Search found 4166 matches

by syzygy
Tue Jul 24, 2012 11:32 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: pawn enumeration
Replies: 130
Views: 10327

Re: pawn enumeration

If you ever want a chance for having 7 men - storing more than win/draw/loss is gonna be a problem. With current hardware 7-piece tables are just too big to handle in any reasonable way, let alone to use them during search. When I can generate them in RAM, I will start to get interested. The 5 men ...
by syzygy
Tue Jul 24, 2012 11:08 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: pawn enumeration
Replies: 130
Views: 10327

Re: pawn enumeration

Please don't tell me you give a course programming somewhere :) Except the: "How my methods can slow you down at modern hardware - let's start writing in the same cacheline with different threads" My threads aren't writing in the same cacheline, unless by chance. p.s. Forward generating is of cours...
by syzygy
Tue Jul 24, 2012 11:01 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: pawn enumeration
Replies: 130
Views: 10327

Re: pawn enumeration

DTM is not so clever nor anything similar. it's 1 or 2 bytes. If you use 50 move rule, you have still a 7 bits need an entry. That's way too much. For up to 5 pieces, my WDL50+ tables are 438MB and my DTZ50+ tables are 668MB. The DTZ50+ tables are half-sided and make use of the WDL50+ information. ...
by syzygy
Tue Jul 24, 2012 10:47 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: pawn enumeration
Replies: 130
Views: 10327

Re: pawn enumeration

File slicing helps in two ways compared to exhaustive slicing. a) size is manageable and always 8 b) The slices are self-sufficient since you don't need to load other slices when pawn moves. The latter will require you to load the necessary helper slices manually. But exhaustive slicing could be ad...
by syzygy
Tue Jul 24, 2012 10:25 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: pawn enumeration
Replies: 130
Views: 10327

Re: pawn enumeration

This is much more difficult than you think. Even enumerating n alike pieces requires some weird math. Look in my other post for a code snippet. Now merging this for two groups of white and black pawns, you see it is a nightmare. Even though I think it may be possible to find an enumeration function...
by syzygy
Tue Jul 24, 2012 7:01 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: pawn enumeration
Replies: 130
Views: 10327

Re: pawn enumeration

I am trying to generate 6 men bitbases with pawns. The pawnless need about 4G of RAM but those with single pawn need as much as 13G without pawn slicing. In that case dividing the pawn table into 4 files (based on the leading pawn) would reduce 13G to 3.25G < 4G, which should be sufficient. My simp...
by syzygy
Tue Jul 24, 2012 1:36 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: pawn enumeration
Replies: 130
Views: 10327

Re: pawn enumeration

The problem is how to enumerate the pawns in closer to promotion order so that they can be sliced. During generation, what you need is an indexing function with the property that if position A can be reached by pawn moves (and piece moves) from position B, then index of position A < index of positi...
by syzygy
Thu Jul 19, 2012 11:44 am
Forum: Computer Chess Club: General Topics
Topic: Komodo and Don...
Replies: 10
Views: 2497

Re: Komodo and Don...

MM wrote:What i know precisely is that for several different reasons, me, that i am a customer and i made a donation to Komodo project, i don't have anything after so many months and after having read so many news about Komodo that never transfer to releases.
Did you pay for something you did not get?
by syzygy
Wed Jul 18, 2012 11:55 am
Forum: Computer Chess Club: General Topics
Topic: Longest Mate
Replies: 14
Views: 2325

Re: Longest Mate

Letting the number of moves depend on available egtb knowledge seems silly to me. Not to me. It means that your rule is not artificially interfering with the game. If a contrived rule is needed to ensure the game ends, IMO it should be as permissive as possible, so that the other rules take priorit...
by syzygy
Wed Jul 18, 2012 11:46 am
Forum: Computer Chess Club: General Topics
Topic: Longest Mate
Replies: 14
Views: 2325

Re: Definitive explanation.

Then, I opened the merged file made with Merge_PGN.bat with Notepad and searched one by one, until I found a checkmate in 1090 half-moves: [d]8/6R1/8/6N1/3k1K2/1B6/7n/7q w - - I was thinking this can't be right, since the longest win in this tablebase is well-known to be 517 (full) moves (see e.g. ...