Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: HW based Crafty (Boule's thesis)

Author: Tom Kerrigan

Date: 01:26:37 04/03/02

Go up one level in this thread


On April 02, 2002 at 23:41:05, Robert Hyatt wrote:

>This depends on what is implemented. IE I suggested including a FPGA
>MakeMove() and UnmakeMove() function to take care of the hash updates.

If you're talking about 64-bit hash keys, that would be a SHITLOAD of space on
an FPGA. It would take 3072 CLBs just to store an _int64[2][6][64], which puts
it clear out of the realm of any hobbyist FPGAs, and moreover, the busses around
the CLBs are designed such that it would take several cycles to read out any of
those _int64s, not that you would want to be adding them in parallel anyway...

>does the search.  That would result in minimal PCI traffic...  and let the
>hardware run at full FPGA speed, which might be well beyond 30M nodes per
>second easily...

I think the best clockspeed you could possibly hope for with chess logic is
~50MHz. Even the fully custom DB chips required several cycles per node, so 30M
NPS is completely out of the question.

>Deep Thought's evaluation was based directly on Belle's...  Deep Thought
>was simply a "belle on a single chip" where belle was a huge batch of FPGA's

I don't think FPGAs had been invented at the time.

-Tom



This page took 0.03 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.