Author: Steven Edwards
Date: 15:34:35 06/08/05
Symbolic: Status report 2005.06.08 I made a pass through Symbolic's Lisp source narration code and made a number of naming and structural changes to enable a greater consistency between the HTML and text output formatters. These routines now share as much code where possible, otherwise they at least share a very common structure and naming convention. The changes effectively removed the kruft that had accumulated during the deletion of the voice narration and the addition of the HTML narration. Also, a healthy number of Lisp "assert" function calls were added to assuage coder's paranoia. The new StoryToText function now outputs position diagrams in monochrome suitable for a typical interactive text pager. Prior to the modification position diagrams were output in color using ANSI escape sequences; unfortunately, these are changed into unreadable trash when passed through a Unix more command. The new StoryToHTML function retains the prior HTML position diagram encoding capability. StoryToText also knows about CCC style FEN diagram encoding and will generate diagrams in this style instead of the default monochrome ASCII board diagram, if commanded by a run time option. StoryToHTML, if activated by option, produces the HTML narrative on the file "Story.html". StoryToText, if activated by option, produces the text narrative on the file "Story.txt". The WriteStatReport function, if activated by option, produces a brief text statistics report on the file "Report.txt". All three file names are defaults and can be changed; indeed, for a whole game, changes are needed as otherwise the files are overwritten after each search. The regular real time text narrative, if activated by option, appears on the console. It uses ANSI color position diagram encoding by default. -------- The ChessLisp source file that had most of the GA subproject code (SexLib.lsp) has been split with all of the code needed for actual chess search real time support deposited into the new file BioLib.lsp. Also, the initialization and application of the MateAttack move suggesting organism has been integrated into the main code. Testing of this using the MateIn10 BWTC.0031 position along the 19 ply main line results in some encouraging data: in nine of the ten choice points for the first mover, the MateAttack suggester ranked the key move as best of all the available moves; in the tenth choice point, the key move was also ranked as best but had several equally scored alternatives. This is a good result, although it's tempered a bit by the fact that it's somewhat specialized. More testing is needed to see if the above suggester works as well for other MateAttack theme instances. Less encouraging is the amount of time needed to apply the suggester organism to a complicated position: about a full second when running on a 400 MHz PPC. I have some ideas on how this can be improved, but simply moving to a faster machine is seen as a last resort. -------- As you may already know, I've been a big fan of the PPC architecture vs. the thirty years kludges of the iAPX86 alternative. So it was a big disappointment to hear of Apple Computer switching from PPC to iAPX86 starting next year. This will not affect Symbolic's development; it has only a single line of PPC assembly language that can be easily changed or elided, also, Symbolic's binary data is already normalized for endian status. Also to ensure portablilty, Symbolic has been tested from day one on iAPX86 Linux platforms. If only IBM has devoted the resources to follow up on its promises of a 3 GHz PPC in 2004 and a 4 GHz PPC in 2005, then Apple would surely not have had to switch. Maybe there is still hope of a switch back to PPC if IBM can catch up and pass Intel. -------- A recent visit to the doctor has revealed a new heart problem and so tomorrow I get to spend some time with the Cardiac Care prognosticators at the local hospital for further testing. Depending on the results, it may be a while until the next update.
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.