Author: Magnus Homann
Date: 08:19:34 12/22/99
Go up one level in this thread
On December 20, 1999 at 15:51:32, Greg Lindahl wrote: >On December 19, 1999 at 10:18:32, Magnus Homann wrote: > >>The problem with using the FPGA as a custom "co-processor" is that >>communicatuion CPU-FPGA likely swamps any other speedup you might get. > >This is a worry, but less of one than you think. It takes about 1usec to grab >the PCI bus and communicate a few bytes to a card. However, once you're talking >to a PCI card and everyone is obeying the rules for write combining, etc, you >can communicate a fair amount of information in a short time. And how long does it take to read the data back? You'll get about 1-10 MB/s on the PCI bus, unless you watch out a bit. >Array processors are a classic example: it used to be worthwhile to move real >numbers over the bus in order to add them. Now it isn't. Encryption is another >example where a fairly large number of bytes have to be moved. Chess invovles >moving much less data than that. Really? So how much data do you typically move? So far, I have no clue what you expect the FPGA to do. Homann
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.