George Bosilca
7adf74c617
As discussed on the devel mailing list in
...
http://www.open-mpi.org/community/lists/devel/2015/01/16820.php ,
coll ML has two pending issues: a deadlock and a performance critical
on every communicator creation. After confirmation over IM from
Pasha, the ML collective module will be disabled until it is
fixed. Token to Pasha.
2015-01-27 16:27:12 -05:00
Jeff Squyres
2d5b92157f
hierarch: with Edgar's blessing, remove the coll hierarch module
2015-01-27 13:25:27 -06:00
Yohann Burette
a741c44035
mtl/ofi: fix compiler warnings.
2015-01-27 11:14:40 -08:00
Ralph Castain
fcec24b2a4
Minor cleanups to handle comm_spawn and singletons
2015-01-27 09:29:42 -06:00
Devendar Bureddy
036e687d9c
HCOLL: Do not block hcoll progress in finalize
2015-01-27 17:01:00 +02:00
Yohann Burette
a4c1faae37
mtl/ofi: Add OFI provider option.
...
The user can now specify which OFI provider to use with the MTL.
e.g. --mca mtl ofi --mca mtl_ofi_provider psm
2015-01-26 08:38:11 -08:00
Edgar Gabriel
ea8051f8e0
add code for the named semaphores. Configure logic to decide whether to use named or unnamed semaphores still missing.
2015-01-26 10:27:28 -06:00
Nathan Hjelm
ff7be586dd
Merge pull request #355 from ggouaillardet/topic/mpi_get_hang_in_fence
...
osc/pt2pt: invoke ompi_osc_signal_outgoing only once per fragment
2015-01-23 18:33:14 -06:00
George Bosilca
438b6e136b
Edgar noticed that the commit 3960817 broke the MPI_Type_create_resized.
...
So, I revert it until I figure out a better way.
2015-01-23 18:12:26 -05:00
Yohann Burette
3c06fd77db
mtl/ofi: remove unneeded FI_REMOTE_COMPLETE flag.
2015-01-23 10:55:03 -08:00
Yohann Burette
b88708bf68
mtl/ofi: use fi_ep_bind().
2015-01-23 10:50:10 -08:00
Bert Wesarg
92f976bdf1
Remove the actual VampirTrace sources.
2015-01-22 08:08:53 +01:00
Bert Wesarg
0d0a754c42
Remove VampirTrace.
2015-01-22 08:08:07 +01:00
Gilles Gouaillardet
9be4dfb152
osc/pt2pt: invoke ompi_osc_signal_outgoing only once per fragment
2015-01-22 13:43:44 +09:00
Yohann Burette
1ad188206b
Add OFI MTL to CM PML.
...
This allows the CM PML to be picked when the OFI MTL is selected.
2015-01-20 10:50:14 -08:00
George Bosilca
ded4cbf20f
Correctly set the upper and lower bound for the subarray and darray.
2015-01-19 02:26:14 -05:00
George Bosilca
39608176db
If we want the resized data to have the correct LB and UB (both
...
soft and hard markers) we should force an add instead of
relying on the OPAL datatype resize operation.
2015-01-19 02:24:36 -05:00
George Bosilca
df0512550e
The extent of the datatype is irrelevant for deciding to do an immediate
...
send as long as we have to pack.
2015-01-19 02:23:12 -05:00
Gilles Gouaillardet
d14daf40d0
ob1: correctly handle types in which size > extent
...
do not send inline if extent*count *OR* size*count are greater than 256
2015-01-19 14:07:23 +09:00
George Bosilca
19c96465f3
Fix the use of hard markers MPI_LB and MPI_UB in the creation of
...
subarray and darray. Thanks to Gus Correa for pointing to the
MPICH bug report.
2015-01-17 23:11:07 -05:00
Gilles Gouaillardet
661c35ca67
cleanup dead code caused by the removal of the --with-threads configure option
2015-01-16 19:13:59 +09:00
Jeff Squyres
9cc60b9e12
ofi mtl: update to new libfabric constant name
2015-01-15 07:12:39 -08:00
Yohann Burette
bc93e04604
Fixed code around fi_av_insert().
2015-01-14 08:43:57 -08:00
Gilles Gouaillardet
6d84afc660
romio: update configure for Solaris 10 build
2015-01-14 18:25:55 +09:00
Gilles Gouaillardet
8c1698ae4a
coll/libnbc: enhance fix for MPI_Ireduce_scatter on single task communicator
...
this improves open-mpi/ompi@b9349d2eb9
2015-01-09 13:44:01 +09:00
mjbhaskar
39f9880759
Fixed the data type argument in an all reduce operation to fix a bug
...
seen on 32 bit machines.
