next up previous contents
Next: MPI_Reduce_Bcast-length Up: Collective-pattern Previous: MPI_Reduce-length   Contents

MPI_Reduce-nodes

Here me measure the time MPI_Reduce consumes. This operation performes a global operation (here: bit-wise or) on all participating processes. We vary over the number of nodes.
\epsfig{file=col_MPI_Reduce-nodes.eps}
  number of nodes
X axis scale linear
Param. refinement no automatic x wide adaption
Argument range 2 - 8 units (s. below)
Argument stepwidth 1.000000


Per Ekman 2002-01-29