mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Aliquot Sequences (https://www.mersenneforum.org/forumdisplay.php?f=90)
-   -   Reserved for MF - Sequence 4788 (https://www.mersenneforum.org/showthread.php?t=11615)

 akruppa 2012-04-21 19:19

Did a p40 test on c121 of i3099.

 Gimarel 2012-04-21 22:01

I found a P44 with ecm.

 Gimarel 2012-04-22 12:59

I'm running NFS on c134 of i3105.

 akruppa 2012-04-23 10:20

[strike]Dodged a bullet there.[/strike]Oh wait, 2^3. No way to pick up 3.

 rajula 2012-04-23 14:31

Currently we have
i3122 : c137 = 2[SUP]2[/SUP] * 3188501 * c130
I have done 1000@3e6 and 600@11e6 on the c130.

 jrk 2012-04-24 04:48

I did 4000 curves @ B1=4e6, B2=11415205630 on line 3122. NF

 Dubslow 2012-04-24 05:11

Well, throw in the 1240 that I'm in the middle of. Switching to 12e6.

 Dubslow 2012-04-25 01:12

I did 2000@12M, so unless someone says otherwise, I'm starting NFS.

 Dubslow 2012-04-25 18:54

[QUOTE=Dubslow;297248]I did 2000@12M, so unless someone says otherwise, I'm starting NFS.[/QUOTE][code]
best poly:
# norm 2.057751e-12 alpha -6.954778 e 9.219e-11 rroots 3
n: 4833255194656435798989282761701030777827347310176551818145684509524676795349953441183066946922208707091444600773709014656974920141
skew: 379585.76
c0: -3195115069735432888433729580240
c1: -466643288161459274073147012
c2: 1689131675299165859068
c3: 3227902073791165
c4: -281776706
c5: 10920
Y0: -13464909112784320378890631
Y1: 52247598201083
nfs: commencing algebraic side lattice sieving over range: 3000000 - 3040000
total yield: 89686, q=3040003 (0.01299 sec/rel)
found 89686 relations, need at least 9036311, continuing with sieving ...
nfs: commencing algebraic side lattice sieving over range: 3040000 - 3080000
total yield: 90123, q=3080017 (0.01314 sec/rel)
found 179809 relations, need at least 9036311, continuing with sieving ...
nfs: commencing algebraic side lattice sieving over range: 3080000 - 3120000
total yield: 89735, q=3120011 (0.01315 sec/rel)
found 269544 relations, need at least 9036311, continuing with sieving ...
nfs: commencing algebraic side lattice sieving over range: 3120000 - 3160000
total yield: 88212, q=3160007 (0.01320 sec/rel)
found 357756 relations, need at least 9036311, continuing with sieving ...
nfs: commencing algebraic side lattice sieving over range: 3160000 - 3200000
total yield: 87851, q=3200003 (0.01314 sec/rel)
found 445607 relations, need at least 9036311, continuing with sieving ...
nfs: commencing algebraic side lattice sieving over range: 3200000 - 3240000
total yield: 96121, q=3240019 (0.01287 sec/rel)
found 541728 relations, need at least 9036311, continuing with sieving ...
nfs: commencing algebraic side lattice sieving over range: 3240000 - 3280000
total yield: 92727, q=3280003 (0.01280 sec/rel)
found 634455 relations, need at least 9036311, continuing with sieving ...
nfs: commencing algebraic side lattice sieving over range: 3280000 - 3320000
total yield: 86908, q=3320029 (0.01283 sec/rel)
found 721363 relations, need at least 9036311, continuing with sieving ...
nfs: commencing algebraic side lattice sieving over range: 3320000 - 3360000[/code]At [i]least[/i] three or four days, probably more.

 Batalov 2012-04-25 21:21

Are the lpbr/lpba values set to 27 in your poly?
If 28, then you are more likely to need at least ~2e7 relations.

 Dubslow 2012-04-25 21:49

[QUOTE=Batalov;297348]Are the lpbr/lpba values set to 27 in your poly?
[/QUOTE]Yes. I'm just letting yafu automate everything, so the guesses are all made by it, not me.
Edit: My estimate was off. In the ~3 hours between posts, I'm up to 1.5M rels... that's only 36 hours for sieving; 30 hrs from now, so maybe 1.5-2 days to complete the factorization.

 bsquared 2012-04-25 21:55

Yeah, it should be lpbr/a = 27 and 13e.

But I see about 0.013 sec/rel * 9036311 rels ~= 117000 sec = 32 hours, give [STRIKE]or take[/STRIKE] a handful of hours. I.e., not 72 or 96 hours... or do you not run your system 24 hours a day?

 Dubslow 2012-04-25 22:02

[QUOTE=bsquared;297352]Yeah, it should be lpbr/a = 27 and 13e.

But I see about 0.013 sec/rel * 9036311 rels ~= 117000 sec = 32 hours, give [STRIKE]or take[/STRIKE] a handful of hours. I.e., not 72 or 96 hours... or do you not run your system 24 hours a day?[/QUOTE]

That was me being stupid, ignoring those numbers, and trying to extrapolate from previous NFS jobs. Lesson: Don't do those things. You'll see I'd edited my previous post right after you posted (I hadn't refreshed the thread).

