1
1
Граф коммитов

9 Коммитов

Автор SHA1 Сообщение Дата
Devendar Bureddy
228772ae81 hcoll gatherv support
cmr=v1.8.2:reviewer=jladd

This commit was SVN r32097.
2014-06-26 18:14:41 +00:00
Mike Dubman
cadc1485ff HCOLL: register memory release hook to avoid races
fixed by Devender, reviewed by Miked

cmr=v1.8.2:reviewer=ompi-rm1.8

This commit was SVN r31809.
2014-05-17 19:49:43 +00:00
Mike Dubman
a4990de055 mca: track external lib version (runtime/compiletime) for mca component
based on thread: http://www.open-mpi.org/community/lists/devel/2014/04/14505.php

Create mca parameter to track runtime/compiletime ext lib version for component.

cmr=v1.8.2:reviewer=ompi-rm1.8

This commit was SVN r31487.
2014-04-22 18:02:26 +00:00
Mike Dubman
b8550a55a7 HCOLL: many fixes
Adds coll_hcoll_np mca parameter similar to that of fca component (defaults to 32). Those who use hcoll be aware that from now on the communicators less than 32 procs will run w/o hcoll by default. - Resolves fallback issue in case libhcoll runs out of allowed contexts. The solution is moving hcoll_context_create from comm_enable to comm_query. Shortly, comm_enable should never return OMPI_ERROR in the coll component with highest priority (hcoll). Otherwise the ompi coll_base_select will unselect the coll funtion pointers and module references leaving the communicator w/o coll pointer. This will cause the fail. Same behavior can be reproduced even with tuned if one would hardcore some "return OMPI_ERROR" into it's module_enable funtion. - Additionally, removed all the dead code under #if 0; removed unused variables (path for library, active_modules list) and classes (module list wrapper)

Fixed by Val, Reviewed by Devendar/Josh/Miked

cmr=v1.7.4:reviewer=ompi-rm1.7

This commit was SVN r30341.
2014-01-21 12:19:47 +00:00
Mike Dubman
0fae2caef3 Create a comm keyval for hcoll component with delete callback function.
Set comm attribute with keyval.
Wait for pending hcoll module tasks in comm delete callback where PML
still valid on the communicator. safely destroy hcoll context during
hcoll module destructor.

Author: Devendar Bureddy 
reviewed by miked

cmr=v1.7.4:reviewer=ompi-rm1.7

This commit was SVN r30175.
2014-01-09 11:27:24 +00:00
Devendar Bureddy
4554770ee4 hcol fixes
cmr=v1.7.4:reviewer=jladd

This commit was SVN r29787.
2013-12-03 20:21:40 +00:00
Mike Dubman
19748e6957 fix race condition which can happen on finalize
1. Change in rte api implementation: now comm_world used to do p2p. 
This allows to not worry about other comms being destroyed.

2. added a notification mechanism with a help of which runtime can say libhcoll that RTE api can not be used any longer. 
pass a pointer to a flag, and its size to libhcoll.
The flag changes when the RTE is no longer available. 
Currently this flag is just ompi_mpi_finalized global bool value.

cmr=v1.7.3:reviewer=jladd

This commit was SVN r29331.
2013-10-02 13:38:47 +00:00
Joshua Ladd
82e092db1b Adding interface changes in hcoll component to support non-blocking collectives in libhcoll. This was added by Elena Elkina and reviewed by Josh Ladd.
cmr:v1.7.3:reviewer=jladd:subject=Add support for non-blocking collectives in hcoll

This commit was SVN r29244.
2013-09-25 16:14:59 +00:00
Joshua Ladd
d7a50343bf Per the details and schedule outlined in the attached RFC, Mellanox Technologies would like to CMR the new 'coll/hcoll' component. This component enables Mellanox Technologies' latest HPC middleware offering - 'Hcoll'. 'Hcoll' is a high-performance, standalone collectives library with support for truly asynchronous, non-blocking, hierarchical collectives via hardware offload on supporting Mellanox HCAs (ConnectX-3 and above.) To build the component, libhcoll must first be installed on your system, then you must configure OMPI with the configure flag: '--with-hcoll=/path/to/libhcoll'. Subsequent to installing, you may select the 'coll/hcoll' component at runtime as you would any other coll component, e.g. '-mca coll hcoll,tuned,libnbc'. This has been reviewed by Josh Ladd and should be added to cmr:v1.7:reviewer=jladd
This commit was SVN r28694.
2013-07-01 22:39:43 +00:00