You always know "m". It is determined by the number of output arguments you call the function with.
For example length of the output randperm(n) is n. How to make a matrix nxk in which each column is a different permutation of 1..n?
This is a different question. You would do
n=4;k=3;
outCell=arrayfun(@(p) randperm(n).', 1:k,'uni',0);
matrix=[outCell{:}],
but note that for each p=1,...k, the function randperm is being called with just a single output argument. In other words, m is known and is equal to 1.
