topic Array Notation : passing array to function in Software Archive

Array Notation : passing array to function

sheliak — Mon, 24 Oct 2011 03:00:20 GMT

Dear all,

recently I started to use the Array Notation in the new compiler 12.1.0, which is amazing, and I am glad to finally have (dynamic) multi-dimensional array support in C++ too :).

Unfortunately documentationis still a bit rare, so I am confused about following code (here an example)

*****************

#include

typedef std::complex cmplxd;

typedef cmplxd(*A2c)[];

int Nx, Ny, NxLlD, NyLlD;

void goCilk(cmplxd A[NxLlD,Nx][NyLlD,Ny])

{

for(int x = NxLlD; x <= NxLlD+Nx-1; x++) { for(int y = NyLlD; y <= NyLlD+Ny-1; y++) {

std::cout << x << " " << y << " " << A << std::endl;

}}

}

int main()

{

NxLlD = 1; Nx = 4;

NyLlD = 1; Ny = 3;

cmplxd *A = new cmplxd[Nx*Ny];

for(int x = 0; x < Nx*Ny; x++) A = 1.;

A[0]=2.;

goCilk((A2c) A);

}

*****************

The code compiles and runs without error. Now concerning the "cmplxd goCilk(cmplxd A[NxLlD,Nx][NyLlD,Ny])" functions, I tought that A[NxLlD, Nx] defines the starting value, and the length of this dimension. Similar like fortran has with A(NxLlD:NxLlD+Nx-1), but it seems that even tough the code compiles, the NxLlD value has no meaning as I checked with using some random numbers for NxLlD/NyLlD.

Could you please point me to the documentation, where I can find out the correct syntax ?

And is it actually possible todefine an offset to an array (only in notation like in Fortran), e.g. right now A[0][0] points to the first value, but I would like thatA[NxLlD][NyLlD] to point to the lowest value (it makes life far more simpler for MPI (domain decomposition) programs).

Thanks a lot and best wishes,

Paul

Array Notation : passing array to function

Balaji_I_Intel — Mon, 24 Oct 2011 03:24:20 GMT

Hello Paul,
I am not sure if you are aware but this website has the entire Cilk Plus language specification: http://software.intel.com/en-us/articles/intel-cilk-plus-specification/

The Array-triplets are representedusing the following sytax

Array [ ]

For example, Array[0:5:2] implies we are going to touch A[0], A[2], A[4], A[6], A[8]

I hope this helps!

Thanks,

Balaji V. Iyer.

Array Notation : passing array to function

sheliak — Mon, 24 Oct 2011 03:53:34 GMT

Dear Balaji,

thanks a lot for your quick reply. Yes, I know about the specifictation, I looked through it and couldn't find

anything related. So what confused me,is that

WORKS : void goCilk(cmplxd A[NxLlD,Nx][NyLlD,Ny]) (<--- note the comma NxLLD, Nx

behaviour is not documented, and seems not be a valid feature, yet the compiler accepts it without warning.

Is this a bug or am I missing somehting ?

Thanks

Array Notation : passing array to function

sheliak — Mon, 24 Oct 2011 04:08:18 GMT

also, it has a pretty different behavirout compared to Fortran,

Assuming I pass an array in Fortran, I define the index position of this array. E.g.

subroutine (A)

double precison :: A(Start_X:End_X, Start_Y:End_Y)

in the function deklaration. While cilk plus seems to always assume that my index starts from A[0].

Is there a way to tell cilk where to start counting ?

E.g.

double (int Start_X, int Length_X, A[Start_X, Length_X])

calling Start_X=5, and Length_X=10, then A[5] refers to the zeroth element (A'[0] in the previous view).

In Fortran it is pretty easy, in Cilk I hope this feature exists, or may be included ?