Author: Heiner Marxen
Date: 15:54:14 11/10/99
Go up one level in this thread
On November 10, 1999 at 13:55:51, Dann Corbit wrote: >On November 10, 1999 at 04:20:21, Bas Hamstra wrote: >>It looks like threads will be continually created and destroyed. Doesn't that >>give a lot of overhead? >Actually, the threads will not ever be destroyed and all are created at startup. > The pool of threads will use event synchronization to change tasks. The main >thread will always have a full CPU, so it will have the full benefit a standard >engine would. It will act like a conventional chess program, but it will get >special information contributed by the other threads. A second thread (mate >solver) will always have a full CPU, so it will find any mate a mate solver >would find. I will ask Heiner Marxen about using his mate engine at some point, Now, that starts to be quite interesting. I had quite a similar idea for using a mate solver as part of a playing engine as early as 1980. Then there were no multi-CPU machines to use, and not many chess programs I could embed something into, so I lost interest. The basic idea seems to be sound to me. Whether a mate solver has to contribute much, and how often, would need lots of experiments. The mate solver thread would also need to be redirected, when the main engine decides for a move (or receives one from the opponent). >otherwise I will have to write my own. Anyway, his mate solver is light years >ahead of all the rest. The other threads will get new assignments, continue, or >abandon tasks as scheduled. I don't know how to explain it better, but I see >the image clearly in my mind. Besides pondering the actual position, a mate solver might be used to look at "promising" sacs. Just an idea. -- Heiner Marxen heiner@drb.insel.de http://www.drb.insel.de/~heiner/
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.