Author: Andrew Wagner
Date: 08:06:44 03/20/04
Go up one level in this thread
On March 20, 2004 at 10:02:04, Steven Edwards wrote: >Symbolic: Progress report 2004.03.20 > >A few more hours into Symbolic were expended with the following progress: > >1. Symbolic now reports its PV and various search statistics back on the stream >interface in addition to the selected move. There is now enough data passed >back to support the EPD acd, acn, acs, pm, and pv opcodes, thus making running >EPD test suites a bit more informative. Testing this with BWTC shows that the >total per position search overhead is under one second on a modest notebook. > >2. The narration facility has been improved and now supports paragraph formation >with word wrap on both the real time output and the post search narrative file >output. This is a minor feature, but it helps make the debugging process >easier. I've included options to format positions in narrative for: > >-- Color using ANSI escape sequences (default) >-- Monochrome >-- CCC style FEN display strings Sounds to me like you're doing an awful lot of formatting work. On a project with such a radical foundation, don't you think making sure the general idea works first should be a priority? > >A sample unedited narrative file is copied to the end of this post. > >A planned future option is to generate narrative output as HTML files with >hyperlinks connecting search nodes and plans. I'm going to put this on the >deferred list for now, though. > >3. More thought has been put into the pattern representation topic. I may have >some more of the pattern instances be generated directly by functions than by >matching; this is basically an efficiency issue. Work on pattern stuff should >fill the next several weeks of effort. > >The road is load and sometimes appears to be getting longer. Keep the fight! I can't wait to see what you come up with! > >[Sample narrative] > >The selection process is started at 2004.03.20 09:37:12. > >I created the search tree with it's root node and candidate set. > >There are twenty candidate moves available. > >The root position: > >[D] rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1 > >An opening book candidate move is chosen. > >The result position: > >[D] rnbqkbnr/pppppppp/8/8/4P3/8/PPPP1PPP/RNBQKBNR b KQkq - 0 1 > >I've selected the move e4. No expectation is available. The predicted >variation is: (e4). I've used 0.09 processor seconds in making the selection. >There are twenty-one total nodes in the search tree. There are twenty leaf >nodes in the search tree. There is one interior node in the search tree. My >selection rate is 233.333 nodes per second. My average node processing time is >4.28571 milliseconds. No info on what it based its selection of e4 on? What is "expectation"? 233 nps? Do you think that's a typical result? Seems slow to me. > >The selection process is ended at 2004.03.20 09:37:12.
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.