Author: Dann Corbit
Date: 12:20:34 08/23/05
Go up one level in this thread
On August 23, 2005 at 15:13:18, Tord Romstad wrote: >On August 22, 2005 at 12:59:58, Robert Hyatt wrote: > >>Remember 7-8 years ago when everyone was saying "Bob, you are wasting your time >>on that parallel search stuff, that is for expensive computers, not for PCs. We >>are designing for PCs." > >Yes. Most programmers develop for the computers they have now, and not for >computers a decade into the future. > >>Flash forward. By the end of 2006, most any new computer will have at least two >>cores (two processors). And AMD at least will be offering a quad-core processor >>early in 06. >> >>This "SMP stuff" doesn't seem so useless any longer, does it? Anyone that >>ignores it, is just ignoring free performance that will be on _every_ machine >>within a couple of years... > >I don't believe this at all. It will take *much* longer than a couple of years >before the majority of computers have multiple CPUs. People in this forum >vastly overestimate how often people replace their computers. Even if no >single CPU computer is ever sold after the end of this year, I expect that we >will have to wait at least five years before every machine has more than one >CPU (even for very weak values of "every"). > >I also don't think it is true that most chess programmers ignore SMP >stuff. We just can't start working on it yet because our computers have only >one CPU. When I get a dual CPU computer, I will probably start working on >an SMP version of my program immediately. Unforunately this isn't likely >to happen until some time in 2007. My current computer is less than a >year old, and won't be replaced any time soon. I know someone who wrote an SMP version of his program without a multiple CPU machine (had me test it).
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.