search

fplll / g6k

The General Sieve Kernel
GNU General Public License v2.0
102 stars 31 forks source link

Crashes with full_sieve.py #70

Open joerowell opened 3 years ago

joerowell commented 3 years ago

When running ./full_sieve.py 130 --threads 24 on a machine with 32GB of RAM and 24 cores, I get a crash. The reason for this is that the Siever attempts to reserve the amount of RAM it would need to sieve in dimension 130 during construction - which clearly I don't have (>256 GB would be needed).

However, perhaps bizarrely, the siever then relinquishes this memory when the workout begins. This seems undesirable: perhaps we could just issue an error instead.

lducas commented 3 years ago

Oh, sorry for the delay, I missed that. It's probably solved by now, but there is a hardcoded maxdim (set to 128 IIRC) at compile time. You can adjust this during rebuild.sh with the option -m|--maxsievingdim.

Can you confirm this solves it ?