matrix needs more columns than rows

[wreck]

When I factoring Aliquot sequence 1005700 index 994, it meets this problem, and here is the log

[wreck@nkstar2 msieve-1.43]$ ./msieve 14896140456396261111298775277182256

sieving complete, commencing postprocessing
[wreck@nkstar2 msieve-1.43]$ cat msieve.log
Fri May 28 16:24:23 2010  
Fri May 28 16:24:23 2010  
Fri May 28 16:24:23 2010  Msieve v. 1.43
Fri May 28 16:24:23 2010  random seeds: c94f4f14 05197ce2
Fri May 28 16:24:23 2010  factoring 14896140456396261111298775277182256 (35 digits)
Fri May 28 16:24:23 2010  no P-1/P+1/ECM available, skipping
Fri May 28 16:24:23 2010  commencing quadratic sieve (31-digit input)
Fri May 28 16:24:23 2010  using multiplier of 1
Fri May 28 16:24:23 2010  using 64kb Pentium 4 sieve core
Fri May 28 16:24:23 2010  sieve interval: 2 blocks of size 65536
Fri May 28 16:24:23 2010  processing polynomials in batches of 51
Fri May 28 16:24:23 2010  using a sieve bound of 4253 (302 primes)
Fri May 28 16:24:23 2010  using large prime bound of 170120 (17 bits)
Fri May 28 16:24:23 2010  polynomial 'A' values have 3 factors
Fri May 28 16:24:23 2010  restarting with 248 full and 1832 partial relations
Fri May 28 16:24:23 2010  1240 relations (248 full + 992 combined from 1832 partial), need 398
Fri May 28 16:24:23 2010  begin with 2080 relations
Fri May 28 16:24:23 2010  reduce to 2080 relations in 1 passes
Fri May 28 16:24:23 2010  attempting to read 2080 relations
Fri May 28 16:24:23 2010  recovered 2080 relations
Fri May 28 16:24:23 2010  recovered 8 polynomials
Fri May 28 16:24:23 2010  freed 1040 duplicate relations
Fri May 28 16:24:23 2010  attempting to build 200 cycles
Fri May 28 16:24:23 2010  found 200 cycles in 1 passes
Fri May 28 16:24:23 2010  distribution of cycle lengths:
Fri May 28 16:24:23 2010     length 1 : 124
Fri May 28 16:24:23 2010     length 2 : 76
Fri May 28 16:24:23 2010  largest cycle: 2 relations
Fri May 28 16:24:23 2010  matrix is 302 x 200 (0.0 MB) with weight 2311 (11.55/col)
Fri May 28 16:24:23 2010  sparse part has weight 2311 (11.55/col)
Fri May 28 16:24:23 2010  filtering completed in 2 passes
Fri May 28 16:24:23 2010  matrix is 150 x 87 (0.0 MB) with weight 1009 (11.60/col)
Fri May 28 16:24:23 2010  sparse part has weight 1009 (11.60/col)
Fri May 28 16:24:23 2010  matrix needs more columns than rows; try adding 2-3Invalid argumentore relations
[wreck@nkstar2 msieve-1.43]$

[10metreh]

This error happens in NFS jobs, but I've never seen it in QS before. I suggest just re-running the job, as it's so small. (Or use yafu.) This factorization works fine for me:

p1 factor: 2
p1 factor: 2
p1 factor: 2
p1 factor: 2
p1 factor: 3
p1 factor: 7
p2 factor: 19
p10 factor: 8308967087
prp21 factor: 280823754651675772307

BTW, aliqueit is faster for running aliquot sequences: it automatically runs ECM, QS, NFS etc.

[wreck]

Originally Posted by 10metreh

This error happens in NFS jobs, but I've never seen it in QS before. I suggest just re-running the job, as it's so small. (Or use yafu.) This factorization works fine for me:

p1 factor: 2
p1 factor: 2
p1 factor: 2
p1 factor: 2
p1 factor: 3
p1 factor: 7
p2 factor: 19
p10 factor: 8308967087
prp21 factor: 280823754651675772307

BTW, aliqueit is faster for running aliquot sequences: it automatically runs ECM, QS, NFS etc.

Maybe this is because I change the makefile, since this binary is compiled on our school's cluster, and there are many system info cann't be obtained automaticaly. And I just notice this problem is not occur at one of my classmate's PC.

I have already downloaded the aliqueit sourcefile and its Window binary, since I only test factoring those small numbers, it is pain to upload thousands of small factors. I also cann't obtain bsquare's yafu binary since the site of sites.google.com is forbidden in China.

[retina]

Originally Posted by wreck

.... sites.google.com is forbidden in China.

Ouch!

Proxy?

Tor?

[jasonp]

Originally Posted by wreck

When I factoring Aliquot sequence 1005700 index 994, it meets this problem, and here is the log

Fri May 28 16:24:23 2010  attempting to read 2080 relations
Fri May 28 16:24:23 2010  recovered 2080 relations
Fri May 28 16:24:23 2010  recovered 8 polynomials
Fri May 28 16:24:23 2010  freed 1040 duplicate relations

This is the problem; half of your dataset is a copy of the other half. Get rid of it and you don't have enough relations to proceed.

[wreck]

Originally Posted by jasonp

This is the problem; half of your dataset is a copy of the other half. Get rid of it and you don't have enough relations to proceed.

Oh, after I delete the msieve.dat this problem not occur again, thanks for your hint.

[chris2be8]

From lanczos.c:

                 logprintf(obj, "matrix needs more columns than rows; "
                 "try adding 2-3% more relations\n");

Bu the message came out as:
matrix needs more columns than rows; try adding 2-3Invalid argumentore relations

Is the compiler misinterpreting the % character?

Chris K

[Brian Gladman]

Originally Posted by chris2be8

From lanczos.c:

                 logprintf(obj, "matrix needs more columns than rows; "
                 "try adding 2-3% more relations\n");

Bu the message came out as:
matrix needs more columns than rows; try adding 2-3Invalid argumentore relations

Is the compiler misinterpreting the % character?

Chris K

Hi Chris,

Yes, your analysis is correct - I have corrected this in the MSIEVE SVN.

Brian Gladman