Name:unittest::distributed_rank_invariant_collect_assert_or_die - Checks whether all ranks produce equal results, independent of the number of MPI processes
Synopsis:
array proc distributed_rank_invariant_collect_assert_or_die -> -
Description:
The array specifies a list of numbers of
jobs. distributed_rank_invariant_collect_assert_or_die executes the
procedure specified as the second argument in parallel for all of the
numbers of jobs given in the arrays. This means that
distributed_rank_invariant_collect_assert_or_die carries out as many
distributed simulations as there are entries in the array.
In each of the simulations all jobs are expected to return a value on
the stack. The order in which the jobs are completed is of no
importance. After completion of all simulations
distributed_rank_invariant_collect_assert_or_die requires that the
contents of the results of all ranks are identical, independent of
the number of processes.
See distributed_collect_assert_or_die for further documentation and
implementation details and distributed_process_invariant_collect_assert_or_die for
a version requiring identical output from each rank.
Author:
Diesmann
FirstVersion:
100925
SeeAlso:
Source:/opt/conda/conda-bld/nest_1512397208563/work/nest-2.14.0/lib/sli/unittest.sli