HP-MPI V2.3 for Linux Release Note
Using -ha:infra does not allow a convenient way to terminate all ranks associated
with the application. It is the responsibility of the user to have a mechanism for
application teardown.
The -ha:infra option is available only on HP hardware. Usage on non-HP hardware
will result in an error message.
1.2.7.7.3 Using MPI_Comm_connect and MPI_Comm_accept
MPI_Comm_connect and MPI_Comm_accept can now be used without the -spawn
option to mpirun. This allows applications launched using the -ha:infra option to
call these routines. When using high-availability mode, these routines do not deadlock
even if the remote process exits before, during, or after the call.
1.2.7.7.4 Using MPI_Comm_disconnect
In high-availability mode, MPI_Comm_disconnect is collective only across the local
group of the calling process. This enables a process group to independently break a
connection to the remote group in an intercommunicator without synchronizing with
those processes. Unreceived messages on the remote side are buffered and might be
received until the remote side calls MPI_Comm_disconnect.
Receive calls which cannot be satisfied by a buffered message will fail on the remote
processes after the local processes have called MPI_Comm_disconnect. Send calls on
either side of the intercommunicator fail after either side has called
MPI_Comm_disconnect.
1.2.7.7.5 Instrumentation and High-Availability Mode
HP-MPI's lightweight instrumentation is now supported when using -ha and singletons.
In the event that some ranks terminate during or before MPI_Finalize(), then the
lowest rank id in MPI_COMM_WORLD produces the instrumentation output file on behalf
of the application and instrumentation data for the exited ranks will not be included.
See “Expanded Lightweight Instrumentation” (page 21) for other enhancements to
instrumentation in this release.
The use of -ha and -i is available only on HP hardware. Usage on non-HP hardware
will result in an error message.
1.2.7.7.6 Failure Recover (-ha:recover)
Fault-Tolerant MPI_Comm_dup() That Excludes Failed Ranks
When using -ha:recover, the functionality of MPI_Comm_dup() is augmented to
provide a meaningful and useful functionality to allow an application to attempt to
recover from errors.
1.2 What’s in This Version 17