Resultoutputsolver in parallel

Franz Pichler · Post by **Franz Pichler** » 05 Sep 2013, 15:46

Hello dear Elmer firiends,

i think i found a bug and its solution:

the resultoutputsolver has trouble writing pvtu files in parallel.
if a partition is not having a matrix in a solver then the parenv%activecomm is set so that tzhe partition is not taking part in the reduction operations.
and in the resultoutputsolver in writepvtufile there happens exactly that when the line:

Code: Select all

NoActive = NINT( ParallelReduction( 1.0_dp * Active ) )

is calculating the number of involved partitions
Adding a

Code: Select all

parenv%activecomm=MPI_COMM_WORLD

right in front of that solves the problem in my own resultoutputsolver. it would be nice if someone could double check that .

best regards
Franz