Author: Robert Hyatt
Date: 09:54:44 01/23/00
Go up one level in this thread
On January 23, 2000 at 12:07:11, Tom Kerrigan wrote: >On January 23, 2000 at 10:14:09, Robert Hyatt wrote: > >>On January 23, 2000 at 03:01:38, Tom Kerrigan wrote: >> >>>On January 22, 2000 at 23:50:46, Robert Hyatt wrote: >>>>in the great scheme of things very much. You want to make silly suggestions, >>>>and then don't like it when someone points out why they are silly. As far as >>> >>>How about my silly suggestion that DB would run reasonably fast on a PC? Based >>>on FHH's own estimates, I guess it's not that silly, eh? >>> >>>I've tried to reason with you, others have tried to reason with you, it just >>>doesn't work. Don't expect me to post to this thread any more. >>> >>>-Tom >> >>I won't follow up any more either, as 'shouting' doesn't make something >>true. But you have _never_ explained why Hsu would invest a year or more >>to convert his code to the PC (probably more like 2 years). IE I'm not >>going to spend any time trying to get my code to run on a blender's cpu. >>There is no reason to do so... > >*sigh* I can't help myself... > >I actually DID explain why Hsu would invest a year or more to convert his code >to the PC: money. talk to the current commercial chess programmers. What is the probability that a new program would sell enough to pay Hsu at least $100-150K per year, his likely salary requirement after working for IBM? Nearly zero. Why not write the world's greatest tax-preparation software, as he could make a lot more money doing that. But it would be somewhat boring, don't you think? At least my opinion... I think he would view the PC rewrite as less of a challenge and more of a gamble and certainly a drudge... Who wants a thing that says "I can do everything the Intel pentium cpu can do, on your cheap old 486 cpu. Oh yes, it runs 6-9 orders of magnitude slower, but it is the same instruction set, etc... > >Here's what you _never_ explained: why would it take 2 years to make DB/PC? I >can throw a good chess program together in a month, and probably less if I >already knew which terms and weights I wanted in the evaluation function. Even >if FHH didn't want to start from scratch, what's to keep him from directly >porting the DB software? What makes the SP so different from a 4-way Pentium >Xeon? Or a 1-way Celeron, for that matter? Message passing? Using an array of 16 chess processors? Using the hardware in the chess processors to hold the board position, make/unmake moves, etc? And then after designing the DB eval to run in hardware, where everything can be asynchronous and parallel, rewriting all of this for a sequential computer where performance is going to be a critical issue? _lots_ of problems. _lots_ of time. Practically _no_ chance of it paying him for doing so. >Zugzwang is a pretty huge program >that usually runs on a massively multiprocessor PowerPC computer, but it only >takes a few minutes to get Zugzwang to run on a PC, too. Are you trying to say >that the DB software is so poorly engineered (perhaps riddled with SP assembly) >that porting it would take years? I know that MUCH bigger programs have been >ported from the PC to the Mac in a matter of weeks... > >-Tom I am trying to say, apparently unsuccessfully, that rewriting a big chunk of hardware stuff to run in software is a big project. The design decisions would be totally different to make it efficient. Not just a "port". A total re-write. I did this in the late 70's somewhere, taking a communications multiplexor designed in hardware, and re-wrote it to run on a microprocessor-based system. I basically started from scratch, because what you can do in hardware and what you can do in a programming language are light-years apart when speed becomes an issue. Special purpose hardware will _always_ dominate general-purpose stuff in terms of speed. Always has, always will.
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.