`salvus.flow.executors.mpi.utils`

Run simulations on MPI distributed meshes.

Functions

`compute_halo_selection()`

def compute_halo_selection(
    core_global_element_ids: npt.NDArray, halo_width_in_meters: float
) -> collections.abc.Callable[[Mesh], npt.NDArray]: ...

Compute a selection function that selects all elements within a halo around the given core global element IDs.

Parameters

core_global_element_ids npt.NDArray — The global element IDs of the core region.
halo_width_in_meters float — The width of the halo in meters.

Returns collections.abc.Callable[[Mesh], npt.NDArray]

`get_solver_precision()`

def get_solver_precision(
    comm: int, execution_config: MPIExecutionConfiguration
) -> typing.Literal["single", "double"]: ...

Get the solver precision for a given MPI execution configuration.

Collective operations.

This function will only check it on rank 0 and then broadcast the result. In addition it will be cached so that it is only computed once per binary. This should make the cost of this function negligible.

Parameters

comm int — The MPI communicator.
execution_config MPIExecutionConfiguration — The MPI execution configuration. The assumption is that it is the same across all ranks.

Returns typing.Literal['single', 'double']