c186s done
 

[SIZE=2](1487539^31-1)/(1487539-1)[/SIZE]
[CODE]prp89 factor: 65279918221169399842005834890844295717403144110010131583479279256157017712932874292215947
prp97 factor: 2287028170875584392066157958304694037516156360566692371116408306749319795963181125166558345423983
[/CODE][SIZE=2](1583653^31-1)/(1583653-1)[/SIZE]
[CODE]prp51 factor: 641249890018878631415427212389248712887669613706791
prp136 factor: 1523259184413435411061904735610857799990838471854734281961077989150231222015971385272325352613904802697413791519755807916139198382396053
[/CODE](factordb updated)

(5087^53-1)/(5087-1) done:

r1=361855304806180224028379449770289081354201995026046508773679154343710540029506320008347284602637 (pp96)
r2=15050841937328234009408402107798176885715540226192635023658979555897956767195730045269338456936453 (pp98)

Chris

[QUOTE=Dubslow;312450]I've decided that the ECM I cited is sufficiently boring as to be put on hold and interject with doing one of these. As such, I'll do [URL="http://www.mersenneforum.org/showthread.php?p=311744#post311744"](2957^53-1)/(2957-1) C181[/URL]. [/QUOTE]

Done.
[code]PRP58 = 2382612345133882678285245521423097247041686799500841597461
PRP124 = 1280449286529161765644917223654085148188166236518860398623399516700591892697984896248877872831603977217311135165628898320441[/code]
The C15* should be done soon.

[QUOTE=Dubslow;312970]The C15* should be done soon.[/QUOTE]

Too soon, it seems.

I'll start on [URL="http://www.mersenneforum.org/showthread.php?p=311744#post311744"](5011^47-1)/(5011-1) C171[/URL].

This should be a quintic, right?
[code]c5: 25110121 # 5011^2
c0: -1
m: 1992138940215095135188602407592691 # 5011^9
skew: 30.197414[/code]

Dubslow,

Using your quintic: SNFS difficulty 174 ~1.2e7 relations needed
[CODE]Trial sieving with [TEX]\left(\frac{c_n}{c_0}\right)^{\frac{1}{n}}[/TEX]:
Skew: 30.197414
total yield: 54, q=2800103 (0.20022 sec/rel)[/CODE][CODE]Trial sieving with [TEX]\left(\frac{c_0}{c_n}\right)^{\frac{1}{n}}[/TEX]:
Skew: 0.03311541
total yield: 207, q=2800079 (0.03973 sec/rel)[/CODE]Using a sextic: SNFS difficulty 178 ~1.4e7 relations needed
[CODE]Trial sieving with [TEX]\left(\frac{c_0}{c_n}\right)^{\frac{1}{n}}[/TEX]:
skew: 4.13670
total yield: 238, q=3250099 (0.03560 sec/rel)[/CODE][CODE]Trial sieving with [TEX]\left(\frac{c_n}{c_0}\right)^{\frac{1}{n}}[/TEX]:
skew: 0.2417385
total yield: 126, q=3250081 (0.06119 sec/rel)[/CODE]Note:
Use absolute values for the coefficients

Please try on your machine.

[QUOTE=Mathew;313004]Dubslow,

Using your quintic: SNFS difficulty 174 ~1.2e7 relations needed
[CODE]Trial sieving with [TEX]\left(\frac{c_n}{c_0}\right)^{\frac{1}{n}}[/TEX]:
Skew: 30.197414
total yield: 54, q=2800103 (0.20022 sec/rel)[/CODE][CODE]Trial sieving with [TEX]\left(\frac{c_0}{c_n}\right)^{\frac{1}{n}}[/TEX]:
Skew: 0.03311541
total yield: 207, q=2800079 (0.03973 sec/rel)[/CODE]Using a sextic: SNFS difficulty 178 ~1.4e7 relations needed
[CODE]Trial sieving with [TEX]\left(\frac{c_0}{c_n}\right)^{\frac{1}{n}}[/TEX]:
skew: 4.13670
total yield: 238, q=3250099 (0.03560 sec/rel)[/CODE][CODE]Trial sieving with [TEX]\left(\frac{c_n}{c_0}\right)^{\frac{1}{n}}[/TEX]:
skew: 0.2417385
total yield: 126, q=3250081 (0.06119 sec/rel)[/CODE]Note:
Use absolute values for the coefficients

Please try on your machine.[/QUOTE]

I had no idea the skew could be more than one value. How do you know which is right, besides trial sieving?

