quick question about search cost breakdown

Discussion of chess software programming and technical issues.

Moderators: hgm, Rebel, chrisw

oldman

quick question about search cost breakdown

Post by oldman »

Hi all,

Correct me if I'm wrong, but the majority of the time spent in chess search is in move generation and evaluation, correct? What are typical time percentage breakdowns for some common chess engines? I know I could break out Crafty and do some timing, but surely someone here can spout these numbers off the top of their heads.

Thanks much,
Allan
bob
Posts: 20943
Joined: Mon Feb 27, 2006 7:30 pm
Location: Birmingham, AL

Re: quick question about search cost breakdown

Post by bob »

oldman wrote:Hi all,

Correct me if I'm wrong, but the majority of the time spent in chess search is in move generation and evaluation, correct? What are typical time percentage breakdowns for some common chess engines? I know I could break out Crafty and do some timing, but surely someone here can spout these numbers off the top of their heads.

Thanks much,
Allan
Not for me. Last time I profiled, movgen was under 10%. Evaluation, is a different topic and it is over 50% in Crafty. Search itself is the next biggest time user.
Mincho Georgiev
Posts: 454
Joined: Sat Apr 04, 2009 6:44 pm
Location: Bulgaria

Re: quick question about search cost breakdown

Post by Mincho Georgiev »

Same here. The heaviest function in my code once was 'mobility'. Right after it was the pawn evaluation, then piece evaluation, then attack gen, move generation was way under.
User avatar
Onno Garms
Posts: 224
Joined: Mon Mar 12, 2007 7:31 pm
Location: Bonn, Germany

Re: quick question about search cost breakdown

Post by Onno Garms »

I have currently 32% in eval, of which about 10 percent points in mobility. Move generation including mvv-lva-ordering is about 13 percent.
The heaviest function in my code once was 'mobility'
What did you do to get mobility down?
Right after it was the pawn evaluation,
Try pawn hash, worked fine for me.
Mincho Georgiev
Posts: 454
Joined: Sat Apr 04, 2009 6:44 pm
Location: Bulgaria

Re: quick question about search cost breakdown

Post by Mincho Georgiev »

Onno Garms wrote:I have currently 32% in eval, of which about 10 percent points in mobility. Move generation including mvv-lva-ordering is about 13 percent.
The heaviest function in my code once was 'mobility'
What did you do to get mobility down?
Right after it was the pawn evaluation,
Try pawn hash, worked fine for me.
Split up for each piece instead of looping trough piece lists worked fine for the mobility.
I am using pawn hash, that was not the problem. The problem was inserting the info to the pawn fields, which I was able to get down, using bitboards at the current development scheme.
LoopList

Re: quick question about search cost breakdown

Post by LoopList »

oldman wrote:Hi all,

Correct me if I'm wrong, but the majority of the time spent in chess search is in move generation and evaluation, correct? What are typical time percentage breakdowns for some common chess engines? I know I could break out Crafty and do some timing, but surely someone here can spout these numbers off the top of their heads.

Thanks much,
Allan
Look at my thesis on page 13 and 92. Perhaps this could help you. :)

www.loopchess.com

Regards
F. Reul
oldman

Re: quick question about search cost breakdown

Post by oldman »

Great! Thanks for the replies.

I must say that it impresses me that the topic of chess search is so interesting that it seems to always have a decently large community. (my 3 year old daughter wants me to add the mad face: :twisted:)