10+ table

[Raman]

Originally Posted by bsquared

10,236+ factors as:

 
prp68 factor: 32394937980403032343941480318880086318307142570248235727782772042697
prp125 factor: 36837015057313006096634536795587748230336921494253255771361450341226561526611901358724245001042863132758337028749847212746321

After about 15k cpu hours sieving using gnfs-lasieveI15e and 265 hrs of linear algebra using msieve 1.38 on 4 cpus on a 9.1M square matrix.

- ben.

So, I doubt if Prof. Sam received your mail on 10,236+
Still, it is not yet added up in his webpage, and 2,836+ and
7,391+ (a new candidate) are already included.

Try sending again?

[bsquared]

Originally Posted by Raman

So, I doubt if Prof. Sam received your mail on 10,236+
Still, it is not yet added up in his webpage, and 2,836+ and
7,391+ (a new candidate) are already included.

Try sending again?

I would be surprised if he had... I just sent it an hour ago ;)

[Raman]

Oh, that means that you didn't send immediately after you posted
in the forum.
So, we have to wait for some more time for the entry.

So, I hope that the factor base limits should be ~ 40 million. Right?

[bsquared]

I used larger FB limits, similar to others I've done with difficulty ~ 240. I didn't spend a lot of time determining if they were optimal.

rlim: 85000000
alim: 75000000
lpbr: 30
lpba: 30
mfbr: 60
mfba: 60
rlambda: 2.6
alambda: 2.6

[Batalov]

10,393+ (diff.262) c253 splits as p119 . p134
Not a <snip>... well, this joke is overused these days.

p119 = 22470645744200057762885095342697894721605325430609487291715500041029950763944163993319007373686738769124162721892380653

Sieved on both sides (but more on -r) to a good excess by Bruce Dodson at lehigh.edu with 15e and finished with msieve 1.42 (with a w<=30 tweak); the matrix was 13.5M and took 17 days to solve on a Phenom X2 940 (4threads). Log is attached.

What was unusual was that even with very many (>60) restarts (due to unstable memory controller in the summer heat; many discussion boards say that one can use 1066MHz or 8Gb of memory but not both with a Phenom) on a home computer, it was possible to just let the BL run in a bootstrapped loop with periodic restarts from an earlier checkpoint as soon as orthogonality test failed. When after 12 days (and ~60 restarts, all of them between 4 and 6pm, at the hottest time of the day), I finally gave in and set the 1066MHz 5-5-5-15 memory to work as 800MHz 4-4-4-12 memory -- the BL went entirely uninterrupted (albeit 15% slower) until the end. Note that it wasn't 33% slower (1066/800); 4-4-4 somewhat helps.

sqrt was done with the latest SVN 40 version of msieve and seemed very fast for this size (2 hours 20 min only, solved on the 1st dependency).

Thanks to Bruce ...and to Jason for the great additions to the msieve code (orthogonality checks with prompt bailout).

B+D snfs

[bsquared]

Congrats, that's a monster!

And beats my personal record of a P116 cofactor.

Nice job!

[Batalov]

10,241+ (diff.241) c175 splits by snfs as p62 . p114
Not a miss, because it was done to t55+ but surely not t60.

B+D

[Andi_HB]

I try to factorize 10,286+ c161
The polynomial search was done im 5 days - now runs the sieving till some days.
I try to complete this factorization on my home PC - hope it will work.
In use is the 15e siever with this polynomial - beginning at Q 25000000.

n: 15764689982344536517771021186273406748173928006967550504357382938426798303527007683249463226418352501943599358801522375151414371266800385235667781728011088393709
skew: 10887480.70
Y0: -13968842428054971532371343005271
Y1: 406095970085689673
c0: 6396747705471302698957036275292308554400
c1: 238115329123966363474521251154152
c2: -1278255642788702958890606164
c3: 40617525259704032030
c4: 12074212525095
c5: 29640
rlim: 60000000
alim: 26000000
lpbr: 30
lpba: 30
mfbr: 60
mfba: 60
rlambda: 2.65
alambda: 2.65

Hope i can tell about access in ~4month

[bdodson]

Originally Posted by Andi_HB

I try to factorize 10,286+ c161
The polynomial search was done im 5 days - now runs the sieving till some days.
...
Hope i can tell about access in ~4month

Nice choice, Serge has this as:

161     10      286     +       260     0.619231        /11

so with ratio .619 << .70. Also a worthy computation,
better than my patience. -Bruce

PS - You've sent off a reservation with Sam?

[Andi_HB]

I have send Sam an E-Mail for Reservation now.

[bdodson]

Originally Posted by Batalov

10,241+ (diff.241) c175 splits by snfs as p62 . p114
Not a miss, because it was done to t55+ but surely not t60.

B+D

What is it with this 10+ list? We're happy to report the resumption of
B+D numbers, from Serge's email

10,244+ cofactor splits as c215 = p56 . p159
 
p56 factor: 89376848303370085832547257811113680017983140270444916137
p159 factor: 441160056784956066935585043996395720339746991856908200056965652272217569686164467898219919069368304598422319868082307673772600748575192615252243347701807248793
Not quite a miss. Borderline.

First, this is a rather large p56, nearer p57 than p55. These numbers
are (have been) below the diff 250 range where NFS@Home has been (is)
working. I've been restraining myself from testing past 7t50 in this range.
So ecm had it's full chance, to 62%, to find this factor; but well below
the effort to remove a p55 (to 80%, at 2t55 = 11t50). None of the numbers
above difficulty 250, especially those in 255-270, have been tested
below 8t50, with a target of removing these p55/p56's.

Actually, I'm somewhat happier to see this prime factor than the
p90*p_large and p100*p_large for almost half of the NFS@Home
numbers. Even p75*p_large and p80*p_large are mostly hopeless
as ecm targets, so finding the occasional prime in [p56, p65] by
sieving is at least an indication of some hope of finding a factor.
I'd have been happier yet to have found this one by ecm, but I'd
gladly trade this one (somewhat late) for one of those early p60's.
As one expects from ecm, over time.

-Bruce (for B+D).