Re: Linux c++ opmization--- linux runs at half the speed of windows?

[Andrew Haley <aph@redhat.com>]

Shane R writes:

 > I hope this is the appropriate forum. Please direct me to the correct one if 
 > it is not.
 > 
 > I am trying to optimize a c++ application that I ported from a windows 
 > system to Linux.

OK.

 > The app is a terminal based application that does some one time file io at 
 > the start then runs completely in memory. After the one time io the app runs 
 > successive timed epochs on the same data in Windows as Linux. The app is a 
 > program that runs some code for doing non-linear optmization (math stuff).
 > 
 > The reason why I am posting is that I timed the time it takes for the 
 > application to complete an epoch. It take twice as long in Linux as 
 > windows?!?!

That is fairly unusual.

 > My system is an Intel Centrino Duo with 2gigs of ram. The application is 
 > only using a fraction of available memory in windows and linux. The 
 > application is single-threaded in both.
 > 
 > I am using Visual Studio 2003 in Windows and when I type gcc -v I get:
 > Target: i486-linux-gnu
 > Configured with: ../src/configure -v 
 > --enable-languages=c,c++,fortran,objc,obj-c++,treelang --prefix=/usr 
 > --enable-shared --with-system-zlib --libexecdir=/usr/lib 
 > --without-included-gettext --enable-threads=posix --enable-nls 
 > --program-suffix=-4.1 --enable-__cxa_atexit --enable-clocale=gnu 
 > --enable-libstdcxx-debug --enable-mpfr --enable-checking=release 
 > i486-linux-gnu
 > Thread model: posix
 > gcc version 4.1.2 20060928 (prerelease) (Ubuntu 4.1.1-13ubuntu5)
 > 
 > 
 > 
 > I am currently using these g++ options:
 > 
 > CFLAGS = -o3 -O3 -march=pentium4 -ffast-math  -funroll-loops -Wall 
 > -Wno-return-type
 > 
 > But I have tried every permutation of the above options to virtually no 
 > effect
 > 
 > The average run time of an epoch in windows is about 3000 milliseconds while 
 > the average run time of an epoch in Linux is 6000!
 > 
 > I don't know if it matters but I am doing calls to the rand() function in 
 > both my windows and linux apps.

It might matter, yes.

I suspect that the right approach is to do some profiling.  Make sure
oprofile is installed, then

sudo opcontrol --init
sudo opcontrol --start
<do your thing>
sudo opcontrol --stop
opreport -l 

Let us know how you get along.  You might be surprised where the time
goes.

Andrew.