[code]total yield: 53, q=2800103 (0.06698 sec/rel)

bill@Gravemind:~/yafu/SNFS/trunk∰∂ cat nfs.job
n: 157256716361364298776164081109168157704586591070408028843581358832587249647653648659697514795580530588448260948401612520796605607658503832066063053924318509732043555092357
c5: 25110121 # 5011^2
c0: -1
m: 1992138940215095135188602407592691 # 5011^9
skew: 30.197414
type: snfs
rlim: 5500000
alim: 5500000
lpbr: 27
lpba: 27
mfbr: 54
mfba: 54
rlambda: 2.5
alambda: 2.5[/code]
[code]total yield: 242, q=2800103 (0.01450 sec/rel)

bill@Gravemind:~/yafu/SNFS/trunk∰∂ cat nfs.job
n: 157256716361364298776164081109168157704586591070408028843581358832587249647653648659697514795580530588448260948401612520796605607658503832066063053924318509732043555092357
c5: 25110121 # 5011^2
c0: -1
m: 1992138940215095135188602407592691 # 5011^9
skew: 0.03311541
type: snfs
rlim: 5500000
alim: 5500000
lpbr: 27
lpba: 27
mfbr: 54
mfba: 54
rlambda: 2.5
alambda: 2.5[/code]
Why the different sq for the sextic?
[code]total yield: 158, q=3250109 (0.02038 sec/rel)

bill@Gravemind:~/yafu/SNFS/trunk∰∂ cat nfs.job
n: 157256716361364298776164081109168157704586591070408028843581358832587249647653648659697514795580530588448260948401612520796605607658503832066063053924318509732043555092357
c6: 1
c0: -5011
m: 397553171066672347872401198881 # 5011^8
skew: 4.13670038868
type: snfs
rlim: 5500000
alim: 5500000
lpbr: 27
lpba: 27
mfbr: 54
mfba: 54
rlambda: 2.5
alambda: 2.5[/code]
[code]total yield: 101, q=3250109 (0.03089 sec/rel)

bill@Gravemind:~/yafu/SNFS/trunk∰∂ cat nfs.job
n: 157256716361364298776164081109168157704586591070408028843581358832587249647653648659697514795580530588448260948401612520796605607658503832066063053924318509732043555092357
c6: 1
c0: -5011
m: 397553171066672347872401198881 # 5011^8
skew: 0.24173856
type: snfs
rlim: 5500000
alim: 5500000
lpbr: 27
lpba: 27
mfbr: 54
mfba: 54
rlambda: 2.5
alambda: 2.5[/code]
I guess I still have a lot to learn about SNFS.

[QUOTE=Dubslow;313010]
Why the different sq for the sextic?
[code]total yield: 158, q=3250109 (0.02038 sec/rel)

bill@Gravemind:~/yafu/SNFS/trunk∰∂ cat nfs.job
n: 157256716361364298776164081109168157704586591070408028843581358832587249647653648659697514795580530588448260948401612520796605607658503832066063053924318509732043555092357
c6: 1
c0: -5011
m: 397553171066672347872401198881 # 5011^8
skew: 4.13670038868
type: snfs
rlim: 5500000
alim: 5500000
lpbr: 27
lpba: 27
mfbr: 54
mfba: 54
rlambda: 2.5
alambda: 2.5[/code]
[code]total yield: 101, q=3250109 (0.03089 sec/rel)
[/QUOTE]
Have you tried sieving the algebraic side on the sextic?

[QUOTE=axn;313016]Have you tried sieving the algebraic side on the sextic?[/QUOTE]

No, but I'll gladly try. How do you determine when algebraic might be better for SNFS? And while we're at it, how do you know it cn/c0 or c0/cn should be the skew? (Or is that trial-only?)

[QUOTE=Dubslow;312970]The C15* should be done soon.[/QUOTE]

[QUOTE=Dubslow;312995]Too soon, it seems.[/QUOTE]
[code]PRP104 = 24368901202433510671499255871775679719188986024478308679997354317057833986308168190236540683244839782291
PRP54 = 273792115925621996806134435690492438372348877550882851
[/code]
That makes me three for three in having factors of the form P5- * P1-- (the other two were P54*P103 and P58*P124).


