Factoring for a publication - Page 38

thyrex · 2021-07-22, 14:48

So I call the stage at which the nfs.dat.p file is filled and the best poly is searched

charybdis · 2021-07-22, 14:59

Ah, so by preNFS you really mean the first stage of NFS. I suppose if we're being really strict about the meaning of the word "sieve" then polynomial selection is pre-NFS and postprocessing is post-NFS, but most people use NFS to refer to the whole algorithm rather than just the sieving stage.

I think going straight to GNFS after t50 at this size is suboptimal, but of course everyone's free to do what they want. If you have the resources to do the GNFS yourself, then you also have the resources to do the optimal amount of ECM.

thyrex · 2021-07-22, 15:35

Quote:

Originally Posted by charybdis

I think going straight to GNFS after t50 at this size is suboptimal

This is not I started with the GNFS stage, but yafu itself after the completion of the t50 stage

charybdis · 2021-07-22, 16:28

So you ran t50 and then called yafu factor(), which presumably ran at least another t50?

thyrex · 2021-07-22, 17:21

I ran yafu -work 45 and after completing this stage automatically started pre-NFS stage for c168 and c169

LaurV · 2021-07-23, 03:11

You can tune how much ecm yafu will do, by using a "pretest_ratio=..." in yafu.ini file. I usually go with 0.334 (33.4%, which means, for a C168, at least a t55 will be done, and this way have a large chance to avoid a 3-way split).

bur · 2021-07-23, 07:09

Will the data generated here be used in the paper that is currently published at arxiv or will it be part of a new publication? Do you know when it will be submitted?

wombatman · 2021-07-27, 23:16

The C212/SNFS215 for (13,2) factors as:

Code:

p68 factor: 65470839132891587378451398003085663493468996034812587842969753001001
p144 factor: 272999581728907243481927568287887505420486663480088950983800638257617799930179269423084728938601529213529914520447747676995095997912408303807661

thyrex · 2021-07-29, 10:34

Quote:

Originally Posted by LaurV

You can tune how much ecm yafu will do, by using a "pretest_ratio=..." in yafu.ini file. I usually go with 0.334 (33.4%, which means, for a C168, at least a t55 will be done, and this way have a large chance to avoid a 3-way split).

Can you attach your yafu.ini?

In factor.log I see for example

Quote:

07/27/21 18:48:38 v1.34.5 @ UUNP-HV05N4, Starting factorization of 3247903177530905810688738828962481204542302302414955042770879837334793211816425127548527769179079565260917709105074137909390860953109380133800537359422487798480789987383771707902131083201
07/27/21 18:48:38 v1.34.5 @ UUNP-HV05N4, using pretesting plan: custom
07/27/21 18:48:38 v1.34.5 @ UUNP-HV05N4, custom pretest ratio is: 0.3340
07/27/21 18:48:38 v1.34.5 @ UUNP-HV05N4, no tune info: using qs/gnfs crossover of 95 digits
07/27/21 18:48:38 v1.34.5 @ UUNP-HV05N4, input indicated to have been pretested to t45.00
07/27/21 18:48:38 v1.34.5 @ UUNP-HV05N4, ****************************
07/27/21 18:48:38 v1.34.5 @ UUNP-HV05N4, rho: x^2 + 3, starting 1000 iterations on C187
07/27/21 18:48:38 v1.34.5 @ UUNP-HV05N4, rho: x^2 + 2, starting 1000 iterations on C187
07/27/21 18:48:38 v1.34.5 @ UUNP-HV05N4, rho: x^2 + 1, starting 1000 iterations on C187
07/27/21 18:48:38 v1.34.5 @ UUNP-HV05N4, current ECM pretesting depth: 45.67
07/27/21 18:48:38 v1.34.5 @ UUNP-HV05N4, scheduled 7553 curves at B1=43000000 toward target pretesting depth of 62.46
07/29/21 06:31:17 v1.34.5 @ UUNP-HV05N4, Finished 7560 curves using Lenstra ECM method on C187 input, B1=43M, B2=gmp-ecm default
07/29/21 06:31:17 v1.34.5 @ UUNP-HV05N4, final ECM pretested depth: 50.85
07/29/21 06:31:17 v1.34.5 @ UUNP-HV05N4, scheduler: switching to sieve method
07/29/21 06:31:17 v1.34.5 @ UUNP-HV05N4, nfs: commencing nfs on c187

I don't know why yafu skip at least B1=110M stage

EdH · 2021-07-29, 12:54

If I'm reading the table correctly, the following are available for GNFS:

line 123 (9, -7) c168
line 166 (10, 10) c169
line 180 (2, 11) c163
line 183 (5, 11) c162

You may book all available of the four to me.

thyrex · 2021-07-29, 13:49

Quote:

Originally Posted by EdH

line 180 (2, 11) c163
line 183 (5, 11) c162

I will take soon please

c168 and c169 are free to book

2021-07-29, 12:54	#417
EdH "Ed Hall" Dec 2009 Adirondack Mtns 2·19·101 Posts	If I'm reading the table correctly, the following are available for GNFS: line 123 (9, -7) c168 line 166 (10, 10) c169 line 180 (2, 11) c163 line 183 (5, 11) c162 You may book all available of the four to me. Last fiddled with by EdH on 2021-07-29 at 12:55

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
factoring 2ⁿ-2 equivalent to factoring 2ⁿ-1(I think)	baih	Miscellaneous Math	9	2020-09-21 07:11
OpenCL GPU P-1 Factoring and ECM Factoring	xx005fs	GPU Computing	3	2018-10-27 14:49

2021-07-22, 14:48	#408
thyrex Jun 2021 3² Posts	So I call the stage at which the nfs.dat.p file is filled and the best poly is searched

2021-07-22, 14:59	#409
charybdis Apr 2020 379 Posts	Ah, so by preNFS you really mean the first stage of NFS. I suppose if we're being really strict about the meaning of the word "sieve" then polynomial selection is pre-NFS and postprocessing is post-NFS, but most people use NFS to refer to the whole algorithm rather than just the sieving stage. I think going straight to GNFS after t50 at this size is suboptimal, but of course everyone's free to do what they want. If you have the resources to do the GNFS yourself, then you also have the resources to do the optimal amount of ECM.

2021-07-22, 16:28	#411
charybdis Apr 2020 379 Posts	So you ran t50 and then called yafu factor(), which presumably ran at least another t50?

2021-07-22, 17:21	#412
thyrex Jun 2021 9₁₆ Posts	I ran yafu -work 45 and after completing this stage automatically started pre-NFS stage for c168 and c169

2021-07-23, 03:11	#413
LaurV Romulan Interpreter Jun 2011 Thailand 3·3,221 Posts	You can tune how much ecm yafu will do, by using a "pretest_ratio=..." in yafu.ini file. I usually go with 0.334 (33.4%, which means, for a C168, at least a t55 will be done, and this way have a large chance to avoid a 3-way split).

2021-07-23, 07:09	#414
bur Aug 2020 2²×7×11 Posts	Will the data generated here be used in the paper that is currently published at arxiv or will it be part of a new publication? Do you know when it will be submitted?