for 20000<n<1000000, and at p=13.3*10^9 there are ~11400 candidates left. I plan to take it till p=100*10^9.[/QUOTE]

If you are planning to check this k-value to 1 million for n, you need to sieve MUCH deeper than 100B. That is a fairly high-weight number, with 11 candidates per 1000, and should likely be sieved to at least 1000B by the time you are testing numbers around n=400k.

To save yourself time, consider LLRing n=20,000 to n=300,000 after sieving to 100B, then sieving the rest to 1 trillion before LLRing the rest. You'll find the sieve removes candidates faster than LLR would for a number around 600,000.
-Curtis

Cruelty 2006-05-12 10:54

VBCurtis - thanks for advice, I have used LLR till n~125000 and switched to NewPGen from p~66*10^9 till... today afternoon :smile:

Thomas11 2006-06-20 16:27

k=59493015971
 
k=59493015971 has reached n=4M now! :showoff:
Taking a break... :sleep:

Kosmaj 2006-06-21 07:56

n=4M !!! :w00t: :shock:
Sounds like clucking bell! I know you are aiming high, but I had no idea it's that high. With a prime form such heights we could certainly move further up by score :cool:

Can you give us some numbers, for example how many tests in the 3-4M range, and FFT lengths & exe times in the vicinity of 4M. Thanks!

Thomas11 2006-06-21 08:58

Okay, here is some information:

The following FFT lengths had been in use so far:
[CODE] fftlen nmax
-----------------------
327680 3259650
393216 3884481
458752 4522811[/CODE]
Note, that zero-padding is used at this size of k. This means slightly different running times when compared to the "ordinary" non-zero-padded FFT lengths. For much smaller k (e.g. 321search's k=3) the FFT lengths would be only half as large (and running times too, of course).

On average there are "only" 88 LLR tests to be done per 1 million n range.
The tests were done on a 1.7GHz P4, and in the vicinity of n=4M each test lasts about 110,000 secs (= 30.5 hours).

kar_bon 2006-07-12 06:43

status
 
k n
20141 760.000
27253 757.000
31511 466.000
no new primes

Templus 2006-07-12 08:52

I reserved k=411793, it is tested until n=260k with 5000 tests left until 2M.
Found primes are:

411793 35
411793 683
411793 9419
411793 233819

grobie 2006-07-26 23:58

Is Joss still updating the low weight status page? [url]http://www.15k.org/lowweight.htm[/url] it seem there are some k's being tested but does not reflect the reservations!

grobie 2006-07-27 02:52

Reserving k=256453
 
I would like to reserve k=256453

tested n<=50k with proth.exe

Primes Found:
256453*2^607-1
256453*2^41359-1

Sieving up to 1M in progress.

Kosmaj 2006-07-27 04:57

Curtis is now the editor in charge of the low weight stats page, and I'm sure he will update it shortly. Just sit back, relax, and keep on crunching. :smile:

VBCurtis 2006-07-27 08:58

Thanks for the reminder.. two months is too long to go without a web update. I've emailed Kosmaj the updated page, and it should be up within a day or two.

Marcin has not reported updates on his litany of reservations-- if anyone is in contact with him, it would be nice to see how he has progressed. Some of the reservations on the page have not seen reported progress in over a year.

-Curtis

grobie 2006-07-27 21:57

k=256453
 
Just FYI sieved to 50G with 2359 tests remain to 1M

Sieving continues!!

grobie 2006-07-28 22:19

Wierd
 
I was looking through my results file for k=256453 and noticed 1 test showed as not prime but the time for the test was low, so I tested it again with same result:

256453*2^152347-1 is not prime. LLR Res64: EF434C86322A6770 Time : 118.682 sec.
256453*2^152959-1 is not prime. LLR Res64: 192C92DE09137ED8 Time : 63.728 sec.


for that n time is normally average 118 sec for a test, but see above n=152959 took only 63 sec.

Can someone please check this out & let me know what you think!!!:ermm:

Here are more results:

256453*2^149683-1 is not prime. LLR Res64: 6FAF994E84D10193 Time : 113.957 sec.
256453*2^149827-1 is not prime. LLR Res64: 05ABC34E977B29C5 Time : 115.395 sec.
256453*2^150223-1 is not prime. LLR Res64: 5B5D4DFD6A60202F Time : 114.247 sec.
256453*2^150979-1 is not prime. LLR Res64: CCEE9050A4141A0C Time : 114.259 sec.
256453*2^151303-1 is not prime. LLR Res64: 7B10E45F788A8BD8 Time : 115.284 sec.
256453*2^151879-1 is not prime. LLR Res64: 2861E51DA81F118F Time : 115.928 sec.
256453*2^152347-1 is not prime. LLR Res64: EF434C86322A6770 Time : 118.682 sec.
256453*2^152959-1 is not prime. LLR Res64: 192C92DE09137ED8 Time : 63.728 sec.
256453*2^154579-1 is not prime. LLR Res64: 0883147ABCC5A0DE Time : 25.787 sec.
256453*2^154903-1 is not prime. LLR Res64: A5EFEE2FD6C63A93 Time : 70.105 sec.
256453*2^155263-1 is not prime. LLR Res64: B3AD7C1039F3F71A Time : 76.081 sec.
256453*2^156559-1 is not prime. LLR Res64: 57C0AAA6D3C7E645 Time : 71.904 sec.
256453*2^157639-1 is not prime. LLR Res64: 8A581EE475CC303C Time : 71.456 sec.
256453*2^157927-1 is not prime. LLR Res64: 25071AADA88C146E Time : 73.447 sec.

