MATLAB: Is vectorizing way slower than the for-next loop it replaced

for loopMATLABperformancespeedvectorization

You're recomputing the logical tests multiple times for one thing. I count nine (9) times overall and as many as three (3) of the same result.

Sometimes for loops are just as fast or faster; in the end the compile code has to come down to looping construct for actual execution.

I would think the optimizer would detect those in the same line, not sure it can/will do so over the full function.

iz=(z<=200);
ea(iz)=0;
ea(~iz)=eds(~iz).*(z(~iz)-200)/100;
iz=(z<=100);
td(~iz)=tdo;
td(iz)=tdo*z(iz)/100;

Don't know if will be appreciably faster or not…

Best Answer

The for-loop you've shown is only 4 iterations long (i=1:4). I suspect that's because you abbreviated the loop for testing purposes and forgot to revert back when comparing to the vectorized version. If I'm correct, you won't see an accurate performance comparison until you use the true loop length. If it really is a 4-step loop, then there really isn't enough computation there to significantly improve upon. You need to look at the rest of the code for opportunities for speed-up.

Related Solutions

MATLAB: Why slow 5 times when tic or toc and expression in one line

Funny enough I also get horrible timings. I am using Matlab 2014a and I get:

Elapsed time is 1.053381 seconds.
Elapsed time is 0.216281 seconds.

I will try to explain what I think is going on.

Matlab is interpreted and because of this it has to accelerate loops using JIT (Just-In-Time) compilation. The compiler is not perfect, because it is done rather quickly (it is done "just in time"). It usually manages to convert loops to vectorized code, which is very fast and more comparable to the speed you would get from a compiled language. JIT manages to successfully accelerate the inner loops of a nested loop situation pretty well.

In this case we have only one loop, so it should be very fast. By simply wrapping the for and end statement with tic and toc, we get slow code. This should be very simple for JIT to realize what is going on.

Let us just check a couple of things. First I remove the call to plus, this is not important. Now, let's see

%0.83 s
tic; for k = 1:9e7; end; toc
%0.19 s
tic
for k = 1:9e7; end;
toc

Let's try to arrange the tic and toc a bit

%0.98 s
tic;for k = 1:9e7;
end; toc

Ok, so we agree that it seems that if you place a tic before a for AND a toc right after an end you will not get a accelerated code. You have to move both tic and toc out of the way for JIT to realize what you are really doing.

My guess is that at some point between Matlab 2010 and Matlab 2014, the JIT accelerator rules were changed to allow for this weird behavior.

Also please note that I am not doing anything in the loop, still the code is slow, this confirms that JIT is not that smart ;-)

MATLAB: Surf plot using meshgrid

As z is an array, the following two statements

if z<100

and

elseif z>=100 & z<200

will only be true if every element within z satisfies the conditions, so you always end up with the third result (k=z-100).

To apply the if elseif statements to every element within z, try

x=0:300;
y=x;
[xx,yy]=meshgrid(x,y);
z=(xx+yy)/2;
figure
surf(x,y,z,'EdgeColor','none')
k = z;
k(z>=100) = z(z>=100)-100;
k(z>=100&z<200) = 100;
figure
surf(x,y,k,'EdgeColor','none')

Best Answer

Related Solutions

MATLAB: Why slow 5 times when tic or toc and expression in one line

MATLAB: Surf plot using meshgrid

Related Question