Search found 433 matches

by Aleks Peshkov
Mon Apr 11, 2016 9:05 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: New open source engine - Pigeon
Replies: 8
Views: 2869

Re: New open source engine - Pigeon

Unpacking a move map (which is bit masks) into a variable-length move list is still a scalar operation, and the main flow in Engine::NegaMax() is not SIMD-aware yet. I'm working on hooking it up now, and then we'll see how well things scale. Fingers crossed. (Does that make sense?) I think there is...
by Aleks Peshkov
Wed Mar 30, 2016 9:08 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Some questions from a beginner
Replies: 27
Views: 9409

Re: Some questions from a beginner

Your board representation is bitboards+array, so you do not synchronize them but update both independently each time your make move on the board. There are other data (like position hash key) you have to update incrementally during each move because it is faster to update then regenerate it again an...
by Aleks Peshkov
Fri Feb 26, 2016 8:51 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Best way to debug perft?
Replies: 26
Views: 7351

Re: Best way to debug perft?

hgm wrote:or return 1 at depth==0.
I am not sure that '1' is correct sum of empty set.
by Aleks Peshkov
Fri Feb 26, 2016 6:38 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Best way to debug perft?
Replies: 26
Views: 7351

Re: Best way to debug perft?

Perft is recursive number of legal moves.

Perft for depth 1 is number of legal moves from the given position.
Perft for depth N is sum of legal moves for all positions of Perft N-1.
by Aleks Peshkov
Mon Feb 08, 2016 8:45 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Invalid move
Replies: 17
Views: 3724

Re: Invalid move

I cannot say better then Bob had said already.

Your nearest goal should be not to make another mediocre chess engine but too study and practice programming skills.
by Aleks Peshkov
Thu Feb 04, 2016 10:10 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Buckets for pawn hash?
Replies: 25
Views: 5803

Re: Buckets for pawn hash?

bob wrote:The purpose for buckets is to provide better replacement decisions.
I do not know any better then LRU replacement scheme for pawn hash buckets. And LRU works badly with recursive nature of search tree.
by Aleks Peshkov
Thu Feb 04, 2016 1:13 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Buckets for pawn hash?
Replies: 25
Views: 5803

Re: Buckets for pawn hash?

I personally doubt that pawn hash is good idea from the beginning. It is the hash with the highest hitrate (by far) and saves you to collect all those backward, passed, isolated pawns over and over again. Because the hitrate is so high the pawn hash size can be small so using buckets is probably no...
by Aleks Peshkov
Thu Feb 04, 2016 8:48 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Buckets for pawn hash?
Replies: 25
Views: 5803

Re: Buckets for pawn hash?

I personally doubt that pawn hash is good idea from the beginning.
by Aleks Peshkov
Wed Feb 03, 2016 2:22 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Transposition Age Tutorial
Replies: 95
Views: 16883

Re: Transposition Age Tutorial

A “significant performance drop” as compared to what? An infinite size table? Different positions have drastically different success hit rate. No single number can be given. I discovered that my 3+2 slot scheme becomes remarkably worse against standard 4-slot bucket scheme at the moment when averag...
by Aleks Peshkov
Fri Jan 29, 2016 1:12 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Transposition Age Tutorial
Replies: 95
Views: 16883

Re: Transposition Age Tutorial

So what he's saying is if you're designing your engine for say a 4 GB hash when playing 40 moves in 120 minutes and your test games are running at say 40 moves in 1 minute, then you should probably be testing with only a 32MB hash table to get comparable hash pressure to compare the schemes. Exactl...