Wierd, only thing I can think of something was running in the background:(
Tests are running through 100%

amphoria 2006-07-29 07:32

[QUOTE=grobie]I was looking through my results file for k=256453 and noticed 1 test showed as not prime but the time for the test was low, so I tested it again with same result:

256453*2^152347-1 is not prime. LLR Res64: EF434C86322A6770 Time : 118.682 sec.
256453*2^152959-1 is not prime. LLR Res64: 192C92DE09137ED8 Time : 63.728 sec.
[/QUOTE]

Well this is what I got which is what I was expecting to see:

256453*2^152347-1 is not prime. LLR Res64: EF434C86322A6770 Time : 42.971 sec.
256453*2^152959-1 is not prime. LLR Res64: 192C92DE09137ED8 Time : 43.137 sec.
256453*2^149683-1 is not prime. LLR Res64: 6FAF994E84D10193 Time : 42.200 sec.
256453*2^149827-1 is not prime. LLR Res64: 05ABC34E977B29C5 Time : 42.346 sec.
256453*2^150223-1 is not prime. LLR Res64: 5B5D4DFD6A60202F Time : 42.320 sec.
256453*2^150979-1 is not prime. LLR Res64: CCEE9050A4141A0C Time : 42.473 sec.
256453*2^151879-1 is not prime. LLR Res64: 2861E51DA81F118F Time : 42.859 sec.
256453*2^152347-1 is not prime. LLR Res64: EF434C86322A6770 Time : 42.975 sec.
256453*2^152959-1 is not prime. LLR Res64: 192C92DE09137ED8 Time : 43.055 sec.
256453*2^154579-1 is not prime. LLR Res64: 0883147ABCC5A0DE Time : 43.534 sec.
256453*2^154903-1 is not prime. LLR Res64: A5EFEE2FD6C63A93 Time : 43.961 sec.
256453*2^155263-1 is not prime. LLR Res64: B3AD7C1039F3F71A Time : 43.654 sec.
256453*2^156559-1 is not prime. LLR Res64: 57C0AAA6D3C7E645 Time : 44.147 sec.
256453*2^157639-1 is not prime. LLR Res64: 8A581EE475CC303C Time : 44.393 sec.
256453*2^157927-1 is not prime. LLR Res64: 25071AADA88C146E Time : 44.397 sec.

You either have another process stealing CPU cycles or your processor is overheating and reducing its speed at times.

grobie 2006-07-29 09:42

[QUOTE=amphoria]Well this is what I got which is what I was expecting to see:

256453*2^152347-1 is not prime. LLR Res64: EF434C86322A6770 Time : 42.971 sec.
256453*2^152959-1 is not prime. LLR Res64: 192C92DE09137ED8 Time : 43.137 sec.
256453*2^149683-1 is not prime. LLR Res64: 6FAF994E84D10193 Time : 42.200 sec.
256453*2^149827-1 is not prime. LLR Res64: 05ABC34E977B29C5 Time : 42.346 sec.
256453*2^150223-1 is not prime. LLR Res64: 5B5D4DFD6A60202F Time : 42.320 sec.
256453*2^150979-1 is not prime. LLR Res64: CCEE9050A4141A0C Time : 42.473 sec.
256453*2^151879-1 is not prime. LLR Res64: 2861E51DA81F118F Time : 42.859 sec.
256453*2^152347-1 is not prime. LLR Res64: EF434C86322A6770 Time : 42.975 sec.
256453*2^152959-1 is not prime. LLR Res64: 192C92DE09137ED8 Time : 43.055 sec.
256453*2^154579-1 is not prime. LLR Res64: 0883147ABCC5A0DE Time : 43.534 sec.
256453*2^154903-1 is not prime. LLR Res64: A5EFEE2FD6C63A93 Time : 43.961 sec.
256453*2^155263-1 is not prime. LLR Res64: B3AD7C1039F3F71A Time : 43.654 sec.
256453*2^156559-1 is not prime. LLR Res64: 57C0AAA6D3C7E645 Time : 44.147 sec.
256453*2^157639-1 is not prime. LLR Res64: 8A581EE475CC303C Time : 44.393 sec.
256453*2^157927-1 is not prime. LLR Res64: 25071AADA88C146E Time : 44.397 sec.

You either have another process stealing CPU cycles or your processor is overheating and reducing its speed at times.[/QUOTE]
Ok great, I just wanted to make sure!!! Think I will open box up & clean inside just incase.

Cruelty 2006-07-29 10:25

k=864316301 update
 
I have reached n=500000 for this k and will continue till n=1000000 :smile:

grobie 2006-07-29 14:26

[QUOTE=grobie]I would like to reserve k=256453

tested n<=50k with proth.exe

Primes Found:
256453*2^607-1
256453*2^41359-1

Sieving up to 1M in progress.[/QUOTE]
Tested to 244k

new prime at n=233779 (too small for top 5000)

Citrix 2006-08-04 21:46

Can I reserve

[code]
1665624349782373
10075614324682349
12782924755172441
15335906269828439
23009979551311559
30729699810869707
41360993926068041
49050506465852977
49791527843569597
50670759245200691
72292129717621199
(11 k)

[/code]

These are from the file Thomas11 posted with k's with weight less than 5. Thanks to Thomas 11 for posting the files. If anyone has already worked on these or is working, please let me know.

Thanks!

Kosmaj 2006-08-04 22:39

I see no reason why not. Has anybody worked on them? I have no idea, maybe Thomas or Curtis can tell you more. BTW, I see that the file contains more than 5000 Ks, how did you select exactly these 11 out of 5000?

And I just tried sieving k=15335906269828439. In the n=1-10k range no [I]n[/I] survived sieving to 85M. In the n=10-200k range only 12 did!

Citrix 2006-08-05 06:26

[QUOTE=Kosmaj]I see no reason why not. Has anybody worked on them? I have no idea, maybe Thomas or Curtis can tell you more. BTW, I see that the file contains more than 5000 Ks, how did you select exactly these 11 out of 5000?

