🚀 go-pugleaf

Sorry about the delay.  I was thinking about my design and realized that I 
would not get the full advantage of Alpha/Beta reductions.  With some 
refinements I think it can still work.  I am also interested in any ideas and 
improvements that you could provide.  Here is the basics of what I have so 
far.  

The idea that I had was to use the transposition tables as message boards for 
evaluations from other systems.  
When a deep evaluation is started on a position, a request will be sent to the 
master to look up the position in the global transposition file.  If the master 
has an evaluation it would be returned otherwise a tag would be placed on 
the position to indicate it is under evaluation.  Any subsequent requests for 
the same position and depth would get an "In-progress" response.  
When a deep evaluation is completed the score is sent to the master which 
updates the global table and forwards the score to all systems that had 
requested the position.
A slave system receiving an evaluation from the master can look up the 
position in the local transposition table to determine what action is required. 
If the position is locked it is on the current evaluation tree so the evaluator 
should roll back to the position and use the provided score.  If the position 
is marked "In-progress" the evaluator will be working on an alternate path.  
The score would be backed up to update the score of the parent position and 
determine if an Alpha/Beta cut should be applyed.  An unmarked position entry
is just updated with the new (deeper) score for future reference.

What I still need to work on is how to divide the work assignments between 
the slave nodes and picking an estimated Alpha/Beta window to start the 
evaluations before the true values are known.

	Dan Oetting	<oetting@gldfs.cr.usgs.gov>
<.sig not found>