|
2009-01-04, 20:21
|
#12 |
Nov 2008
1001000100102 Posts |
Same error again on 2^293-673. This time, the problem was "matrix not dense enough, retrying" every time. I got 568325 relations, how many do I need?
|
|
|
|
2009-01-04, 20:53
|
#13 |
|
Tribal Bullet
Oct 2004
3,541 Posts |
Try with 3/4 of the relations. If that doesn't work try with 1/2 the relations.
I'll put back the code that can limit the number of relations read from the .dat file, since suddenly everybody seems to need it. |
|
|
|
|
2009-01-04, 21:33
|
#14 |
Apr 2007
Spessart/Germany
16210 Posts |
Hello,
no luck with GMP-ECM this time, but msieve 1.38 had no problem again to factor the number: Code:
Sun Jan 04 22:22:40 2009 Msieve v. 1.38 Sun Jan 04 22:22:40 2009 random seeds: 458c0840 d969bc98 Sun Jan 04 22:22:40 2009 factoring 15914343565113172548972231940698266883214596825515126958094847260581103904401068017057119 (89 digits) Sun Jan 04 22:22:41 2009 searching for 15-digit factors Sun Jan 04 22:22:42 2009 commencing quadratic sieve (89-digit input) Sun Jan 04 22:22:42 2009 using multiplier of 1 Sun Jan 04 22:22:42 2009 using 32kb Intel Core sieve core Sun Jan 04 22:22:42 2009 sieve interval: 28 blocks of size 32768 Sun Jan 04 22:22:42 2009 processing polynomials in batches of 8 Sun Jan 04 22:22:42 2009 using a sieve bound of 1535717 (58333 primes) Sun Jan 04 22:22:42 2009 using large prime bound of 122857360 (26 bits) Sun Jan 04 22:22:42 2009 using double large prime bound of 363849688796320 (42-49 bits) Sun Jan 04 22:22:42 2009 using trial factoring cutoff of 49 bits Sun Jan 04 22:22:42 2009 polynomial 'A' values have 11 factors Sun Jan 04 22:22:43 2009 restarting with 16259 full and 644376 partial relations Sun Jan 04 22:22:43 2009 63733 relations (16259 full + 47474 combined from 644376 partial), need 58429 Sun Jan 04 22:22:43 2009 begin with 660635 relations Sun Jan 04 22:22:44 2009 reduce to 155875 relations in 10 passes Sun Jan 04 22:22:44 2009 attempting to read 155875 relations Sun Jan 04 22:22:45 2009 recovered 155875 relations Sun Jan 04 22:22:45 2009 recovered 132451 polynomials Sun Jan 04 22:22:45 2009 attempting to build 63733 cycles Sun Jan 04 22:22:45 2009 found 63733 cycles in 6 passes Sun Jan 04 22:22:45 2009 distribution of cycle lengths: Sun Jan 04 22:22:45 2009 length 1 : 16259 Sun Jan 04 22:22:45 2009 length 2 : 11892 Sun Jan 04 22:22:45 2009 length 3 : 11364 Sun Jan 04 22:22:45 2009 length 4 : 8734 Sun Jan 04 22:22:45 2009 length 5 : 6363 Sun Jan 04 22:22:45 2009 length 6 : 4038 Sun Jan 04 22:22:45 2009 length 7 : 2336 Sun Jan 04 22:22:45 2009 length 9+: 2747 Sun Jan 04 22:22:45 2009 largest cycle: 21 relations Sun Jan 04 22:22:45 2009 matrix is 58333 x 63733 (15.7 MB) with weight 3869123 (60.71/col) Sun Jan 04 22:22:45 2009 sparse part has weight 3869123 (60.71/col) Sun Jan 04 22:22:46 2009 filtering completed in 4 passes Sun Jan 04 22:22:46 2009 matrix is 53548 x 53612 (12.2 MB) with weight 2992533 (55.82/col) Sun Jan 04 22:22:46 2009 sparse part has weight 2992533 (55.82/col) Sun Jan 04 22:22:46 2009 saving the first 48 matrix rows for later Sun Jan 04 22:22:46 2009 matrix is 53500 x 53612 (8.6 MB) with weight 2425560 (45.24/col) Sun Jan 04 22:22:46 2009 sparse part has weight 1923237 (35.87/col) Sun Jan 04 22:22:46 2009 matrix includes 64 packed rows Sun Jan 04 22:22:46 2009 using block size 21444 for processor cache size 4096 kB Sun Jan 04 22:22:46 2009 commencing Lanczos iteration Sun Jan 04 22:22:46 2009 memory use: 8.1 MB Sun Jan 04 22:23:01 2009 lanczos halted after 847 iterations (dim = 53495) Sun Jan 04 22:23:01 2009 recovered 14 nontrivial dependencies Sun Jan 04 22:23:01 2009 prp40 factor: 7941690361259552691442291169877013099153 Sun Jan 04 22:23:01 2009 prp49 factor: 2003898772324983526166134218491278011857331882223 Sun Jan 04 22:23:01 2009 elapsed time 00:00:21 Best regards, Matthias |
|
|
|
|
2009-01-05, 08:29
|
#15 | |
|
Nov 2008
2×33×43 Posts |
Quote:
|
|
|
|
|
|
2009-01-05, 09:31
|
#16 |
|
Nov 2008
44228 Posts |
I knocked my number of relations down to about 3/4 of what it was before, and msieve neede more relations. I kept on sieving more and re-running msieve, but when msieve had enough relations, it proceeded to keep on saying "matrix not dense enough, retrying" as before. I think this occurs becuase msieve has not been optimized for small SNFS numbers, as a normal version will only do the postprocessing for >=97 digits. This means that the "to do" things for msieve should be:
1. Put the option to read relations x to y from msieve.dat back 2. Remove the 97 digit minimum for NFS 3. Optimize for small SNFS's The same job finished fine with GGNFS's postprocessing. Last fiddled with by 10metreh on 2009-01-05 at 09:44 |
|
|
|
|
2009-01-05, 12:22
|
#17 |
|
Just call me Henry
"David"
Sep 2007
Cambridge (GMT/BST)
10110111110002 Posts |
i have never had ggnfs's postprocessing fail because of oversieving except for running out of memory on my 2gb system
i did an experiement once on a small number ~80 digits and got 10M relations before i did the postprocessing i had to reduce to ~7M relations before i stopped running out of memory a matrix made with 5M relations ran in 3/4 of the time that the 7M one ran msieve needed the number of relations ruduced loads before it would succeed the ggnfs postprocessing for that number of relations took 2 hours each time |
|
|
|
|
2009-01-05, 14:58
|
#18 | |
|
Tribal Bullet
Oct 2004
67258 Posts |
You can give GGNFS tremendous amounts of excess relations because it will delete relations it doesn't need. I guess msieve needs the same thing, since even the largest jobs sometimes require it.
Quote:
Last fiddled with by jasonp on 2009-01-05 at 14:58 |
|
|
|
|
 |
| Thread Tools | |
Similar Threads
|
||||
| Thread | Thread Starter | Forum | Replies | Last Post |
| Bug/oddity reports | fivemack | NFS@Home | 1 | 2015-01-30 11:34 |
| Attempts vs. Successes oddity | Rodrigo | GPU Computing | 8 | 2014-09-19 08:44 |
| Another server(?) oddity | Graff | PrimeNet | 2 | 2010-01-25 17:00 |
| Enormous filtering run oddity | frmky | Msieve | 23 | 2009-11-03 18:52 |
| Self-test oddity | JuanTutors | Software | 1 | 2007-03-16 12:33 |
