Help for optimizing a code

Hi

I'm trying to optimize a code with nested for loops (originally written in another language ), and I try to get it in the proper "matlab" form... I've already asked many questions in this forum but now I don't know if I'm going the proper way or not. I've attached to this post a m file with the original loops, and the way I think optimized...

Could anyone of you have a look and tell me if it's just rubbish or if I'm going the good way ?

Thanks...

[Geoff Hayes]

Stéphane - your m file isn't attached. Make sure you press the Attach File button after you have chosen it. Note that you can just edit your above question and use the paperclip button to add the file.

[Sean de Wolski]

Why do you want to remove the loops? other than looks is there a reason? For performance, there's no guarantee a vectorized method is going to be faster and it might take longer than the total running time to vectorize it.

[Stéphane]

Geoff > Thanks... now it's attached :-)

Sean > It's only for speed as it takes really long time with small matrices and I need to increase their size by a lot...

[Geoff Hayes]

Stéphane - does your testLoop.m code contain the small matrix that takes a really long time to evaluate? When I ran the attached, tic and toc indicated that the elapsed time was 0.056486 seconds which is pretty quick…

Note that there did seem to be some differences between dm and dm2. Have you observed the same?

[Stéphane]

That's the problem when change a code without carrefully checking...

I've uploaded a slighty different version, differences are on the edges (if I'm not mistaken) but I don't use them.

The time is indeed of 0.056 sec for the "no-loop" version, but I need to use this function 10^6 times for now, and I'll have to increase the matrices dimension by a factor at least 100... So you can imagine that any split second save would be great...

Thanks for your help :-)

Stéphane