And I just tried sieving k=15335906269828439. In the n=1-10k range no [I]n[/I] survived sieving to 85M. In the n=10-200k range only 12 did![/QUOTE]

I took all the ~5250 k's (to be more precise) and sieved them in the nmin=1 nmax=1M range upto pmax=10000. Then I selected all the k's with less than 128 candidates left. I have the number of candidates remaining for the rest of the k's, if anyone wants them.

Why I am working on these extremely low weight k's?

Looking at most of large megabit primes found in the top 5000 DB, they are either mersenne numbers or other numbers that can be tested really fast, like k<32 for proth and riesel. Then there are k's that have low weight.

Consider for example all the huge SOB primes. These SOB k have such low weight that they should not have found a prime in the n range they were searched for. Yet projects like rieselsieve, PSP and SOB continue to beat the odds and find new megabit primes.

So I think, that if we choose a bunch of k's and work of them, irrespective of weight, one of them would eventually beat the odds and spit out a prime. To test this hypothesis, I am reserving the lowest 11k's that I know of, to test them to a large n value.

The following k's had 0 values left under 1M. I think these are riesel number and not low wieght numbers, but I do not know how to prove it. Is there a way to decide on their identity?
1531543060309739
24495411634339871
27697189380823769
31153994867516291
61624064322011353
61670219307871063
79259968463949613


Thanks

Kosmaj 2006-08-05 16:28

I checked k=1531543060309739 using the output of ksieve with p<2000 and found the covering set to be [3, 5, 7, 13, 17, 19, 73, 193, 257, 1153] and modulus=288. Here is the relevant part of the output:
[CODE]
3 | 1531543060309739*2^1-1 +2...
5 | 1531543060309739*2^2-1 +4...
7 | 1531543060309739*2^0-1 +3...
13 | 1531543060309739*2^4-1 +12...
17 | 1531543060309739*2^0-1 +8...
19 | 1531543060309739*2^2-1 +18...
73 | 1531543060309739*2^5-1 +9...
193 | 1531543060309739*2^44-1 +96...
257 | 1531543060309739*2^4-1 +16...
1153 | 1531543060309739*2^188-1 +288...
[/CODE]
(I couldn't reduce it further.) The above primes can cover any slot between 1 and 288, therefore k=1531543060309739 is Riesel number. Most likely the same can be shown for other k's on your list.

Citrix 2006-08-10 08:36

[QUOTE=Citrix]Can I reserve

[code]
1665624349782373
10075614324682349
12782924755172441
15335906269828439
23009979551311559
30729699810869707
41360993926068041
49050506465852977
49791527843569597
50670759245200691
72292129717621199
(11 k)

[/code]

[/QUOTE]

Complete to 1M, no primes, unreserving.
I think my hypothesis was wrong, the k's should have some substantial wieght else, no prime will be found.

Thanks.

grobie 2006-08-11 04:06

Reserving k=41343149

grobie 2006-08-13 10:33

[QUOTE=grobie]Reserving k=41343149[/QUOTE]
Tested to 475k (No Primes) Releasing this one.

Templus 2006-08-23 17:11

k=10453199 tested until n=440000, no primes. Releasing

grobie 2006-08-24 02:35

Status Update on k=256453:

Primes at:

n=607
n=41359
n=233779

Tested to 810k

Continuing to 1M

VBCurtis 2006-09-02 01:07

[QUOTE=Citrix;84606]
Consider for example all the huge SOB primes. These SOB k have such low weight that they should not have found a prime in the n range they were searched for. Yet projects like rieselsieve, PSP and SOB continue to beat the odds and find new megabit primes.

So I think, that if we choose a bunch of k's and work of them, irrespective of weight, one of them would eventually beat the odds and spit out a prime. To test this hypothesis, I am reserving the lowest 11k's that I know of, to test them to a large n value.
Thanks[/QUOTE]

There's two issues here you missed. First, SOB has done a HUGE amount of work to find those primes-- take a glance at their production stats for an idea of just how much work. I don't know what makes you say they are beating the odds in finding those primes.

Second, the phrase "eventually beat the odds" makes no sense. If you search long enough, the odds say you WILL find a prime. You can hope to get lucky, which is really what the low-weight search is about; but saying weight has no relevance, or needs to be above a certain cutoff (which you concluded after striking out to 1M on 11 k-values, which was what, 250 tests?), is folly. I think you're confusing the depth of a search with the time taken TO search. Your original idea to test the very lowest-weight k's was fine, but to give up after a week and 11 candidates proved nothing. If you hope to find a 200k digit prime, it's going to take 3-6 Ghz-months on average, no matter how to try to do the search.

Hope this helps.
-curtis

humanoverlord 2006-09-16 01:45

I would like to reserve the following k (I have indicated limits I have already tested them to and primes found):

k=131069 (to 290K), no additional primes beyond n=16 and 112 found by Joss
k=9705763 (to 635K), prime at n=49335

I have also done testing on some other low weight k's not listed (if yall are interested):

k=792729463 (to 330K), prime at n=4271, 24671
k=29330981023 (600K), no primes yet

I am not really interested in these latter two k's, so if you add them to the list, they can be considered available for reservation.



-Andrew

Kosmaj 2006-09-16 02:42

Andrew
 
Welcome to RPS!

I just updated our [URL="http://www.15k.org/lowweight.htm"]low weight stats page [/URL]with your results, it's normally updated by Mike but I was just "in the office" updating our k<300 page so I did it this time.

BTW, maybe you have already noticed, we are currently processing 36 low weight Ks in our [URL="http://www.mersenneforum.org/showthread.php?t=5789"]3rd Drive[/URL], we are now at n=1,570,000. If you have some spare cpu cycles you are welcome to join. All candidates are sieved to 4.2T.

If you have any questions, please don't hesitate to ask.

