Author: Klaus Friedel
Date: 11:09:22 12/15/04
Go up one level in this thread
Hello, both Snitch and Hagrid use about the same model. One controller-interface two implementations. One for UCI one for XBoard. The controller runs in its own thread, receives the commands from the GUI, parses them and calls the right methods from search. Search in return calls the controller to send search-infos (PV etc.). So search is "protocoll free", knows nothing about UCI or XBoard. Regards, Klaus On December 15, 2004 at 09:44:20, Volker Böhm wrote: >Hi Uri, > >I like the "controller"-model most. In Practice Spike has a Winboard-Class that >is the controller. The Chess-Player itself doesn´t know anything about winboard >but only add´s an interface for the winboard-class. > >If Spike gets a winboard command, the winboard - class receives it and decides >what to do. Example: If it receives a "go" command it will send call >"IterativeSearch" of the "PonderSearch" Class. If it receives a stop it will >call "Stop" of the "PonderSearch" Class. > >The UCI-Interface uses exactly the same interface of "PonderSearch". > >The PonderSearch Class is statefull (not a good design but easy to implement) >thus it knows about the last moves. > >Spike breaks the rule when printing informations. PonderSearch knows about the >format the UCI or Winboard Info-Interface. This is not a good design. A callback >interface of Winboard is a better design but more compicated to implement. Here >we have choosen the easier design, not the more clear one. To implement this, >"PonderSearch" has a flag for "Winboard" or "UCI" that is set by the Winboard >class - the drawback of this design. > >Greetings Volker
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.