Keeping relations

[fivemack]

At the moment I keep copies of the relations for all of the Cunningham-table numbers that mersenneforum has sieved.

This is taking hundreds of gigabytes of disc, and I am starting to grow short of space. It seems a pity to throw away data corresponding to tens of thousands of CPU-hours of effort, but is there actually any use in keeping these hundreds of millions of (x,y:Rfactors:Afactors) lines once the matrix has been solved and the factors submitted?

I keep the msieve.log files because there's still the potential of more tuning for msieve, and it's nice to have some idea of what the timing looks like and what sort of number of relations were required for problems of various size, but that's 100k per number rather than 10G.

[R.D. Silverman]

Originally Posted by fivemack

At the moment I keep copies of the relations for all of the Cunningham-table numbers that mersenneforum has sieved.

This is taking hundreds of gigabytes of disc, and I am starting to grow short of space. It seems a pity to throw away data corresponding to tens of thousands of CPU-hours of effort, but is there actually any use in keeping these hundreds of millions of (x,y:Rfactors:Afactors) lines once the matrix has been solved and the factors submitted?

I keep the msieve.log files because there's still the potential of more tuning for msieve, and it's nice to have some idea of what the timing looks like and what sort of number of relations were required for problems of various size, but that's 100k per number rather than 10G.

The factorizations can be reconstructed. Just keep the (a,b) lattice
locations. Perhaps the special q as well. Nothing else.