Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Does 64 bit architecture have any benefit on chess programs?

Author: Vincent Diepeveen

Date: 05:00:26 10/01/02

Go up one level in this thread


On October 01, 2002 at 02:16:34, emerson tan wrote:

>Does 64 bit architecture have any benefit on chess programs? If it does, how
>much faster is it compared to 32 bit programs?

there is basically 2 types of chess programs nowadays
  - bitboards
  - non-bitboards

Obviously the bitboard programs will profit. We can guess from alpha
specs that it's about 33% profit.

For the non-bitboarders other things are more interesting.

Such as in random order:
  - is the L1 cache bigger
  - is the BTB (branch target buffer or branch prediction table) bigger
  - is the cpu higher clocked
  - does the cpu have more registers
  - does the compiler support generating instructions that remove branches
  - does the cpu do more instructions a clock, or bigger bundles

A default compile from DIEP at an Itanium2/McKinley for example it is
33% faster at it than it is at a K7 at the same clock.

That's *very* good. Obviously it's a mixture of Level caches and especially
doing bundles of 6 instructions a clock.

Obviously bitboarders profit even more from this mckinley. It's a great
supercomputer CPU!









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.