[tstanisl]

There are multiple false myths about VLAs. Please known that VLA is about the typing not about the storage.

The line:

    typedef int T[n];

is the essence of VLA-ness, not `int A[n]`. The array of VLA type can have any kind of storage one wants.

It can be stack

    T a;

It can be heap

    T *a = malloc(sizeof *a);

It can be even infamous alloca()

    T *a = alloca(sizeof *a);

Please stop talking about this "VLA is stack-base vector" crap because it means that one does not understand what VLAs are about. I admin that automatic VLAs are pretty much always wrong but this use case is a tiny bit of the realy functionality of VLA types.

VLA were added to language to handle multidimensional arrays.

And the really shine at this task. Even C++ has not good alternative for it. Vector of vectors is a really crappy data structure.

A few examples when working with square matrices:

- allocation on stack

    float A[n][n];

- allocation on heap

    float (*A)[n] = malloc(sizeof(int[n][n]));
    ...
    free(A);

- indexing

    A[i][j]

- passing to functions:

    void add(int n, float A[static n][n], float B[static n][n], float RES[restrict static n][n]);

    ...

Please show me something as simple, effective, self-documenting and elegant in C++.