Search found 17071 matches

by bob
Thu Aug 30, 2007 2:44 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: SMP thread goes here
Replies: 39
Views: 16568

Re: SMP thread goes here

My apologies, I always read in chronological order. :) Here's my reply from the other thread: I don't like the stack idea although I had originally considered it. The problem is this. A single thread splits at A, then B, then C, then D, then E. And while it is working on E, the other thread working...
by bob
Wed Aug 29, 2007 11:54 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: SMP thread goes here
Replies: 39
Views: 16568

SMP thread goes here

Let's re-start the SMP thread as things are now all on the right-hand side, and with several joining in, it is hard to see what goes with what...
by bob
Wed Aug 29, 2007 11:52 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Frustrations of a mediocre chess programmer. What next?
Replies: 34
Views: 15809

Re: Frustrations of a mediocre chess programmer. What next?

I am not sure Winboard would be much faster; it might all be OS delays. I always use Winboard_x under PSWBTM to set up gauntlets. There you can see the results while the match is running. I do not know what is PSWBTM practically I am not interested in book in my testing so my testing is from the no...
by bob
Wed Aug 29, 2007 11:47 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: re-inventing the SMP wheel
Replies: 81
Views: 37358

Re: re-inventing the SMP wheel

I have a thing called a "split block" that represents the local tree state a particular thread is working on. Any split block represents a point in the tree where two or more processors are working together. So I have those split blocks linked "horizontally" (sibling or brothers/sisters links). Tha...
by bob
Wed Aug 29, 2007 10:01 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: re-inventing the SMP wheel
Replies: 81
Views: 37358

Re: re-inventing the SMP wheel

The problem is that the description of the state-of-the-art methods is either unduly long, or rather vague. The questions I am left with are mostly about implementation details, the various strategies are clear enough. E.g., if it says that a thread, finding a cut-move in a split point, has to abor...
by bob
Wed Aug 29, 2007 9:52 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Testing the waters for opening move
Replies: 6
Views: 2800

Re: Testing the waters for opening move

Hi all, I have created an opening book out of gm2600. Earlier my program exited from the opening phase with an inferior position at times, it did not understand how to proceed and did not recover. To beat this I now allow my program to search with the book move for a small time of about 1/5th of re...
by bob
Wed Aug 29, 2007 5:52 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Testing the waters for opening move
Replies: 6
Views: 2800

Re: Testing the waters for opening move

Hi all, I have created an opening book out of gm2600. Earlier my program exited from the opening phase with an inferior position at times, it did not understand how to proceed and did not recover. To beat this I now allow my program to search with the book move for a small time of about 1/5th of re...
by bob
Wed Aug 29, 2007 5:51 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Testing the waters for opening move
Replies: 6
Views: 2800

Re: Testing the waters for opening move

Hi all, I have created an opening book out of gm2600. Earlier my program exited from the opening phase with an inferior position at times, it did not understand how to proceed and did not recover. To beat this I now allow my program to search with the book move for a small time of about 1/5th of re...
by bob
Wed Aug 29, 2007 5:20 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: re-inventing the SMP wheel
Replies: 81
Views: 37358

Re: re-inventing the SMP wheel

Well, the main point is actually how best to do SMP search.. :? So far the idea of having the owner of split-nodes hand out a limited number of moves to would-be slaves via the hash bucket seems a nice way to do it. Only the polling to know if the current search should be aborted might become a pai...
by bob
Wed Aug 29, 2007 3:29 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: re-inventing the SMP wheel
Replies: 81
Views: 37358

Re: re-inventing the SMP wheel

Well, it isn't that bad, because usually the TLB misses are satisfied from L2. Usually the data misses L2 before you start missing the TLB, as manufacturers are careful to provide at least enough TLB entries to address all of L2. (I guess an increase of L2 size would hardly show up in the benchmark...