Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Spoke to soon <s>

Author: Robert Hyatt

Date: 20:19:36 08/10/01

Go up one level in this thread


On August 10, 2001 at 14:37:50, Les Fernandez wrote:

>On August 10, 2001 at 12:47:01, Shep wrote:
>
>>On August 10, 2001 at 08:03:37, Les Fernandez wrote:
>>>Strictly from a speed point of view the above computer is approxiamtely 50,000
>>>times faster then the hardware that was used in the Deeper Blue hardware!
>>
>>Sorry Les, but this is utterly wrong.
>>
>>Deeper Blue had a speed of 200 - 1,000 million _nodes_per_second_ in chess.
>>This is something completely different from saying it had 0.2 - 1 billion
>>instructions per second.
>>
>>Your calculation would be right if "node" == "instruction", but no computer (not
>>even a highly specialized one as DB) can examine one node in one instruction.
>>
>>---
>>Shep
>
>Yes Shep you are correct, my statement would only be true if node=intruction.
>While we are on the subject how many instructions was Deeper Blue's hardware
>able to do? Just curious.
>
>Thx
>
>Les


It could do at least the following:

1.  search.  Given the current hardware position, search N plies using normal
moves, then do a quiescence search and a static eval at the endpoints.

2.  make/unmake.  Update the current hardware position given the from and
to square and so forth.

3.  generate attacks to or from a specific square.  I'm not sure this was in
all versions, but it was used both for debugging and as a hardware move
generator used by the software search to increase speed by avoiding a software
move generator.

4.  give a static evaluation of the current hardware position.

There were a few others to load the evaluation term matrix, etc...

Nothing like a general purpose machine as you can tell..



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.