MPIThreadedDistribution::Config
Config
Section titled “Config”#include <sbio/execution/mpi_threaded.hh>struct ConfigDefined in src/lib/sbio/execution/mpi_threaded.hh:95
Public Attributes
Section titled “Public Attributes”| Return | Name | Description |
|---|---|---|
MPI_Comm |
communicator |
|
std::vector< int > |
active_ranks |
|
int |
main_rank |
|
bool |
main_rank_loops |
|
std::size_t |
num_threads |
|
std::vector< int > |
cpu_affinities |
|
int |
ranks_per_node |
Number of (active) ranks per node, mainly for testing. |
communicator
Section titled “communicator”MPI_Comm communicator { MPI_COMM_WORLD }Defined in src/lib/sbio/execution/mpi_threaded.hh:96
active_ranks
Section titled “active_ranks”std::vector< int > active_ranks {}Defined in src/lib/sbio/execution/mpi_threaded.hh:97
main_rank
Section titled “main_rank”int main_rank { 0 }Defined in src/lib/sbio/execution/mpi_threaded.hh:98
main_rank_loops
Section titled “main_rank_loops”bool main_rank_loops { true }Defined in src/lib/sbio/execution/mpi_threaded.hh:99
num_threads
Section titled “num_threads”std::size_t num_threads { 0 }Defined in src/lib/sbio/execution/mpi_threaded.hh:101
cpu_affinities
Section titled “cpu_affinities”std::vector< int > cpu_affinities {}Defined in src/lib/sbio/execution/mpi_threaded.hh:102
ranks_per_node
Section titled “ranks_per_node”int ranks_per_node { 0 }Defined in src/lib/sbio/execution/mpi_threaded.hh:112
Number of (active) ranks per node, mainly for testing.
0 (the default) detects the nodes with MPI_COMM_TYPE_SHARED. A positive value splits the active ranks into consecutive groups of this size instead, which must still be able to share memory (e.g. to emulate several nodes on one host).