Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: branch misprediction

Author: Dan Andersson

Date: 12:36:48 07/14/04

Go up one level in this thread


 I recommend you use wikipedia insted to get up to date entries. Throw some
money at them when you get rich.
 http://en.wikipedia.org/wiki/Main_Page
 http://en.wikipedia.org/wiki/Branch_prediction
 Notice the similarity ;)
 As for what to do to reduce mispredictions there is some work done in
Abstract/Virtual Machine research. See for example:
 M. Anton Ertl and David Gregg.  The behaviour of efficient virtual machine
interpreters on modern architectures.
 http://www.complang.tuwien.ac.at/papers/ertl%26gregg01.ps.gz
 And Ertls home page: http://www.complang.tuwien.ac.at/anton/home.html
 Interpreter/AM/VM designers have been working on this problem for a long time.
But the short answer is to get as good move order as possible and have unique
code segments with conditional branches for each type of node closest to the
leaf node.
 This will of course blow up the working code size and may cause cache trashing
and thus not lead to speed gain. So it is neccessary to benchmark different
compiles. And it might well turn out to be too much of a burden adding yet more
compile options.

MvH Dan Andersson



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.