The VXL Project

Hi all,

after long time I remembered that I wrote this speed Subject. 
So I did some speed tests with variable declerations within
loops and outside loops. Unfortunately I am a beginner regarding
vxl and thus a freshman on this point. Well, I used  vul.timer e.g.
vul_timer tic;
....
tic.print(vcl_cerr)

I calculated this on a stereo program I wrote myself
which runs around 0.035 sec/frame on VGA images with
ZMCC as correlation function  from the vcl package. 
So there are a lot of variable declerations and loops. 

 
I must say that I only had the impression that the outside
loop  decleration is faster. This is because the vul_timer 
speed measurement is not very exact so that I can not give
you an objective value. It has only three digits after zero and 
it changes from one calculation to the other.

Best Regards

Emrullah

Peter.Vanroose@... schrieb am 09.03.04 12:30:30:
> 
> > Isn't it faster to do the declarations out of the loop, for example
> > instead of
> > for (unsigned p=0;p<np;++p,planeA += pstepA,planeB += pstepB,planeS += pstepS)
> > {
> > const aT* rowA = planeA;
> ...
> > using following technique:
> > unsigned p;
> > const aT* rowA;
> > for ( p=0;p<np;++p,planeA += pstepA,planeB += pstepB,planeS += pstepS)
> > {
> > ...
> 
> Modern C++ compilers should implement both ways identically, certainly
> when optimisation is on (but probably also without optimisation).
> I have never tested this, though.
> 
> 
> -- Peter.


__________________________________________________________________
Zeigen Sie Emotionen mit der WEB.DE Bild-SMS! Das Bild ist gratis,
Sie bezahlen nur den Versand. http://freemail.web.de/?mc=021196