Who is adding all these N^a-1?

didgogns · 2014-03-31, 08:15

Distribution of composize numbers <= 120 digits
Size Count
54 1
56 1
59 1
65 1,604
66 3,294
67 3,886
68 4,320
69 4,862
71 362
72 1,390
73 3,268
74 4,095
75 3,953
76 4,094
77 3,643
78 3,710
79 3,805
80 3,586
81 3,441
82 3,911
83 4,042
84 3,803
85 3,791
86 3,495
87 3,384
88 3,062
89 2,621
90 1,317
91 135
92 698
93 1,128
94 1,254
95 1,205
96 1,015
97 428
98 944
99 631
100 958
101 928
102 685
103 376
104 402
105 304
106 257
107 244
108 150
109 260
110 359
111 312
112 512
113 220
114 350
115 9,262
116 19,078
117 18,400
118 18,712
119 18,375
120 17,443
Total: 197,767

Oh please...

Wick · 2014-03-31, 14:35

Still seems to be rising... 208,783 in total now

chris2be8 · 2014-03-31, 15:54

I've factored over 5000 in the last few days. And factordb will have done quite a few as well. But that's like trying to empty a bucket 1 drop at a time.

It would help if a few people ran yafu.pl (from http://mersenneforum.org/showthread.php?t=16359&page=6 post 65). It won't run out of work for a few days!

The Odd Perfect Number project is interested in p^q-1 numbers. But I'm not saying that's more than a possible source of this surge.

Chris

danaj · 2014-04-01, 03:38

I've been running my script, but I have to throttle it to stop from overflowing the allowed requests per hour. It makes a dent in the 70-72 range, but then they come back again.

EdH · 2014-04-01, 03:58

Quote:

Originally Posted by chris2be8

...
It would help if a few people ran yafu.pl (from http://mersenneforum.org/showthread.php?t=16359&page=6 post 65). It won't run out of work for a few days!
...
Chris

I've turned some machines toward this endeavour...

ChristianB · 2014-04-01, 08:14

I also have a 2-Core desktop working on the C73 and soon C74. I modified the yafu.pl to download 5 composites with a random start offset to not run into the query-limit problem.

danaj · 2014-04-01, 14:33

Quote:

Originally Posted by ChristianB

I also have a 2-Core desktop working on the C73 and soon C74. I modified the yafu.pl to download 5 composites with a random start offset to not run into the query-limit problem.

I do 40 with a random 1000 offset, but for each 40 I do simple factoring using a Perl GMP factoring library first, before selecting one for yafu. With all these new factors, it whips through many of the new factors in hardly any time. Is there a single call to report factors of multiple numbers? That would help immensely.

Edit: Moving to size 80 so at least when yafu runs it takes 50 seconds or so, helps keep the rate reasonable. The pre-factoring keeps yafu working on non-trivial composites.

wombatman · 2014-04-01, 17:30

If anybody is interested, I have a (very) simple and crude python script that converts Yafu's factor.log to a text file in the "multiple factors per line in base 10" format for factordb.

ChristianB · 2014-04-01, 18:02

Quote:

Originally Posted by wombatman

If anybody is interested, I have a (very) simple and crude python script that converts Yafu's factor.log to a text file in the "multiple factors per line in base 10" format for factordb.

Is it possible to report factors for multiple numbers via this format so we only have to do one call to http://factordb.com/report.php and report factors for all numbers?

wombatman · 2014-04-01, 18:08

Yeah. You have the number being factored and its factors on one line like this:

Code:

30 = 6 * 5

and you can have as many as you want (or at least up to 1000 at once). It takes a little longer for factordb to process everything, but you get the same listing of factors known and accepted and all that.

I should note that the python script isn't set up to do that at this point, although it no doubt could be.

wblipp · 2014-04-02, 14:21

I'm the cause. I recently finished a six month ECM pass on 2.2 million composites of 115-120 digits from my Odd Perfect factor database. I'm processing these factors into my database and adding them to factordb. Part of the problem is that I partitioned the work flow badly - some numbers are getting their found factors several days apart instead of on one line. But most of the problem is that ECM has often left 70-100 digit composites.

I've previously reasoned that OPN researchers - some known to me now, others to come along in the future - will be interested in these factors, so it's a help to add them in this place they are likely to find them. But I've previously processed much smaller batches that were not so disruptive - usually tens of thousands instead of millions.

