So it looks 'roughly' like the hash table gives you about a 2/1 speed up at around depth 10....or about 1 to 2 extra ply for the same time.
Have i interpreted this correctly ?
Laurie
How to speed up my engine
Moderators: hgm, Rebel, chrisw
-
- Posts: 199
- Joined: Sun Nov 03, 2013 9:32 am
Re: How to speed up my engine
So at depth 10 the hash version takes about half the time. Do people think this and the previous post is a typical result (all things being equal).
If i implement a hash table my engine would reach the same depth in about half the time.
If it did, it would still not have time to search the next ply fully since my branching factor would impeed this ???
Laurie
If i implement a hash table my engine would reach the same depth in about half the time.
If it did, it would still not have time to search the next ply fully since my branching factor would impeed this ???
Laurie
-
- Posts: 199
- Joined: Sun Nov 03, 2013 9:32 am
Re: How to speed up my engine
So at depth 10 the hash version takes about half the time. Do people think this and the previous post is a typical result (all things being equal).
If i implement a hash table my engine would reach the same depth in about half the time.
If it did, it would still not have time to search the next ply fully since my branching factor would impeed this ???
Laurie
If i implement a hash table my engine would reach the same depth in about half the time.
If it did, it would still not have time to search the next ply fully since my branching factor would impeed this ???
Laurie
-
- Posts: 97
- Joined: Mon Jun 25, 2012 10:16 pm
- Location: Forks, WA
- Full name: Ben Nye
Re: How to speed up my engine
All of the rest is minor details. Transposition tables are important.lauriet wrote: No transposition table.
Laurie
Ben
-
- Posts: 4675
- Joined: Mon Mar 13, 2006 7:43 pm
Move generation for check evasion
Move generation for check evasion
I may have missed the answer, but have you implemented move generation explicitly designed for check evasion? This can provide a tremendous speed-up for nodes where the king is in check with only a very few legal moves but otherwise with dozens of illegal moves which have to be made and unmade -- that a big waste.
I may have missed the answer, but have you implemented move generation explicitly designed for check evasion? This can provide a tremendous speed-up for nodes where the king is in check with only a very few legal moves but otherwise with dozens of illegal moves which have to be made and unmade -- that a big waste.
-
- Posts: 199
- Joined: Sun Nov 03, 2013 9:32 am
Re: Move generation for check evasion
Thats a good point. At the moment I have a legal move generator, which checks each generated move to see if the king is still in check.......this appears to be dumb, so I will work on changing this AND then your suggestion will help from there on in.
Laurie.
Laurie.
-
- Posts: 27808
- Joined: Fri Mar 10, 2006 10:06 am
- Location: Amsterdam
- Full name: H G Muller
Re: Move generation for check evasion
Actually the fraction of nodes where you are in check is quite negligible, so whatever you do there will hardly impact overall performance. But you might still have to do some extra testing for checks if your normal test for seeing if you can capture the King uses shortcuts that would not be justified if the King was already under attack.
-
- Posts: 7220
- Joined: Mon May 27, 2013 10:31 am
Re: How to speed up my engine
Why do you want a faster engine ? There are already fast engines enough.
-
- Posts: 7220
- Joined: Mon May 27, 2013 10:31 am
Re: How to speed up my engine
Maybe there should be more tournaments for slow engines that beginner or moderate chess players can watch.
-
- Posts: 27808
- Joined: Fri Mar 10, 2006 10:06 am
- Location: Amsterdam
- Full name: H G Muller
Re: How to speed up my engine
I agree. The ChessWar promo division was the most fun of all to watch. Things like TCEC are utterly boring, if you are not an engine yourself.