What do you think is wrong? You'll have a random number of odd/even entries every time (unless you reset RNG to produce reproducible result; see the documentation for how to do that for testing/debugging) and the sequence will be in whatever order the values are in the original vector vect since you didn't sort them.
Given the HW restriction to use only for and if constructs, it's as reasonable an approach as any, other than you could have preallocated the two results vectors and then trimmed them to size for a slight runtime performance gain, but for tiny array sizes such as this it's not worth the extra code if you recognize the issue for larger problems.
The
numel(vect_even)+numel(vect_odd)
always totals 50 doesn't it? Or either by inspection or running a check on the results all elements of each are even/odd as supposed to be aren't they?
