Data Flow in Fortran MEX Files

Showing Data Input and Output

Suppose your MEX-file myFunction has two input arguments and one output argument. The MATLAB® syntax is [X] = myFunction(Y, Z). To call myFunction from MATLAB, type:

X = myFunction(Y, Z);

The MATLAB interpreter calls mexFunction, the gateway routine to myFunction, with the following arguments:

Your input is prhs, a two-element array (nrhs = 2). The first element is a pointer to an mxArray named Y and the second element is a pointer to an mxArray named Z.

Your output is plhs, a one-element array (nlhs = 1) where the single element is a null pointer. The parameter plhs points at nothing because the output X is not created until the subroutine executes.

The gateway routine creates the output array and sets a pointer to it in plhs[0]. If the routine does not assign a value to plhs[0] but you assign an output value to the function when you call it, MATLAB generates an error.


It is possible to return an output value even if nlhs = 0, which corresponds to returning the result in the ans variable.

Gateway Routine Data Flow Diagram

The following MEX Cycle diagram shows how inputs enter a MEX-file, what functions the gateway routine performs, and how outputs return to MATLAB.

In this example, the syntax of the MEX-file func is [C, D] = func(A,B). In the figure, a call to func tells MATLAB to pass variables A and B to your MEX-file. C and D are left unassigned.

The gateway routine uses the mxCreate* functions to create the MATLAB arrays for your output arguments. It sets plhs[0] and plhs[1] to the pointers to the newly created MATLAB arrays. It uses the mxGet* functions to extract your data from your input arguments prhs[0] and prhs[1]. Finally, it calls your computational routine, passing the input and output data pointers as function parameters.

MATLAB assigns plhs[0] to C and plhs[1] to D.

Fortran MEX Cycle