Author: Ernst A. Heinz
Date: 07:25:15 05/21/98
Go up one level in this thread
On May 21, 1998 at 09:25:33, Robert Hyatt wrote: >On May 21, 1998 at 09:14:21, Ernst A. Heinz wrote: > >>On May 21, 1998 at 08:29:51, Robert Hyatt wrote: >> >>>I see two problems to solve. First you can get a hash hit, but no >>>suggested >>>move... >> >>Just don't store entries without a suggested best move in the table. > >you do realize that 1/2 of the positions fit this category? IE *every* >fail low node (where you store the flag "UPPER_BOUND" has no "best" move >associated with it, since *every* move was bad.... So I'm not sure I >follow what you mean here... Fail-soft alpha-beta also suggests "best moves" at fail-low nodes. These moves are usually good enough for move ordering such that we skip internal iterative deepening in these cases. >My memory requirements seem modest, based on tests on my quad-alr... >which >doesn't slow down at all whether I run 1 crafty or 4 identical copies >doing >the same search, while other programs can take a 20% hit quite easily >when >I run two of them at once... I did not mean "main memory" but "L1 cache" requirements. But you know "Crafty" better than I do, of course (the last time I looked into your sources was roughly one year ago). =Ernst=
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.