Linux versus Windows....

Discussion of chess software programming and technical issues.

Moderators: hgm, Rebel, chrisw

Michel
Posts: 2272
Joined: Mon Sep 29, 2008 1:50 am

Linux versus Windows....

Post by Michel »

Hi,

I wrote a little generic bench marking utility for UCI engines. See

http://alpha.uhasselt.be/Research/Algeb ... nch/devel/

The idea was to be able to quickly and routinely compare different OS'es, CPU's, compilers etc (for the same engine). This came out of some discussions at the Toga development forum.

I ran some tests in Linux and Windows. You can find the results on the above URL. The tests were run on a laptop with a Pentium M@1.6GHz.

Two noteworthy conclusions.

(1) A Windows engine generally runs at the same speed under wine as on native Windows.

(2) I made a Linux binary of Toga II 1.4b5c with the Intel compiler. Strangely enough it seems to run slower than the corresponding Windows binary which is made with the same compiler. I suspect this is an issue of compiler options. I still have to investigate this (although any pointers to "magic" compiler options would be welcome).

I am kind of curious what the situation on OSX is.

Regards,
Michel
Michael Sherwin
Posts: 3196
Joined: Fri May 26, 2006 3:00 am
Location: WY, USA
Full name: Michael Sherwin

Re: Linux versus Windows....

Post by Michael Sherwin »

How a program lines up in memory from one compile to the next makes a difference. If a single line of code in a non time critical section is added the program may run slower. Add a line of code that is not cheap to a time critical section and the program might run faster!
If you are on a sidewalk and the covid goes beep beep
Just step aside or you might have a bit of heat
Covid covid runs through the town all day
Can the people ever change their ways
Sherwin the covid's after you
Sherwin if it catches you you're through
Michel
Posts: 2272
Joined: Mon Sep 29, 2008 1:50 am

Re: Linux versus Windows....

Post by Michel »

(2) I made a Linux binary of Toga II 1.4b5c with the Intel compiler. Strangely enough it seems to run slower than the corresponding Windows binary which is made with the same compiler. I suspect this is an issue of compiler options. I still have to investigate this (although any pointers to "magic" compiler options would be welcome).


It appears that the secret is profiling. Using the profiling facilities of the Intel compiler I could make a Linux binary of Toga II 1.4b5c which seems to run faster on my system (a Pentium M@1.6GHz) than the standard Windows binary (as measured by my utility). See

http://alpha.uhasselt.be/Research/Algeb ... _versions/

The numbers are

Code: Select all

Toga II       1.4b5c      windows               584k
Toga II       1.4b5c       wine                 580k
Toga II       1.4b5c    linux/intel             599k
Toga II       1.4b5c    linux/gcc 4.2           484k
Regards,
Michel