LinuxQuestions.org - Strange Perforamce

Hello.
I am using fedora core 1, programming a large system in c++.
Recently I have noticed that one part of my program is taking too much time, and I have no idea how to fix it. The part is responsible for making difference image between 2 images.
The code is:
-------------------------------------------------------------------------
void
Frame::MakeDiff() {
//substract frame1 from frame 0
int w,h,bpp;
int i,len;
unsigned short *s0,*s1, *sd;

mFr0->GetSizes(w,h,bpp);//w=640, h=480, bpp=12
len = w*h;
s0 = (unsigned short*)mFr0->GetPtr();//pointer to im1 data
s1 = (unsigned short*)mFr1->GetPtr(); pointer to im2 data
sd = (unsigned short*)mFrD->GetPtr();pointer to im dif data
for (i=0; i<len; i++,s0++,s1++,sd++)
*sd = *s0 - *s1;
}
-------------------------------------------------------------------------
running the code takes 10 msec.
If I change last 2 lines of the function to:
-------------------------------------------------------------------------
for(i=0; i<len; i++)
sd[i] = s0[i] - s1[i];
-------------------------------------------------------------------------
the function takes 4 msec. It is strange for me too, because seems that the same job is done.
Originally the function should take and took 1 msec on my system. I believe that maybe the problem is in adding other parts of my system which are not connected to that function but I have no idea how to find and deal it.
Please help me if You have any ideas what could cause such problems and/or You know how to deal it, or maybe any programs with memory which will guide me to the root of the problem.
Thanks in Advance,
Leonid