Topic/v1.10/coll fixes #1262

ggouaillardet · 2016-07-11T08:23:13Z

No description provided.

some of the collective modules. Added a new function opan_datatype_span, to compute the memory span of count number of datatype, excluding the gaps in the beginning and at the end. If a memory allocation is made using the returned value, the gap (also returned) should be removed from the allocated pointer. (back-ported from commit open-mpi/ompi@4d00c59)

(cherry picked from commit open-mpi/ompi@688108c)

Offset temporary buffer when a non zero lower bound datatype is used. Thanks Hristo Iliev for the report (back-ported from commit open-mpi/ompi@0e39319)

…duce_intra_basic_linear() Thanks Yuki Matsumoto for the report (back-ported from commit open-mpi/ompi@c06fb04)

- correctly handle non zero lower bound ddt - correctly handle ddt with size > extent Thanks Yuki Matsumoto for the report (back-ported from commit open-mpi/ompi@488d037)

- correctly handle non zero lower bound ddt - correctly handle ddt with size > extent Thanks Yuki Matsumoto for the report (back-ported from commit open-mpi/ompi@3e559a1)

- correctly handle non commutative operators - correctly handle non zero lower bound ddt - correctly handle ddt with size > extent - revamp NBC_Sched_op so it takes two buffers and matches ompi_op_reduce semantic - various fix for inter communicators Thanks Yuki Matsumoto for the report (back-ported from commit open-mpi/ompi@678d086)

as reported by Coverity with CIDs 1363349-1363362 (back-ported from commit open-mpi/ompi@7b8094a)

…e_alltoallv_intra_basic_inplace() (back-ported from commit open-mpi/ompi@a55d574)

…edoubling err handler Fix CID 1362630 Fixes open-mpi/ompi@0e39319 (back-ported from commit open-mpi/ompi@80e362d)

…reduce_scatter_inter() this is now useless since the scatter is done via the local communicator (back-ported from commit open-mpi/ompi@1462450)

mellanox-github · 2016-07-11T09:26:41Z

Test FAILed.
See http://bgate.mellanox.com/jenkins/job/gh-ompi-release-pr/1842/ for details.

ggouaillardet · 2016-07-11T09:37:21Z

:bot:assign: @bosilca
:bot🏷️bug
:bot:milestone:v1.10.4

bosilca · 2016-07-11T10:17:33Z

👍

jsquyres · 2016-07-11T11:54:11Z

@jladd-mlnx @artpol84 Looks like Mellanox Jenkins timed out on the thread tests. But I see that the --bind-to none fix was applied there. Are the thread tests relevant to run on the v1.10 branch?

artpol84 · 2016-07-11T12:04:34Z

Yes for the yalla I think.
I noticed that some tests are really slow progressing and the runtime was
close to timeout value. I can increase it for now. With to check the reason
of the slowness when I'll have the time.

понедельник, 11 июля 2016 г. пользователь Jeff Squyres написал:

@jladd-mlnx https://github.com/jladd-mlnx @artpol84
https://github.com/artpol84 Looks like Mellanox Jenkins timed out on
the thread tests. But I see that the --bind-to none fix was applied
there. Are the thread tests relevant to run on the v1.10 branch?

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
#1262 (comment),
or mute the thread
https://github.com/notifications/unsubscribe/AHL5PmZiP72a7drQ9MDWmrW0LhR65zGNks5qUi7kgaJpZM4JJIeS
.

