[r-t] B****r Re: Big searches
mark at snowtiger.net
Sun Jan 6 00:40:27 UTC 2008
> I forgot that the method I was using wasn't Yorkshire, it was Rutland
Quite all right Don, makes your figures look even better, anyway!
Here's SMC32 for Rutland:
283,637,000 nodes, 4140 ms, 68,511,000 nodes/s, 34250 nodes/Mcycle
So your fast search is some 100 times quicker - good stuff. The benefits of
a tuned algorithm are clear; well assuming you can find a bigger search
space to show them off in. Any plans...?
I've just realised there is one optimisation I can turn on in SMC32 for
this search - restricting the number of Befores to 3 per part max. This
133,009,000 nodes, 3156 ms, 42,145,000 nodes/s, 21070 nodes/Mcycle
A slight improvement.
More information about the ringing-theory