 Dubslow 2012-04-27 03:32

[code]found 9050697 relations, need at least 9036311, proceeding with filtering ...
nfs: commencing msieve filtering
4833255194656435798989282761701030777827347310176551818145684509524676795349953441183066946922208707091444600773709014656974920141

commencing relation filtering
estimated available RAM is 12019.7 MB
commencing duplicate removal, pass 1
found 1126211 hash collisions in 9050697 relations
added 62796 free relations
commencing duplicate removal, pass 2
found 987520 duplicates and 8125973 unique relations
memory use: 49.3 MB
reading ideals above 100000
commencing singleton removal, initial pass
memory use: 344.5 MB
reading all ideals from disk
memory use: 301.7 MB
keeping 10442169 ideals with weight <= 200, target excess is 43059
commencing in-memory singleton removal
begin with 8125973 relations and 10442169 unique ideals
reduce to 137 relations and 0 ideals in 13 passes
max relations containing the same ideal: 0
nfs: commencing algebraic side lattice sieving over range: 7080000 - 7120000[/code]
Is there any way to show the total number of unique rels after all filtering? I saw the "8125973 unique relations", but that was above 50% of the output. (Then again, and changes would have to be Msieve, right? Blergh.)

Either way, it'll probably be <12 hours until completion, depending on LinAlg.

 Dubslow 2012-04-27 18:12

Okay... I think it's time someone inducted me into the group of people who can ballpark rels required to build a matrix. I hit 9.036M rels, started filtering, and have since soared passed it. Now I'm at 10.478M unique rels and it's still sieving. Should I try building the matrix by hand?

 bsquared 2012-04-27 18:26

No - the rels estimate is just that, an estimate. Specifically, an estimate of the first point at which might be possible to produce a matrix. It will build the matrix when it is able to and no sooner. Patience...

 Batalov 2012-04-27 18:27

9M is a minimorum ballpark for "easy" 27-bit jobs. Here, 27-bit is pushed to its limit, so you may need 11-12M - just let it go. You are watching the paint dry. (Sometimes I do the same, but we really shouldn't.)

 fivemack 2012-04-27 18:28

This is a 130-digit number with 27/13e; my giant database of such things suggests that 10.6 million unique relations might be enough and 11.2 million unique relations is almost sure to be. So it will be done soon. It is arguably a bit inefficient for the script to be running a filtering pass about every 80,000 relations.

 bsquared 2012-04-27 18:35

[QUOTE=fivemack;297603]It is arguably a bit inefficient for the script to be running a filtering pass about every 80,000 relations.[/QUOTE]

True... I'll have to adjust those at some point.

 Batalov 2012-04-27 18:39

I've usually done this size with 28-bit.
Also, a small hint (for D.): for jobs like this, a line
qintsize: 400000
(or 200000 or whatever) can be added to the .poly to override the 80000.

That is of course if the script (or the abstraction of the script which is inside yafu) is used. One may want to wean from the vanilla script around this size. Or not. :-)

 Dubslow 2012-04-27 19:06

While I would love to get this working on my own, the problem is my introduction to NFS/anything-besides-GIMPS has been entirely through this forum, where 90%+ of the participants knew what they were doing before they joined, as opposed to the other way 'round (like me/LaurV). The end result is that there's no "How-to", and I have no idea how to go about learning to use the software without asking what I know is a "dumb" question to most people here. I've gotten along so far only because YAFU automates things.

I had no idea what 27-bit meant, until I connected it to that lpb* thing, but then again, I really don't have a clue about that either. I know that you use different la-sievers for different size jobs, but I have no clue what the cutoffs are, and worse, I've yet to find a -h or equiv. for even using the sievers, much the same as everything else NFS, 'cause everyone knows it already. (LaurV, I think, is much like me, except being somewhat less averse to asking the "dumb" questions :razz:) I will make an effort to learn what NFS actually does (as opposed to just learning the practicalities of using the software) thanks to the paper someone linked in the Factoring forum, but I don't really have time ATM so it will probably have to wait until summer. Is there any way I can get access to your DB, fivemack, or else is it possible for someone to summarize such things in a post/PM?

Edit: @Batalov: Yafu doesn't have a .poly, but I guessed you meant in the same file as the poly, which is .job for me. I've put it there -- we'll see if it has any effect (do I need to restart yafu first?)

 Dubslow 2012-04-27 20:59

[code]matrix is 1174349 x 1174579 (345.0 MB) with weight 90222308 (76.81/col)
linear algebra at 0.1%, ETA 4h17m1174579 dimensions (0.1%, ETA 4h17m)
checkpointing every 280000 dimensions579 dimensions (0.2%, ETA 4h20m)
[/code]
ETA's actually 4hr even, but I can't copy the last line :unsure:
Edit: Not quite an hour later and it's 33% through and ETA is 2.25 hrs... :huh: *shrug*

 bsquared 2012-04-27 21:20

[QUOTE=Dubslow;297618]
Edit: @Batalov: Yafu doesn't have a .poly, but I guessed you meant in the same file as the poly, which is .job for me. I've put it there -- we'll see if it has any effect (do I need to restart yafu first?)[/QUOTE]

Should not need to restart, the ggnfs sievers will see the new file the next time they are invoked. Looks like it's moot now, though.

 fivemack 2012-04-28 00:01

1 Attachment(s)
[QUOTE=Dubslow;297618]Is there any way I can get access to your DB, fivemack, or else is it possible for someone to summarize such things in a post/PM?[/QUOTE]

[url]http://www.mersenneforum.org/showpost.php?p=189863&postcount=7[/url] contains an old version in gnumeric format; I've attached the current version in Excel 2007 for Mac format.

 Dubslow 2012-04-28 00:06

Stupid question, incoming!

[QUOTE=fivemack;297675][url]http://www.mersenneforum.org/showpost.php?p=189863&postcount=7[/url] contains an old version in gnumeric format; I've attached the current version in Excel 2007 for Mac format.[/QUOTE]

Cool, thanks.

