The new, faster program used up so much memory that it got really slow due to swapping virtual memory on disk. It ran for 11 days and generated a little over 213,000 graphs, and then my boss rebooted the computer again.
Next time, I’ll have to just put the graphs into a database instead of holding everything in memory. And use a different computer.
I’m giving this low priority for now, though.
Oct 20, 2006, 05:35AM PDT | 2 comments
A few weeks ago, I wrote a Python program to generate all the order-9 graphs. I ran it on my boss’s dual-CPU G5. It ran for a week, until my boss rebooted the computer. It had generated about 180,000 graphs. I think that’s almost all of them.
I modified the program to use more “hashing and caching”, a better way of spotting non-isomorphs, and to auto-include complements of graphs with fewer than 36 lines. That sped up generation of all 12,346 order-8 graphs from 125 minutes to 9 minutes. I started it running on the order-9 graphs yesterday. It generated the first 50,000 or so very quickly, but now the memory usage is at 1.5 GB, so I think the process is doing a lot of swapping. It’s completed 114,000 graphs now, and is generating about 1000 graphs per hour.
Oct 06, 2006, 04:06PM PDT | 0 comments