Team sieve #20: c170 from 4788:2549

[schickel]

We're tackling a c170 from aliquot sequence 4788, index 2549.

Post-processing will be done by em99010pepe.
Post-processing started 02 Nov 2010.
Estimated completion 08 Nov 2010
Completed. Factored as p55*p115.

Poly for this job:

n: 31312681856591762976684825948611411853648506930718408786922952694966820760977803934965693400499529851375037912983239126649207012737394067178518887401399185420987310860499
# norm 1.089488e-16 alpha -8.144343 e 2.852e-13 rroots 5
skew: 2848875.02
c0:  3449752476991340409710593294897864121568
c1:  3679068299798050640485923392826756
c2: -55695962181000919237307903624
c3:  5626603601260154351771
c4:  6624408564633178
c5:  64149120
Y0: -217612366832903667482868415880575
Y1:  542063406908335635313
rlim: 33550000
alim: 33550000
lpbr: 30
lpba: 30
mfbr: 60
mfba: 87
rlambda: 2.6
alambda: 3.6
type: gnfs

This poly allows for 3 large primes on the algebraic side, resulting in a slightly faster job.

Sieve using the 15e siever, starting from 30M on the algrebraic side with the following command line:

gnfs-lasieve4I15e -a 4788_2549.poly -o <output_file_name> -f <start_of_range> -c <length_of_range>

Per usual, please check downthread for reservations not transferred up here yet.

N.B.: the 15e siever will use up to 200MB of RAM per instance.

On my 2Ghz dual-cores, blocks of 10k special-q take ~3 hours.

Relations needed: ~97M unique

Relations received: 94.2M unique (~97.1%)

Reservations:

  15  M- 15.5M   Andi47 (done, 1101038 relations)
  15.5M- 16  M   Andi47 (done, 1091882 relations)
  16  M- 30  M   schickel (done, just over 24000000 relations)
  30  M- 31  M   bchaffin (done, 2040094 unique relations)
  31  M- 32  M   smh (done, 1965787 relations)
  32  M- 40  M   bsquared (done, 15138914 relations)
  40  M- 47.2M   bsquared (done, 12742177 relations)
  47.2M- 50  M   bchaffin (done)
  50  M- 52  M   bchaffin (done)
  52  M- 57  M   bsquared (done, 8143236 relations)
  57  M- 60  M   bsquared (done, 4778150 relations)
  60  M- 62  M   bchaffin (done)
  62  M- 64  M   bchaffin (done)
  64  M- 65  M   em99010pepe (done)
  65  M- 68  M   bchaffin (done)
  68  M- 70  M   bsquared (done, 3004347 relations)
  70  M- 72  M   bchaffin (done)
  72  M- 75  M   bsquared (done, 4409292 relations)
  75  M- 77  M   bchaffin (done)
  77  M- 80  M   bsquared (done, 4290164 relations)
  80  M- 92  M   em99010pepe (done)

Estimated completion at present speed: November, 2010 Much better!

[schickel]

For the relations on this job, shall we use a file-share service, or should I see if I can set up a home FTP server?

[em99010pepe]

Originally Posted by schickel

For the relations on this job, shall we use a file-share service, or should I see if I can set up a home FTP server?

How fast is your connection?

[Mini-Geek]

http://www.sendspace.com/ is what we usually use, but if it looks like a home FTP setup would be better, we can go with that.

[jrk]

Originally Posted by schickel

qintsize: 10000

Take this line out of the poly file. It'll override the -c count parameter on the command line.

[schickel]

Originally Posted by em99010pepe

How fast is your connection?

It usually tests at around 1Mb down/.1Mb up. So I can collect 'em pretty fast, but I'll be piss poor sending 'em.

Originally Posted by Mini-Geek

http://www.sendspace.com/ is what we usually use, but if it looks like a home FTP setup would be better, we can go with that.

I guess it would make more sense if I did all the post-processing (filtering and LA), but it might make sense to do it with an FTP server if I did the filtering and cycle-counting, then pass the relations and .cyc files to em99010pepe.

[schickel]

Originally Posted by jrk

Take this line out of the poly file. It'll override the -c count parameter on the command line.

Done. I copied the poly from my work directory. I thought that was only used by the scripts, I didn't realize that the siever used that line....

[em99010pepe]

Originally Posted by schickel

It usually tests at around 1Mb down/.1Mb up. So I can collect 'em pretty fast, but I'll be piss poor sending 'em.

I guess it would make more sense if I did all the post-processing (filtering and LA), but it might make sense to do it with an FTP server if I did the filtering and cycle-counting, then pass the relations and .cyc files to em99010pepe.

I just don't know one step, after downloading the relations how to merge them into msieve dat file. The rest is peaceful...I hope...

msieve.exe -i xxx.ini -s xxx.dat -nf xxx.fb -nc -v -t 4

or

having msieve.fb, worktodo.ini and msieve.dat, run msieve.exe -nc -v -t 4

Post-processing will be done on a core i5 750@3.7 GHz with 8GB.

[Batalov]

Originally Posted by em99010pepe

I just don't know one step, after downloading the relations how to merge them into msieve dat file. The rest is peaceful...I hope...

You can start with a msieve.dat file with one line:

N 31312681856591762976684825948611411853648506930718408786922952694966820760977803934965693400499529851375037912983239126649207012737394067178518887401399185420987310860499

then gzip msieve.dat, and then, when you get a new chunk from collaborators,
1. gzip -tv new_file.gz to check for errors (or if bzip2'd, then bunzip and regzip)
2. cat new_file.gz >> msieve.dat.gz

That will save you some disk space.

[em99010pepe]

On Jeff Gilchrist's webpage he has 4 versions of ggnfs, SVN 374 (32 bits and 64 bits) and SVN 360 (32 bits and 64 bits). Which combination is fastest for gnfs-lasieve4I15e under windows 7 64 bits?

[schickel]

Originally Posted by em99010pepe

On Jeff Gilchrist's webpage he has 4 versions of ggnfs, SVN 374 (32 bits and 64 bits) and SVN 360 (32 bits and 64 bits). Which combination is fastest for gnfs-lasieve4I15e under windows 7 64 bits?

Unless things have changed, the 32-bit sievers are faster on 64-bit Windows than the 64-bit sievers.