matrix needs more columns than rows

[ryanp]

Hi,

I seem to be hitting this problem again with some larger jobs: the -nc1 step completes and produces a cycles file (I'm using target_density=100, if that matters) but linear algebra fails with "matrix needs more columns than rows; try adding 2-3% more relations". Also of note: I'm using lasieve5, which seems to run faster and can sieve q > 2^30. But perhaps it's somehow producing more duplicate/bad relations (just speculating)?

Here's the full output from -nc2:

Msieve v. 1.52 (SVN 879M)
Wed May  1 01:36:55 2013
random seeds: 525dfdb8 ab67515a
factoring 9957223967396174026852789233155627891038068723057068144296080631729864465804898206811407977824400956027486754034269650484028731374105774063596653234136909784394071881184497315147696834136314854683032111 (202 digits)
no P-1/P+1/ECM available, skipping
commencing number field sieve (202-digit input)
R0: -72890483685103052142902866787761839379440139451
R1: 1
A0: 11
A1: 0
A2: 0
A3: 0
A4: 0
A5: 0
A6: 1
skew 1.49, size 8.535e-14, alpha 1.243, combined = 1.772e-14 rroots = 0

commencing linear algebra
read 26605987 cycles
cycles contain 64802006 unique relations
read 64802006 relations
using 20 quadratic characters above 4294917296
warning: zero character
warning: zero character
building initial matrix
memory use: 9061.1 MB
read 26605987 cycles
matrix is 26638102 x 26605987 (7946.8 MB) with weight 2332379792 (87.66/col)
sparse part has weight 1790552060 (67.30/col)
filtering completed in 2 passes
matrix is 26606739 x 26589934 (7942.8 MB) with weight 2331168695 (87.67/col)
sparse part has weight 1789667895 (67.31/col)
matrix starts at (0, 0)
matrix is 26606739 x 26589934 (7942.8 MB) with weight 2331168695 (87.67/col)
sparse part has weight 1789667895 (67.31/col)
matrix needs more columns than rows; try adding 2-3% more relations

[henryzz]

You are on the edge of the needed number of relations. Sieve some more to fix this particular error.

[frmky]

How many unique relations do you have? What size large primes are you using?