fun with msieve

[michaf]

Taking 341552*5^146-1 for some snfs practice

[Phil MjX]

Originally Posted by michaf

I knew these were the easy ones, but you have to start somewhere :)
Sieving nearing the end now.

The reason why I wanted to plunge into snfs, is that I have a ton of composites, all from cyclic Smarandache numbers (or reverse), and I wanted to see if these were snfs-able.

I have read someplace that someone had done a reverse smarandache with snfs, but I can't find where anymore :(

The type of numbers I'd love to do are:

smarandache: 123456789101112...
reverse smarandache: ...121110987654321
cyclic smarandache : 4567891011...9899123
reverse cyclic smarandache: 1110987654321999897...141312

No idea whether these are even doable with snfs, but spelled decimally, they are easily expressed.
If anyone has any clues as to wether it is possible to get poly's, please give me a few hints, and I'll go delve some deeper.

For now, gnfs will still keep me busy for a year to come (If only snfs could make that 1/4 year :> )

Hi !

I know there is a way to find snfs polys for smarandache numbers.
Sean has done it : http://www.worldofnumbers.com/factorlist.htm
but I don't know how many of these composites have been factorized by him yet.

quote "Further, after studying Backstrom's work on Rsm76 I am now able to generate
SNFS polynomials for all the remaining Sm numbers below 100. It would have
been much faster to do Sm78 by SNFS, but I had already started it before
working out how to apply SNFS to the number. It should be possible to complete
all values up to Sm(100) by SNFS, although a few will be quite difficult runs."


finally quoting Bob (ggnfs mailing list, the link you have pointed) :
"This is not one of your "serious" xyyxf numbers - it is one of the
Reverse Smarandache numbers. In fact, Rsm76 which is the number
76757473...1110987654321 - BUT amenable to GGNFS attack.

If you multiply this number by 9801 (99^2), you get the interesting
pattern:

752300000000000000000000000000000000000000000000000000000000000000
000000000000000000000000000000000000000000000000000000000000000000
008790000000121

which is then obviously: 7523 * 10^143 + 8790000000121.

Just right for SNFS! (Pity about the large constant coeff.)"

Hope this helps further.

Best regards

philippe Strohl

[michaf]

Yes, Philippe, thats about what I meant.

In the meantime, I'm wondering what should be preferred in terms of coefficient-size:

for example: 341552*5^146-1 (108 digits)
I can get c4=8538800 and c0=-1
or c4=341552 and c0=-25

In my initial tests they don't differ too much, but a bit later on, the first seems to get slower.
Is there a way to get an educated guess on which is better, or is c4*c0 the measure and they should be about the same speed?

[michaf]

If people are still doing these numbers gnfs-style,
snfs is about 9 times quicker then gnfs!

My average c108 takes 15.24 hours

Number: 341552_5_146m1
N=382893034469655936166513349305469947576828821417446171900136149199300739098816848127171397209167480468749999
( 108 digits)
SNFS difficulty: 107 digits.
Divisors found:
r1=542728113867830244067138216512326913122257037 (pp45)
r2=705496960054111545782648015642403172848460309065685675669130027 (pp63)
Version: GGNFS-0.77.1-20060513-athlon-xp
Total time: 1.65 hours.

BTW Masser, I seem to be missing the sierpinski candidates from about ^142 to ^172
Do they happen to be all factored, or something went wrong?

I'll be reserving 49568*5^148-1 and 151026*5^149-1

[masser]

Originally Posted by michaf

BTW Masser, I seem to be missing the sierpinski candidates from about ^142 to ^172
Do they happen to be all factored, or something went wrong?

I believe they have all been factored, but I am not positive. All of the recent dat files have had a gap in candidates there.

EDIT: Starting from scratch, I just factored all but two (the n=142 and n=144 above) Sierpinski candidates with 140 <= n <= 170. The last factor was found with ECM and was only 25 digits. So previous factoring attempts (there was a thread devoted to ECM factoring some time ago) must have created the gap between n=144 and n=172.

[tnerual]

Originally Posted by masser

I believe they have all been factored, but I am not positive. All of the recent dat files have had a gap in candidates there.

maybe robin benson sieved them ! he have a very high score in sieving but a low number of factors ...

[masser]

329584*5^167-1 has a factor: 5241057957852627561388071314039
143632*5^227-1 has a factor: 39380516835266032923819254377
114986*5^234-1 has a factor: 5097636932117932459931728843
155056*5^293-1 has a factor: 101097048320526348991729337
205252*5^307-1 has a factor: 110445601445085491300123
76322*5^308-1 has a factor: 7727473801031924267
104944*5^311-1 has a factor: 79729296171006575373037591
109862*5^312-1 has a factor: 14428713366621346757
22934*5^312-1 has a factor: 6988885143763825087991
206894*5^316-1 has a factor: 2603362749054918005269
63838*5^327-1 has a factor: 275562839414284746497
325918*5^327-1 has a factor: 179645473089901164798541
98038*5^333-1 has a factor: 353205211253651802011
109838*5^334-1 has a factor: 15813457908908687623464607
71146*5^349-1 has a factor: 763903809065288092895949683
170446*5^363-1 has a factor: 161269426616459344710431
267298*5^365-1 has a factor: 1360525285548580770217
35816*5^366-1 has a factor: 127843758973780683109
190468*5^369-1 has a factor: 3141538091184105926063
45742*5^371-1 has a factor: 320921721264702801389807
58882*5^381-1 has a factor: 10942852665279604928611
35816*5^386-1 has a factor: 381777473044414476307
194368*5^389-1 has a factor: 62994924827984419087
70082*5^392-1 has a factor: 17808578538039401533
245114*5^396-1 has a factor: 1522698524372299551210416557079
159388*5^399-1 has a factor: 23131333649190044628391123

Finished 210 ECM curves at 25 digit level:
5374*5^217-1 has a factor: 185177445213802226446514023
282316*5^323-1 has a factor: 18310907376007652351411
105464*5^313+1 has a factor: 6769055423154911818818569
59912*5^285+1 has a factor: 37868416470466955581862833
282316*5^151-1 has a factor: 1374473098898080022419909213
155056*5^269-1 has a factor: 493732343824713867325957181
162668*5^312-1 has a factor: 3084883048607561252740138889
104944*5^319-1 has a factor: 23976057507966204858652489

[ltd]

I lost track a little bit and i do not have the time to find everything out on my own.

Which results have been returned to sieve submission and which not?


Cheers,

Lars

[masser]

I believe the factors found by Andi_HB and fivemack are the only factors that have not been submitted yet. They have never submitted results, so they won't be able to use the sieveimport page.

[fivemack]

Originally Posted by CedricVonck

By following your instructions I came up with following poly file

n: 74229775530521455488007758492787522898438620227229981378006965553757821319102732125294955012329012333793798461556434631347656251
type: snfs
skew: 1
c4: 71098
c0: 1
Y0: 5684341886080801486968994140625
Y1: -1

GGNFS ran smoothly but then crashed when processing the relations?
Can anyone enlighten me?

Thx

I'm not certain about this, but it might help to negate both Y0 and Y1; I think the ggnfs relation-processing code might prefer Y1 to be positive. I saw a couple of crashes like this when trying to figure out the polynomials.

[fivemack]

I've found the large constant (or X^5 for the non-reverse case) coefficient distinctly troublesome for the Smarandache numbers; I set up for an attack on Sm94 thinking it would be a matter of a couple of weeks (180 digits SNFS), but couldn't find parameters which came up as less than several months relation-collection; it seems maybe even a little harder than Fib(1009), which was 210 digits but with a really convenient polynomial.