Edit: Um... I have no idea how to view XML files. Just double clicking opened the source in Chrome; Libre Office just opened the source in Writer (i.e. Word), even if I tried to open them from inside Calc. ...How do I view them in a readable fashion? Or am I stuck with the source? I did Google this, and I've figured out that XML contains no styling of its own, and that ODF is just compressed XML, but... that didn't really help.

 fivemack 2012-04-28 00:24

[QUOTE=Dubslow;297677]Cool, thanks.

Edit: Um... I have no idea how to view XML files. Just double clicking opened the source in Chrome; Libre Office just opened the source in Writer (i.e. Word), even if I tried to open them from inside Calc. ...How do I view them in a readable fashion? Or am I stuck with the source? I did Google this, and I've figured out that XML contains no styling of its own, and that ODF is just compressed XML, but... that didn't really help.[/QUOTE]

It's not an XML file, it's a Microsoft Excel file (with a slightly odd extension: rename to .xlsx rather than .xlsx. after uncompressing it). File .. Open from a spreadsheet ought to open it, or just get hold of a copy of Excel.

 Dubslow 2012-04-28 00:33

[QUOTE=fivemack;297684]It's not an XML file, it's a Microsoft Excel file (with a slightly odd extension: rename to .xlsx rather than .xlsx. after uncompressing it). File .. Open from a spreadsheet ought to open it, or just get hold of a copy of Excel.[/QUOTE]That makes much more sense, especially considering what I just said myself about the OpenDocumentFormat :smile:
(Also, sqrt is commencing)

Edit: Is the third matrix the weight, and if so, what is that? What's pmatrix?
(First GCD is 1, as is the 2nd GAH so is the 3rd wtf CRAP the 4th gcd is N son of a... 5th is one again... I thought it was like 1/2 or 2/3s odds for a factor? is this just [i]really[/i] bad luck? 6th is one... 7th is N...)

 Batalov 2012-04-28 00:53

[QUOTE=Dubslow;297688](First GCD is 1, as is the 2nd GAH so is the 3rd wtf CRAP the 4th gcd is N son of a...)[/QUOTE]
Har-dee-har-har! It's always like that when you are actually watching. Turn away, and it will be done on the 1st dep. :rolleyes:

 akruppa 2012-04-28 00:58

A watched pot and such. And yes, sqrt is computationally the shortest step of NFS but it usually feels like the longest.

 Dubslow 2012-04-28 01:00

[QUOTE=Batalov;297692]Har-dee-har-har! It's always like that when you are actually watching. Turn away, and it will be done on the 1st dep. :rolleyes:[/QUOTE]

I haven't even been watching, I just check it every three minutes (just to see another failure...)

(First GCD is 1, as is the 2nd GAH so is the 3rd wtf CRAP the 4th gcd is N son of a... 5th is one again... I thought it was like 1/2 or 2/3s odds for a factor? is this just [i]really[/i] bad luck? 6th is one... 7th is N... 8th is 1... Guess what 9 is? Hint: It's not p or q.)

YAY!! It only took 10 tries!!! :headdesk: P57*P73.

The next one's already at t30, and plowing into 1M. I suppose I'll just let it run through NFS.

 Batalov 2012-04-28 01:10

[QUOTE=akruppa;297694]A watched pot and such. And yes, sqrt is computationally the shortest step of NFS but it usually feels like the longest.[/QUOTE]
"A watched boil never pops."

 Batalov 2012-04-28 01:26

"Anytwo five elevenis?"
(translation: Time to run 11e6s.)

 Dubslow 2012-04-28 01:57

[QUOTE=Batalov;297701]"Anytwo five elevenis?"
(translation: Time to run 11e6s.)[/QUOTE]

Indeed. Starting 2K@11M.

 LaurV 2012-04-28 02:24

[QUOTE=Batalov;297692]Har-dee-har-har! It's always like that when you are actually watching. Turn away, and it will be done on the 1st dep. :rolleyes:[/QUOTE]
:tu: Subscribe 101%!

 Batalov 2012-04-28 02:58

[QUOTE=Dubslow;297704]Indeed. Starting 2K @11M.[/QUOTE]
+1.5K @11M. Time to NFS.

 Dubslow 2012-04-28 03:47

[QUOTE=Batalov;297712]+1.5K @11M. Time to NFS.[/QUOTE]You wanna take it or should I? (Btw, I'm only at 240, but was aiming for 2K.)

 Batalov 2012-04-28 03:56

My vote is for you. If I had a cent for each c120 I factored...
You OTOH have to collect your first dollar ;-)

(ECM was enough already when we finished 3e6s. gnfs-120 * 1/3 = t40, i.e. 2400 3e6s; 11e6s were for dessert.)

 Andi47 2012-04-28 05:47

Has anyone started NFS on the c120 yet? I have just started, but I can stop it if anybody else hast started NFS before.

 Dubslow 2012-04-28 05:54

[QUOTE=Batalov;297717]My vote is for you. If I had a cent for each c120 I factored...
You OTOH have to collect your first dollar ;-)

(ECM was enough already when we finished 3e6s. gnfs-120 * 1/3 = t40, i.e. 2400 3e6s; 11e6s were for dessert.)[/QUOTE]I hadn't done any 3e6. (I have a script set up, so I knew what we were supposed to be aiming for; I figured you just like to overshoot, like I do :smile:)

[QUOTE=Andi47;297723]Has anyone started NFS on the c120 yet? I have just started, but I can stop it if anybody else hast started NFS before.[/QUOTE]
Go ahead. Wanna race? :wink::smile: (If so, I'd only be putting one core on it.)

