Author: Robert Hyatt
Date: 20:21:09 02/10/00
Go up one level in this thread
On February 10, 2000 at 22:03:24, Tom Kerrigan wrote: >On February 10, 2000 at 17:50:54, Robert Hyatt wrote: > >>There are pros and cons for all implementations. The pros for bitmaps don't >>become apparent until they are run on a 'native' architecture (64 bit machines). >>When all is said and done, it is likely that both approaches are going to end >>up pretty much equal, except for the 'data width' problem. IE an 0x88 program > >But you can look at the "data width" problem from the other direction: bitmaps >don't run efficiently on non-64-bit processors. > >I like having a program that runs reasonably well on a small processor with >small memory... > >-Tom Two important points: (1) 64 bits are the future, not 32 bits, So in 5 years, 32 bits will probably be like 286's today. (2) super-scalar architectures have a problem keeping both pipes full. Bitmap programs aren't as inefficient as first suspected, as they offer thousands of places where two 32 bit and/or/xor/etc operations are needed to complete a 64 bit operation. Super-scalar eats that alive and makes the penalty much less than expected. IE I'll bet a bitmap program actually executes 2 instructions per clock way more than a non-bitmap program, which means part of the 'loss' is covered by clever hardware...
This page took 0.01 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.