Author: Robert Hyatt
Date: 09:00:57 04/10/02
Go up one level in this thread
On April 09, 2002 at 16:29:11, Tom Kerrigan wrote: >On April 09, 2002 at 10:03:49, Robert Hyatt wrote: > >>with a bit of care... My number hasn't changed one iota... Still well >>over 3000 at least... > >And how do you figure that? > >-Tom It is my number, obviously. Would you like some .S output for Search, InCheck, GenerateMoves, MakeMove, UnmakeMove, NextMove and so forth. All of those are done for most every node. Then there is the code in Evaluate that is done for 10-30% (typically) of all nodes searched. I just "add"... On the Cray, I just "observed" by using the hpm program... either way works for me. Cray Blitz required 7,000 instructions per node. Computed by running searches for a minute or two, then dividing total nodes searched into the total number of instructions executed. 7,000 is not a "great" number because quite a few of those "instructions" were vector operations that produced several results for one instruction. But it is a number that can be measured. I suspect it could be measured on the Intel machine, I just haven't looked into the machine-specific register stuff closely enough to see how to capture the number of instructions retired over a given interval...
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.