 LaurV 2012-04-28 06:07

[QUOTE=Dubslow;297726]
Go ahead. Wanna race?[/QUOTE]I am sure that was a joke :D, it would make no sense HERE. Better exchange the poly with him and then your single core can contribute with some more relations, shortening the time.

 Dubslow 2012-04-28 06:09

[QUOTE=LaurV;297728]I am sure that was a joke :D, it would make no sense HERE. Better exchange the poly with him and then your single core can contribute with some more relations, shortening the time.[/QUOTE]

C130 only took three days :razz:

 Andi47 2012-04-28 08:33

I have put 6 cores on the c120 - should be done within one day

edit: estimated ETA 8-9 hours from now. I will post the factors either around midnight or tomorrow in the morning (central european time)

 Andi47 2012-04-28 22:19

[QUOTE=Andi47;297739]I have put 6 cores on the c120 - should be done within one day

edit: estimated ETA 8-9 hours from now. I will post the factors either around midnight or tomorrow in the morning (central european time)[/QUOTE]

factored (see factordb). gnfs on alq4788.3126 c115 in progress, I will post the factors in the morning (CEDT).

 jrk 2012-04-28 23:33

[QUOTE=Andi47;297808]factored (see factordb). gnfs on alq4788.3126 c115 in progress, I will post the factors in the morning (CEDT).[/QUOTE]

Line 3126 was already factored.

 Dubslow 2012-04-28 23:39

[QUOTE=Andi47;297808]factored (see factordb). gnfs on alq4788.3126 c115 in progress, I will post the factors in the morning (CEDT).[/QUOTE]

Dude, ECM. P40. Edit: Whoops, took my sweet time posting :razz:

Edit2: The C109 has stood up to a significant amount of ECM, so I'll just let yafu run NFS. Should be pretty quick.

 Batalov 2012-04-29 01:29

i3135 2^2 * c136
Why not 2^2 * p136 ? :razz:

 Dubslow 2012-04-29 02:41

[QUOTE=Batalov;297822]i3135 2^2 * c136
Why not 2^2 * p136 ? :razz:[/QUOTE]

Huh -- guess I just got unlucky with the ECM.

 Gimarel 2012-04-29 19:37

I had a lucky hit with ecm and got a p48.

 Dubslow 2012-04-29 21:38

The C123 has been ECMd through t35; I've now started 2K@12M.

 jrk 2012-04-29 21:55

[QUOTE=Dubslow;297903]The C123 has been ECMd through t35; I've now started 2K@12M.[/QUOTE]

p39: 718068333241085723448923908459684536389

 Dubslow 2012-04-29 22:38

We appear to have picked up a 2^2*17 guide/thingy; we're still (slowly) dropping.

 science_man_88 2012-04-29 22:59

[QUOTE=Dubslow;297914]We appear to have picked up a 2^2*17 guide/thingy; we're still (slowly) dropping.[/QUOTE]

to requote from: [QUOTE="http://www.mersenneforum.org/showpost.php?p=174259&postcount=20"]Define a guide to be [TEX]2^a[/TEX], together with a subset of the prime factors of [TEX]\sigma (2^a)[/TEX]. A driver is defined as a number [TEX]2^a v[/TEX]with a>0, v odd, [TEX]v| \sigma (2^a)[/TEX] and [TEX]2^{a-1} | \sigma (v)[/TEX] . The last requirement is included so that the power of the prime 2 will tend to persist at least as well as it does for the driver 2 itself, for which the condition is trivially satisfied.[/QUOTE] 2^2*17 does not work for this.

 axn 2012-04-29 23:09

[QUOTE=Dubslow;297914]We appear to have picked up a 2^2*17 guide/thingy; we're still (slowly) dropping.[/QUOTE]

Appearances can be deceiving. aka Coincidence :smile:

 Dubslow 2012-04-30 00:26

[QUOTE=science_man_88;297916]to requote from: 2^2*17 does not work for this.[/QUOTE]

[QUOTE=axn;297918]Appearances can be deceiving. aka Coincidence :smile:[/QUOTE]

Hence the "thingy"; I guess I really meant that the 17 is persisting across lines (Edit: Just like [URL="http://www.mersenneforum.org/showpost.php?p=174250&postcount=18"]schickel[/URL]), and that we are (incidentally) still going down. :razz: (Now that I've seen a concise definition of driver/guide, I won't misuse them again. :smile:)

Edit2:
Okay, a question and reality check, the latter first: $$\sigma\left(2^\alpha\right)/2=2^\alpha- 1$$, correct? (Derf thanks sm88)
And, with that definition, a driver requires that $$2^{\alpha-1}|\sigma(v)$$, but, if v is prime, then $$\sigma(v)$$ is 1, and then that condition doesn't hold, so 2^2*3 isn't a driver? What am I doing wrong?

 science_man_88 2012-04-30 01:23

[QUOTE=Dubslow;297927]Hence the "thingy"; I guess I really meant that the 17 is persisting across lines (Edit: Just like [URL="http://www.mersenneforum.org/showpost.php?p=174250&postcount=18"]schickel[/URL]), and that we are (incidentally) still going down. :razz: (Now that I've seen a concise definition of driver/guide, I won't misuse them again. :smile:)

Edit2:
Okay, a question and reality check, the latter first: $$\sigma\left(2^\alpha\right)=2^\alpha- 1$$, correct?
And, with that definition, a driver requires that $$2^{\alpha-1}|\sigma(v)$$, but, if v is prime, then $$\sigma(v)$$ is 1, and then that condition doesn't hold, so 2^2*3 isn't a driver? What am I doing wrong?[/QUOTE]

[CODE](22:05)>sigma(2^3)
%5 = 15
(22:17)>sigma(2^4)
%6 = 31
(22:18)>sigma(2^5)
%7 = 63
(22:18)>sigma(2^2)
%8 = 7
(22:18)>sigma(2^1)
%9 = 3[/CODE]

sigma is a sum of divisors but it can also be used as a sum of proper divisors and I believe you use it both ways, the sum of proper divisors of a prime is 1 the sum of divisors = prime+1. the sigma Pari uses is the sum of divisors, proper divisors don't include the number itself so we have to subtract that.