Sextic/algebraic: strange. The range of 100 gives literally zero rels, but a range of 500 was faster than anything else.
[code]syscmd: /home/bill/yafu/ggnfs/gnfs-lasieve4I13e -a nfs.job -f 3250000 -c 500 -o rels0.dat -n 0
total yield: 708, q=3250517 (0.01815 sec/rel)
...
syscmd: /home/bill/yafu/ggnfs/gnfs-lasieve4I13e -a nfs.job -f 3250000 -c 100 -o rels0.dat -n 0
total yield: 0, q=3250109 (nan sec/rel)[/code]

In an extended test sieve, however, the rational/quintic still comes out slightly faster.
[code]syscmd: /home/bill/yafu/ggnfs/gnfs-lasieve4I13e -a nfs.job -f 3250000 -c 1000 -o rels0.dat -n 0
total yield: 1642, q=3251009 (0.01773 sec/rel)
...
syscmd: /home/bill/yafu/ggnfs/gnfs-lasieve4I13e -r nfs.job -f 2800000 -c 1000 -o rels0.dat -n 0
total yield: 2203, q=2801003 (0.01650 sec/rel) [/code]

[QUOTE=Dubslow;313023]And while we're at it, how do you know it cn/c0 or c0/cn should be the skew? [/QUOTE]

It is always (c0/cn)^(1/n). [ignoring the signs of c0 & cn]

[QUOTE=axn;313029]It is always (c0/cn)^(1/n). [ignoring the signs of c0 & cn][/QUOTE]

Hmm... pinhodecarlos' files show it [STRIKE]either way[/STRIKE] the other way once:
[code]~/yafu/SNFS∰∂ grep abs *
exponent 11.poly:skew: abs(c0/c5)^(1/degree)
exponent 29.poly:skew: abs(-base/c5)^(1/degree)
exponent 43.poly:skew: abs(c0/c6)^(1/degree)
exponent 53.poly:skew: abs(c6/c0)^(1/degree)
exponent 61.poly:skew: abs(c0/c5)^(1/degree)[/code]

And RDS [URL="http://www.mersenneforum.org/showthread.php?p=266499#post266499"]said[/URL] sometimes the inverse depending on the code, and as Mathew demonstrated, GGNFS will take it either way, and in the case of the sextic is about as fast (meaning within order of magnitude) either way.

To add to it, [URL="http://ggnfs.svn.sourceforge.net/viewvc/ggnfs/trunk/src/experimental/lasieve4_64/cweb.INSTALL.and.USE?revision=409&view=markup"]this GGNFS file[/URL] has this to say about skew (I happened to look at this a minute ago, great timing :smile:):
[quote]The '-S' parameter of the lattice siever is the skewness of your polynomial.
For gnfs polynomials, it is calculated by the gnfs-qpoq or (better) gnfs-qpoq1
programs. For SNFS projects, it normally close to 1. In some cases, values
between 1 and 10 (or less than one if one of the polynomials has a dominating
leading coefficient) offer a (slight) improvement. You may want to try this
out before starting your SNFS project.[/quote]

I guess I'll stick with c0/cn as the default then, unless the inverse is significantly closer to the range the file mentions. Thanks!

[QUOTE=Dubslow;313030]Hmm... pinhodecarlos' files show it [STRIKE]either way[/STRIKE] the other way once:
[code]~/yafu/SNFS∰∂ grep abs *
exponent 11.poly:skew: abs(c0/c5)^(1/degree)
exponent 29.poly:skew: abs(-base/c5)^(1/degree)
exponent 43.poly:skew: abs(c0/c6)^(1/degree)
exponent 53.poly:skew: abs(c6/c0)^(1/degree)
exponent 61.poly:skew: abs(c0/c5)^(1/degree)[/code][/quote]
I don't know what this means. If it was used the other way for some factorization, then it almost certainly was the wrong choice.

[QUOTE=Dubslow;313030]And RDS [URL="http://www.mersenneforum.org/showthread.php?p=266499#post266499"]said[/URL] sometimes the inverse depending on the code[/quote]
Right. And the sievers that we use requires it to be this way.

[QUOTE=Dubslow;313030]and as Mathew demonstrated, GGNFS will take it either way[/quote]
It will "take" it. But that doesn't mean anything. Skew affects the shape of the sieve region. It only affects the efficiency of sieving. You could give 100, 10, 1, 0.1, 0.01 or any positive value -- it'll "take" it.

[QUOTE=Dubslow;313030]and in the case of the sextic is about as fast (meaning within order of magnitude) either way.[/quote]
Sure. Still, the "legal" one was better. In fact, a slightly higher skew might be even better in that particular case.

Regarding the comment from the documentation, isn't that the exact thing that you observe when you compute (c0/cn)^(1/n)? That is the natural result of c0 & cn being small values. When they're not small, that observation would not hold (but the formula remains the same).


[QUOTE=Dubslow;313030]I guess I'll stick with c0/cn as the default then, [STRIKE]unless the inverse is significantly closer to the range the file mentions.[/STRIKE] Thanks![/QUOTE]
Good choice.