|
pomerol
2.2
|
Data Structures | |
| struct | pMPI::MPIWorker |
| Abstraction of an MPI worker process. More... | |
| struct | pMPI::MPIMaster |
| Abstraction of an MPI master process. More... | |
| struct | pMPI::ComputeWrap< PartType > |
| Wrapper around a computable object that calls the compute() method of the wrapped object and carries information about the complexity of a call to that method. More... | |
| struct | pMPI::PrepareWrap< PartType > |
| Wrapper around a computable object that calls the prepare() method of the wrapped object and carries information about the complexity of a call to that method. More... | |
| struct | pMPI::mpi_skel< WrapType > |
| This structure carries a list of wrappers and uses the mpi_dispatcher mechanism to distribute the wrappers over MPI ranks and to call run() for all of them in parallel. More... | |
Typedefs | |
| using | pMPI::JobId = int |
| ID of a job. More... | |
| using | pMPI::WorkerId = int |
| ID of a worker process. More... | |
Enumerations | |
| enum | pMPI::WorkerTag : int { pMPI::Pending , pMPI::Work , pMPI::Finish } |
| MPI message tags used in communications between the master and its workers. More... | |
Functions | |
| int | pMPI::size (MPI_Comm const &Comm) |
| int | pMPI::rank (MPI_Comm const &Comm) |
| std::map< pMPI::JobId, pMPI::WorkerId > | pMPI::mpi_skel< WrapType >::run (MPI_Comm const &Comm, bool VerboseOutput=true) |
| using pMPI::JobId = typedef int |
ID of a job.
Definition at line 39 of file mpi_dispatcher.hpp.
| using pMPI::WorkerId = typedef int |
ID of a worker process.
Definition at line 41 of file mpi_dispatcher.hpp.
| enum pMPI::WorkerTag : int |
MPI message tags used in communications between the master and its workers.
| Enumerator | |
|---|---|
| Pending | A worker is waiting for a new job. |
| Work | Request a worker to do a job. |
| Finish | Order a worker to shut down. |
Definition at line 32 of file mpi_dispatcher.hpp.
|
inline |
| std::map< pMPI::JobId, pMPI::WorkerId > pMPI::mpi_skel< WrapType >::run | ( | MPI_Comm const & | Comm, |
| bool | VerboseOutput = true |
||
| ) |
Distribute the stored wrappers over MPI ranks according to their complexity and call run() for each of the wrappers.
| [in] | Comm | MPI communicator. |
| [in] | VerboseOutput | Print extra information about the parallelization process. |
Definition at line 85 of file mpi_skel.hpp.