This was a non-parallel version of perft(), with no hash table. It was run on a single core.
It took 16.5 days averaging 52.5 million nodes per second. I think it is safe to say, this is the limit of perft() for checkers on a single core with no additional enhancements.
I would be interested in learning more about parallelizing perft(), and also implementing a hash table for it. I'm not sure how to go about either at the moment.
The good news is: all of the divide() numbers agree with Aart Bik's postings on his page here:
http://www.aartbik.com/MISC/checkers.html
I also have capture and crown data for my runs, and await confirmation from other programmers for this also.
Code: Select all
TOTAL NODES = 74545030871553
TOTAL JUMPS = 12980932483933
TOTAL CROWNS = 2680251548251
TOTAL TIME = 1419377.5390 seconds
TOTAL SPEED = 52519522.6945 nodes/second.