Search found 643 matches

by smatovic
Thu Jun 16, 2011 8:15 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Possible Board Presentation and Move Generation for GPUs?
Replies: 14
Views: 3732

Re: Possible Board Presentation and Move Generation for GPUs

For the record: i have implemented a Kogge-Stone move generator with 8 Threads (8 directions) in parallel on a GPU but it could not outperform a Magic-Bitboard Move Generator with only one thread. So i will use a Quad-Bitboard Board Representation with Magic Bitboards for sliders and precalculated A...
by smatovic
Sat Jun 11, 2011 9:13 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Chess Evaluation
Replies: 3
Views: 995

Re: Chess Evaluation

So does some engines have better eval than others? Of course, Zeta Dva for example has a poor evaluation, when it plays against another engine which searches at the same depth it generally looses the game. You can make some tests on your own and let Engines play on a fix search depth against each o...
by smatovic
Fri Jun 10, 2011 7:33 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: max amount of moves from a position?
Replies: 33
Views: 12038

Re: max amount of moves from a position?

i think about an parallel processing scheme for all possible moves of an position...maybe i will use 128 an build some kind of overflow handling in.

--
Srdja
by smatovic
Fri Jun 10, 2011 7:22 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: max amount of moves from a position?
Replies: 33
Views: 12038

Re: max amount of moves from a position?

and that he had managed to construct a position with 218 legal moves:
hmm, does "construct" also mean that this position can be reached during a legal game or is it only a theoretical set of pieces?

--
Srdja
by smatovic
Fri Jun 10, 2011 4:39 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: max amount of moves from a position?
Replies: 33
Views: 12038

max amount of moves from a position?

a basic question:

How many possible, legal moves can i get from one board position?

I thought it were 218, but maybe it is less?

--
Srdja
by smatovic
Wed Jun 08, 2011 12:42 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: GPUs better for chess than CPUs?
Replies: 31
Views: 6992

Re: GPUs better for chess than CPUs?

to pain, frustration, and failure. :) As an alternative to YBW i thought on using a depth*128x128 board array. If you run 128 threads and assume 128 max possible moves you get an array of boards[max_searchdepth][128 Threads][128 possible boards] With such an array you could move up and down in bloc...
by smatovic
Wed Jun 08, 2011 3:20 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: GPUs better for chess than CPUs?
Replies: 31
Views: 6992

Re: GPUs better for chess than CPUs?

Hello Srdja Sad to hear you stopped working on it. What was your major difficulty ? Can you elaborate on your experience a bit ? I read that you have used it for move generation but I do not quite get what you have tried regarding search.. I don't think recursion is a problem because alpha-beta can...
by smatovic
Tue Jun 07, 2011 4:49 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: GPUs better for chess than CPUs?
Replies: 31
Views: 6992

Re: GPUs better for chess than CPUs?

The entire concept of alpha-beta pruning was designed in the era of slow serial processing cpu's. With today's enormous, much faster, AND CHEAP parallel processing power available, alternative methods to alpha-beta pruning should be considered. You are right, but consider: Modern high-end GPUs have...
by smatovic
Tue Jun 07, 2011 4:13 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: GPUs better for chess than CPUs?
Replies: 31
Views: 6992

Re: GPUs better for chess than CPUs?

The point is to feed in a parallel-alpha-beta search thousands of threads with boards considering the known GPU-limitations (no recursion, no sub-threads). "Holger Hopp" describes in his german Diplom Thesis a way to implement YBCW for an synthetic game tree on an Mas-Par SIMD computer. But the Mas-...
by smatovic
Sat Jun 04, 2011 12:16 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Zeta, a chess engine in OpenCL
Replies: 14
Views: 5440

Re: Zeta, a chess engine in OpenCL

Project is canceled.

SIMD friendly move generation is one thing, but I see no way to implement an SIMD friendly search algorithm in OpenCL which is able to feed thousands of threads...

--
Srdja