Happy hunting! :smile:

kar_bon 2006-09-21 06:43

status
 
here the actual parameters:
20141, sieved to 800G, act. n=793k , 760 n's left
27253, 490G, 781k, 815n
31511, 150G, 745k, 803n

tried new ones:
353437897, 99G, 622k, 406n
353880067, 42G, 859k, 91n

no new primes found.

kar_bon 2006-09-22 06:18

P.S.
 
sorry, forgot a 'little' prime to list:

353437897*2^9533-1 is prime!

-------------
BTW: can't edit my previous post. no edit-icon. don't know why.
options say i may edit!

Kosmaj 2006-09-22 06:56

You can edit your own post only within one hour (or 20 minutes? not sure) after posting it. But it's all right, we got your messages.

BTW, in your previous post can you clarify the meaning of respecitve numbers, for example
27253, 490G, 781k, 815n

means, 27253*2^n-1 sieved to p=490G, tested to n=781000 (no primes found?), but what does "815n" mean? 815 numbers to test to reach n=1M?

Thanks.

kar_bon 2006-09-25 09:55

meaning
 
[QUOTE=Kosmaj;87716]
BTW, in your previous post can you clarify the meaning of respecitve numbers, for example
27253, 490G, 781k, 815n
means, 27253*2^n-1 sieved to p=490G, tested to n=781000 (no primes found?), but what does "815n" mean? 815 numbers to test to reach n=1M?
Thanks.[/QUOTE]

yep, your're right. and '815n' is correct too. i only sieved up to n=1M and this is the number of remaining n to test.

Karsten

humanoverlord 2006-10-01 16:25

Returning k=131069, tested up to n=500K, no new primes.
Reserving k=56251213.

Status update: have run k=9705763 up to about 700K with no new primes, should have it at n=1M in another week or two.

Andrew

grobie 2006-10-03 23:19

Status update:

k=256453 at 961k still chugging along.

grobie 2006-10-10 22:00

k=256453 completed to 1M (No Additional Primes)

607
41359
233779

Releasing this K
Does anyone need the lresults file?

humanoverlord 2006-11-15 23:30

Status Update:
k=9705763 run to n=1M, no additional primes
k=56251213 run to n=1.12M, no additional primes

VBCurtis 2006-11-16 01:18

grobie-- no, we don't collect the results files. The stats page is mostly to make sure nobody duplicates work someone else has already done or reserved.

h.overlord: thanks for the update. I'll update the stats page in a week or two, after someone else posts their status.
-Curtis

VBCurtis 2006-11-28 21:23

My Status:
399239 finished to 1M, released.
24384313 finished to 2M, released.
13900807 at 2.1M, continuing.
3 other k's on hiatus until some machines free up.

Please post progress updates-- I'll update the page shortly. Posting every 4-6 weeks on a k you are actively searching is appropriate for this search.
-curtis

kar_bon 2006-11-29 14:39

353437897 done to n=1M: no new primes
353880067 done to n=1M: no new primes
sieved further to n=5M: 655n from 1M to 2M and 1886 to 5M in all, and sieving

20141 at 905k
27253 at 825k
31511 at 798k
50227 at 356k
57943 at 320k
352744289 at 297k (879n's to n=1M)

Cruelty 2006-12-14 06:44

k=864316301 tested till n=10^6, no primes to report. I'm still working on it.

Kosmaj 2006-12-14 07:03

Cruelty,

Nice new avatar (icon)! But what happened to your old one with Intel/AMD logos? You don't need cpu's any more? Switching to Cell?? :grin:

Xyzzy 2006-12-14 07:18

[quote]But what happened to your old one with Intel/AMD logos?[/quote]
[URL]http://www.mersenneforum.org/showpost.php?p=94024&postcount=11[/URL]

Cruelty 2006-12-14 11:35

[QUOTE=Kosmaj;94025]Switching to Cell?? :grin:[/QUOTE]Is there any sieving or factoring or LL testing software for Cell? I don't thinks so :rolleyes: I have simply decided that it was a time to change avatar and this new one looked cool and still does :cool:

humanoverlord 2006-12-20 10:20

status update: k=56251213 now run up to n=1.39M, no new primes but still going

kar_bon 2006-12-22 08:45

k=20141 is done to n=1M
no new primes!

btw, updates on [url]http://www.15k.org/[/url] are very rare!

VBCurtis 2006-12-23 23:42

I have not updated since early november. I'll do it when I return from Holiday, before 1 jan. I have my own progress to update also.
-Curtis

Templus 2007-01-07 18:01

k=10453199 tested until n=780000, 100 n's to 1M, sieved until 650bn.
I am working on it until 1M, then I switch to another.

VBCurtis 2007-01-10 05:59

Update complete, emailed to Kosmaj. All my reservations are active, so I'll have another update around the end of the month.
-Curtis

Kosmaj 2007-01-14 12:38

The latest update of the low-weight status page can be found [URL="http://www.mersenneforum.org/rps/lowweight.htm"]here[/URL].

kar_bon 2007-01-31 10:14

Status:
27253 done to n=1M: no new primes
31511 at n=982k
50227 at n=521k
57943 at n=531k
352744289 at n=691k

kar_bon 2007-02-02 13:34

k=31511 done to n=1M: no prime

VBCurtis 2007-02-03 05:08

I have recorded your updates, Karbon. Do you plan to continue any of your reservations beyond n=1M? I have them still marked reserved to you. If you do not plan to go past 1M, please let me know.

I'll post the updated page to the website in a few days, after I update my own low-weight work.

VBCurtis 2007-02-13 18:48

131707 at 1.8M, continuing to 3M.
167021 at 1.8M, continuing to 2M.
9770317 at 2M, complete and released.
13900807 at 2.35M, continuing to 4M.
Updated page coming this weekend-- post any updates you have before then!
-Curtis

humanoverlord 2007-02-16 13:41

Status Update:
k=56251213 now at 1.62M, but still no primes... will keep at it though
Have not run k=9705763 any further beyond my last update.

