Nathan Hjelm 2b57f4227e ob1: optimize blocking send and receive paths ...

Per RFC. There are two optimizations in this commit:

 - Allocate requests for blocking sends and receives on the stack. This
   bypasses the request free list and saves two atomics on the critical path.
   This change improves the small message ping-pong by 50-200ns on both AMD
   and Intel CPUs.

 - For small messages try to use the btl sendi function before intializing a
   send request. If the sendi fails or the btl does not have a sendi function
   silently fallback on the standard send path.

cmr=v1.7.5:reviewer=brbarret

This commit was SVN r30343.

2014-01-21 15:16:21 +00:00

..

configure.m4

Remove the proc_pml and proc_bml fields from ompi_proc_t and replace with a

2013-08-30 16:54:55 +00:00

help-mpi-pml-ob1.txt

Add empty line that was inadvertently removed in message.

2013-12-30 18:38:07 +00:00

Makefile.am

Fix longstanding issue with our multi-project support. Rather than using

2014-01-07 22:11:15 +00:00

pml_ob1_comm.c

Keep track of the last probe on each communicator, so we can probe all

2011-10-11 20:24:54 +00:00

pml_ob1_comm.h

Keep track of the last probe on each communicator, so we can probe all

2011-10-11 20:24:54 +00:00

pml_ob1_component.c

Fix warnings in ob1 introduced by the pvar commit

2013-07-17 03:41:05 +00:00

pml_ob1_component.h

- Replace combinations of

2009-08-20 11:42:18 +00:00

pml_ob1_cuda.c

Chnage some CUDA configure code and macro names per review request by jsquyres in ticket #3880.

2013-12-06 14:35:10 +00:00

pml_ob1_hdr.h

Add the MPI 2.2 MPI_Dist_graph functionality.

2013-07-01 12:40:08 +00:00

pml_ob1_iprobe.c

Our macros for the OMPI-level free list had one extra argument, a possible return

2013-07-04 08:34:37 +00:00

pml_ob1_irecv.c

ob1: optimize blocking send and receive paths

2014-01-21 15:16:21 +00:00

pml_ob1_isend.c

ob1: optimize blocking send and receive paths

2014-01-21 15:16:21 +00:00

pml_ob1_progress.c

Remove redundant macro. This was from reviewed of earlier ticket.

2013-11-01 12:19:40 +00:00

pml_ob1_rdma.c

A little bit of cleanup, the main logic is still the same.

2009-10-01 14:05:25 +00:00

pml_ob1_rdma.h

Schedule SEND traffic of pipeline protocol between BTLs in accordance with

2007-07-01 11:34:23 +00:00

pml_ob1_rdmafrag.c

make ompi_free_list_item_t a class..

2006-06-12 16:44:00 +00:00

pml_ob1_rdmafrag.h

rename ompi_free_list operations to _mt, as per discussions at last face to face meeting

2013-07-08 22:07:52 +00:00

pml_ob1_recvfrag.c

Remove redundant macro. This was from reviewed of earlier ticket.

2013-11-01 12:19:40 +00:00

pml_ob1_recvfrag.h

rename ompi_free_list operations to _mt, as per discussions at last face to face meeting

2013-07-08 22:07:52 +00:00

pml_ob1_recvreq.c

ob1: optimize blocking send and receive paths

2014-01-21 15:16:21 +00:00

pml_ob1_recvreq.h

Remove redundant macro. This was from reviewed of earlier ticket.

2013-11-01 12:19:40 +00:00

pml_ob1_sendreq.c

Remove redundant macro. This was from reviewed of earlier ticket.

2013-11-01 12:19:40 +00:00

pml_ob1_sendreq.h

ob1: optimize blocking send and receive paths

2014-01-21 15:16:21 +00:00

pml_ob1_start.c

Move variable declaration at beginning of code block to avoid warnings.

2010-11-17 18:01:56 +00:00

pml_ob1.c

Fix various items discovered by review of ticket #3951.

2013-12-13 21:25:07 +00:00

pml_ob1.h

rename ompi_free_list operations to _mt, as per discussions at last face to face meeting

2013-07-08 22:07:52 +00:00

post_configure.sh

* include the correct file if we are doing the component bypass thing with ob1

2006-02-22 16:16:38 +00:00