GitLab

[svn-r9034] Added parameter to allow netmods to force ready sends to use rendezvous for large messages.  Reviewed by goodell@

[svn-r8965] Add VC_EVENT_TERMINATED event calls when a conn is teared down. This fix corresponds to changes made in r8201 in the tcp netmod. Without the fix programs hang in MPI_Finalize

[svn-r8861] adds a new environment variable MPICH_NETWORK_IFACE to allow the user to select a pseudo-ethernet interface to use (e.g., eth1, ib0, etc).  This commit also adds a loop around gethostbyname to retry the hostname lookup a few times before giving up.  Reviewed by balaji

As of this change, setting "HYDRA_IFACE=ib0" will actually work just like
the documentation on the wiki suggests.

Reviewed by buntinas@.

Thanks to Cray for reporting the bad logic.

No reviewer.

This fixes MPIX_Ialltoallw as used in the coll/icalltoallw test.

No reviewer.

Users will never encounter this directly because MPI_Type_indexed
and friends perform checking, but MPIR_Type_indexed_impl and friends
do not.  Without this check, much more confusing assertions will trigger
later deep in nemesis fastbox code when negative blocklens are provided.

No reviewer.

The code did not originally track per-entry communicators, so it was
correct in the past to use a single comm stored in the schedule.

This fixes most intercommunicator collectives, which were broken because
intercomm collectives commonly use multiple comms (the real comm and the
local comm).  SMP collectives similarly were broken.

Reviewed by buntinas@.

I don't think that this has any ABI implications, since we don't
actually have "external32" support and it doesn't change any sizes for
any of the actual routines or symbols that the library exposes to
applications.

No reviewer.

No reviewer.

MPICH_NO_LOCAL was already handled by the MPIR_Param system, but this
parameter was not for some reason.

No reviewer.

No reviewer.

No reviewer.

This makes sorting out truncation errors easier because it becomes clear
exactly where and how the truncation is occurring.  This code is disabled
in fast builds.

Reviewed by buntinas@.

This refactors the add_ref/release for datatypes in mpid_sched.c and
additionally fixes a leak when copy schedule entries are used with
non-builtin datatype arguments.

Reviewed by buntinas@.

[svn-r8201] Make sure we send all queued sends before terminating a VC.  Fixes #945. Reviewed by goodell@

This is needed in order to parallel the blocking form of MPIX_Igatherv,
and potentially for other NBC flow control scenarios.

Reviewed by buntinas@.

[svn-r8149] fixed tcp bug where if the sockconn table is resized when processing received messages, an sc pointer becomes invalid.

[svn-r8113] chain signal handler, rather than aborting if a sighandler has already been installed by the user.  reviewed by balaji@

is enabled.

Reviewed by buntinas.