 Dubslow 2012-04-30 01:28

Derf, I managed to screw up the edit. So what I really meant was that $$\sigma(2^\alpha)=2^{\alpha+1}-1$$. Then for 2^2*3, alpha=2, and 2^1=2 | sigma(3)=4.

Also, I'm about halfway through bringing the C129 up to t40, so I'll let yafu run it up to t39.69 (a bit low perhaps) and start NFS. Somebody should probably run at least a few curves at a higher bound.

 Dubslow 2012-04-30 04:01

[QUOTE=Dubslow;297933]
Also, I'm about halfway through bringing the C129 up to t40, so I'll let yafu run it up to t39.69 (a bit low perhaps) and start NFS. Somebody should probably run at least a few curves at a higher bound.[/QUOTE]

Poly select started. (It will take a total of ~13 hours.)

 jrk 2012-04-30 17:12

I did 5000 curves @ B1=4e6, B2=11415205630 on line 3140. NF

 Dubslow 2012-05-01 23:45

Bizarre thing with sieving:
[code]found [U]9207128[/U] relations, need at least 9036311, proceeding with filtering ...
nfs: commencing msieve filtering
182406447322336820309560461032830305181092209591614813564635719276441105985665782458025218176998282272069123864041991524112937073

commencing relation filtering
estimated available RAM is 12019.7 MB
commencing duplicate removal, pass 1
error -15 reading relation 1566313
error -9 reading relation 3239082
found 3354104 hash collisions in 9207127 relations
added 55898 free relations
commencing duplicate removal, pass 2
[U]found 5904462 duplicates and 3358563 unique relations[/U]
memory use: 45.3 MB
reading ideals above 100000
commencing singleton removal, initial pass
memory use: 172.2 MB
reading all ideals from disk
memory use: 122.7 MB
keeping 6591825 ideals with weight <= 200, target excess is 20050
commencing in-memory singleton removal
begin with [U]3358563 relations[/U] and 6591825 unique ideals
reduce to 84 relations and 0 ideals in 5 passes
max relations containing the same ideal: 0
nfs: commencing algebraic side lattice sieving over range: 4054711 - 4094711[/code]

Since when are 2/3rds of the total rel count duplicates?
This also means the C129 might take longer than expected.

 Batalov 2012-05-02 00:10

[QUOTE=Dubslow;298124]Since when are 2/3rds of the total rel count duplicates?[/QUOTE]
This doesn't happen when done right. Have you stopped and restarted or any other details you may consider sharing? You need to get B[SUP]2[/SUP]'s help and debug.

 Dubslow 2012-05-02 01:00

[QUOTE=Dubslow;298124]Bizarre thing with sieving:
[code]<reproduced below>[/code]

Since when are 2/3rds of the total rel count duplicates?
This also means the C129 might take longer than expected.[/QUOTE]

[QUOTE=Batalov;298129]This doesn't happen when done right. Have you stopped and restarted or any other details you may consider sharing? You need to get B[SUP]2[/SUP]'s help and debug.[/QUOTE]

Yes, I had stopped and restarted a few times in poly select (not previously caused problems) and twice during previous sieving, hence the funny bounds.

[code]found [U]9207128[/U] relations, need at least 9036311, proceeding with filtering ...
nfs: commencing msieve filtering
182406447322336820309560461032830305181092209591614813564635719276441105985665782458025218176998282272069123864041991524112937073

commencing relation filtering
estimated available RAM is 12019.7 MB
commencing duplicate removal, pass 1
[U]error -15 reading relation 1566313
error -9 reading relation 3239082[/U]
found 3354104 hash collisions in 9207127 relations
added 55898 free relations
commencing duplicate removal, pass 2
[U]found 5904462 duplicates and 3358563 unique relations[/U]
memory use: 45.3 MB
reading ideals above 100000
commencing singleton removal, initial pass
memory use: 172.2 MB
reading all ideals from disk
memory use: 122.7 MB
keeping 6591825 ideals with weight <= 200, target excess is 20050
commencing in-memory singleton removal
begin with [U]3358563 relations[/U] and 6591825 unique ideals
reduce to 84 relations and 0 ideals in 5 passes
max relations containing the same ideal: 0
nfs: commencing algebraic side lattice sieving over range: 4054711 - 4094711
total yield: 535430, q=4254739 (0.01050 sec/rel)
[I]<The above is from my previous post, and here's what's happened since. Of course, it's only now that I notice errors both above and below.>[/I]
found [U]9742558 relations, need at least 9036311[/U], proceeding with filtering ...
nfs: commencing msieve filtering
182406447322336820309560461032830305181092209591614813564635719276441105985665782458025218176998282272069123864041991524112937073

commencing relation filtering
estimated available RAM is 12019.7 MB
commencing duplicate removal, pass 1
[U]error -15 reading relation 1566313
error -9 reading relation 3239082[/U]
found 3553246 hash collisions in 9798455 relations
added 392 free relations
commencing duplicate removal, pass 2
[U]found 6339694 duplicates and 3459153 unique relations[/U]
memory use: 53.3 MB
reading ideals above 100000
commencing singleton removal, initial pass
memory use: 172.2 MB
reading all ideals from disk
memory use: 126.4 MB
keeping 6707590 ideals with weight <= 200, target excess is 20329
commencing in-memory singleton removal
begin with 3459153 relations and 6707590 unique ideals
reduce to 84 relations and 0 ideals in 5 passes
max relations containing the same ideal: 0
nfs: commencing algebraic side lattice sieving over range: 4094711 - 4134711
[/code] It seems that it keeps reproducing the duplicate rels each time.

