Ciao, i have a question about setting stockfish and multiple core.
If you have more core/cpu, you can change the Uci option Threads.
but if you have 4 or 8 or 16 or 32 threads is it possible to set a different parameter for
Razoring
Futility Pruning
ProbCut
or some other variable that seems a constant like 4 or 7 or 10 (*oneply) in source.
In order to process an higher number of nodes, being slower than a common multi/core that simply adds threads?
is it possible that having so many threads, being faster, could be not efficient as being more accurate?
If stockfish already has an algorithm to do that, can you tell me where is (source/variable?)
ciao e grazie
Quello che vorrei capire è se esiste un numero di tread oltre il quale non più senso andare in profondità , quanto piuttosto, rendere più precisa possibile la valutazione.
P.S. per gli strafalcioni in inglese, è colpa di google.
stockfish and horizon effects question
Moderators: hgm, Rebel, chrisw
-
- Posts: 22
- Joined: Tue Feb 04, 2014 9:52 pm
- Location: Italy Milan
-
- Posts: 20943
- Joined: Mon Feb 27, 2006 7:30 pm
- Location: Birmingham, AL
Re: stockfish and horizon effects question
Why would you want to do that? The goal for parallel search is to gain speed to go deeper. If you become less aggressive with pruning or reductions, you are actually losing speed. Even though the overall engine is faster with N threads and less aggressive stuff than it is with 1 thread and normal settings, what would be the justification for becoming less selective? And if there is a justification, why does it not apply equally as well to the single-thread version?acconcio wrote:Ciao, i have a question about setting stockfish and multiple core.
If you have more core/cpu, you can change the Uci option Threads.
but if you have 4 or 8 or 16 or 32 threads is it possible to set a different parameter for
Razoring
Futility Pruning
ProbCut
or some other variable that seems a constant like 4 or 7 or 10 (*oneply) in source.
In order to process an higher number of nodes, being slower than a common multi/core that simply adds threads?
is it possible that having so many threads, being faster, could be not efficient as being more accurate?
If stockfish already has an algorithm to do that, can you tell me where is (source/variable?)
ciao e grazie
Quello che vorrei capire è se esiste un numero di tread oltre il quale non più senso andare in profondità , quanto piuttosto, rendere più precisa possibile la valutazione.
P.S. per gli strafalcioni in inglese, è colpa di google.
-
- Posts: 22
- Joined: Tue Feb 04, 2014 9:52 pm
- Location: Italy Milan
Re: stockfish and horizon effects question
thanks you for the answer, it is very clear.
Now I have to ruminate over.
ciao and thank you very much
Now I have to ruminate over.
ciao and thank you very much