NxN random matrix much cheaper than N^2 random scalars

[Daniel Molina García]

Hello,

I am using GNU Octave, version 3.8.1. I found that generating a matrix,
for example as

for i=1:1
  normrnd(0, 1, [n, 1]);
endfor

is much fast than

for i=1:n
  normrnd(0, 1);
endfor

Why? Does it happen also in Matlab? It is motivating me to save the
exact huge amount of random numbers that I need in a very big array
before iterating, even considering the associated cost of memory.

Thanks,
Daniel

By the way, I found that specifying the 2 dimensions with a vector (as
in the above example) instead of using two parameters is quicker.