1
1
Граф коммитов

30674 Коммитов

Автор SHA1 Сообщение Дата
Ralph Castain
ebd164b4c1
Update PMIx and PRRTE
Signed-off-by: Ralph Castain <rhc@pmix.org>
2020-05-06 12:40:11 -07:00
Jeff Squyres
b66e27d3ca
Merge pull request #7671 from hjelmn/add_support_for_component_aliasing_and_reluctantly_rename_vader_to_btl_sm
Add support for component aliasing
2020-05-05 10:57:16 -04:00
Nathan Hjelm
9d8f634044 btl/vader: rename vader -> sm
Now that the old sm btl has been gone for some time there was a request
to rename vader to sm. This commit does just that (reluctantly).

An alias has been generated so specifying vader in the btl selection
variable or specifying vader parameters will continue to work.

Signed-off-by: Nathan Hjelm <hjelmn@google.com>
2020-05-05 06:43:19 -07:00
Nathan Hjelm
9fae5bfdf3 mca/base: add support for component aliasing
This commit adds support for aliasing component names. A component
name alias is created by calling: mca_base_alias_register. The name
of the project and framework are optional. The component name and
component alias are required. Once an alias is registered all
variables registered after the alias creation will have synonyms
also registered. For example:

```c
mca_base_alias_register("opal", "btl", "vader", "sm", false);
```

would cause all of the variables registered by btl/vader to have
aliases that start with btl_sm. Ex: btl_vader_single_copy_mechanism
would have the synonym: btl_sm_single_copy_mechanism.

