For a C function that inputs an mxArray*, one can simply:
calllib(mylib, 'myinputfun', myarray);
If the C function instead inputs a void* (reinterpreted within the function as an mxArray*), then this works:
calllib(mylib, 'myinputfun', libpointer('MATLAB array', myarray));
Similarly, if the C function returns an mxArray*:
myarray = calllib(mylib, 'myoutputfun');
When returning a void* (that actually is an mxArray*), a libpointer with a voidPtr type is returned, which makes sense. However I'm having difficulty accessing the underlying value. I would expect something like this to work, but I cannot seem to cast the pointer to a 'MATLAB array'.
mypointer = calllib(mylib, 'myoutputfun');
setdatatype(mypointer, 'MATLAB array', size);
myarray = mypointer.Value;
I am aware that it would be simpler to work with native types like double*. The reason I require this is because the functions I have are actually creating and returning mxGPUArray*, and I need to preserve the array structure in order preserve its data pointer to GPU memory (hope that makes sense).
