1
1

29236 Коммитов

Автор SHA1 Сообщение Дата
Howard Pritchard
af7a7f5da1
Merge pull request #6216 from abouteiller/export4x/overspawn
v4.0.x: Correctly propagate the oversubscribe flag to the spawnees
2018-12-21 15:33:05 -07:00
Howard Pritchard
fd2259774d
Merge pull request #6167 from ggouaillardet/topic/v4.0.x/btl_uct_fix_warning
btl/uct: fix misc warnings
2018-12-21 15:31:16 -07:00
Howard Pritchard
63315250d8
Merge pull request #6204 from jsquyres/pr/v4.0.x/another-rank-file-fix
v4.0.x: another rank file fix
2018-12-21 15:30:42 -07:00
Aurélien Bouteiller
d9b0dad828
Correctly propagate the oversubscribe flag to the spawnees
This is a cherry-pick of master (2820aef). The propagation is intended to resolve issue #6130

Signed-off-by: Aurélien Bouteiller <bouteill@icl.utk.edu>
2018-12-21 14:53:25 -05:00
Ralph Castain
2d9c936082 If job is fully described, there will be no ppn string to unpack
Signed-off-by: Ralph Castain <rhc@open-mpi.org>
(cherry picked from commit d7283807410c1d37d966180847debe1d43361e92)
2018-12-18 08:27:39 -08:00
Howard Pritchard
4be4282312
Merge pull request #6128 from ggouaillardet/topic/v4.0.x/mpiext_short_path
mpiext: keep paths short
2018-12-17 13:22:19 -07:00
Howard Pritchard
71b83e8a09
Merge pull request #6193 from kawashima-fj/pr/v4.0.x/fix-type-create-f90
v4.0.x: mpi/c: Fix MPI_TYPE_CREATE_F90_{REAL,COMPLEX}
2018-12-17 13:21:21 -07:00
Howard Pritchard
d39d68e06d
Merge pull request #6169 from matcabral/remove_psm2_lower_p_v4
MTL/PSM2: Do not lower the priority when all processes are local.
2018-12-15 09:41:53 -07:00
KAWASHIMA Takahiro
8eb90ae9aa mpi/c: Fix MPI_TYPE_CREATE_F90_{REAL,COMPLEX}
This commit fixes edge cases of `r = 38` and `r = 308`.

As defined in the MPI standard, `TYPE_CREATE_F90_REAL` and
`TYPE_CREATE_F90_COMPLEX` must be consistent with the Fortran
`SELECTED_REAL_KIND` function. The `SELECTED_REAL_KIND` function is
defined based on the `RANGE` function. The `RANGE` function returns
`INT(MIN(LOG10(HUGE(X)), -LOG10(TINY(X))))` for a real value `X`.

The old code considers only `INT(LOG10(HUGE(X)))` using `*_MAX_10_EXP`.
This commit adds `INT(-LOG10(TINY(X)))` part using `*_MIN_10_EXP`.

This bug affected the following `p`-`r` combinations.

| p             | r   | expected  | returned  | expected  | returned  |
| :------------ | --: | :-------- | :-------- | :-------  | :-------- |
| MPI_UNDEFINED |  38 | REAL8     | REAL4     | COMPLEX16 | COMPLEX8  |
| 0 <= p <= 6   |  38 | REAL8     | REAL4     | COMPLEX16 | COMPLEX8  |
| MPI_UNDEFINED | 308 | REAL16    | REAL8     | COMPLEX32 | COMPLEX16 |
| 0 <= p <= 15  | 308 | REAL16    | REAL8     | COMPLEX32 | COMPLEX16 |

MPICH returns the same result as Open MPI with this fix.

Signed-off-by: KAWASHIMA Takahiro <t-kawashima@jp.fujitsu.com>
(cherry picked from commit 6fb01f64fe2bcdb4668e520eb458ffd3477e5e6f)
2018-12-13 16:01:56 +09:00
Jeff Squyres
ec8b2c47a9 ompi_ext.m4: ensure ext name <= 26 characters
Ensure that MPI extensions with mpif.h bindings have names that are
<=26 characters long.  26 is the magic number that still allows us to
have an "include ..." line in the user-facing mpif-ext.h header file
that includes this extension's header file without going over 72
characters.

Signed-off-by: Jeff Squyres <jsquyres@cisco.com>
Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>

(cherry picked from commit open-mpi/ompi@c0faf34855)
2018-12-12 09:25:12 +09:00
Gilles Gouaillardet
a79ce7d17f mpiext: updates for header file locations
Per discussion on https://github.com/open-mpi/ompi/pull/6030
and https://github.com/open-mpi/ompi/pull/6145, move
around where MPI extension header files are installed (specifically:
the installation tree path does not need to match the source tree
path).

For reference, header files were installed like this :

 - <prefix>/include/openmpi/ompi/mpiext/pcollreq/mpif-h/mpiext_pcollreq_mpifh.h
 - <prefix>/include/openmpi/ompi/mpiext/pcollreq/c/mpiext_pcollreq_c.h

and they are now installed like this :

 - <prefix>/include/openmpi/mpiext/mpiext_pcollreq_mpifh.h
 - <prefix>/include/openmpi/mpiext/mpiext_pcollreq_c.h

Signed-off-by: Jeff Squyres <jsquyres@cisco.com>
Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>

(cherry picked from commit open-mpi/ompi@975e3cd0c9)
2018-12-12 09:24:45 +09:00
Howard Pritchard
4cce2b84fa
Merge pull request #6176 from ggouaillardet/topic/v4.0.x/uct_configury
v4.0.x: btl/uct: fix a typo in configure.m4
2018-12-11 09:19:06 -07:00
Howard Pritchard
ebd3421e53
Merge pull request #6165 from ggouaillardet/topic/v4.0.x/pmix_atomics
pmix/pmix3x: fix macros usage in embedded pmix3x
2018-12-11 09:17:17 -07:00
Howard Pritchard
7ad025e92b
Merge pull request #6160 from hjelmn/v4.0.x_btl_uct_fix_usage_when_on_ugni_even_though_we_really_dont_want_to_support_that_usage
v4.0.x: fix btl/uct usage on Cray hardware
2018-12-11 09:14:09 -07:00
Howard Pritchard
552834b9a6
Merge pull request #6110 from hoopoepg/topic/amo-int32-64-datatypes-v4.0
OSHMEM/AMO: added int/uint/32/64 atomics calls - v4.0
2018-12-11 06:43:05 -07:00
Gilles Gouaillardet
f446472f06 btl/uct: fix a typo in configure.m4
remove whitespace around '=' when setting btl_uct_LIBS

Thanks Ake Sandgren for reporting this

Refs. open-mpi/ompi#6173

Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>

(cherry picked from commit open-mpi/ompi@b89deeb1bb)
2018-12-11 15:15:12 +09:00
Matias A Cabral
b2327049c1 MTL/PSM2: add missing default priority
Missing default priority after PR #6153

Signed-off-by: Matias Cabral <matias.a.cabral@intel.com>
(cherry picked from commit c76c6d8b2801ca43ba33168a0b92522786c7c5bb)
2018-12-07 16:22:59 -08:00
Matias A Cabral
80113a368f MTL/PSM2: Do not lower the priority when all processes are local.
The intention of lowering the priority when all processes are local
was to favor Vader BTL. However, in builds including the OFI MTL it
gets selected instead.

Reviewed-by: Spruit, Neil R <neil.r.spruit@intel.com>
Reviewed-by: Gopalakrishnan, Aravind <aravind.gopalakrishnan@intel.com>
Signed-off-by: Matias Cabral <matias.a.cabral@intel.com>
(cherry picked from commit fc8582c5606b7a3d1b711f8f7b6144808290a48f)
2018-12-07 11:11:43 -08:00
Gilles Gouaillardet
dd2b1ce49b btl/uct: fix a warning
Use the PRIsize_t macro to correctly print a size_t

Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>

(cherry picked from commit open-mpi/ompi@78aa6fdd1d)
2018-12-07 17:02:16 +09:00
Gilles Gouaillardet
ffbe85c65f btl/uct: fix AC_CHECK_DECLS usage
AC_CHECK_DECLS take a comma separated list of macros/symbols,
so replace the whitespace separator with a comma.

Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>

(cherry picked from commit open-mpi/ompi@b715dd2657)
2018-12-07 16:27:57 +09:00
Gilles Gouaillardet
195a07d03d pmix/pmix3x: fix macros usage in embedded pmix3x
Use PMIX_* macros instead of OPAL_* macros
master does things differently, so this is a one-off commit

Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>
2018-12-07 16:07:07 +09:00
Nathan Hjelm
0957861689 btl/uct: fix some issues when using UCX over ugni
Though not a recommended configuration it is possible to use Open MPI
over UCX over uGNI. This configuration had some issues related to the
connection management and tl selection. This commit fixes those
issues.

Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
(cherry picked from commit e07a64c52d92adf51732ea78e17b679f6deffa12)
Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
2018-12-06 10:57:59 -07:00
Gilles Gouaillardet
057118dbe6 btl/uct: fix AC_CHECK_DECLS usage
AC_CHECK_DECLS take a comma separated list of macros/symbols,
so replace the whitespace separator with a comma.

Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>
(cherry picked from commit b715dd26572ab18fdba92f06143456c0f9d6380a)
Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
2018-12-06 10:56:45 -07:00
Howard Pritchard
4f7f6d7061
Merge pull request #6155 from ggouaillardet/topic/4.0.1a1
roll to v4.0.1a1
2018-12-05 22:16:25 -07:00
Gilles Gouaillardet
0c3e76ec90 roll to v4.0.1a1
fixes previous commit open-mpi/ompi@116a140be8
that incorrectly rolled to v4.1.0a1

Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>
2018-12-06 11:29:04 +09:00
Howard Pritchard
804f65f247
Merge pull request #6035 from ggouaillardet/topic/v4.0.x/mpiext_cuda
mpiext/cuda: do not include automatically generated file into dist ta…
2018-12-04 09:26:55 -07:00
Geoff Paulsen
752bbd195f
Merge pull request #6102 from hoopoepg/topic/set-osc-ucx-level-200-v4.0
OSC: set UCX module used by default - v4.0
2018-12-04 10:26:37 -06:00
Howard Pritchard
9fc4782b36
Merge pull request #6124 from hoopoepg/topic/oshmem-zero-len-coll-v4.0
OSHMEM: added processing of zero-length collectives - v4.0
2018-12-04 05:44:16 -07:00
Geoff Paulsen
03cf3e4400
Merge pull request #6112 from kawashima-fj/pr/v4.0.x/update-pcoll-doc
v4.0.x: README & man: Update pcollreq documentation
2018-11-30 13:58:34 -06:00
Geoff Paulsen
240582f600
Merge pull request #6121 from ggouaillardet/topic/v4.0.x/ompi_ext_fortran
v4.0.x: configury: fix auto-generated mpi-[f08-]ext-module.F90
2018-11-30 13:57:29 -06:00
Geoff Paulsen
bd2990f502
Merge pull request #6131 from devreal/rdma-plug-memleak-v4.0.x
v4.0.x: Plug two memory leaks in rdma osc
2018-11-30 13:54:51 -06:00
Geoff Paulsen
51d20915fd
Merge pull request #6139 from rhc54/cmr40/rmap
v4.0.x: Fix typo for rmaps_base_oversubscribe
2018-11-30 13:51:36 -06:00
Geoff Paulsen
937cf86077
Merge pull request #6135 from jsquyres/pr/v4.0.x/README-typo-fix
v4.0.x: README: Fix a typo
2018-11-29 14:05:08 -06:00
Ralph Castain
98c8492057 Fix typo for rmaps_base_oversubscribe
Causes the MCA param to be ignored, while the cmd line option still
works.

Thanks to @iassiour for the report!

Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2018-11-29 07:40:51 -08:00
Joseph Schuchart
c5346751e6 Plug two memory leaks in rdma osc
Signed-off-by: Joseph Schuchart <schuchart@hlrs.de>
(cherry picked from commit 91885f5876129aa4fb43ed4b3404c9d1ca7e08b8)
2018-11-29 10:19:26 -05:00
Jeff Squyres
e56c179d58 README: Fix a typo
Signed-off-by: Jeff Squyres <jsquyres@cisco.com>
(cherry picked from commit e6241eaf6ba6bffeb6b85def68e420a7ab66dce8)
2018-11-28 14:52:08 -08:00
Sergey Oblomov
6651672711 OSC/UCX: set max level value to 60
Signed-off-by: Sergey Oblomov <sergeyo@mellanox.com>
(cherry picked from commit 2d230b3aacce0185f0d46e69f608071b670eeb3c)
2018-11-27 20:35:30 +02:00
Sergey Oblomov
5838760a3a OSHMEM/COLL/BCAST: removed unnecessary bcast call
- removed unnecessary bcast call on zero-length request

Signed-off-by: Sergey Oblomov <sergeyo@mellanox.com>
(cherry picked from commit c93927e27a8e4241236d82c0d34ea445aa619aff)
2018-11-27 14:26:56 +02:00
Sergey Oblomov
0a064d8c8d OSHMEM/COLL: optimization on zero-length ops
- removed barrier call on zero-length operations

