19. Semantic Terms

When discussing MPI procedures the following semantic terms are used.

{ nonblocking}

A procedure is nonblocking if the procedure may return before the operation completes, and before the user is allowed to reuse resources (such as buffers) specified in the call. A nonblocking request is started by the call that initiates it, e.g., MPI_ISEND. The word complete is used with respect to operations, requests, and communications. An operation completes when the user is allowed to reuse resources, and any output buffers have been updated; i.e. a call to MPI_TEST will return flag = true. A request is completed by a call to wait, which returns, or a test or get status call which returns flag = true. This completing call has two effects: the status is extracted from the request; in the case of test and wait, if the request was nonpersistent, it is freed, and becomes inactive if it was persistent. A communication completes when all participating operations complete.

{ blocking}

A procedure is blocking if return from the procedure indicates the user is allowed to reuse resources specified in the call.

{ local}

A procedure is local if completion of the procedure depends only on the local executing process.

{ non-local}

A procedure is non-local if completion of the operation may require the execution of some MPI procedure on another process. Such an operation may require communication occurring with another user process.

{ collective}

A procedure is collective if all processes in a process group need to invoke the procedure. A collective call may or may not be synchronizing. Collective calls over the same communicator must be executed in the same order by all members of the process group.

{ predefined}

A predefined datatype is a datatype with a predefined (constant) name (such as MPI_INT, MPI_FLOAT_INT, or MPI_UB) or a datatype constructed with MPI_TYPE_CREATE_F90_INTEGER, MPI_TYPE_CREATE_F90_REAL, or MPI_TYPE_CREATE_F90_COMPLEX. The former are named whereas the latter are unnamed.

{ derived}

A derived datatype is any datatype that is not predefined.

{ portable}

A datatype is portable, if it is a predefined datatype, or it is derived from a portable datatype using only the type constructors MPI_TYPE_CONTIGUOUS, MPI_TYPE_VECTOR, MPI_TYPE_INDEXED, MPI_TYPE_CREATE_INDEXED_BLOCK, MPI_TYPE_CREATE_SUBARRAY, MPI_TYPE_DUP, and MPI_TYPE_CREATE_DARRAY. Such a datatype is portable because all displacements in the datatype are in terms of extents of one predefined datatype. Therefore, if such a datatype fits a data layout in one memory, it will fit the corresponding data layout in another memory, if the same declarations were used, even if the two systems have different architectures. On the other hand, if a datatype was constructed using MPI_TYPE_CREATE_HINDEXED, MPI_TYPE_CREATE_HVECTOR or MPI_TYPE_CREATE_STRUCT, then the datatype contains explicit byte displacements (e.g., providing padding to meet alignment restrictions). These displacements are unlikely to be chosen correctly if they fit data layout on one memory, but are used for data layouts on another process, running on a processor with a different architecture.

{ equivalent}

Two datatypes are equivalent if they appear to have been created with the same sequence of calls (and arguments) and thus have the same typemap. Two equivalent datatypes do not necessarily have the same cached attributes or the same names.