Search found 594 matches

by petero2
Mon Jan 06, 2020 4:11 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: RMO - Randomized Move Order - yet another Lazy SMP derivate
Replies: 113
Views: 6062

Re: RMO - Randomized Move Order - yet another Lazy SMP derivate

Ok I have made history and refutation tables thread-local now but it seems there is no change at least for depth=8 will test 14 later. Score of scorpio-sht-8-thread vs scorpio-sht-1-thread: 66 - 65 - 68 [0.503] 199 My changes are very simple: make the tables local by uncommenting static, and clear ...
by petero2
Sun Jan 05, 2020 10:36 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: RMO - Randomized Move Order - yet another Lazy SMP derivate
Replies: 113
Views: 6062

Re: RMO - Randomized Move Order - yet another Lazy SMP derivate

Edit My result after the fix is not as great as Peter's but it is still there Score of stockfish-8-thread vs stockfish-1-thread: 3629 - 1130 - 1380 [0.704] 6139 +151 elo compared to +251 of Peter. Stockfish git hash I used is from Jul 14, 2019: commit 7090d2561ae9ce3803bbc04319c4c93f I tried that v...
by petero2
Sun Jan 05, 2020 8:39 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: RMO - Randomized Move Order - yet another Lazy SMP derivate
Replies: 113
Views: 6062

Re: RMO - Randomized Move Order - yet another Lazy SMP derivate

// Iterative deepening loop until requested to stop or the target depth is reached while ( (rootDepth += ONE_PLY) < DEPTH_MAX && !Threads.stop && !(Limits.depth && mainThread && rootDepth / ONE_PLY > Limits.depth)) Perhaps this is the culprit? If I read it correctly then the depth limit is only res...
by petero2
Sun Jan 05, 2020 5:56 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: RMO - Randomized Move Order - yet another Lazy SMP derivate
Replies: 113
Views: 6062

Re: RMO - Randomized Move Order - yet another Lazy SMP derivate

I modified this logic to make it never exceed the fixed depth limit and ran 8 threads vs 1 thread again at fixed depth 10. The result was +42.5 elo after 5536 games. Thanks. Interesting! This points indeed to a bit of widening. But I would argue that the elo difference is quite small for 3 thread d...
by petero2
Sun Jan 05, 2020 10:55 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: RMO - Randomized Move Order - yet another Lazy SMP derivate
Replies: 113
Views: 6062

Re: RMO - Randomized Move Order - yet another Lazy SMP derivate

Peter, have you seen my latest results. ... You said it is a mix of ABDADA + Lazy so that d/d+1 is a potential problem ... My implementation does not use the classical d/d+1 scheme: * I disabled the lazy SMP "depth + 1" trick, as I don't know if that makes sense in combination with ABDADA. Disablin...
by petero2
Sun Jan 05, 2020 1:14 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: RMO - Randomized Move Order - yet another Lazy SMP derivate
Replies: 113
Views: 6062

Re: RMO - Randomized Move Order - yet another Lazy SMP derivate

I think it still needs to be established that Lazy indeed searches a substantially wider tree (this is testable to some extent). It is not very obvious to me which mechanism might cause this as all the threads use basically the same search. Has anyone ever tested SHT-2 threads vs SHT-1 thread at fi...
by petero2
Mon Dec 30, 2019 2:56 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Trying to understand check evasion move generation.
Replies: 4
Views: 552

Re: Trying to understand check evasion move generation.

I'm trying to understand Texel's check evasion move generation so I can implement something similar in my engine. https://github.com/B4dT0bi/texel/blob/master/src/moveGen.cpp I think I mostly understand it but there are a couple of things I'm not sure about. validTargets |= pos.pieceTypeBB(OtherCol...
by petero2
Mon Oct 28, 2019 8:15 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Another take on DTS?
Replies: 20
Views: 2849

Re: Another take on DTS?

BTW, for the record, this was the idea from Harry Nelson (one of the Cray Blitz team) way way back... Tim Mann re-discovered it when porting Crafty to the Alpha. Not sure whether he found Harry's old paper that mentioned it, or if he re-discovered it on his own. In either case, It was not "my" idea...
by petero2
Mon Oct 28, 2019 6:15 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Another take on DTS?
Replies: 20
Views: 2849

Re: Another take on DTS?

* Each cluster node has its own local transposition table, which works the same as in a normal SMP engine. In Texel's case I use your lockless hashing algorithm. Thanks for inventing that. What's this lockless hashing algorithm? I'm definitely interested in it. Lockless hashing has also been referr...
by petero2
Sun Oct 27, 2019 10:02 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Another take on DTS?
Replies: 20
Views: 2849

Re: Another take on DTS?

Nothing wrong with that. EXCEPT that you are accepting a significant performance degradation (no global transposition table) ... The nice thing about SMP/NUMA implementations is that the latency is near zero for SMP and not a whole lot higher for reasonable NUMA configurations. It is certainly true...