Support of Multi-dimensional arrays

Hi,
as OpenCL will is getting popular within the scientific community. I wonder if
for the next release support for (dynamic) multi-dimensional arrays are planned or could
be included - similar to what Fortran supports. e.g.

Fortran- 90 : double precision :: A(1:Nx, 1:Ny, 1:Nz)

could be done in C/C++ lilke

void F(int Nx, int Ny, A[1:Nx][1:Ny])

and access like A[1+x][y] …, with x,y some integeer variables withing arrays bounds.

or some similar notation (e.g. Intel Cilk+ Array notation)
[ C-99 has something similar with VLA arrays, but seems to work only with
static arrays for me, also are VLA arrays in C-99 are only heap allocated) .

Having this array features similar to Fortran’s would be awesome, and will give OpenCL a great boost in the science community .

Thanks a lot

Yes we definitely need C99 VLA. It is a pity that current OpenCL is based on C99 but dropped VLA…
Parallelism is already hard stuff, so if we can avoid messing up with dirty array linearization… :frowning:

It would also add an advantage on CUDA. :slight_smile: