GitLab

application returns a zero exit code.

we don't have to maintain some of the patches for hwloc that we used
to.

only available on MPI_COMM_WORLD.

Also correspondingly update the only current usage of these macros in
mpid_sched.c.

No reviewer.

No reviewer.

This is a straightforward port of the algorithms in MPI_Alltoallv.

No reviewer.

This version only provides a binomial reduce algorithm that was lifted
from the existing MPI_Reduce code.

No reviewer.

No reviewer.

This is just a direct translation of the algorithm used in MPI_Barrier.

No reviewer.

Reviewed by balaji@.

Reviewed by balaji@.

Only supports a binomial broadcast at this time.  Override functions
will be honored though.

Reviewed by balaji@.

This change consists of four main parts:

 1. Updating all of the MPIR_Ifoo_impl routines to assume that all
    nonblocking collective functions are set in the coll_fns field.

 2. Defining the MPID_Sched_ portion of the ADI in mpir_nbc.h.

 3. Implementing that interface in src/mpid/common/sched

 4. Modifications to ch3:nemesis and ch3:sock in order to call the
    progress engine hook for the scheduling mechanism, as well as
    correctly including the various scheduling headers in CH3 headers.

Reviewed by balaji@.

This is needed to provide always-set collops functions for the nonblocking
collective operations.

Reviewed by balaji@.

This should reduce the amount of time we spend coding, reviewing,
and testing ad hoc linked-list implementations throughout MPICH2.
It should also improve readability in code that uses it.

Partially reviewed by balaji@.

Same as the MALLOC variant, but it zeroes the memory first instead.

Reviewed by balaji@.

This should simplify logic for any coll_ops overrides by the device.

Reviewed by balaji@.

Reviewed by balaji@.

The logic is duplicated in at least two places, leading to a potential error
in the future when new criteria are added.

Reviewed by balaji@.

Reviewed by balaji@.

Reviewed by balaji@.

Reviewed by balaji@.

This provides a single, simple way to invoke reduction operators from
within the library.

Reviewed by balaji@.

Reviewed by balaji@.

It was causing variable shadowing in a source file that used it.

Reviewed by balaji@.

[svn-r7766] Reverted r7739 r7745 r7746 and r7747.  These were an attempt to eliminate the need to send zero-byte messages in collectives.  However the problem is more subtle that it looks.

query management.

(required for FTB support). Only implemented for PBS for now.

Reviewed by buntinas.

libraries to cleanup resources.

MPI_UNWEIGHTED==NULL, so NULL actually is a valid value for sourceweights.
This change also adds a weak check for MPI_UNWEIGHTED in the distgraph1
test.

No reviewer.

whether to send/recv a message or not. This occurs if an alltoallv/w
uses zero-count on one process and non-zero-count, but zero size
datatype (MPI_DATATYPE_NULL) on another process.

Reviewed by buntinas.