Accelerating MATLAB Performance

35Proling MATLAB

®

Performance

The reason for this is that while subFunc is now being run only once instead of 100

times, it is actually only the rst iteration which takes any signicant amount of time (at

least now that we have implemented our data-load caching mechanism). We still spend

nearly 0.9 s to read our data le, something that cannot be avoided.

Let’s turn our attentions to the loop’s remaining code line. Here we are recomputing the

maximal value of a constant data matrix, and then adding a nonconstant (random) value

to it. Instead, we can move the constant calculation term outside the loop:

function perfTest

newData = subFunc(); ...