Signed-off-by: Sergey Oblomov <sergeyo@mellanox.com>
(cherry picked from commit ff2fd0679eb4b31bfd840395d73746612e2670f4)
2018-11-27 14:26:52 +02:00
Sergey Oblomov
dea9cf6b63 OSHMEM: added processing of zero-length collectives
- according spec 1.4, annex C shmem collectives should process
  calls where number of elements is zero independently from pointer
  value
- added zero-count processing - it just call barrier to
  sync ranks

Signed-off-by: Sergey Oblomov <sergeyo@mellanox.com>
(cherry picked from commit 9de128afaf5224193a88f00c897f5d5c94336f99)
2018-11-27 14:26:44 +02:00
Gilles Gouaillardet
cb6841ac64 configury: fix auto-generated mpi-[f08-]ext-module.F90
prefer #include vs include in order to correctly handle long Fortran lines.
We use the full path, and it can be very long, this is why
it cannot be passed to the Fortran compiler.

Thanks Igor Andriyash and Axel Huebl for reporting this issue.

Refs open-mpi/ompi#6106

Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>
(cherry picked from commit 475d0355d7ce2d31b582944c685fae13d711b3c9)
2018-11-27 09:31:56 +09:00
Howard Pritchard
7fc0841791
Merge pull request #6117 from yosefe/topic/pml-ucx-init-req_mpi_object-v4.0.x
pml_ucx: initialize req_mpi_object.comm for error handler
2018-11-26 13:30:53 -07:00
Howard Pritchard
176206fe8c
Merge pull request #6098 from jjhursey/enh/v4.0.x/vpid-unpack
Add OPAL_VPID to unpacking
2018-11-26 13:30:20 -07:00
Yossi Itigin
a112d10c93 pml_ucx: initialize req_mpi_object.comm for error handler
without this fix, an error handler invoked on pml_ucx request would
segfault while trying to dereference requests[i]->req_mpi_object.comm

(picked from master f36eeef)

Signed-off-by: Yossi Itigin <yosefe@mellanox.com>
2018-11-26 11:57:34 +02:00
KAWASHIMA Takahiro
6f68483fd5 README & man: Update pcollreq documentation
The feature of persistent collectives is approved in the Sept. 2018
MPI Forum meeting and 2018 Draft Specification of the MPI standard is
published during SC18.

Signed-off-by: KAWASHIMA Takahiro <t-kawashima@jp.fujitsu.com>
(cherry picked from commit 5f0fcf0f456aaa43f36215859412f33f79ac91f9)
2018-11-26 18:28:08 +09:00
Sergey Oblomov
63cbe36cab OSHMEM/AMO: added int/uint/32/64 atomics calls
- added int/uint/32/64 atomics calls
- added SHMEM_SYNC_SIZE macro

Signed-off-by: Sergey Oblomov <sergeyo@mellanox.com>
(cherry picked from commit 4c071da565b152dbceb8f48368327bd0ae218942)
2018-11-26 09:29:55 +02:00
Sergey Oblomov
38a4953707 OSC/UCX: added UCX version evaluation
- added UCX version evaluation to set OSC UCX priority

Signed-off-by: Sergey Oblomov <sergeyo@mellanox.com>
(cherry picked from commit e91f214982391b8e1b26be39147c357d32b8380e)
2018-11-22 11:31:53 +02:00
Sergey Oblomov
012e27af77 OSC: set UCX module used by default
- OSC/UCX module set priority to 200 to be used by default

Signed-off-by: Sergey Oblomov <sergeyo@mellanox.com>
(cherry picked from commit 36934a8bb2484c3d27d14683d65012ff422334f4)
2018-11-22 10:59:43 +02:00
Joshua Hursey
e1f75d5ff1 Add OPAL_VPID to unpacking
* Needed to properly read PMIx job data like the following
   - `OPAL_PMIX_LOCALLDR`
   - `OPAL_PMIX_RANK`
   - `OPAL_PMIX_GLOBAL_RANK`
   - `OPAL_PMIX_APPLDR`
   - `OPAL_PMIX_APP_RANK`

Signed-off-by: Joshua Hursey <jhursey@us.ibm.com>
(cherry picked from commit a557c4130c42a5a41aba5c08e606e7129d0bcb6d)
2018-11-21 11:48:58 -06:00
Geoff Paulsen
206b574bb3
Merge pull request #6076 from hppritcha/topic/on_to_v4.0.1
roll to v4.0.1a1
2018-11-20 09:55:55 -06:00