Author: Dann Corbit
Date: 17:23:42 04/05/99
Go up one level in this thread
On April 05, 1999 at 20:12:38, Hristo wrote:
>On April 05, 1999 at 18:01:54, Dann Corbit wrote:
>
>>On April 05, 1999 at 17:46:28, Hristo wrote:
>>
>>>On April 05, 1999 at 12:47:59, Joeri wrote:
>>>
>>>>I want this program when will it be available ?
>>>>I stress that i heard about it in EUROPE ECHECS !!! I want to have this program
>>>>i really am fed up with all those updates or the bulk of commercial programs I
>>>>want the program the software that beated kasparov just to have it ! I am tired
>>>>before the 2000 year i want it badly and only that !!!
>>>>Am i mad for this - give me some thoughts ?
>>>>Joeri.Ramon@rug.ac.be
>>>
>>>You are mad .. :)))) !!!
>>>But so am I ! :)))
>>>I wish they(IBM) would release the source code(at least). In todays world this
>>>is not too much to ask. It will probably never happen.
>>
>>Source code won't do you a lick of good!!!
>>
>>When CPU hardware instructions are things like: Bg7 Nf3 Nc6 Bc4 d6 d3 Nf6 e5
>>dxe5 Nxe5
>>!!!
>
>Bummer .... Didn't think of that. :)))
>Did they actualy write the entire thing in Assembler ?! If they did all I can
>say is *DAMN*!
Actually, the source code portion is largly in C. However, the C is used to
control custom CPU's that have chess moves as their native opcodes. So if you
get ahold of that source code, what will you do with it?
>>So why would you even want it!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
>>;-)
>
>At some point they must have an Eval function of some sort or more likely a
>multitude of them bound in function table(s) which correspond to a given type of
>position. So, we can possibly see, how does the order of evaluation(extensions)
>changes with the different type of positions. Perhaps even see the individual
>evluations(extensions).
>Then again, is this possible?!?
I suspect strongly that the eval function was software/firmware combination
rather than fully in Silicon (Gallium Arsenide?). But again, their algorithm
might say {rough possible outline}, "Use brute force out to 10 plies, then null
move out to 20 plies and then selective search from 20 plies forward. Have
these three CPU's doing positional evaluation, and these 15 doing pure tactics
and these 10 using our special extensions, look for killers. Meanwhile, use the
other 2 CPU's to communicate with the other clusters of CPU's and process
special requests."
How will you use something like this without their CPU clusters?
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.