Introducing Weak

[pocopito]

Straight away your suggestions have uncovered some bugs Very useful + very much appreciated, thanks!

Nice!


About clang and your unexpected good performance, my compilation knowledge is, at best, quite weak, and my suggestion had to do with the comments in this post:
http://talkchess.com/forum/viewtopic.ph ... ight=clang

In any case I guess I'll take a deeper look to your Make file

Best regards

E Diaz

[lorenzo]

Ok, so a shift to gcc 4.7 has resulted in a significant speed up (after finally finding a way of getting it to work on mac!), though I am experiencing a lot of bugs at the moment that I thought I'd squashed :-S however I am seeing nice improvements in perft performance, which I shall fully quantify once all the bugs are ironed out.

[Don]

Well, though move generation is ok, for now the search code is very basic + buggy, so the name is accurate for the time being but the hope is it will eventually become ironic. A bit of British humour...

I head a friend who once named his Doberman dog "Puff" - it was similar humor.

[lorenzo]

Okay, so I fixed a big load of bugs and I'm getting correct perft results both for the perft result tests on the chess programming wiki and the hardcore perft tests E Diaz linked here.

This batch of testing has helped uncover numerous bugs which was awesome . I intend to introduce a number of unit tests (I already have an extensive system test which tests consistency of cached data vs. calculated results) to make regression testing easier in future.

I am now using GCC 4.7, which has improved speed both for weak and stockfish, and I am getting very close Mnps speeds to Stockfish (e.g. perft 6 from standard initial position is ~ 53 Mnps on my 2.13GHz core 2 duo macbook air).

Next steps - improving the currently terrible hacky search code, create a non-skeletal evaluation function, add UCI support, add a transition table... oh the list goes on some way from v1, but getting there!