home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: gnu.chess
- Path: sparky!uunet!retix.com!dspencer
- From: dspencer@spock.retix.com (David Spencer)
- Subject: Re: Iterative Deepening
- Message-ID: <1993Jan25.214918.18646@spock.retix.com>
- Organization: Retix, Santa Monica, CA USA
- References: <C15r40.D1r@sju.edu> <21659@ucdavis.ucdavis.edu> <JON.93Jan25110340@cas.org>
- Distribution: gnu
- Date: Mon, 25 Jan 1993 21:49:18 GMT
- Lines: 36
-
- In article <JON.93Jan25110340@cas.org> jon@cas.org (Jon Vander Hill) writes:
- >>>>>> On 22 Jan 93 07:34:07 GMT, lim@toadflax.cs.ucdavis.edu (Lloyd Lim) said:
- >
- >> In article <C15r40.D1r@sju.edu> tmoody@sju.edu (T. Moody) writes:
- >>>
- >>>My real problem is with (b). Levy and Newborn say
- >>>something about searching at a given level and storing lots of positions
- >>>in a hash table before moving to the next level. Is this the only way
- >>>to do it? I don't know a thing about hash tables and, from what I have
- >>>read it looks like it's over my head.
- >>>
- >>>So my question is, can iterative deepening be usefully (if not
- >>>optimally) accomplished without using hash tables?
- >
- >> Iterative deepening is commonly used with transposition (hash) tables,
- >> but it doesn't require them. Iterative deepening just means increasing
- >> the search depth by one each time (i.e. - searching to a depth of 1,
- >> then 2, then 3, ...). You don't need to use transposition tables.
- >> They are just used to speed things up, because many of the same nodes
- >> are encountered again when you search to depth n+1.
- >
- >One the the more expensive routines in a chess programs is the
- >move generation function. The transposition table limits the
- >number of times that move generation is called to once per position.
- ^^^^^
- Um, I don't think that this is so. From the aritcles I've read and
- the source code I've seen, the "typical" transposition table entry
- has the best move for a position, but *not* all the pseudo-legal
- moves for that position. So it's possible that the stored bounds
- and stored move doesn't cause a cutoff, thus the legal moves for
- the current position may have to be generated *again*.
-
- I guess that in general the move generation function may not be called
- more that once per pos, but it certainly doesn't limit the calls to
- once per position.
-
-