 Batalov 2012-05-02 02:28

I did not [I]suggest[/I] stopping and restarting -- this is not "Windows". :rolleyes:

(I may suspect that if you restart the process and the {equivalent of the .job} file was non existent, then the sieving could start from the initial q0 value and as a result all rels from the 2nd run will be duplications of the first one until the sievers will get to a yet unplowed range. But I don't know what yafu does. Ask Ben.)

Relations are not some magic smoke - if they are there, they will stay there, and of course if they were redundant, they will stay redundant. yafu (and msieve which is inside of it) don't rewrite the relations file with a non-redundant version of it. The errors -9 and -15 are jammed lines in the file; they are harmles per se, but they will be reported over and over at each filtering; no surprise here.

The amount of relations in your two passes seems growing, though. If you don't know what to do, do nothing - it may plow through all by itself.

 Dubslow 2012-05-02 02:47

[QUOTE=Batalov;298149]
(I may suspect that if you restart the process and the {equivalent of the .job} file was non existent, then the sieving could start from the initial q0 value and as a result all rels from the 2nd run will be duplications of the first one until the sievers will get to a yet unplowed range. But I don't know what yafu does. Ask Ben.)[/quote]This wasn't the case.
[code]bill@Gravemind:~/yafu∰∂ yafu "nfs(182406447322336820309560461032830305181092209591614813564635719276441105985665782458025218176998282272069123864041991524112937073)" [U]-R[/U]
>> nfs: checking for job file - number in job file matches input
nfs: checking for data file
nfs: commencing NFS restart
nfs: previous data file found - commencing search for last special-q
line 0 = -4916171347,5713:7310311,2c20d,d2adf,da19d:753557,3874f39,1afb,56371,6181f,8a50f,4cd,3727f7

line 1 = -5103316103,6761:64c7953,172f927,81013,cd477:147a0ab,6afc657,1f75,5273,6b59,126f7,ca29b,3727f7

line 2 = 789282735,1489:3055,8e15,c99d,36a61,fed3d:5bfd9b,68836bb,758b,18c7d,16ff33,3727f7

line 3 = 2007379955,883:17a9fdd,7fccd9,2b049,5963dd:1a6551d,58bdbbf,1525,276b,43a5,10ccd,14bf3,3727f7

nfs: parsing special-q
parsing rat side spq from 789282735,1489:3055,8e15,c99d,36a61,fed3d:5bfd9b,68836bb,758b,18c7d,16ff33,3727f7

found fed3d
parsing alg side spq from 789282735,1489:3055,8e15,c99d,36a61,fed3d:5bfd9b,68836bb,758b,18c7d,16ff33,3727f7

found 3727f7
parsing rat side spq from 2007379955,883:17a9fdd,7fccd9,2b049,5963dd:1a6551d,58bdbbf,1525,276b,43a5,10ccd,14bf3,3727f7

found 5963dd
parsing alg side spq from 2007379955,883:17a9fdd,7fccd9,2b049,5963dd:1a6551d,58bdbbf,1525,276b,43a5,10ccd,14bf3,3727f7

found 3727f7
parsing rat side spq from -4916171347,5713:7310311,2c20d,d2adf,da19d:753557,3874f39,1afb,56371,6181f,8a50f,4cd,3727f7

found da19d
parsing alg side spq from -4916171347,5713:7310311,2c20d,d2adf,da19d:753557,3874f39,1afb,56371,6181f,8a50f,4cd,3727f7

found 3727f7
nfs: commencing gnfs on c129: 182406447322336820309560461032830305181092209591614813564635719276441105985665782458025218176998282272069123864041991524112937073
nfs: found 3239081 relations, continuing job at specialq = 3614711
found 3239081 relations, need at least 9036311, continuing with sieving ...
nfs: commencing algebraic side lattice sieving over range: 3614711 - 3654711
[/code]
The low end of the q range is in the middle of the q range it was working on when I stopped, so I presume it recorded where it stopped and was not doing duplicate work.
[QUOTE=Batalov;298149]
Relations are not some magic smoke - if they are there, they will stay there, and of course if they were redundant, they will stay redundant. yafu (and msieve which is inside of it) don't rewrite the relations file with a non-redundant version of it.[/quote]So what you're saying is that it doesn't remove the duplicates from the "total"? Why shouldn't that be done? [QUOTE=Batalov;298149] The errors -9 and -15 are jammed lines in the file; they are harmles per se, but they will be reported over and over at each filtering; no surprise here.
[/quote]Okay, I didn't know the errors are persistent/will appear each time. Either way though, they still happened.[QUOTE=Batalov;298149]
The amount of relations in your two passes seems growing, though. If you don't know what to do, do nothing - it may plow through all by itself.[/QUOTE]Yes, I suppose so. I'll try not to mess with it :razz:

I suppose I should also mention that I had set "qintsize: 300000" in nfs.job, and then later changed it to "qintsize: 200000". Knowing how these things work, this is probably the source of the problem, isn't it?

