Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: What is the speed difference that programs earn from 64 bit systems

Author: Dann Corbit

Date: 12:39:00 05/02/05

Go up one level in this thread


On May 02, 2005 at 12:20:42, Matthew Hull wrote:

>On May 02, 2005 at 12:03:56, Daniel Pineo wrote:
>
>>On May 01, 2005 at 23:27:40, Eugene Nalimov wrote:
>>
>>>On May 01, 2005 at 21:01:17, Daniel Pineo wrote:
>>>
>>>>On April 30, 2005 at 20:22:22, Christos Gitsis wrote:
>>>>
>>>>>On April 29, 2005 at 16:26:28, Uri Blass wrote:
>>>>>
>>>>>>I read that Shredder9 for 64 bit is released and my question is what is the
>>>>>>speed difference that program earn from compiling for 64 bits when you need also
>>>>>>64 bit window to run it.
>>>>>>
>>>>>>Can somebody tell me about the speed gain in nodes per second of all the free
>>>>>>source code programs including programs that do not use bitboard like tscp or
>>>>>>fruit and bitboard programs like Crafty.
>>>>>>
>>>>>>Uri
>>>>>
>>>>>I have tested two different versions of Crafty-19.19 using my Athlon 64 3400+:
>>>>>
>>>>>- Under 32-bit Windows XP the fastest version I could find in the Internet
>>>>>(among about 8 different executables) runs at an average of about 1.5 million
>>>>>NPS and has reached a maximum of 1.7 MNPS.
>>>>>
>>>>>- Under 64-bit Gentoo Linux it runs at an average of 2.2 million NPS and has
>>>>>reached a maximum of 2.48 MNPS.
>>>>>
>>>>>The increase in speed is 40~50%.
>>>>>
>>>>>I hope this helps
>>>>
>>>>
>>>>Don't you think switching from windows to linux might have something to do with
>>>>that 40~50% increase?
>>>
>>>Last time I asked AMD people to check, 64-bit Visual C generated better code
>>>than 64-bit GCC 4.0.
>>
>>I would expect so, microsoft makes quality stuff nowadays.
>>
>>
>>>
>>>So on 64-bit Windows increase would be even larger.
>>>
>>>Thanks,
>>>Eugene
>>
>>I was just pointing out that giving the nps increase from 32bit windows to 64bit
>>linux doesn't really pinpoint the benefit of goint to 64bit since some of the
>>increase may be due to the OS change.
>
>
>I think the OS has very little to do with it.  The compiler has a lot to do with
>it and the program itself in terms of it's ability to profit from 64-bit
>architecture.  AFAIK, bitboard programs tend to benefit more than others.  The
>OS counts for next to nothing by comparison.

I think for sure that this is the correct analysis.

I would be interested to know the version of GCC used.

I have heard interesting projections about GCC 4.0, but the version I built has
problems finding the include files so I cannot use it.




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.