I have an `SMatrix`

of dimension `(D, k)`

. What I want to do is to normalize each *column* of this matrix, as if they were individual vectors.

What I do at the moment is always carry around with me a `Vector[SVector]`

, and do:

```
using StaticArrays
D = 3; k = 3
dummy = [rand(SVector{D}) for i in 1:k]
A = rand(SMatrix{D, k})
for i in 1:k
dummy[i] = normalize(A[:, i])
end
A = hcat(dummy...)
```

But that doesn’t seem very efficient…

Both `D, k`

can be known at compile time, maybe there is a more performant way involving metaprogramming?