Edit: Third round of filtering, which picks up directly after the previous post's code block.
[code]nfs: commencing algebraic side lattice sieving over range: 4094711 - 4134711
total yield: 532657, q=4294723 (0.01082 sec/rel)
found 10275215 relations, need at least 9036311, proceeding with filtering ...
nfs: commencing msieve filtering
182406447322336820309560461032830305181092209591614813564635719276441105985665782458025218176998282272069123864041991524112937073

commencing relation filtering
estimated available RAM is 12019.7 MB
commencing duplicate removal, pass 1
error -15 reading relation 1566313
error -9 reading relation 3239082
found 3778211 hash collisions in 10331504 relations
added 340 free relations
commencing duplicate removal, pass 2
found 6776760 duplicates and 3555084 unique relations
memory use: 53.3 MB
reading ideals above 100000
commencing singleton removal, initial pass
memory use: 188.2 MB
reading all ideals from disk
memory use: 130.0 MB
keeping 6816144 ideals with weight <= 200, target excess is 20618
commencing in-memory singleton removal
begin with 3555084 relations and 6816144 unique ideals
reduce to 84 relations and 0 ideals in 5 passes
max relations containing the same ideal: 0
nfs: commencing algebraic side lattice sieving over range: 4134711 - 4174711[/code]

 Batalov 2012-05-02 03:03

You will need Ben. -R parsing seems ok-ish if a bit verbosiousiousious.

Could be the parameters to blame. If the lims or bits are too low, then you will never get enough relations. (if it were just the "run once and then run the same over again", that would be a lesser weevil - it would have righted itself, but if the parameters are off, then the siever will just tread water.)

 firejuggler 2012-05-02 03:04

did you start with a certain number of thread then reduced it after? ( by modifi-ing factMsieve?=

 bsquared 2012-05-02 12:20

[QUOTE=Dubslow;298151]
I suppose I should also mention that I had set "qintsize: 300000" in nfs.job,[/QUOTE]

This is the source of the problem... see the yafu bugs thread. If you stop, remove that line, and restart again, you should relatively breeze through to the finish.

 Dubslow 2012-05-03 21:04

LinAlg finally started. ETA 3.5hrs.

 Dubslow 2012-05-04 00:51

[QUOTE=Dubslow;297927]Hence the "thingy"; I guess I really meant that the 17 is persisting across lines [/QUOTE]

Hmm, I guess now I see that isn't really true either. :smile:

C131 i3143 is almost at t30. The C129 was a 61*68 split.
Edit: I'll take the C131 to t35 and then I'm outta here. (TF2!)

 Batalov 2012-05-04 05:32

It is ready for NFS. Don't ask, don't tell.

 LaurV 2012-05-04 06:00

[QUOTE=Dubslow;298359](TF2!)[/QUOTE]Bhwaa haha, you got me, I googled it, expecting to find some sort of math exam, I was thinking to second semester, third or forth term session of exams, this kinda variation, like you say "calculus 2" or "geometry 3", and imagine you are going to read/learn. Never imagined you can leave factoring alq4788 to go play a game...:razz:

 schickel 2012-05-04 06:47

[QUOTE=Batalov;298393]It is ready for NFS. Don't ask, don't tell.[/QUOTE]Does that mean you're doing it, or shall I?

 Dubslow 2012-05-04 06:52

[QUOTE=LaurV;298396]Bhwaa haha, you got me, I googled it, expecting to find some sort of math exam, I was thinking to second semester, third or forth term session of exams, this kinda variation, like you say "calculus 2" or "geometry 3", and imagine you are going to read/learn. Never imagined you can leave factoring alq4788 to go play a game...:razz:[/QUOTE]

The problem is, nobody makes games for Linux, and the only thing I have set up in Windows is Prime95.

 Batalov 2012-05-04 07:20

[QUOTE=schickel;298404]Does that mean you're doing it, or shall I?[/QUOTE]
"No time for love, Dr. Jones."

(Meaning: You.)

 schickel 2012-05-04 10:39

[QUOTE=Batalov;298408]"No time for love, Dr. Jones."

(Meaning: You.)[/QUOTE]Aye, Cap'n! I'm givin' it all she's got. I canna git more than warp 8....

 akruppa 2012-05-05 13:57

C131 got factored somehow, we're now at i3155 c123. Edit: started NFS.

 schickel 2012-05-05 15:39

[QUOTE=akruppa;298537]C131 got factored somehow, we're now at i3155 c123.[/QUOTE] :grumble: And I was [I]almost[/I] there....

 akruppa 2012-05-05 22:39

It got factored. Now i3168 c128.

 akruppa 2012-05-06 15:10

Now i3184 c105. Edit: did p35.

 akruppa 2012-05-06 18:19

Done, [strike]c123[/strike]c125 now.

 akruppa 2012-05-07 16:35

Well, well, well. What have we here?

 Dubslow 2012-05-07 16:43

[QUOTE=akruppa;298693]Well, well, well. What have we here?[/QUOTE]

Arggggg

Btw, I'm doing NFS on the C101.

