Search found 4424 matches

by syzygy
Sun Dec 09, 2012 3:54 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Speaking of the hash table
Replies: 89
Views: 16174

Re: Speaking of the hash table

Note that in case of an i7 (not sandy bridge), it shows itself as having 64 bytes cacheline, however the memory controller only stores 192 bytes at a time. Sandy Bridge stores 256 bytes at once. You are forgetting that cache lines can be stolen by other cores between two instructions and even halfw...
by syzygy
Sun Dec 09, 2012 3:24 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Speaking of the hash table
Replies: 89
Views: 16174

Re: Speaking of the hash table

AFAIK hash key collisions and multithreading are the only source of illegal moves in TT. You forget to mention bugs. Not 100% sure, but I think that cache line alignment doesn't protect you from the issue. You are correct, except that OS interruptions are not even needed for this to happen. If one ...
by syzygy
Wed Dec 05, 2012 12:47 am
Forum: Computer Chess Club: General Topics
Topic: Great game played in London
Replies: 6
Views: 1892

Re: Great game played in London

I dont know is there is a serious problem on what people thinks about what an engine can do, or if its a language problem. A comp+program just count beans, and can give a good result, more precise than a human in some (most) cases. But they don't understand nothing. I think it is safe to assume tha...
by syzygy
Tue Dec 04, 2012 11:32 pm
Forum: Computer Chess Club: General Topics
Topic: Great game played in London
Replies: 6
Views: 1892

Re: Great game played in London

I dont know is there is a serious problem on what people thinks about what an engine can do, or if its a language problem. A comp+program just count beans, and can give a good result, more precise than a human in some (most) cases. But they don't understand nothing. I think it is safe to assume tha...
by syzygy
Tue Dec 04, 2012 8:04 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: True iterative search...
Replies: 16
Views: 3253

Re: True iterative search...

I don't think triple null moves do much damage. Null moves that fail low do not affect anything, except that they wasted a little time. And two consecutive moves cannot both fail high. So the third null move will never resort in any effect. And by the time you are there, three null-move reductions ...
by syzygy
Tue Dec 04, 2012 7:43 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Transposition Table - Replacement and PV
Replies: 42
Views: 8331

Re: Transposition Table - Replacement and PV

This is what I use in DiscoCheck currently: - if the keys are different - or (if the keys are equal, but) the depth is >= current entry depth - then overwrite If I understood you right you don't replace an entry for the same position when you encounter this position stored with greater depth in the...
by syzygy
Tue Dec 04, 2012 2:29 am
Forum: Computer Chess Club: Tournaments and Matches
Topic: Ponziani Swiss 8CPU
Replies: 65
Views: 10513

Re: Ponziani Swiss 8CPU

I suppose the idea is to test how well engines do on 8 cores. Some engines are better at using 8 threads than others. This depends on the quality of their multi-processing implementation.
by syzygy
Mon Dec 03, 2012 10:50 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: True iterative search...
Replies: 16
Views: 3253

Re: True iterative search...

Sven Schüle wrote:
syzygy wrote:you might want to forbid two consecutive nullmoves.
Wait until Vincent comes here ... ;-)
I know, and I wanted to add that double nullmove can be OK if you know what you're doing and you're not allowing triple nullmoves. But Vincent might agree that it's best to start simple :)
by syzygy
Mon Dec 03, 2012 10:18 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: True iterative search...
Replies: 16
Views: 3253

Re: True iterative search...

I do store the remaining depht. But as far as I remember I tried with several nullmoves in a line, and perhaps this is causing problems... It might be worth to have a look at it again... Several nullmoves in a line should be fine, but you might want to forbid two consecutive nullmoves. I suppose nu...
by syzygy
Mon Dec 03, 2012 12:06 am
Forum: Computer Chess Club: Tournaments and Matches
Topic: Houdini 3 vs ClusterRybka Game 3 ...
Replies: 34
Views: 5316

Re: Houdini 3 vs ClusterRybka Game 3 ...

S.Taylor wrote:Now that we see Grahams tournament, and Rybka is on top of Houdini with equal haredware, how can it be that the cluster isn't said to be WEAKER than a normal quad?
If Carlsen does not win the tournament he is currently competing in, will you say that he is a WEAKER player than whoever wins it?