MPI Derived Types

Modern programming languages provide data structures that may be called “structs,” or “classes,” or “types.” These data structures permit grouping of different quantities under a single variable name.

MPI also provides a general type that enables programmer-defined datatypes. Unlike arrays, which must be adjacent in memory, MPI derived datatypes may consist of elements in non-contiguous locations in memory.

While more general derived MPI datatypes are available, one of the most commonly used is the MPI_TYPE_VECTOR. This creates a group of elements of size blocklength separated by a constant interval, called the stride, in memory. Examples would be generating a type for columns in a row-major-oriented language, or rows in a column-major-oriented language.

C++

MPI_Datatype newtype;
MPI_Type_vector(ncount, blocklength, stride, oldtype, newtype);

Fortran

integer newtype
!code
call MPI_TYPE_VECTOR(ncount, blocklength, stride, oldtype, newtype, ierr)

For both C++ and Fortran, ncount, blocklength, and stride must be integers. The oldtype is a pre-existing type, usually a built-in MPI Type such as MPI_FLOAT or MPI_REAL. For C++ the new type would be declared as an MPI_Datatype, unless it corresponds to an existing built-in type. For Fortran oldtype would be an integer if not a built-in type. The newtype is a name chosen by the programmer.

Python

newtype = oldtype.Create_vector(ncount, blocklength, stride)

A derived type must be committed before it can be used.

MPI_Type_commit(newtype)

Fortran

call MPI_TYPE_COMMIT(newtype,ierr)

Python

newtype.Commit()

To use our newly committed type in an MPI communication function, we must pass it the starting position of the data to be placed into the type.

C++

MPI_Send(&a[0][i],1,newtype,i,MPI_COMM_WORLD);
//We need to pass the first element by reference because an array element
//is not a pointer

Fortran

MPI_Send(a(1)(i),1,newtype,i,MPI_COMM_WORLD,ierr)