3221^73-1 sieving - Page 6

fivemack · 2008-12-29, 03:27

And indeed, on the sixth dependency:

Mon Dec 29 02:46:29 2008 prp70 factor: 4268789244276949228762861690430742945199326935775238201407340720225441
Mon Dec 29 02:46:29 2008 prp88 factor: 7393720864497209547797418741987855153929288725078597776812623352014398844851509314050533
Mon Dec 29 02:46:29 2008 prp96 factor: 119206318557610125536465637952294120218272312333630042840751434567237099832651324905880534881861

I'm writing this at 3:30am my time, since the cold is somewhat keeping me from sleeping.

Batalov · 2008-12-29, 05:29

Congratulations!

Andi47 · 2008-12-29, 07:44

nice!

10metreh · 2008-12-29, 09:15

I always like triple splits when they come

Congrats to you Serge for correctly guessing this one!
Has anyone ever got a quadruple split?

jasonp · 2008-12-29, 14:47

Quote:

Originally Posted by Batalov

I've found that I better like parallel single dependency sqrts - on 8 cpus you are then usually done after 1*8 sqrt

This is the way to go if you have the CPUs for it. Note that if there are three factors A, B, C and two dependencies find A*B and B*C respectively, the code is smart enough to stop there. Too bad it had to keep going in this case...

I've never seen a 4-way split for a nontrivial-size number; the odds of it happening (i.e. a huge number that is the product of 4 medium size numbers) are very low.

bsquared · 2008-12-29, 15:16

Very nice! Congrats!

Zeta-Flux · 2008-12-29, 15:29

fivemack,

Thank you for putting this all together, and I hope you feel better soon.

wblipp · 2008-12-29, 16:08

WHOO HOO!

Thanks Tom and everybody else. I've updated the news and status at http://oddperfect.org

William

FactorEyes · 2008-12-29, 22:50

Nice to see diversity: factors of every race, color, and creed.

And, yes: don't hold your breath waiting for a 4-way split, at least while the unfactored composites are under 280 digits.

Batalov · 2008-12-30, 04:18

Well, four way splits do happen, but very rarely and arguably as an ECM miss. E.g. p40.p43.p43.p61 http://hpcgi2.nifty.com/m_kamada/f/c.cgi?q=51111_186 (personally, I am sure that this number begs for 50% of 45-digit ECM, at least, and then - voila - you're done or extremely unlucky... but not everyone has tons of CPUs, right?) and this is the only example I've seen.

Most reasonable Cunnigham numbers are ECMd to 25% of "size", aren't they, so there we will have to wait for c260s, true (where the ratio will fall under 25%; ...however next year people will start Cell-ECMing to t65 and this logic will fail). And even then, we'll need to see thousands of them. In smaller projects people use the 2/9 rule (even if!), then after thousands of factored numbers a superfecta split could emerge. Which could be defended as not an ECM miss: c220 = p55.p55.p55.p55 (if one ECM'd to 50-digits).

I had a p55.p55.p114 once.

10metreh · 2008-12-30, 08:18

Quote:

Originally Posted by Batalov

I had a p55.p55.p114 once.

As I have seen in the Gratuitous Factors Thread. How far had that one been ECM'd? (I guess 50 digits.)

2008-12-29, 09:15	#59
10metreh Nov 2008 100100010010₂ Posts	I always like triple splits when they come Congrats to you Serge for correctly guessing this one! Has anyone ever got a quadruple split? Last fiddled with by 10metreh on 2008-12-29 at 09:17

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Going over 100% during sieving	wombatman	Msieve	4	2013-07-11 15:41
NFS sieving?	Dubslow	Factoring	8	2012-09-28 06:47
Line sieving vs. lattice sieving	JHansen	NFSNET Discussion	9	2010-06-09 19:25
10^420 + 1 sieving	juno1369	Factoring	20	2010-04-28 01:11
Sieving	OmbooHankvald	Prime Sierpinski Project	4	2005-06-30 07:51

2008-12-29, 03:27	#56
fivemack (loop (#_fork)) Feb 2006 Cambridge, England 2³×11×73 Posts	And indeed, on the sixth dependency: Mon Dec 29 02:46:29 2008 prp70 factor: 4268789244276949228762861690430742945199326935775238201407340720225441 Mon Dec 29 02:46:29 2008 prp88 factor: 7393720864497209547797418741987855153929288725078597776812623352014398844851509314050533 Mon Dec 29 02:46:29 2008 prp96 factor: 119206318557610125536465637952294120218272312333630042840751434567237099832651324905880534881861 I'm writing this at 3:30am my time, since the cold is somewhat keeping me from sleeping.

2008-12-29, 05:29	#57
Batalov "Serge" Mar 2008 Phi(4,2^7658614+1)/2 9,497 Posts	Congratulations!

2008-12-29, 07:44	#58
Andi47 Oct 2004 Austria 2·17·73 Posts	nice!

2008-12-29, 15:16	#61
bsquared "Ben" Feb 2007 110111000001₂ Posts	Very nice! Congrats!

2008-12-29, 15:29	#62
Zeta-Flux May 2003 3013₈ Posts	fivemack, Thank you for putting this all together, and I hope you feel better soon.

2008-12-29, 16:08	#63
wblipp "William" May 2003 New Haven 4476₈ Posts	WHOO HOO! Thanks Tom and everybody else. I've updated the news and status at http://oddperfect.org William

2008-12-29, 22:50	#64
FactorEyes Oct 2006 vomit_frame_pointer 2³×3²×5 Posts	Nice to see diversity: factors of every race, color, and creed. And, yes: don't hold your breath waiting for a 4-way split, at least while the unfactored composites are under 280 digits.

2008-12-30, 04:18	#65
Batalov "Serge" Mar 2008 Phi(4,2^7658614+1)/2 9497₁₀ Posts	Well, four way splits do happen, but very rarely and arguably as an ECM miss. E.g. p40.p43.p43.p61 http://hpcgi2.nifty.com/m_kamada/f/c.cgi?q=51111_186 (personally, I am sure that this number begs for 50% of 45-digit ECM, at least, and then - voila - you're done or extremely unlucky... but not everyone has tons of CPUs, right?) and this is the only example I've seen. Most reasonable Cunnigham numbers are ECMd to 25% of "size", aren't they, so there we will have to wait for c260s, true (where the ratio will fall under 25%; ...however next year people will start Cell-ECMing to t65 and this logic will fail). And even then, we'll need to see thousands of them. In smaller projects people use the 2/9 rule (even if!), then after thousands of factored numbers a superfecta split could emerge. Which could be defended as not an ECM miss: c220 = p55.p55.p55.p55 (if one ECM'd to 50-digits). I had a p55.p55.p114 once.