Odd perfect related road blocks II - Page 25

wblipp · 2010-04-09, 16:27

Quote:

Originally Posted by alexhiggins732

I sent an email to reserve c < 100 digits on that list, no response but here they are:

That would be me. The reservation email never arrived. Thanks for the factors.

William

chris2be8 · 2010-04-09, 19:59

6 more results:
sigma(18750312275474698293380477984859211^4)
r1=3955084213996480368187397039669603605259103800371 (pp49)
r2=2993654893573563190431934488554477666944329734311027266211 (pp58)

sigma(94129832004310937469012032249717^6)
r1=1155271394700042415542902599256142825068119547 (pp46)
r2=4258088522579234847554549401695421527482182690820149 (pp52)

sigma(21004289493395452945341335064971627611657694767086058109903^2)
r1=8098410394545925697561201774725804855650757 (pp43)
r2=1784029171128098382481688512907724611281227944444193481 (pp55)

sigma(248554705490239^16)
r1=689733267382795759310719135242387592479450943 (pp45)
r2=40620748314807508666050475872956194123971798625230547 (pp53)

sigma(2591973653217041466735433632439900254868117872281608803776120270483^2)
r1=154145127160645854264098262299807743 (pp36)
r2=326450849485870867930032229652008840837378399434998179455160597 (pp63)

sigma(2884158155431699570670577666164571913820879317217155062743042256541^2)
r1=1279024418180109084399675711877457679822121 (pp43)
r2=60433982589247930874807922186221409909685964933571485689 (pp56)

I'll reserve the 3 numbers in t380.txt next, they should take a few days each:
sigma(280257881^18)
sigma(10177286401^16)
sigma(3158528101^18)

Chris K

wblipp · 2010-04-09, 22:25

Quote:

Originally Posted by alexhiggins732

Code:

starting SIQS on c93: 804014963263071354744224934141558692779975868958332820133596413337953111689850608888500792487

==== sieving in progress ( 4 threads):   77228 relations needed ====
====            Press ctrl-c to abort and save state            ====

SIQS elapsed time = 1793.2854 seconds.


***factors found***


ans = 804014963263071354744224934141558692779975868958332820133596413337953111689850608888500792487

This one for 239713^23-1 doesn't seem to have factors.

William

fivemack · 2010-04-09, 23:07

2801^79-1 is entirely practical, I'm optimising the sieving parameters over the weekend. The sieving will take about 200 million seconds on a single 2.5GHz K8 CPU running 64-bit Linux, and it needs siever 15e which I don't think RSALS's software can yet handle.

So it'll take a little while; I could do it on my own but it would take three to four months depending how many company CPUs I can use over the weekends. NFS@home as presently constituted could do the sieving inside a week, but has thirteen projects of about this size queued already. Is anyone reading this interested and in possession of non-trivial (say, one quad-core 64-bit-Linux system whose idle cycles for a month they'd be willing to donate) compute power?

schickel · 2010-04-09, 23:49

Quote:

Originally Posted by fivemack

2801^79-1 is entirely practical, I'm optimising the sieving parameters over the weekend. The sieving will take about 200 million seconds on a single 2.5GHz K8 CPU running 64-bit Linux, and it needs siever 15e which I don't think RSALS's software can yet handle.

Is anyone reading this interested and in possession of non-trivial (say, one quad-core 64-bit-Linux system whose idle cycles for a month they'd be willing to donate) compute power?

I'd be willing to donate some cycles to this. Unfortunately, I only have 2 dual-core Vista systems (32-bit & 64-bit) that would be up to the task....

Batalov · 2010-04-09, 23:53

I second that it is fairly doable. I suspect that it may run better with 16e coupled with lower FBLIMs (because it needs less range). For a similar complexity 3,568+ (difficulty is within ~1 digit), we chose 16e even though it is ~10% slower than 15e. With 15e, the sieving range looked like we'd need traveling into mid-300Ms (that's on one side. Both sides, I loathe - from the limited experience ...and from what I read). With 16e, we will probably go only to 190M on one side. From Bruce, I hear that half of the range is done and redundancy is on a leash (~19% now; expecting to reach maybe 30% by the end). Just a thought.

I can volunteer a backup algebra time. Only if it will happen to be needed.

axn · 2010-04-10, 00:45

could someone get the 64 bit sievers working on windows, please?

alexhiggins732 · 2010-04-10, 00:53

Quote:

Originally Posted by wblipp

This one for 239713^23-1 doesn't seem to have factors.

William

Huh?? Bug in Yafu perhaps??

Xeon 3220 @2.40GHz, 4 cores, but its win32. Benchmark is here: primehttp://www.mersenneforum.org/showpost.php?p=210363&postcount=38

Running through Yafu now to double check. If it fails again, will run through MSIEVE

alexhiggins732 · 2010-04-10, 01:19

Does the sieving need to be done on one machine? I mean we can split the job across multiple users and just have one centralized placed for everyone to dump their spairs.add.[ClientId] files.

Now the matrix step and poly selection is a difference story, but that multiple cores doesn't help there any way.

alexhiggins732 · 2010-04-10, 01:20

Quote:

Originally Posted by wblipp

This one for 239713^23-1 doesn't seem to have factors.

