Changes

Jump to: navigation, search

Perft

1,365 bytes added, 00:17, 13 June 2020
no edit summary
<span id="Divide"></span>
* [ Perft speed and depth questions] by [[Mark Buisseret]], [[CCC]], June 12, 2020
 
=Quotes=
by [[Robert Hayatt]] in a reply to [[Mark Buisseret]], June 12, 2020 <ref>[http://talkchess.com/forum3/viewtopic.php?f=7&t=74153 Re: Perft speed and depth questions] by [[Mark Buisseret]], [[Computer Chess Forums|CCRL Discussion Board]], June 12, 2020</ref> :
I believe I was the first to use this. Back in the 80's. We rewrote the move generator in Cray Blitz in assembly language. It was a pain to debug. I decided on the "perft" approach solely to test/debug the move generator. We'd run two versions, one FORTRAN, one assembly, and we tested and debugged until they matched.
I carried this over into Crafty as early versions went through several different approaches on move generation. Starting with the Slate/Atkin approach, then rotated bit boards (which took some time to debug), and the magic. It was really intended solely for that purpose. Then several started to use it as a benchmark for speed. I never followed that path since move generation is a very small part of the overall CPU time burned.
Speed here is not so important. I doubt anyone's move generator takes more than 10% of total search time, which means a 20% improvement in perft numbers is only a 2% overall speed gain. I would not worry about anything but matching the node counts exactly...
=Results=

Navigation menu