10 Jul
2008
10 Jul
'08
7:07 p.m.
On Thu, 2008-07-10 at 18:32 -0400, Ronald Guida wrote:
Your ratios are about 1 : 3 : 8. That pretty close to quadratic growth, 1 : 4 : 9, so I think all is well.
Maybe, but 96MB of resident memory for a 1000-node graph looks bad, especially considering p is low. Is the internal representation of inductive graphs perhaps not very memory-efficient? I still haven't read Erwig's paper... I know this is probably not fair, but I'm comparing these numbers with a C implementation which uses Ruby's C API for its complex data structures, and a 10,000 nodes graph uses around 6MB of memory. Thanks, Andre