1
1
Граф коммитов

818 Коммитов

Автор SHA1 Сообщение Дата
George Bosilca
0445670bb9 Fix the automatic handling of communicator associated requests.
If the array doesn't exist, or if it's size is not adequate then
we reallocate it. Otherwise just keep using the same array of requests.
2015-02-26 15:52:18 -05:00
George Bosilca
67d01bd8cd Redirect most of the basic module functions to base. 2015-02-26 15:52:00 -05:00
George Bosilca
211f05fb09 Complete the dismantle of the tuned module. 2015-02-26 15:50:55 -05:00
George Bosilca
aa019e239e Rename the base header file containing the prototypes of the collective
functions.
2015-02-26 15:50:29 -05:00
George Bosilca
8fbcdf685d Split the tuned framework in two. Move all the functions down in the
base, so that they can now be used by all modules. Keep the decision
functions in tuned.
2015-02-26 15:46:13 -05:00
George Bosilca
004f65a865 Fix issue with the error reporting as suggested by Gilles. 2015-02-26 13:01:13 -05:00
Todd Kordenbrock
c73e4fd98b coll-portals4: fix incomplete free list conversion 2015-02-26 10:53:45 -06:00
Gilles Gouaillardet
05140df1e6 coll/tuned: regression fix
fix the regression introduced in open-mpi/ompi@004160f8da
2015-02-26 13:58:06 +09:00
Nysal Jan K.A
ded408f485 Fix a crash while closing libnbc
If the free list initialization fails in libnbc_open()
mca_coll_libnbc_component.active_requests remain uninitialized,
resulting in a crash while closing the component
2015-02-25 17:26:28 +05:30
Jeff Squyres
a85a392896 Merge pull request #422 from jsquyres/topic/coverity-fixes
Some Coverity fixes
2015-02-24 17:00:10 -05:00
Jeff Squyres
1c3cf068a4 nbc ireduce_scatter: ensure to check the correct return code
This was CID 1196644 and 1196621
2015-02-24 15:24:11 -05:00
Jeff Squyres
e9980654a8 nbc ireduce_scatter_block: ensure to check the correct return code
This was CID 1196643 and 1196615
2015-02-24 15:24:11 -05:00
Jeff Squyres
b35eb6fe10 nbc ireduce_scatter_block: ensure to check the correct return code
This was CID 709594 and 709592
2015-02-24 15:24:10 -05:00
Jeff Squyres
b0acef6f2d nbc_ireduce_scatter: ensure to check the proper return code
This was CID 709229 and 709224.
2015-02-24 15:24:10 -05:00
Jeff Squyres
1cf197d771 coll_basic_barrier: guard against opal_hibit() returning -1
This was CID 1196606 and 1196607
2015-02-24 15:24:08 -05:00
Nathan Hjelm
5f1254d710 Update code base to use the new opal_free_list_t
Use of the old ompi_free_list_t and ompi_free_list_item_t is
deprecated. These classes will be removed in a future commit.

This commit updates the entire code base to use opal_free_list_t and
opal_free_list_item_t.

Notes:

OMPI_FREE_LIST_*_MT -> opal_free_list_* (uses opal_using_threads ())

Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
2015-02-24 10:05:45 -07:00
Howard Pritchard
c9e81b54fb Merge pull request #412 from hppritcha/topic/owner_files
add owner files to opa/ompi/orte mca directories
2015-02-23 09:48:20 -07:00
Howard Pritchard
61fb62499a hcoll belongs to MLNX and is active 2015-02-23 09:14:03 -07:00
Gilles Gouaillardet
004160f8da coll/tuned: silence CID 1269934 2015-02-23 13:45:23 +09:00
Howard Pritchard
bf89131f9e add owner files to opa/ompi/orte mca directories
This commit adds an owner file in each of the component directories
for each framework.  This allows for a simple script to parse
the contents of the files and generate, among other things, tables
to be used on the project's wiki page.  Currently there are two
"fields" in the file, an owner and a status.  A tool to parse
the files and generate tables for the wiki page will be added
in a subsequent commit.
2015-02-22 15:10:23 -07:00
Igor Ivanov
010dce307a Fix set of coverity issues
List of CIDs (scan.coverity.com):
oshmem:
1269787, 1269907, 1270161, 1270162, 1270977, 1270978
ompi:
1270170, 1270172, 1270173