2015-01-08 14:18:54 -06:00
Gilles Gouaillardet
b746a8f584
romio: compile openmpi mpi-io glue
2015-01-08 14:08:46 +09:00
Ralph Castain
4e592ac434
Fix the tarball by providing the correct list of headers in the Makefile.am
2015-01-07 18:37:26 -08:00
mjbhaskar
2d33b0a745
A fix for memory corruption seen on 32 bit machines
2015-01-07 14:41:44 -06:00
Nathan Hjelm
81dc3a5db9
Merge pull request #335 from hjelmn/osc_updates
...
Osc updates
2015-01-07 11:16:55 -06:00
Ralph Castain
e0927895db
Grrr...how many files did they forget?
2015-01-06 19:40:18 -08:00
Ralph Castain
84c41429e9
Add missing file
2015-01-06 18:41:11 -08:00
Nathan Hjelm
e68ed2876c
osc/pt2pt: threading fixes and code cleanup
2015-01-06 13:39:16 -07:00
Nathan Hjelm
3d79806805
add more internal RMA error codes
2015-01-06 13:39:04 -07:00
Nathan Hjelm
9eba7b9d35
Rename the OSC "rdma" component to pt2p to better reflect that it does not actually use btl rdma
2015-01-06 13:38:55 -07:00
Jeff Squyres
cab1379dfb
Fortran: only emit real16 and complex32 if supported
...
This is the master version of @ggouaillardet's patch from
open-mpi/ompi-release#148 (there was a minor conflict to fix and
several fuzzings of line numbers).
2015-01-06 09:47:26 -08:00
Howard Pritchard
ec632001b1
Merge pull request #329 from ggouaillardet/topic/romio_refresh
...
refresh ROMIO based on v3.2a2-84-gef1cf14
2015-01-06 10:27:20 -07:00
Gilles Gouaillardet
0914de9eae
refresh ROMIO based on v3.2a2-84-gef1cf14
2015-01-06 19:43:58 +09:00
Yohann Burette
f01dd429df
Reset pointer to NULL to prevent double-freeing.
2015-01-05 17:01:37 -08:00
Yohann Burette
1e24da90fe
Fix fi_av_insert return code test.
2015-01-05 17:01:37 -08:00
Yohann Burette
5944c294ad
Add return code testing for fi_mr_reg.
2015-01-05 17:01:37 -08:00
Howard Pritchard
c857cc926c
Merge pull request #327 from hppritcha/topic/async_progress
...
Topic/async progress
2015-01-05 16:20:44 -07:00
Jeff Squyres
ce2008aa88
man pages: update non-blocking send descriptions
...
As noted by Alexander Pozdneev, non-blocking sends are now able to
*access* buffers in pending non-blocking send operations; the buffers
just can't be *modified*.
2015-01-05 15:44:27 -05:00
Devendar Bureddy
e732152304
HCOLL: Fix hcoll supported datatype checks corretcly
2015-01-02 21:18:12 +02:00
Gilles Gouaillardet
e8d084e6b9
fix ABI fix
...
Fix an undeleted line in open-mpi/ompi@24df0ed039
Thanks to Nick Papior Andersen for pointing this.
2014-12-28 18:07:51 +09:00
Gilles Gouaillardet
24df0ed039
MPI_Comm_split_type: fix ABI compatibility
...
ABI compatibility was previously broken in
open-mpi/ompi@3deda3dc82
2014-12-25 19:43:58 +09:00
Howard Pritchard
3fc7b389ff
initial async progress changes for gni
2014-12-24 11:50:23 -07:00
Howard Pritchard
65c4f8d18e
Merge pull request #326 from zerothi/master
...
Enabled COMM_TYPE_SPLIT dependent on locality
2014-12-24 09:13:04 -07:00
Nick Papior Andersen
3deda3dc82
Added several new COMM_TYPE_<> splits
...
Using the underlying hardware identification to split
communicators based on locality has been enabled using
the MPI_Comm_Split_Type function.
Currently implemented split's are:
HWTHREAD
CORE
L1CACHE
L2CACHE
L3CACHE
SOCKET
NUMA
NODE
BOARD
HOST
CU
CLUSTER
However only NODE is defined in the standard which is why the
remaning splits are referred to using the OMPI_ prefix instead
of the standard MPI_ prefix.
I have tested this using --without-hwloc and --with-hwloc=<path>
which both give the same output.
NOTE: I think something fishy is going on in the locality operators.
In my test-program I couldn't get the correct split on these requests:
NUMA, SOCKET, L3CACHE
where I suspected a full communicator but only got one.
2014-12-24 11:21:35 +00:00
Gilles Gouaillardet
b9349d2eb9
coll/libnbc: fix MPI_Ireduce_scatter for single task communicator
...
when MPI_IN_PLACE is not used.
that commit fixes a regression introduced
open-mpi/ompi@49e79a9ade
2014-12-24 12:12:58 +09:00
