Hey,
I got a question about the abdada algorithm described at https://chessprogramming.wikispaces.com/ABDADA.
How would one implement PVS withing the ABDADA-Algorithm-Framework or should I not implement PVS inside ABDADA ?
Is there a good open source implementation of the abdada algorithm that uses PVS (and MoveOrdering) I can have a look at ?
Would appreciate any help.
Thanks guys
abdada algorithm/parallel srchea
Moderators: hgm, Rebel, chrisw
-
- Posts: 4367
- Joined: Fri Mar 10, 2006 5:23 am
- Location: http://www.arasanchess.org
Re: abdada algorithm/parallel srchea
ABDADA is a kind of ancestor of the technique now called "LazySMP" (https://chessprogramming.wikispaces.com/Lazy+SMP). Quite a few programs are using the latter with PVS, including now (I think) Stockfish.
--Jon
--Jon
-
- Posts: 273
- Joined: Wed Aug 24, 2016 9:49 pm
Re: abdada algorithm/parallel srchea
Thats good to know. When using parallel alpha beta, does each thread have history-heuristics,counter-moves,killers or is this shared among all threads ? As far as I understood, alpha, beta values are not shared among threads so each thread would need to save the principal-variation seperately, I am I right ?jdart wrote:ABDADA is a kind of ancestor of the technique now called "LazySMP" (https://chessprogramming.wikispaces.com/Lazy+SMP). Quite a few programs are using the latter with PVS, including now (I think) Stockfish.
--Jon
-
- Posts: 4367
- Joined: Fri Mar 10, 2006 5:23 am
- Location: http://www.arasanchess.org
Re: abdada algorithm/parallel srchea
That is pretty much your choice I think. For me they are per-thread.CheckersGuy wrote:When using parallel alpha beta, does each thread have history-heuristics,counter-moves,killers or is this shared among all threads ?
.As far as I understood, alpha, beta values are not shared among threads
They are shared in the sense that there is always a global beta value that is the upper search window and if any search in any thread exceeds this, you get beta cutoff. And there is also a global alpha, which is the best score found so far in any search.
--Jon