Andrew

kar_bon 2007-03-01 13:49

352744289 llr'ed to n=1M: no new primes!
sieving/llring further to n=5M

SB2 2007-03-11 22:23

Does anyone know if the system setup to allow downloads of pre-sieved ranges still works?

Anyone with access to upload the ranges if it is still working?

Why?

I have a number of idle athlons that I have recently brought back online, started sieving some of the non-reserved ranges (and a few that appear to be abandoned), testing a few of them. I have far more athlons than P4's so thought it would be a good idea to just sieve all the ranges not currently being worked and share them with the group.

Kosmaj 2007-03-11 22:53

I can access 15k.org and upload ranges.

Rihgt now we have available low-weight files from our [URL="http://www.mersenneforum.org/showthread.php?t=5789"]3rd drive[/URL] but the exponent is 1.65M and exe time on Athlon will be 3-4 hrs per test.

SB2 2007-03-12 22:26

I have 8 K's sieved from n=1 to 1,000,000 upto p=1T (236377, 249127, 284579, 294907, 313979, 469933, 1002401, 1005449), but will wait to finish off 2 more for an even 10 before sending to you.


Another 10 K's in the oven that should be done before the weekend.

kar_bon 2007-03-13 08:00

reserving 9787: actual n is 550k and sieved till 170G, 1428n's left to n=1M

SB2 2007-03-13 10:13

Small status report, this is what I have currently running, hopefully not duplicating work of others.

K ----- LLR Progress

3343 ----> 740k
3817 ----> 739k
4813 ----> 541k
5077 ----> 441k
6119 ----> 599k
239857 ----> 427k
248047 ----> 617k
306251 ----> 534k
320107 ----> 592k
334147 ----> 588k
334331 ----> 748k
808477 ----> 407k
10247561 ----> 574k
10346561 ----> 707k
36231101 ----> 864k


K --- Currently Sieving (when finished wil be passed along for upload to server)

273613
278713
279731
323563
332159
362737
370421
384029
412843
434729
441907
451049
464353
464909
465187
700057
700477
1006469
1034503
1210421
10013593
10108837

SB2 2007-03-14 03:49

1 Attachment(s)
First batch of pre-sieved files;

VBCurtis 2007-03-14 04:51

SB2-
Have you considered using sr2sieve to sieve all these at once, instead of sequentially sieving each individually? I imagine you could sieve much more effiiciently, to 2 or 3T, by running a batch of 10 or 12 at once.

I am sieving just 6 k values from 600k to 1.2M, and at 10T (180k candidates in sieve) I'm still removing candidates faster than 1 per 20 min. These low-weight numbers are vastly less dense, but still would benefit from a deeper sieve.

If you like, I can build the sr2data.txt for use from the zip file you uploaded already for those 10, and email it to you.

I'll update your reservations on the low-weight stats page later this week.
-curtis

SB2 2007-03-14 10:39

TBH, I have been just using what I know, NewPGen .... I suppose I could learn to use sr2sieve. Does it have the ability to split the work across numerous machines?

SB2 2007-03-14 16:30

[QUOTE=VBCurtis;100737]SB2-
Have you considered using sr2sieve to sieve all these at once, instead of sequentially sieving each individually? I imagine you could sieve much more effiiciently, to 2 or 3T, by running a batch of 10 or 12 at once.

I am sieving just 6 k values from 600k to 1.2M, and at 10T (180k candidates in sieve) I'm still removing candidates faster than 1 per 20 min. These low-weight numbers are vastly less dense, but still would benefit from a deeper sieve.

If you like, I can build the sr2data.txt for use from the zip file you uploaded already for those 10, and email it to you.

I'll update your reservations on the low-weight stats page later this week.
-curtis[/QUOTE]

If you will send the file with instructions I will give it a shot.

VBCurtis 2007-03-15 03:44

These instructions are scattered around on this forum, and it doesn't hurt to have them in one more post:

Find geoff's thread about new sieving software, and a link to his download site. if you use the first link, you'll have to go up a few levels to get to the sr2sieve and/or sr1sieve folders.

unfortunately, only srsieve can create a new sieve, so you might wish to continue to use NewPGen for that-- I did for a while. Once the sieve is at 1G, save the file.

When you have a batch of files at 1G, use srfile (see helpfile for it, in the srsieve folder) to merge the files into a single file. command format is "srfile -a name1.txt name2.txt (etc)". This will output a single file in ABCD format, ready for sr2sieve. The -a switch controls the save format as ABCD, which is sr2's preferred format (also very small compared to npg format).

sr2sieve sieves over a range of p, without regard to the marker in the sieve file. you can easily split work among machines by telling each machine to sieve a different range-- 1G to 500G on CPU1, 500G to 1T on CPU2, etc. Merging factor files is easy, done rieselsieve style:
Each sieving run will output a factors-found file, named "factorsxxx", where xxx is the starting p-value in billions. Sr2sieve does not alter the input file in any way-- your sieve is unchanged until you remove the factors with srfile. I use "srfile -h" for the list of commands at this point, but the gist is to use the -k switch to remove factors from the sieve file.

Finally, srfile -g splits the combined many-k sieve into individual sieve files, which are saved in NewPGen format ready for LLR testing.

I rambled.. if you have questions, or just want me to build the sr2data file for you the first time, I'm happy to. Email is this name at gmail, or PM me.
-Curtis

SB2 2007-03-18 15:15

Quick Update

Thanks to VBCurtis' help with creating the sieve file, and detailed instructions for Dummies <--me!, these 20 K's are now being sieved from 1 < n < 1M p=2T, ETA to completion 7 days.


