Search found 17076 matches

by bob
Mon Aug 20, 2007 12:07 am
Forum: Computer Chess Club: General Topics
Topic: EGTB's accesed from the internet?
Replies: 14
Views: 7630

Re: EGTB's accesed from the internet?

You can easily do that. But the cost is prohibitive, in that the program will play so incredibly weakly because of the internet latency. Currently you do a probe and get an answer back in a few milliseconds. On the internet, multiply that by 100 or 1000, and it would absolutely kill the search speed.
by bob
Sun Aug 19, 2007 6:49 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: re-inventing the SMP wheel
Replies: 81
Views: 38203

Re: re-inventing the SMP wheel

I am not seeing how this avoids the classic race condition, which would allow multiple processors to search the same exact tree. If you introduce critical sections, I believe their overhead will _far_ exceed the overhead of doing a normal split algorithm like DTS or YBW... When you start a brand new...
by bob
Sun Aug 19, 2007 12:19 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: re-inventing the SMP wheel
Replies: 81
Views: 38203

Re: re-inventing the SMP wheel

The problem is this:

You probe A and miss and do a cache line fill. Now you search all kinds of nodes, and then return to store position A's results. What's the probability it is still in L1/L2?? Pretty thin I would suspect...
by bob
Sun Aug 19, 2007 12:17 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: handling repetition in the hash question
Replies: 10
Views: 4808

Re: handling repetition in the hash question

Yes, bad problem... To avoid this 'hash contamination' Joker only stores scores in hash that the position would have if none of the other tree nodes had been visited yet (but nodes from game history have). This requires the path between first and second occurrence to be run twice, once to deteremen...
by bob
Sat Aug 18, 2007 4:50 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: re-inventing the SMP wheel
Replies: 81
Views: 38203

Re: re-inventing the SMP wheel

Well, I hash-probe every node I search. That requires a read from DRAM. If that is not 'beyond expensive', why would this be? Accessing a table in a shared L2 seems very cheap in comparison with DRAM access. Here's the problem: you probe once, which is not cheap. But now you have to (a) store posit...
by bob
Sat Aug 18, 2007 1:25 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: re-inventing the SMP wheel
Replies: 81
Views: 38203

Re: re-inventing the SMP wheel

We seem to be miscommunicating. In my terminology node = tree node, not thread or processor. In each thread, all nodes in the tree it is searching on the direct path from root to position currentlly begin searched (i.e. the top of its recursion stack) would be open. All other nodes in that tree wou...
by bob
Sat Aug 18, 2007 1:20 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: nondeterministic testing
Replies: 86
Views: 34612

Re: nondeterministic testing

I know that if you repeat the test of Nunn 2 times you obtain different values and also I know that more games and rivals better, in that we agree, but if you have a single computer, to play 2500 games 15 + 10 is impossible. When your engine is weak, I do not believe that you must make such number ...
by bob
Sat Aug 18, 2007 1:18 am
Forum: Computer Chess Club: General Topics
Topic: Something wrong with the testing at CEGT?
Replies: 11
Views: 4286

Re: Something wrong with the testing at CEGT?

Very "nice" headline :evil: What if there is something wrong with the 4CPU implementation inside Deep Fritz 10? Actauly,there are a lot of engines which perform better using 2 cpu's and CEGT is not guilty at all,this phenomena is observed in other rating lists.... It speaks to some testing issues. ...
by bob
Fri Aug 17, 2007 4:58 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: re-inventing the SMP wheel
Replies: 81
Views: 38203

Re: re-inventing the SMP wheel

Well, with 100 threads each searching upto 40 ply deep, there would only be 4000 open nodes max. And as I would prefer split points close to the leaves, most of the lines would largely coincide, so 400 is a more likely number (not to mention that my number of threads is more likely to be 8 than 100...
by bob
Fri Aug 17, 2007 3:38 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: nondeterministic testing
Replies: 86
Views: 34612

Re: nondeterministic testing

I know that if you repeat the test of Nunn 2 times you obtain different values and also I know that more games and rivals better, in that we agree, but if you have a single computer, to play 2500 games 15 + 10 is impossible. When your engine is weak, I do not believe that you must make such number ...