Best regards, Artem Polyakov
(Mobile mail)

artpol84 · 2016-07-11T12:16:20Z

Indeed those are high latencies. Maybe related to non-bounded run:

12:16:11 Size (bytes) Time (us)
12:16:31 0 9906.442500
12:16:50 16 9628.991000
12:17:10 32 9732.075000
12:17:30 48 10289.908500
12:17:51 64 10521.495500
12:18:13 80 10696.196000
12:18:33 96 10311.793000
12:18:54 112 10442.492000
12:19:15 128 10665.487000
12:19:35 144 9607.496000
12:19:54 160 9481.857500
12:20:14 176 10232.132000
12:20:36 192 10754.497500
12:20:56 208 10296.493500
12:21:15 224 9524.908000
12:21:35 240 10106.571500
12:21:57 256 10715.749500
12:22:18 320 10668.747000
12:22:40 384 10953.499000
12:23:02 448 10758.986500
12:23:23 512 10828.990000
12:23:44 576 10571.493500
12:24:06 640 10701.493000
12:24:26 704 10335.995500
12:24:45 768 9506.994000
12:25:07 832 10707.497000
12:25:28 896 10694.498000
12:25:50 960 10679.993000

I'll check ASAP

artpol84 · 2016-07-11T12:17:09Z

Bot:retest

jsquyres · 2016-07-11T12:33:06Z

@artpol84 The thread tests might not be relevant for the v1.10 branch. Only in master / v2.x did we do all the correctness and performance improvements for THREAD_MULTIPLE.

jsquyres · 2016-07-11T12:33:24Z

bot:assign: @bosilca
bot:label:bug
bot:milestone:v1.10.4

ompiteam-bot · 2016-07-11T12:41:27Z

OMPIBot error: Milestone "v1.10.4" does not exist.

mellanox-github · 2016-07-11T13:14:43Z

Test FAILed.
See http://bgate.mellanox.com/jenkins/job/gh-ompi-release-pr/1843/ for details.

artpol84 · 2016-07-15T17:39:33Z

bot:retest

mellanox-github · 2016-07-15T18:21:24Z

Test FAILed.
See http://bgate.mellanox.com/jenkins/job/gh-ompi-release-pr/1884/ for details.

Di0gen · 2016-07-17T07:23:14Z

bot:retest

mellanox-github · 2016-07-17T07:30:18Z

Test FAILed.
See http://bgate.mellanox.com/jenkins/job/gh-ompi-release-pr/1887/ for details.

artpol84 · 2016-07-17T07:58:34Z

bot:retest

artpol84 · 2016-07-17T08:02:01Z

bot:retest

mellanox-github · 2016-07-17T08:21:16Z

Test FAILed.
See http://bgate.mellanox.com/jenkins/job/gh-ompi-release-pr/1889/ for details.

mellanox-github · 2016-07-17T10:56:16Z

Test PASSed.
See http://bgate.mellanox.com/jenkins/job/gh-ompi-release-pr/1890/ for details.

bosilca · 2016-07-19T19:03:51Z

👍

jsquyres · 2016-07-26T15:05:28Z

bot:mellanox:retest

mellanox-github · 2016-07-26T16:17:55Z

Test PASSed.
See http://bgate.mellanox.com/jenkins/job/gh-ompi-release-pr/1974/ for details.

bosilca and others added 11 commits July 11, 2016 09:11

Patch submitted by @ggouaillardet on ticket open-mpi#1091.

3d570ed

(cherry picked from commit open-mpi/ompi@688108c)

coll/base: fix [all]reduce with non zero lower bound datatypes

2207713

Offset temporary buffer when a non zero lower bound datatype is used. Thanks Hristo Iliev for the report (back-ported from commit open-mpi/ompi@0e39319)

coll/base: fix non zero lower bound ddt handling in ompi_coll_base_re…

400ac5d

…duce_intra_basic_linear() Thanks Yuki Matsumoto for the report (back-ported from commit open-mpi/ompi@c06fb04)

coll/basic: fix non standard ddt handling

ca93528

- correctly handle non zero lower bound ddt - correctly handle ddt with size > extent Thanks Yuki Matsumoto for the report (back-ported from commit open-mpi/ompi@488d037)

coll/inter: fix non standard ddt handling

a7a9ae8

- correctly handle non zero lower bound ddt - correctly handle ddt with size > extent Thanks Yuki Matsumoto for the report (back-ported from commit open-mpi/ompi@3e559a1)