236377
249127
279731
284579
294907
313979
323563
362737
370421
384029
412843
469933
700057
1002401
1005449
1006469
1034503
1210421
10013593
10108837


In hopes of not duplicating work, next batch of 20 K's that I have started sieving listed below. This batch will be sieved from 1 < n < 2M.

332159
434729
441907
451049
464353
464909
465187
473279
492787
643843
648433
685183
686711
700477
780427
783073
790841
844559
10284899
37616317


If there are conflicts with work that someone has aready started on one of these K values please let me know so I can stop it and move to another. I realize it is hard to keep up with what is being worked on with so few updates. So from now on I will try to update my progress and ranges being worked on at least once a week.

humanoverlord 2007-03-19 01:58

Status update: k=56251213 now at 1.74M and still going

SB2 2007-03-20 10:41

k=36231101 --->469002<--- LLR Tested from 300k to 1M - releasing

SB2 2007-03-24 10:28

k=278713 tested to n = 1M (no new primes) released

nuggetprime 2007-03-26 17:34

Is k=22183 still available?
If yes, I'm reserving it from 30k

nuggetprime

VBCurtis 2007-03-26 17:52

As far as I can tell, 22183 is open for work. I suggest you sieve it to n=1M, as low-weight numbers have few candidates to test. Something on the order of 9-10 weeks to LLR all the way to 1M on a modern Pentium 4 is reasonable, after 2-3 days of sieving. LLR testing to n=500k should take around 2 weeks.
-Curtis

SB2 2007-03-27 23:26

K=3817 LLR tested to 1M, continuing to 2M

SB2 2007-03-30 22:51

k=334331 LLR tested to n=1M. --> releasing.

334331 78
334331 294
334331 7998

humanoverlord 2007-05-13 01:26

k=56251213 now at n=1.96M... still no primes.

kar_bon 2007-05-14 06:51

Status
 
k (Range) (n's left) (sieved to) (llr'ed to)
9787 (1M) 1324 241G 577K
20141 (5M) 16025 361G 1M
27253 (5M) 15305 463G 1M
31511 (5M) 13137 70G 1M
50227 (1M) 930 640G 775K
57943 (1M) 640 488G 755K
99311355 (1-285k) 21232 327G 148k
99311355 (285k-1M) 110233 347G 295k
352744289 (5M) 4602 282G 1M
353437897 (5M) 4026 600G 1M
353880067 (5M) 1826 1.4T 1.625M
475977645 (1M) 91879 3.36T 381k
372321161 (5M) 1554 506G 1.143M no primes
8922449 (5M) 3629 129G 405k no primes

kar_bon 2007-06-08 07:19

57943
 
llr completed to n=1M: no new primes. (5 for this k mentioned before)
releasing this k.

gd_barnes 2007-07-14 20:07

Testing on 31 low-weight k's, missing primes found
 
1 Attachment(s)
I have tested 31 low-weight k's from k=10K to 200K up to n=100K that are shown on the summary page. This was a verification effort and not a gap-filling effort, although I did partially fill a gap on a couple of them. I have found that there are missing small primes on 9 of them.

Because of the complexity of the different situations on the k's, I won't go into a lot of detail in a message here. Attached are the details of the problems found and some k's that need to be unreserved. On 4 of the k's that had their first prime higher than n=100K, I tested to a little higher than their first prime.

The bottom line is that for every one of the k's that I tested that did not already have a prime listed, I found at least one and sometimes several primes. Of course I did not test Riesel k's as I'm sure that all of those have been double-checked by that effort at all ranges of n.

The main problem that I found is that very small primes are being missed on low-weight k's.

This was originally in a spreadsheet that I cut-and-pasted into a Notepad document, so you may need to maximize it on your screen to see it clearly.


Gary

SB2 2007-07-14 20:54

Much delayed status report;

3343 - Done to 1000k no new primes
3817 - Done to 1000k no new primes
4813 - Done to 1000k no new primes
5077 - Done to 900k 490629, 881829..testing in progress
6119 - Done to 1035k 678080, 1011416..testing in progress
7331 - Done to 477k no new primes..testing in progress
9179 - sieving to 1000k ..in progress
9613 - Done to 630k 464439..testing in progress
9913 - Done to 736k no new primes..testing in progress
10949 - Done to 879k no new primes..testing in progress
11519 - sieving to 1000k ..in progress
12791 - Done to 683k no new primes..testing in progress
17077 - Done to 762k no new primes..testing in progress
19919 - Done to 417k no new primes..testing in progress
21547 - sieving to 1000k ..in progress
21673 - Sieved to 1000k ..Ready for testing
22183 - Sieved to 1000k ..Ready for testing
24067 - Sieved to 1000k ..Ready for testing
24671 - Sieved to 1000k ..Ready for testing
25229 - Sieved to 1000k ..Ready for testing
30727 - Sieved to 1000k ..Ready for testing
43541 - Sieved to 1000k ..Ready for testing
48973 - Sieved to 1000k ..Ready for testing
65279 - sieving to 1000k ..in progress
70079 - sieving to 1000k ..in progress
106303 - sieving to 1000k ..in progress
114613 - sieving to 1000k ..in progress
117037 - sieving to 1000k ..in progress
236337 - Sieved to 2000k ..Ready to test
239857 - Done to 500k no new primes..testing on hold
248047 - done to 512k no new primes..testing on hold
278713 - Done to 1000k no new primes
284579 - Done to 916k 56, 4736..testing in progress
294907 - Sieved to 2000k ..Ready to test
306251 - Done to 550k no new primes..testing on hold
313979 - Sieved to 2000k ..Ready to test
320107 - Done to 592k no new primes..testing on hold
334147 - Done to 600k no new primes..testing on hold
334331 - Done to 1000k 78, 294, 7998
685183 - Sieved to 2000k ..Ready to test
686711 - Sieved to 2000k ..Ready to test
700477 - sieving to 1000k ..in progress
808477 - Done to 856k no new primes..testing in progress
842711 - Sieved to 2000k ..Ready to Test
844559 - Sieved to 2000k ..Ready to test
10284899 - Sieved to 1000k ..Ready to test
10247561 - Done to 727k no new primes..testing in progress
10346561 - Done to 1000k 32714, <released>
37616317 - Sieved to 1000k ..Ready to test
36231101 - Done to 1000k 469002, <released>

