Ralph Castain 01ba861f2a When direct launching applications, we must allow the MPI layer to progress during RTE-level barriers. Neither SLURM nor Cray provide non-blocking fence functions, so push those calls into a separate event thread (use the OPAL async thread for this purpose so we don't create another one) and let the MPI thread sping in wait_for_completion. This also restores the "lazy" completion during MPI_Finalize to minimize cpu utilization. ... Update external as well Revise the change: we still need the MPI_Barrier in MPI_Finalize when we use a blocking fence, but do use the "lazy" wait for completion. Replace the direct logic in MPI_Init with a cleaner macro		2016-05-14 16:37:00 -07:00
..
help-mpi-runtime.txt	init/finalize: extensions	2015-10-15 12:39:15 -04:00
Makefile.am	mpi: infrastructure to gracefully disable MPI dyn procs	2015-10-14 13:42:56 -07:00
mpiruntime.h	init/finalize: extensions	2015-10-15 12:39:15 -04:00
ompi_cr.c	Purge whitespace from the repo	2015-06-23 20:59:57 -07:00
ompi_cr.h	Purge whitespace from the repo	2015-06-23 20:59:57 -07:00
ompi_info_support.c	opal: fix multiple bugs in MCA and opal	2015-04-07 19:13:20 -06:00
ompi_info_support.h	tools: Add oshmem_info utility	2013-10-12 19:03:32 +00:00
ompi_mpi_abort.c	debuggers: remove some useless code	2016-05-05 14:22:55 -07:00
ompi_mpi_dynamics.c	mpi: infrastructure to gracefully disable MPI dyn procs	2015-10-14 13:42:56 -07:00
ompi_mpi_finalize.c	When direct launching applications, we must allow the MPI layer to progress during RTE-level barriers. Neither SLURM nor Cray provide non-blocking fence functions, so push those calls into a separate event thread (use the OPAL async thread for this purpose so we don't create another one) and let the MPI thread sping in wait_for_completion. This also restores the "lazy" completion during MPI_Finalize to minimize cpu utilization.	2016-05-14 16:37:00 -07:00
ompi_mpi_init.c	When direct launching applications, we must allow the MPI layer to progress during RTE-level barriers. Neither SLURM nor Cray provide non-blocking fence functions, so push those calls into a separate event thread (use the OPAL async thread for this purpose so we don't create another one) and let the MPI thread sping in wait_for_completion. This also restores the "lazy" completion during MPI_Finalize to minimize cpu utilization.	2016-05-14 16:37:00 -07:00
ompi_mpi_params.c	ompi_mpi_params.c: set mpi_add_procs_cutoff default to 0	2016-02-09 09:41:36 -08:00
ompi_mpi_preconnect.c	Purge whitespace from the repo	2015-06-23 20:59:57 -07:00
params.h	Remove extraneous declaration.	2015-12-19 01:34:48 -05:00