If aliases are registered before component filtering the alias
can also be used for component selection. For example, if sm is
registered as an alias to vader in the btl framework register
function then ```--mca btl self,sm``` would be equivalent to
```--mca btl self,vader```.

Signed-off-by: Nathan Hjelm <hjelmn@google.com>
2020-05-05 06:43:19 -07:00
Nathan Hjelm
3a036f8486 opal/class: add additional object helper functions
This commit adds two additional helpers to opal/class:

 - OPAL_HASH_TABLE_FOREACH_PTR: Same as OPAL_HASH_TABLE_FOREACH but
   operating on ptr hash tables. This is needed because the _ptr
   iterator functions take an additional argument.

 - OPAL_LIST_FOREACH_DECL: Same as OPAL_LIST_FOREACH but declares
   the variable specified in the first argument.

Signed-off-by: Nathan Hjelm <hjelmn@google.com>
2020-05-05 06:43:19 -07:00
Ralph Castain
1b6622f5d9
Merge pull request #7685 from rhc54/topic/sync
Update PRRTE
2020-05-04 09:13:26 -07:00
Ralph Castain
b60ea7a6ad
Update PRRTE
Protect against systems that forward entire environment

Signed-off-by: Ralph Castain <rhc@pmix.org>
2020-05-04 07:57:22 -07:00
Ralph Castain
0dce61b1fc
Merge pull request #7679 from rhc54/topic/nm
Update PRRTE to resolve naming issues
2020-05-02 06:33:42 -07:00
Ralph Castain
34f3f3bbf2
Update PRRTE to resolve naming issues
Signed-off-by: Ralph Castain <rhc@pmix.org>
2020-05-02 05:31:25 -07:00
Yossi Itigin
b61bf9a00a
Merge pull request #7349 from hoopoepg/topic/ucx-new-api-nbx
OPAL/UCX: enabling new API provided by UCX
2020-05-02 14:30:44 +03:00
Ralph Castain
0e17e5b8e4
Merge pull request #7678 from rhc54/topic/nit
Ensure proper handling of default MCA param files
2020-05-01 15:18:18 -07:00
Ralph Castain
f608575eec
Remove references to numa_rank
Signed-off-by: Ralph Castain <rhc@pmix.org>
2020-05-01 13:32:29 -07:00
Ralph Castain
86709b1c80
Fix PMIx_Fence call signature
Signed-off-by: Ralph Castain <rhc@pmix.org>
2020-05-01 12:27:42 -07:00
Ralph Castain
10c93a10e2
Ensure proper handling of default MCA param files
Update PMIx/PRRTE to ensure we pickup the default system and user MCA
param definitions during PMIx_server_setup_application so they get
propagated. Protect OPAL's MCA var processing so it doesn't try to
process a NULL filename when PMIx provides the params for it.

Signed-off-by: Ralph Castain <rhc@pmix.org>
2020-05-01 12:02:10 -07:00
Sergey Oblomov
75bda25ddb OPAL/UCX: enabling new API provided by UCX
- added detection of new API into configuration
- added tag_send call implemented using new API
- added MPI_Send/MPI_Isend/MPI_Recv/MPI_Irecv implementations

Signed-off-by: Sergey Oblomov <sergeyo@mellanox.com>
2020-05-01 17:58:29 +03:00
Ralph Castain
0fa7ead700
Merge pull request #7662 from rhc54/topic/dpm
Update dpm to handle deprecation of MPI_Info keys
2020-04-29 16:32:50 -07:00
Ralph Castain
bd29ab0ae9
Update dpm to handle deprecation of MPI_Info keys
Deprecate the current OMPI-specific MPI_Info key definitions for
MPI_Comm_spawn and replace them with their PMIx equivalents. Issue a
deprecation/conversion warning as this is done. Also issue deprecation
warnings for options such as "ompi_non_mpi" that are no longer used.

Handle both cases where the user might pass either the PMIx attribute
name itself (e.g., "PMIX_MAPBY") or the string value of the attribute
(e.g., PMIX_MAPBY, which translates to "pmix.mapby"). This can only be
done for PMIx v4 and above, so protect that code.

Silence a couple of Coverity warnings and add a test along the way.

Signed-off-by: Ralph Castain <rhc@pmix.org>
2020-04-29 14:56:38 -07:00
Ralph Castain
4318d41fb4
Merge pull request #7669 from rhc54/topic/niut
Sync PMIx
2020-04-29 09:36:26 -07:00
Ralph Castain
6146d52772
Sync PMIx
Remove pmix_config.h from the tarball. Deal with the case of no local
procs when register_nspace is called.

Signed-off-by: Ralph Castain <rhc@pmix.org>
2020-04-29 09:00:04 -07:00
Ralph Castain
515e8acfe6
Merge pull request #7667 from rhc54/topic/up
Sync PMIx and PRRTE
2020-04-28 08:42:26 -07:00
Ralph Castain
fd098d0eba
Sync PMIx and PRRTE
Remove prrte_config.h from tarball plus misc bug fixes

Signed-off-by: Ralph Castain <rhc@pmix.org>
2020-04-28 07:46:21 -07:00
Brian Barrett
4f03f44ced
Merge pull request #7582 from dipti-kothari/pml_check
mca/pml: PML check for direct modex
2020-04-27 12:29:11 -07:00
Ralph Castain
de0f34eb91
Merge pull request #7665 from rhc54/topic/hwl
Remove no-longer-used hwloc support fns
2020-04-26 07:47:28 -07:00
Ralph Castain
ae90412098
Remove no-longer-used hwloc support fns
Remove a set of functions that were only used by ORTE as they are no
longer required. We can probably remove more of them with a little