William

OK here it is, with factors this time :)

Code:

yafu  siqs(8040149632630713547442249341415586927799758689
58332820133596413337953111689850608888500792487) -threads 4



starting SIQS on c93: 8040149632630713547442249341415586927799758689583328201335
96413337953111689850608888500792487

==== sieving in progress ( 4 threads):   77228 relations needed ====
====            Press ctrl-c to abort and save state            ====
77504 rels found: 21085 full + 56419 from 973558 partial, (602.98 rels/sec)

SIQS elapsed time = 1696.4821 seconds.


***factors found***

PRP40 = 4462049037102822483950369137268688231101
PRP54 = 180189629602348050001139990201386333142411431335316787

ans = 1

10metreh · 2010-04-10, 06:49

Quote:

Originally Posted by alexhiggins732

Does the sieving need to be done on one machine? I mean we can split the job across multiple users and just have one centralized placed for everyone to dump their spairs.add.[ClientId] files.

Now the matrix step and poly selection is a difference story, but that multiple cores doesn't help there any way.

Firstly, there won't be any automated poly selection. It's an SNFS number so very simple polys can be created by hand.

The sieving can be done on many machines and then joined up: here is a c171 GNFS that is being done this way.

For the matrix, it can't be run on multiple machines, but the Lanczos can be done on multiple cores.
The square root can be done in parallel too.

2010-04-09, 19:59	#266
chris2be8 Sep 2009 2⁴×131 Posts	6 more results: sigma(18750312275474698293380477984859211^4) r1=3955084213996480368187397039669603605259103800371 (pp49) r2=2993654893573563190431934488554477666944329734311027266211 (pp58) sigma(94129832004310937469012032249717^6) r1=1155271394700042415542902599256142825068119547 (pp46) r2=4258088522579234847554549401695421527482182690820149 (pp52) sigma(21004289493395452945341335064971627611657694767086058109903^2) r1=8098410394545925697561201774725804855650757 (pp43) r2=1784029171128098382481688512907724611281227944444193481 (pp55) sigma(248554705490239^16) r1=689733267382795759310719135242387592479450943 (pp45) r2=40620748314807508666050475872956194123971798625230547 (pp53) sigma(2591973653217041466735433632439900254868117872281608803776120270483^2) r1=154145127160645854264098262299807743 (pp36) r2=326450849485870867930032229652008840837378399434998179455160597 (pp63) sigma(2884158155431699570670577666164571913820879317217155062743042256541^2) r1=1279024418180109084399675711877457679822121 (pp43) r2=60433982589247930874807922186221409909685964933571485689 (pp56) I'll reserve the 3 numbers in t380.txt next, they should take a few days each: sigma(280257881^18) sigma(10177286401^16) sigma(3158528101^18) Chris K Last fiddled with by chris2be8 on 2010-04-09 at 20:20 Reason: Another just finished.

2010-04-09, 23:07	#268
fivemack (loop (#_fork)) Feb 2006 Cambridge, England 2³·11·73 Posts	2801^79-1 is entirely practical, I'm optimising the sieving parameters over the weekend. The sieving will take about 200 million seconds on a single 2.5GHz K8 CPU running 64-bit Linux, and it needs siever 15e which I don't think RSALS's software can yet handle. So it'll take a little while; I could do it on my own but it would take three to four months depending how many company CPUs I can use over the weekends. NFS@home as presently constituted could do the sieving inside a week, but has thirteen projects of about this size queued already. Is anyone reading this interested and in possession of non-trivial (say, one quad-core 64-bit-Linux system whose idle cycles for a month they'd be willing to donate) compute power? Last fiddled with by fivemack on 2010-04-09 at 23:07

2010-04-10, 01:19	#273
alexhiggins732 Mar 2010 Brick, NJ 67 Posts	Sieving 2801^79-1 Does the sieving need to be done on one machine? I mean we can split the job across multiple users and just have one centralized placed for everyone to dump their spairs.add.[ClientId] files. Now the matrix step and poly selection is a difference story, but that multiple cores doesn't help there any way.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Odd perfect related road blocks	jchein1	Factoring	31	2009-04-29 15:18
Odd perfect related number	Zeta-Flux	Factoring	46	2009-04-24 22:03
Question about triming [code] blocks	schickel	Forum Feedback	4	2009-04-01 03:27
MonoDevelop vs. Code::Blocks	ixfd64	Software	1	2008-03-10 08:30
Intels Intresting Road	moo	Hardware	7	2005-12-13 02:20

2010-04-09, 23:53	#270
Batalov "Serge" Mar 2008 Phi(4,2^7658614+1)/2 9497₁₀ Posts	I second that it is fairly doable. I suspect that it may run better with 16e coupled with lower FBLIMs (because it needs less range). For a similar complexity 3,568+ (difficulty is within ~1 digit), we chose 16e even though it is ~10% slower than 15e. With 15e, the sieving range looked like we'd need traveling into mid-300Ms (that's on one side. Both sides, I loathe - from the limited experience ...and from what I read). With 16e, we will probably go only to 190M on one side. From Bruce, I hear that half of the range is done and redundancy is on a leash (~19% now; expecting to reach maybe 30% by the end). Just a thought. I can volunteer a backup algebra time. Only if it will happen to be needed.

2010-04-10, 00:45	#271
axn Jun 2003 3²·5·113 Posts	could someone get the 64 bit sievers working on windows, please?