Author: Robert Hyatt
Date: 17:04:01 11/19/02
Go up one level in this thread
On November 19, 2002 at 19:49:49, Bob Durrett wrote: >On November 19, 2002 at 18:26:26, Robert Hyatt wrote: > >>On November 19, 2002 at 16:31:43, Bob Durrett wrote: >> >>> >>>I hope this is not the end. There must be something BETTER. >> >>Possibly. Nobody even says bitboards are _the_ answer to every question. And >>quite probably >>as hardware evolves, so will the data structures that take advantage of the new >>hardware features. >>Until we had decent hardware support for bitboards in the old CDC and Cray >>machines, they >>were not very popular, for example... >> >> >>> >>>Maybe changes in computer capabilities and architectures will eventually make >>>bitboards obsolete? And how about invention and innovation? How about TODAY? >>> >>>Over-reliance on bitboards may throw us into another "dark ages." The chess >>>engine gurus need to break out of that cardboard box. Too much "resting on >>>one's laurels." >> >>You miss the point. Using arrays is the _old_ approach. Bitboards are much >>newer. >>So those that have made the jump have changed. It is not unreasonable to expect >>that >>there will be another big "jump" in the future at some point. IE many have >>jumped >>to SMP search, even though "some" have been doing it for 25 years now. Many >>that have jumped to it were saying 5 years ago "that is a waste of time, you >>will never >>have two processors in a home computer." Now, with hyper-threading, you >>_do_.... as >>will most new machines from here on... > >Yes, the CCC thread(s) on hyper-threading are very interesting. As an outsider, >I see this as a product of the general trend toward more complex >microprocessors. Where can this lead? i.e. what is this new technology >pointing toward? Maybe hyper-threading is a fledgling new techology which will >blossom into something really spectacular. Maybe the ways chess engines are >programmed today will have to change a lot? > >I also wonder what this new hyper-technology means to the "world of >multiprocessor computers." [ I hope I'm not asking how to do brain surgery >this time. : ) ] > >Bob D. > One thing it will lead to is better "bitmap instructions". IE the Cray has always had a "popcnt" instruction to count the number of one bits set. IA32 doesn't so we jump thru hoops. One day, maybe not... > >> >> >> >> >>> >>>: ) : ) : ) : ) : ) : ) : ) : ) : ) : ) >>> >>>Bob D.
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.