newb needs help - factoring RSA512 error

[VBCurtis]

Originally Posted by pomodoro

Thanks for the tips, worked.
Deleted the .mat and .cyc files, start msive in phase -nc1, and then -nc2 and finally -nc3. Ended normally.
Probably the .mat and/or .cyc files were corrupted because msieve do not start directly into phase -nc2.
Only still do not know how to use the spairs.save.gz.
Nobreak = Uninterruptible Power Supply (UPS)

[ ] ´s

Great! Glad to hear it. This C133 is a nice size to play with some options in filtering/matrix solving:

The target_density flag during -nc1 phase controls how hard the filtering works to prune relations and make a denser smaller matrix. Default is 70, while maximum is around 125-130. It might interest you to try the filtering step with -target_density=75 or 80, see how much smaller the matrix is.

If you accidentally (or on purpose!) oversieve, where you know you have an excess of relations, you can set target_density at 90 or 96 or 120 to make use of these extra relations. For most sets of relations, there is a maximum density where a matrix will build, higher than that you get the usual "need 1M more relations" msg from msieve and no matrix. I have noticed many of the experienced folks who solve the NFS@home matrices use multiples of 8 or 16 for target_density, but I do not know if there is a reason for this. The NFS@home problems are usually oversieved enough that densities of 90 or higher are possible.

For jobs of size suited to 14e, it is generally held that leaving this setting at default saves more time in sieving than it costs you in matrix-solving, but as jobs grow in size the matrix-step time grows faster than the sieving step (and some of us have memory restrictions requiring matrices be kept below a certain size). Anyway, filtering on a c133 is likely under an hour, so you could tinker a little with your data set and see if you learn anything.

If you have a slow disk, or find that your NFS runs attempt filtering 10+ times before success, you can alter the 'qintsize' in line 565 of the python script. This will change the size of each chunk of sieving. For jobs above 140 digits, I changed mine to 300000 from default 100000, so that once 100% of estimated relations is reached it attempts filtering 1/3 as often. I view failed filtering runs as wasted time, but making this too coarse results in sometimes wasted time oversieving because you overshoot the necessary data need.
-Curtis