 Dubslow 2012-05-07 18:26

[QUOTE=akruppa;298693]Well, well, well. What have we here?[/QUOTE]

Wow. It only lasted 3 lines.

 jrk 2012-05-08 04:13

found a nice p50 for line 3241
[code]Using B1=4000000, B2=11415205630, polynomial Dickson(12), sigma=4391922847262760571
Step 1 took 10625ms
Step 2 took 7114ms
********** Factor found in step 2: 45416124634863473662413395025713212988554679679571
Found probable prime factor of 50 digits: 45416124634863473662413395025713212988554679679571
Probable prime cofactor 25508576380978103091317282772459882271991864252304997022734869257428350983 has 74 digits
[/code]

 Batalov 2012-05-08 05:10

[QUOTE=Dubslow;298703]Wow. It only lasted 3 lines.[/QUOTE]
What only lasted 3 lines? The suspense is killing me! :-)

 Raman 2012-05-08 06:51

[QUOTE=Batalov;298752]What only lasted 3 lines? The suspense is killing me! :-)[/QUOTE]

[CODE]9661· c132 = [B]2[SUP]2[/SUP] · 3[/B] · 11 · 31 · 353 · 613 · 14387 · 643178369650376017 · 991644996925961632533902637367445413049636900849 · 55323601517800262358753763492036420403942500981785229
9662· c132 = [B]2[SUP]2[/SUP] · 3[/B] · 11 · 1229 · 13877 · 27376303 · 11957822966056413250553273496383084681889255122857774658405664459834559776367233827892118406177333463665964452139181
9663· c133 = [B]2[SUP]2[/SUP] · 3[/B] · 27553321 · 431044261 · 3637509037 · 157768746359767876402363891 · 13945548933403944721096352866731110525584844156577476991306762091091154176068813[/CODE]

 schickel 2012-05-08 07:33

[QUOTE=Raman;298756]ERROR: Stack overflow -- too many frickin' [noparse][CODE][/noparse] tags.[/QUOTE]Was that on purpose?[QUOTE=Batalov;298752]What only lasted 3 lines? The suspense is killing me! :-)[/QUOTE]The '3' in 2^2 * 3 only lasted 3 lines.....3201-3203 inclusive.

 Batalov 2012-05-08 07:38

That message was as important as it was artful. :-)
But something was definitely missing -- I am shocked by the absense of color tags.

Rhetorical question, hello?

 LaurV 2012-05-08 07:51

And why should this be curious? Or magnificent? With even powers of 2, usually, 3 is not stable, it is coming and going. That is because no mersenne with odd exponent is divisible by 3. The case we should worry about is a 3 with an odd power of two, as all mersenne with even expos are divisible by 3. See for example 2^1*3, 2^3*3 (with/without *5), or 2^5*3(at ^1 or ^2, with or without 7) etc.

Examples (from my reserved sequences): [URL="http://factorization.ath.cx/sequences.php?se=1&eff=2&aq=225900&action=last20&fr=0&to=100"]225900[/URL], [URL="http://factorization.ath.cx/sequences.php?se=1&eff=2&aq=865152&action=last20&fr=0&to=100"]865152[/URL], [URL="http://factorization.ath.cx/sequences.php?se=1&eff=2&aq=996666&action=last20&fr=0&to=100"]996666[/URL], etc.

 schickel 2012-05-08 08:16

[QUOTE=Batalov;298758]Rhetorical question, hello?[/QUOTE]Goodbye?

 Andi47 2012-05-08 09:15

[QUOTE=LaurV;298759]And why should this be curious? Or magnificent? With even powers of 2, usually, 3 is not stable, it is coming and going. That is because no mersenne with odd exponent is divisible by 3. The case we should worry about is a 3 with an odd power of two, as all mersenne with even expos are divisible by 3. See for example 2^1*3, 2^3*3 (with/without *5), or 2^5*3(at ^1 or ^2, with or without 7) etc.

Examples (from my reserved sequences): [URL="http://factorization.ath.cx/sequences.php?se=1&eff=2&aq=225900&action=last20&fr=0&to=100"]225900[/URL], [URL="http://factorization.ath.cx/sequences.php?se=1&eff=2&aq=865152&action=last20&fr=0&to=100"]865152[/URL], [URL="http://factorization.ath.cx/sequences.php?se=1&eff=2&aq=996666&action=last20&fr=0&to=100"]996666[/URL], etc.[/QUOTE]

...or 500+ lines of the 2*3 driver in sequence 10212. (sometimes mutated to 2*3^n with n>1 for a few lines, but never getting rid of the driver)

 LaurV 2012-05-08 11:24

Exactly. As long as it doesn't get a seven, 2^2 is a beauty! :smile: No matter if 3 or not 3.

 akruppa 2012-05-08 12:16

No, n=2^2*3 is not a driver, but the 3 tends to linger a while (until no 2 (mod 3) prime factor to odd exponent appears in an interation) and σ(n)/n-1=4/3 so you usually gain a digit or two when a 3 appears.

 akruppa 2012-05-08 13:35

Is someone doing NFS on i3242 c124?

Edit: thanks, Ben.

 bsquared 2012-05-08 13:58

[QUOTE=akruppa;298784]Is someone doing NFS on i3242 c124?[/QUOTE]

I'll start on it.

 bsquared 2012-05-08 22:29

Did the c124, a c117, and then another c124, along with assorted other stuff. Stepping out now on i3247 with a c119 that I haven't tested at all.

 akruppa 2012-05-08 22:34

Sweet. Thanks for helping this along.

 Batalov 2012-05-09 06:52

{sarcasm}
OMG, it's already there for three iterations. Last time, it was stuck for six!!*
{/sarcasm}

*..and don't even ask, it moves the sequence UP [SPOILER] = 2^2 · 5[/SPOILER]

All times are UTC. The time now is 10:48.