How many plys without pruning?

[AlvaroBegue]

That sounds kind of slow for modern hardware. You should be able to get to around a million without too much work.

What board representation are you using, how are you representing moves and how are you generating them?

[AlvaroBegue]

I just checked with my new [still nameless] engine and I got depth 11.

Correction: I forgot to disable null-move pruning. Without that I only get 10.

[Ralf Müller]

You must consider that I've counting evals per second and not nodes. How many evals do you have?

[Ralf Müller]

I'm not the main programmer but somewhat like L. Kaufman for Komodo. The problem is that my main programmer has leaved the project and so I'm now alone.

[AlvaroBegue]

You must consider that I've counting evals per second and not nodes. How many evals do you have?

For a depth-14 search from the starting position I get 0.79 evaluations per node visited. So it's still something like 3 million per second.

[Ralf Müller]

I'm using a 10x12-Array with Index from 1 to 120 (square-centric).

[hgm]

I have 150,000 - 400,000 evaluations per second.
So my move generation is the problem?

Well, it could also be evaluation. Does your evaluation do a lot? If so, how many evals/sec can you do when you let the eval return material only (or even just zero)?

And do you genrate pseudo-legal moves, or are all moves tested for legality?

[Ralf Müller]

Thanks for the help.
Without eval I have 285,000 - 400,000 evals/sec, so my eval doesn't much. Move generation I've post tomorrow.

But is it now clear, that this is a problöem with move generation/board representation?

Many thanks in advance!

[AlvaroBegue]

I think you can write a very fast engine using 10x12 representation (although I haven't done it myself).

A quick question: Do you have any dynamic memory allocation (malloc, new...) in your code?

[Ralf Müller]

But in any way I must start again from scratch, right?
Yes, I have dynamic memory allocation.