You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I would assume that if you call a parallel_reduce(0, functor) it would call init and then join for the result, or at least call init for the result. However, this call does nothing.
Therefore, for a reduction of an array has to do the convoluted code listed below with the if statement.
if ( particles_.migrate_count_() == 0 )
for (size_t i=0; i<n_remote_procs_; ++i)
count[i] = 0;
else
// This is the only thing that should be needed
Kokkos::parallel_reduce(particles_.migrate_count_(), counter, &count[0]);
The text was updated successfully, but these errors were encountered:
That was supposed to get fixed with a commit about 5 days ago, which should have made it into Trilinos with the last integration I believe. Is this an issue in the current Head Trilinos?
I would assume that if you call a parallel_reduce(0, functor) it would call init and then join for the result, or at least call init for the result. However, this call does nothing.
Therefore, for a reduction of an array has to do the convoluted code listed below with the if statement.
if ( particles_.migrate_count_() == 0 )
for (size_t i=0; i<n_remote_procs_; ++i)
count[i] = 0;
else
// This is the only thing that should be needed
Kokkos::parallel_reduce(particles_.migrate_count_(), counter, &count[0]);
The text was updated successfully, but these errors were encountered: