View Single Post
Old 2021-07-20, 13:19   #12
charybdis's Avatar
Apr 2020

541 Posts

Originally Posted by bur View Post
So having an SNFS-177 take as long as a GNFS-139 really seems a bit long? I thought it might be due to the large coeffcient but the steep increase surprised me.
It was the 4-digit increase in GNFS equivalent difficulty for a 3-digit increase in SNFS difficulty that surprised me. Judging by the E scores of the polynomials, I would have expected them to sieve more like GNFS-134 and 135, so the second number has taken longer than it should have done. I'm especially surprised given that you used the same parameters for both numbers. Are you sure you didn't have any other processes slowing the second job down?

I'd have used roughly c135 parameters for both jobs given the E scores.

I managed to build a matrix using msieve with 60M relations. From my experience msieve tends to require more relations than cado to build a matrix (why is that?), so for the current factorization I decreased rels_wanted to 60M.
I think the CADO filtering tools are just slightly better tuned than msieve. I've never tried this, but it's possible that playing around with some of the additional arguments for -nc1 might allow msieve to build a matrix with slightly fewer relations.

If the rels_wanted figure was off for the c135 parameters, maybe it was off for the c130 parameters too. That still doesn't explain the discrepancy between the two numbers.

By the way, it is essential to add tasks.sieve.adjust_strategy = 2 whenever you use tasks.A = [even number] rather than tasks.I. There should be a noticeable speedup.
charybdis is offline   Reply With Quote