Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: next deep blue

Author: Robert Hyatt

Date: 19:38:53 01/23/00

Go up one level in this thread


On January 23, 2000 at 19:01:15, Tom Kerrigan wrote:

>On January 23, 2000 at 17:29:38, Robert Hyatt wrote:
>
>>reread the abstract to the IEEE micro article.  First sentence I believe.
>>"the most complex eval of any program..."
>>I believe Hsu believes that.  Since he wrote it.
>
>So how do you explain Hsu's estimate of 40k instructions per node?
>
>I think Hsu is probably right about "40k", and 95% right about "most complex."
>
>-Tom

I personally wouldn't know how to make that 40K estimate.  IE I couldn't
begin to tell you how many "GP instructions" it would take to do my eval in
Cray Blitz.  Without a _lot_ of work.  He said it had about 8K weights.
Maybe he simply said "OK, 8K weights = 8K patterns to match = 8K comparisons,
8K branches, 8K instructions, 8K branches around the else, and 8K instructions
for the else:

If (P) s1
else s2;

that would be a _wild_ guess.  And might be within a factor of 10 or so of
reality.  who knows?

As far as "95% right about the eval" how do you assess that?  Knowing _nothing_
about deep blue II...



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.