cleanup in the rest of the code.

Signed-off-by: Ralph Castain <rhc@pmix.org>
2020-04-25 21:05:07 -07:00
Ralph Castain
1281e2aff4
Merge pull request #7664 from rhc54/topic/sz
Update PMIx
2020-04-25 17:09:23 -07:00
Ralph Castain
e1841fab17
Update PMIx
Fixes #7663

Signed-off-by: Ralph Castain <rhc@pmix.org>
2020-04-25 16:00:12 -07:00
Howard Pritchard
ee9baf0a3c
Merge pull request #7656 from hppritcha/topic/fix_issue_7580
rework check lustre config to avoid rpath lib64
2020-04-24 09:38:42 -06:00
Howard Pritchard
ea690d008b rework check lustre config to avoid rpath lib64
The original configury check for lustre was ending up rpathing in /usr/lib64 in
the compiler wrapper scripts.  This commit fixes that issue.

related to #7580

Signed-off-by: Howard Pritchard <hppritcha@gmail.com>
2020-04-24 08:56:33 -06:00
Austen Lauria
2e22a247bb
Merge pull request #7650 from devreal/fix-7617-oscpt2pt-leak
PT2PT osc: don't extra retain datatype
2020-04-24 08:55:28 -04:00
Austen Lauria
9f2f98e3ec
Merge pull request #7651 from devreal/fix-7617-oscrdma-complete_atomic
RDMA osc: remove extra retain on pending_op
2020-04-24 08:55:08 -04:00
Ralph Castain
becd3052dd
Merge pull request #7657 from rhc54/topic/s2
Sync PMIx to pickup dmodex fix
2020-04-23 15:49:28 -07:00
Ralph Castain
29832798ef
Sync PMIx to pickup dmodex fix
Signed-off-by: Ralph Castain <rhc@pmix.org>
2020-04-23 13:52:32 -07:00
Ralph Castain
6c5e57f2da
Merge pull request #7655 from rhc54/topic/sync
Sync PMIx and PRRTE
2020-04-23 09:34:16 -07:00
Ralph Castain
111f0a53ef
Sync PMIx and PRRTE
Signed-off-by: Ralph Castain <rhc@pmix.org>
2020-04-23 08:46:54 -07:00
Ralph Castain
4167ab6bc5
Merge pull request #7654 from rhc54/topic/dict
Update PMIx to add missing include to tarball
2020-04-22 21:18:55 -07:00
Ralph Castain
3a15ab0ab5
Update PMIx to add missing include to tarball
Signed-off-by: Ralph Castain <rhc@pmix.org>
2020-04-22 20:13:18 -07:00
Ralph Castain
91be01beb2
Merge pull request #7652 from rhc54/topic/het
Cleanup heterogeneous builds
2020-04-22 16:20:06 -07:00
Ralph Castain
60c650e79b
Ensure "mpirun --version" reports as Open MPI
Signed-off-by: Ralph Castain <rhc@pmix.org>
2020-04-22 15:11:21 -07:00
Ralph Castain
6d29bbfde8
Cleanup heterogeneous builds
Consolidate the ompi_process_info and opal_process_info structs to
remove duplicate storage and conversion issues. Unwind some interweaving
of include files using opal.h. Silence a couple of warnings.

For now, set the arch to local if PMIX_ARCH is not found.

Signed-off-by: Ralph Castain <rhc@pmix.org>
2020-04-22 12:46:27 -07:00
Dipti Kothari
5418cc56dd mca/pml: PML check for direct modex
For direct modex, all procs publish the selected pml module
and then at add_procs pml module for each proc is checked
against every other proc in the add_proc call.
For full modex, there is no change in functionality. Only Rank0
publishes its selected pml, all other procs in the add_proc call
check their selected pml against Rank0.
If pml's do not match, throw error and exit.

Signed-off-by: Dipti Kothari <dkothar@amazon.com>
2020-04-22 16:25:01 +00:00
Austen Lauria
53154756e1
Merge pull request #7649 from wckzhang/master
Typo and segfault fixes for coll/tuned
2020-04-22 11:59:17 -04:00
Joseph Schuchart
de67ada442 RDMA osc: remove extra retain on pending_op
Signed-off-by: Joseph Schuchart <schuchart@hlrs.de>
2020-04-21 22:49:48 +02:00
Joseph Schuchart
07d1011afe OSC base: fix typos in documentation
Signed-off-by: Joseph Schuchart <schuchart@hlrs.de>
2020-04-21 21:53:36 +02:00
Joseph Schuchart
154cf571b6 OSC base: do not retain datatype by default
Signed-off-by: Joseph Schuchart <schuchart@hlrs.de>
2020-04-21 21:53:10 +02:00
William Zhang
771f9c011d coll/tuned: Add NULL check to prevent segfault
Signed-off-by: William Zhang <wilzhang@amazon.com>

cr https://code.amazon.com/reviews/CR-23837553
2020-04-21 17:53:46 +00:00
William Zhang
50640402ab coll/tuned: Fix typos
Signed-off-by: William Zhang <wilzhang@amazon.com>
2020-04-21 17:39:37 +00:00
Ralph Castain
5929ee5c1c
Merge pull request #7640 from dancejic/numa_rank
adding NUMA_RANK to process metadata
2020-04-20 20:56:16 -07:00
Nikola Dancejic
3637443454 adding NUMA_RANK to process metadata
adding PMIX_NUMA_RANK info to process metadata so that the local NUMA
rank can be accessed through the opal_process_info object.

Signed-off-by: Nikola Dancejic <dancejic@amazon.com>
2020-04-20 22:02:47 +00:00
Ralph Castain
6539a55708
Merge pull request #7645 from rhc54/topic/up
Sync with PMIx and PRRTE masters
2020-04-20 10:52:32 -07:00
Ralph Castain
336f44ecc3
Sync with PMIx and PRRTE masters
Signed-off-by: Ralph Castain <rhc@pmix.org>
2020-04-20 10:10:16 -07:00