coll/base: silence misc warning

efa0df0

as reported by Coverity with CIDs 1363349-1363362 (back-ported from commit open-mpi/ompi@7b8094a)

coll/base: fix non zero lower bound datatype handling in mca_coll_bas…

ac36c80

…e_alltoallv_intra_basic_inplace() (back-ported from commit open-mpi/ompi@a55d574)

coll/base: fix memory free in ompi_coll_base_allreduce_intra_recursiv…

5dbf964

…edoubling err handler Fix CID 1362630 Fixes open-mpi/ompi@0e39319 (back-ported from commit open-mpi/ompi@80e362d)

coll/libnbc: do not exchange data between roots in ompi_coll_libnbc_i…

a6cf208

…reduce_scatter_inter() this is now useless since the scatter is done via the local communicator (back-ported from commit open-mpi/ompi@1462450)

ompiteam-bot assigned bosilca Jul 11, 2016

ompiteam-bot added the bug label Jul 11, 2016

jsquyres added this to the 1.10.4 milestone Jul 11, 2016

ompiteam-bot added the reviewed label Jul 19, 2016

ggouaillardet mentioned this pull request Jul 26, 2016

Wrong result on non-blocking collectives with user-defined op open-mpi/ompi#1754

Closed

rhc54 merged commit 640bcf6 into open-mpi:v1.10 Jul 28, 2016

rhc54 added the RM-Approved label Jul 28, 2016

jsquyres mentioned this pull request Jul 28, 2016

v1.10 MTT fails on collective tests open-mpi/ompi#1914

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Topic/v1.10/coll fixes #1262

Topic/v1.10/coll fixes #1262

ggouaillardet commented Jul 11, 2016

mellanox-github commented Jul 11, 2016

ggouaillardet commented Jul 11, 2016

bosilca commented Jul 11, 2016

jsquyres commented Jul 11, 2016

artpol84 commented Jul 11, 2016

artpol84 commented Jul 11, 2016 •

edited

Loading

artpol84 commented Jul 11, 2016

jsquyres commented Jul 11, 2016

jsquyres commented Jul 11, 2016

ompiteam-bot commented Jul 11, 2016

mellanox-github commented Jul 11, 2016

artpol84 commented Jul 15, 2016

mellanox-github commented Jul 15, 2016

Di0gen commented Jul 17, 2016

mellanox-github commented Jul 17, 2016

artpol84 commented Jul 17, 2016

artpol84 commented Jul 17, 2016

mellanox-github commented Jul 17, 2016

mellanox-github commented Jul 17, 2016

bosilca commented Jul 19, 2016

jsquyres commented Jul 26, 2016

mellanox-github commented Jul 26, 2016

Topic/v1.10/coll fixes #1262

Topic/v1.10/coll fixes #1262

Conversation

ggouaillardet commented Jul 11, 2016

mellanox-github commented Jul 11, 2016

ggouaillardet commented Jul 11, 2016

bosilca commented Jul 11, 2016

jsquyres commented Jul 11, 2016

artpol84 commented Jul 11, 2016

artpol84 commented Jul 11, 2016 • edited Loading

artpol84 commented Jul 11, 2016

jsquyres commented Jul 11, 2016

jsquyres commented Jul 11, 2016

ompiteam-bot commented Jul 11, 2016

mellanox-github commented Jul 11, 2016

artpol84 commented Jul 15, 2016

mellanox-github commented Jul 15, 2016

Di0gen commented Jul 17, 2016

mellanox-github commented Jul 17, 2016

artpol84 commented Jul 17, 2016

artpol84 commented Jul 17, 2016

mellanox-github commented Jul 17, 2016

mellanox-github commented Jul 17, 2016

bosilca commented Jul 19, 2016

jsquyres commented Jul 26, 2016

mellanox-github commented Jul 26, 2016

artpol84 commented Jul 11, 2016 •

edited

Loading