| MPI_IBARRIER(comm, request) | |
| IN comm | communicator (handle) |
| OUT request | communication request (handle) |
MPI_IBARRIER is a nonblocking version of MPI_BARRIER. By calling MPI_IBARRIER, an MPI process notifies that it has reached the barrier. The call returns immediately, independent of whether other MPI processes have called MPI_IBARRIER. The usual barrier semantics are enforced at the corresponding completion operation (test or wait), which in the intra-communicator case will complete only after all other MPI processes in the communicator have called MPI_IBARRIER. In the inter-communicator case, it will complete when all MPI processes in the remote group have called MPI_IBARRIER.
Advice to users.
A nonblocking barrier can be used to hide latency. Moving
independent computations between the MPI_IBARRIER and the
subsequent completion call can overlap the barrier latency and therefore
shorten possible waiting times. The semantic properties are also useful
when mixing collective operations and point-to-point messages.
( End of advice to users.)