47 - Done to 586k no new primes..testing in progress
53 - Done to 582k 582078,..testing in progress
11025 - Done to 399k 372717..testing in progress

gd_barnes 2007-07-15 08:57

Be sure and check for small primes also
 
As you may know, I have done a lot of gap-filling and some double-checking on this site.

A couple of weeks ago, I noticed the unusual number of low-weight k's that have no primes listed, many that were unreserved, and were not being searched by RieselSieve, so I knew they must have a prime. So last week, I undertook an effort to test 31 low-weight k's from k=10K to 200K from n=0 to 100K that had 2 or less primes, regardless of their current status; i.e. whether they were reserved or not, whether they were being worked on here or not, etc. I only excluded RieselSieve project k's from the effort. I found 9 of them that had unlisted primes, 8 of which whose first primes was n < 1K. 4 of those were explainable because they were either clearly never tested for n < 100K, were never tested for n > 30K, or because they had been reserved by someone who has been inactive for well over a year and had never reported any activity on them. But the other 5 of them appeared that they should have been listed.

This resulted in at least one prime being found for all listed low-weight k's between k=10K and 200K. I posted the missing primes in the 'small primes found' forum yesterday.

I suspect two problems that may have caused this: (1) Most of you may already be aware of this but I should mention it anyway in case someone is not...When using NewPGen for sieving, it will erroneously remove small n's that it should not, especially for larger k values. I always just add back in all n's < 25, regardless of the value of k used. Srsieve does not have this problem. (2) If you pick up a k where someone else left off, say at n=30K or 300K, if there are no low primes listed, you might run a quick sieve and LLR on the range of n=0 to 10K. I think that sometimes people will say they've 'tested to 300K' when in reality they tested BETWEEN n=10K (or more) and n=300K.

A good way to catch this is if the k is below the first known true Riesel number at k=509203, has not had a prime found, and it is not being worked by the RieselSieve project, then it must have a prime that is (1) on the top-5000 site -or- (2) is a relatively small prime. Most likely the latter.

Please know that none of this is to say that the current efforts here are in error. I see many small primes listed above for various k's from SB2 and others. But I just want everyone to be aware of it and watch for it. I suspect some people may have searched a very long time on some k's for that 'first prime' and never found it when in reality its first prime is at n=4 or 8 or 15 or something very irritating like that. :down:

I'm currently checking low-weight k's with 2 primes or less from k=200K to 1M that are not being shown as being worked with the low-weight effort here in the last few status reports that I see. There seem to be plenty of them.


Thanks,
Gary

Kosmaj 2007-07-15 14:18

Gary
 
Thanks for your verification efforts.

Other possible causes of missed primes, besides those you mentioned, were discussed [URL="http://www.mersenneforum.org/showthread.php?t=2985"]here[/URL]. Early versions of LLR were buggy for large k's, and not designed to find primes for small n's. After the bugs were removed nobody bothered to double check previous work.

I agree that we can remove all reservations by Lipinski, and also by Joss and Keller.

Finally, you say
[QUOTE]The following k's are reserved by Marcin, need to be unreserved,
and were found to be missing primes:
[CODE]
k missing prime(s)
38887 1, 25, 33, 225
84277 9
85807 17[/CODE][/QUOTE]
but Marcin never reported anything about them therefore I don't think we can call them missing. Also Joss tested k=179743 to 30k, and now you found a prime at 99k. It's obviously not missing because he never checked beyond 30k, don't you think so?

gd_barnes 2007-07-15 20:44

[quote=Kosmaj;110420]Finally, you say

but Marcin never reported anything about them therefore I don't think we can call them missing. Also Joss tested k=179743 to 30k, and now you found a prime at 99k. It's obviously not missing because he never checked beyond 30k, don't you think so?[/quote]


In the complexity of all of the situations involved in the various k's that I tested, I made a misleading statement in my original post about the missing primes. I attempted to correct myself with the statement in my lastest post below, where I said:
"I found 9 of them that had unlisted primes, 8 of which whose first primes was n < 1K. 4 of those were explainable because they were either clearly never tested for n < 100K, were never tested for n > 30K, or because they had been reserved by someone who has been inactive for well over a year and had never reported any activity on them.".

So you are correct, 4 of them did not have 'missing' primes, they just had not been tested or had not been tested as high as I did. Only 5 of the 9 had 'missing' primes so to speak.

I'm now in the process of verifying many low-weight k's between k=200K and k=1M. I'll let you know what I come up with.

I wish I could test ALL k's this fast! :smile: Low-weight ones are a snap.


Gary

Cruelty 2007-07-15 20:51

864316301*2^1055106-1, after this prime I will not test this k any higher - if someone [U]really[/U] wishes to continue where I left, then I can provide a sieve file till n=2M, p=11T (there are 8513 candidates left) :whistle:

gd_barnes 2007-07-16 18:41

SB2; clarification on starting point of testing
 
[quote=SB2;110378]Much delayed status report;

239857 - Done to 500k no new primes..testing on hold
248047 - done to 512k no new primes..testing on hold
306251 - Done to 550k no new primes..testing on hold
320107 - Done to 592k no new primes..testing on hold
334147 - Done to 600k no new primes..testing on hold
808477 - Done to 856k no new primes..testing in progress
[/quote]

SB2,

Above is a partial quote from your last status report. Can you tell us where you started your testing? I ask this because in addition to how high you've tested, we also need to know where you started your testing. That is, did you start testing the above k's at n=1 or n=1000 or n=100K?

