Author: Tom Kerrigan
Date: 20:25:18 01/26/00
Go up one level in this thread
On January 26, 2000 at 19:54:18, Dave Gomboc wrote: >On January 26, 2000 at 19:47:07, Dann Corbit wrote: > >>On January 26, 2000 at 19:29:45, Tom Kerrigan wrote: >>>On January 26, 2000 at 19:00:43, Dann Corbit wrote: >>> >>>>It certainly does not sound likely, but without the context of what Hsu actually >>>>said, I can't be sure what was meant. But then, I did not read the threads in >>>>question. I think Bob may have reviewed FHH's book, and maybe he got the idea >>>>from there. What is the thread title. I'd like to take a look. >>> >>>From post 90918: (by Bob) >>>>You overlook the hardware point. The equivalent of 40K instructions. But when >>>>done in the hardware design of DB. This does not mean that 40K instructions >>>>on an X86 would replicate this evaluation... >>> >>>I guess my memory isn't that great and I made a mistake. He never directly said >>>you have to send the DB chip 40k instructions to evaluate a node. However, you >>>can see for yourself that what he did say is even more confusing and still not >>>correct. >> >>Sounds like a classical case of missed communication to me. He said the >>"equivalent" of 40K instructions. That could be (in my mind) any number of >>actual instructions, including one. >> >>One problem with the english language is that it has ambiguities. Three people >>can read the same sentence and get three different meanings out of it. > >The meaning I took was that he estimated it would take 40000 general-purpose CPU >instructions to perform the computations that the custom hardware can do in some >small number of cycles. > >Dave I think that is obviously what Hsu intended. (And he says as much in plain English in the abstract.) Hyatt, when presented with this statistic, seems to disagree. He certainly doesn't say, "Sounds good!" My quote of him (above) is very confusing. Anyway, Hsu's estimate shows that the DB algorithm would run reasonably well on a PC. -Tom
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.