Re: computer-go: unmake move?

[David Fotland <fotland@xxxxxxxxxxxxxxx>]

Usually you are unmaking moves near the leafs, so the
old position is still in the cache.  So memory bandwidth
is not a problem.  Even if there is no cache, so copies
go to memory, Pentium PCs should be able to copy over 100 MB/s.
If the state is 1000 bytes, that limits performance to 100,000
positions per second.

For my chess program, I coded an undo function, and the copy/no undo
approach.  The copy approach was faster.

David

At 10:08 AM 6/8/00 +0900, Nobuhiro Yoshimura wrote:
>Don Dailey wrote:
>
>>
>>Note that modern processors are extremely fast at copying data that is
>>layed out consecutively in  memory, believe  me, you  can have  a very
>>fast   program that uses  this  technique.    
>
>Can you aply this rule to PCs ?    I thought that memory bandwidth 
>is very narrow for PC.
>
>
>Can you tell me the size of "POSITION" ?
>
>>
>>
>>/*  nega-max search */
>>/*  --------------- */
>>
>>int  search( position *state,  int depth,  int alpha, int beta )
>>{
>>   POSITION  new_state;   /* allocate state */
>>
>>
>>   foreach move in (0 .. number_of_moves)
>>   {
>>     make_move( state, &new_state, move ); 
>>     score = -search( &new_state, depth-1, -alpha, -beta );
>>
>>     /* No need to unmake!!! */
>>
>>     if (score > best_found_so_far) 
>>     {
>>	...
>>	
>>     }
>>
>>   }	
>>
>>   return( best_found_so_far );
>>
>>}  
>>
>
> 
>Nobuhiro Yoshimura 
> 
>Konami Computer Entertainment Tokyo     The opinions expressed 
>Office: yosimura@xxxxxxxxxxxxxxxxx           here are mine and not 
>  Home: yoshimura@xxxxxxxxxxxxxxxxx               necessarily those of Konami.
>
>