I'm trying to make sure that all ranges of n are getting tested to avoid missing the listing of small primes. I think what may be happening in many cases is that there may be some confusion or assumptions made on how much testing has been done when someone says they've 'tested thru xxx' or 'done thru xxx' without initially giving a starting point of testing.


Thanks,
Gary

SB2 2007-07-16 19:19

All the K's I test are started from the n value reported as last tested by the previous tester on the website. (Mostly from Joss's efforts back in 2004/2005).

If it's not obvious, I am now working straight down the list with the exception of those k values reserved by others. As for the few k's showing testing on hold, I'll restart those once I reach that level in the list. I will be testing all to at least 1000k.

gd_barnes 2007-07-17 04:28

Specific starting point needed
 
[quote=SB2;110502]All the K's I test are started from the n value reported as last tested by the previous tester on the website. (Mostly from Joss's efforts back in 2004/2005).

If it's not obvious, I am now working straight down the list with the exception of those k values reserved by others. As for the few k's showing testing on hold, I'll restart those once I reach that level in the list. I will be testing all to at least 1000k.[/quote]

Yes, it was clear that you're working down your list but that doesn't help us here. What we really need is for you to tell us exactly where you started on each k. We have no way to know where other people left off at because the summary site has already been updated with what you tested through.

Specifically, what we need for you to do is to look back in your LLR results files for the specific k's that I mentioned and tell us exactly where you started your testing.

I'm asking about those particular 6 k's that I brought up before, those being 239857, 248047, 306251, 320107, 334147, and 808477, because in my verification efforts, on all of those, there are missing small primes. The problem is that it appears on all but one of them on the summary site that they have been fully tested from n=1 up to at least 427K.

By getting that info. from you, I can recheck up to the point that you started at and then we'll have accurate and complete lists of primes with as little duplication of effort as possible. This will also help us get to the bottom of what is happening here.


Thanks again,
Gary

gd_barnes 2007-07-17 04:52

Check all low-weight k's from n=0-100K
 
I feel compelled to mention this now because what is causing the problem is becoming increasingly obvious...

I strongly recommend that anyone who begins testing a low-weight k from where someone else left off, IF the k has no primes found below n=100K, then you should probably run a test at least from n=1 to n=10K and if you can, from n=1 to n=100K, to see if there are missing small primes. That is unless you can somehow get ahold of a results file that proves that the range has already been tested.

After finding 5 low weight k's with missing small primes for 10K < k < 200K, I'm finding many more than that for 200K < k < 1M that appear to be tested up to high values of n but that have no (or few) primes listed. I think the above appears to be the large part of what is causing missing small primes on low-weight k's.

I suspect that way back when at some point, people weren't dilligent about testing at low ranges. They were more interested in top-5000 primes. So when they said they 'tested thru 200K (or 300K or whatever), what it really meant is that they probably started at n=100K or at whatever point it took to make the top-5000 list and tested through their aforementioned point.

I will get into the specifics of the problem k's when I've completed my preliminary double-checking, which should be in the next 2-3 days.

By doing the suggested testing, even if it turns out to be double-work, it will help us keep all lists of primes complete and accurate.


Thank you,
Gary

VBCurtis 2007-07-17 06:02

I think it's more a case that some of us simply didn't know NewPGen deleted small n's when sieving. I used to just run NewPGen, then LLR... I had no idea there were missing n's.

On the k I had missing primes (131707), I had them listed in a results file, but had never posted them. Perhaps again because nobody cared... that was my first k ever.

-Curtis

gd_barnes 2007-07-17 16:06

[quote=VBCurtis;110526]I think it's more a case that some of us simply didn't know NewPGen deleted small n's when sieving. I used to just run NewPGen, then LLR... I had no idea there were missing n's.

On the k I had missing primes (131707), I had them listed in a results file, but had never posted them. Perhaps again because nobody cared... that was my first k ever.

-Curtis[/quote]

That makes sense and is a very good point. We didn't have Karsten around at that time to post all of the primes somewhere. Cheers to Karsten! :smile: :flex: So we didn't necessarily post small primes anywhere.

I will say this...It appears to be much more than just NewPgen removing small factors, although that is clearly part of the issue. I'm seeing missing primes up has high as n=100K, with many between 100 and 1000, which is much higher than NewPGen would remove them.

Regardless, I'll keep following up with the double-checks on low-weight k's at low values of n, regardless of reservation status. I doubt anyone will be offended if I find a few small ones and point them out. I only check the ones that don't have a low n and wouldn't post any on a reserved k where someone had not done at least some testing on them, such that it appears that they were missed.

And on another note, these low-weight k's are kind of fun. Tons of them can be tested in short order like SB2 is doing. I may reserve a few myself in the near future to go after a top-5000 prime more quickly. I see Karsten added a bunch of new ones in the last summary update so there are many to go around. They'll be a nice change from the 10-12 heavy-weight k's that I'm testing that seem to take forever!


Gary

Cruelty 2007-07-19 09:00

I think I'll take a look at k=10949 from n=481000... till I find a prime :wink:
geeez, I have missed low-weight thread :blush:
can someone move my post where it belongs? :blush:

SB2 2007-07-19 09:27

[QUOTE=Cruelty;110707]I think I'll take a look at k=10949 from n=481000... till I find a prime :wink:
geeez, I have missed low-weight thread :blush:
can someone move my post where it belongs? :blush:[/QUOTE]

You might want to choose another, 10949 has been tested from n=481000 and is currently at the n=955000 level.


All times are UTC. The time now is 19:38.

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