What do other users of factordb think I should do? Finish pushing these through? Sit on incompletely factored numbers? Dribble the results in over six months?

William

2014-03-31, 08:15	#1
didgogns Mar 2014 South Korea 2⁴ Posts	Who is adding all these N^a-1? Distribution of composize numbers <= 120 digits Size Count 54 1 56 1 59 1 65 1,604 66 3,294 67 3,886 68 4,320 69 4,862 71 362 72 1,390 73 3,268 74 4,095 75 3,953 76 4,094 77 3,643 78 3,710 79 3,805 80 3,586 81 3,441 82 3,911 83 4,042 84 3,803 85 3,791 86 3,495 87 3,384 88 3,062 89 2,621 90 1,317 91 135 92 698 93 1,128 94 1,254 95 1,205 96 1,015 97 428 98 944 99 631 100 958 101 928 102 685 103 376 104 402 105 304 106 257 107 244 108 150 109 260 110 359 111 312 112 512 113 220 114 350 115 9,262 116 19,078 117 18,400 118 18,712 119 18,375 120 17,443 Total: 197,767 Oh please...

2014-04-01, 18:08	#10
wombatman I moo ablest echo power! May 2013 29×61 Posts	Yeah. You have the number being factored and its factors on one line like this: Code: 30 = 6 * 5 and you can have as many as you want (or at least up to 1000 at once). It takes a little longer for factordb to process everything, but you get the same listing of factors known and accepted and all that. I should note that the python script isn't set up to do that at this point, although it no doubt could be.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Adding second CPU to mainboard	Primeinator	Hardware	23	2014-10-28 10:01
Adding a little blurb under your user name	jasonp	Forum Feedback	28	2012-07-06 12:43
Adding New Candidates	wblipp	Operation Billion Digits	6	2011-04-10 17:45
Adding to worktodo.txt in p95v259	JuanTutors	PrimeNet	2	2009-02-09 01:02
Manually adding primes	Dærk	Software	5	2002-11-01 14:01

2014-03-31, 14:35	#2
Wick Nov 2012 1001000₂ Posts	Still seems to be rising... 208,783 in total now

2014-03-31, 15:54	#3
chris2be8 Sep 2009 2·1,039 Posts	I've factored over 5000 in the last few days. And factordb will have done quite a few as well. But that's like trying to empty a bucket 1 drop at a time. It would help if a few people ran yafu.pl (from http://mersenneforum.org/showthread.php?t=16359&page=6 post 65). It won't run out of work for a few days! The Odd Perfect Number project is interested in p^q-1 numbers. But I'm not saying that's more than a possible source of this surge. Chris

2014-04-01, 03:38	#4
danaj "Dana Jacobsen" Feb 2011 Bangkok, TH 908₁₀ Posts	I've been running my script, but I have to throttle it to stop from overflowing the allowed requests per hour. It makes a dent in the 70-72 range, but then they come back again.

2014-04-01, 08:14	#6
ChristianB Apr 2013 Germany 311 Posts	I also have a 2-Core desktop working on the C73 and soon C74. I modified the yafu.pl to download 5 composites with a random start offset to not run into the query-limit problem.

2014-04-01, 17:30	#8
wombatman I moo ablest echo power! May 2013 29×61 Posts	If anybody is interested, I have a (very) simple and crude python script that converts Yafu's factor.log to a text file in the "multiple factors per line in base 10" format for factordb.

2014-04-02, 14:21	#11
wblipp "William" May 2003 New Haven 2·7·13² Posts	I'm the cause. I recently finished a six month ECM pass on 2.2 million composites of 115-120 digits from my Odd Perfect factor database. I'm processing these factors into my database and adding them to factordb. Part of the problem is that I partitioned the work flow badly - some numbers are getting their found factors several days apart instead of on one line. But most of the problem is that ECM has often left 70-100 digit composites. I've previously reasoned that OPN researchers - some known to me now, others to come along in the future - will be interested in these factors, so it's a help to add them in this place they are likely to find them. But I've previously processed much smaller batches that were not so disruptive - usually tens of thousands instead of millions. What do other users of factordb think I should do? Finish pushing these through? Sit on incompletely factored numbers? Dribble the results in over six months? William