Signed-off-by: Igor Ivanov <Igor.Ivanov@itseez.com>
2015-02-20 17:45:46 +04:00
Gilles Gouaillardet
bda8058f39 coll/tuned: fix memory leaks and misc issues
as reported by Coverity with CIDs
70132, 70265, 70267, 70268, 70322, 70400, 70580, 70615,
1269934, 1269944, 1269968, 1269982, 1269983
2015-02-18 16:29:42 +09:00
Jeff Squyres
27a783b1c3 coll_sm_barrier: remove dead code
This was CID 1269978.
2015-02-12 10:24:02 -08:00
Rolf vandeVaart
1f749b0224 Bump priority of coll cuda component so it is higher than self.
Otherwise, get some odd interactions with coll self in CUDA-aware
builds.
2015-02-12 12:29:03 -05:00
Jeff Squyres
a1c521f968 We're .opal_ignore these days, not .ompi_ignore. :-) 2015-02-03 13:56:53 -08:00
Jeff Squyres
30f05bc966 Makefiles: remove unused macros 2015-01-31 04:51:25 -08:00
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
Devendar Bureddy
036e687d9c HCOLL: Do not block hcoll progress in finalize 2015-01-27 17:01:00 +02:00
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
Devendar Bureddy
e732152304 HCOLL: Fix hcoll supported datatype checks corretcly 2015-01-02 21:18:12 +02:00
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
Devendar Bureddy
e398ad6619 HCOLL: Fix OMPI to HCOLL predefined datatypes, Ops mapping 2014-12-23 22:30:29 +02:00
Rolf vandeVaart
3ec9685ee0 Add missing file to sources. Without this, tarball build does not work 2014-12-18 07:17:28 -08:00
George Bosilca
d622db783d Based on https://github.com/open-mpi/ompi/pull/262, we should use
true_lb while computing the lower bound.
2014-11-21 19:16:05 +09:00
Gilles Gouaillardet
705147e98b coll/tuned: fix allgather bruck algorithm 2014-11-21 19:16:05 +09:00
Ralph Castain
780c93ee57 Per the PR and discussion on today's telecon, extend the process name definition as a two-field struct of uint32_t's down to the OPAL layer. This resolves issues created by prior commits that impacted both heterogeneous and SPARC support. This also simplifies the OMPI code base by removing the need for frequent memcpy's when transitioning between the OMPI/ORTE layers and OPAL.
We recognize that this means other users of OPAL will need to "wrap" the opal_process_name_t if they desire to abstract it in some fashion. This is regrettable, and we are looking at possible alternatives that might mitigate that requirement. Meantime, however, we have to put the needs of the OMPI community first, and are taking this step to restore hetero and SPARC support.
2014-11-11 17:00:42 -08:00
bosilca
e7c59e3adb Merge pull request #227 from ggouaillardet/rfc/coll_basic_neighbor
RFC/coll basic neighbor
2014-11-07 11:33:25 -05:00
Gilles Gouaillardet
64c18686b7 fix ompi_request_wait vs ompi_request_wait_all and
MPI_STATUS_IGNORE vs MPI_STATUSES_IGNORE
2014-11-04 12:16:30 +09:00
Jeff Squyres
c22e1ae33b configury: new OPAL_SET_LIB_PREFIX/ORTE_SET_LIB_PREFIX macros
These two macros set the prefix for the OPAL and ORTE libraries,
respectively.  Specifically, the OPAL library will be named
libPREFIXopen-pal.la and the ORTE library will be named
libPREFIXopen-rte.la.

These macros must be called, even if the prefix argument is empty.

The intent is that Open MPI will call these macros with an empty
prefix, but other projects (such as ORCM) will call these macros with
a non-empty prefix.  For example, ORCM libraries can be named
liborcm-open-pal.la and liborcm-open-rte.la.

This scheme is necessary to allow running Open MPI applications under
systems that use their own versions of ORTE and OPAL.  For example,
when running MPI applications under ORTE, if the ORTE and OPAL
libraries between OMPI and ORCM are not identical (which, because they
are released at different times, are likely to be different), we need
to ensure that the OMPI applications link against their ORTE and OPAL
libraries, but the ORCM executables link against their ORTE and OPAL
libraries.
2014-10-22 10:32:19 -07:00
bosilca
d819939841 Merge pull request #233 from ggouaillardet/rfc/coll_module_disable
Provide a symmetric behavior for the activation/deactivation of collective modules.
2014-10-16 09:22:04 -04:00
George Bosilca
7541c03b4c Mark all instances where atomic operations are used but their return value is unnecessary 2014-10-15 21:47:32 -04:00
Gilles Gouaillardet
e3f74aca1c Correctly mote the pointer back by the true_lb.
Fixes #231
2014-10-14 16:26:54 +09:00
Gilles Gouaillardet
0f983d5a4f add a disable function for coll module 2014-10-14 14:46:36 +09:00
Devendar Bureddy
7a6b4c36b0 HCOLL: Update the proc structure dereference
Update the proc structure dereference to reflect the new opal_proc_t
super field
2014-10-13 20:49:19 +03:00
Devendar Bureddy
b8d2a15be9 HCOLL: by default off 2014-10-13 20:49:09 +03:00
Gilles Gouaillardet
8eb2d62919 coll/sm: fix an other memory leak 2014-10-10 19:54:45 +09:00
Gilles Gouaillardet
27e4389259 * comment on communicator creation in mca_topo_base_dist_graph_create(...)
* use accesors to retrieve topo info
2014-10-10 16:07:20 +09:00
Gilles Gouaillardet
5d44a30111 coll/sm: fix minor memory leaks
port 4488.1.patch attached in #196 to master
2014-10-10 14:21:34 +09:00
Gilles Gouaillardet
76204dfafe coll/basic: fix segmentation fault in neighborhood collectives if the degree
of the topology is higher than the communicator size

It is possible to have a topology degree higher than the size of the communicator.
For example, a periodic cartesian communicator on MPI_COMM_SELF. This will leave
the neighborhood collectives with a request buffer that is too small.

This commits introduces a semantic change :
from now, c_topo must be set before invoking coll_select
2014-10-10 11:56:04 +09:00