1
1
Граф коммитов

124 Коммитов

Автор SHA1 Сообщение Дата
Graham Fagg
39207db7cd removed the n-dimmension rule base..
replacing it was simpler code for V1

This commit was SVN r8033.
2005-11-08 03:03:51 +00:00
Graham Fagg
dcd3450e06 simplified the building of different rule sets
(also corrected some prototypes missing 'struct')

This commit was SVN r8003.
2005-11-06 22:05:50 +00:00
Jeff Squyres
42ec26e640 Update the copyright notices for IU and UTK.
This commit was SVN r7999.
2005-11-05 19:57:48 +00:00
Graham Fagg
9547a635a9 snapshot while switching systems
but, dynamic rules from a user defined config file is almost there now

This commit was SVN r7943.
2005-11-01 00:19:05 +00:00
Graham Fagg
fe03e068f2 allow forced algorithms (where the user or *test* suite knows better) to
go through the dynamic decision rule interface.
(forced algorithms are set with MCA params)
fixed some silly verbose output with wrong func name in it etc
updates to fixed dec rules.

This commit was SVN r7940.
2005-10-31 20:45:50 +00:00
Edgar Gabriel
2ec5fa5d24 - The component will remove itself from the list of potential collective
modules, if its priority is zero (the default value). Reason for that is
  + if there is no other module with a priority > 0, the hierarchical
    collective module has a problem anyway, since it has to rely on the coll
    modules of the subcommunicators. On the other hand, if its priority is
    zero, it won't be chosen anyway, and we can simply save the
    allreduce/allgather and comm_split operations which might occur during
    hierarchy detection.
  + to improve the startup times until we have the modex thing which we
    discussed with Jeff and Tim in Knoxville in place

- adding an mca parameter indicating a symmetric configuration. This can 
  speed up startup times, since each process can conclude from its data onto
  the data of the other processes -> no need for the allreduce operations. Per
  default  this parameter is set to "no".

This commit was SVN r7932.
2005-10-30 16:01:13 +00:00
Graham Fagg
5bb0d4a053 enable allreduce to be selected
This commit was SVN r7888.
2005-10-26 23:55:37 +00:00
Graham Fagg
2587d7ade9 added some more linear functions.
minor corrections on naming and debug info

This commit was SVN r7887.
2005-10-26 23:51:56 +00:00
Graham Fagg
c3e1dc410d Started to add basic linear functions
Also started to add the allreduce algorithms as I test them
(i.e. if it goes in its after testing from now on)

This commit was SVN r7886.
2005-10-26 23:11:32 +00:00
Jeff Squyres
d47ce065e9 Minor Makefile.am fix for static builds.
This commit was SVN r7882.
2005-10-26 15:57:58 +00:00
Edgar Gabriel
ba3bf6592f fixing some warnings. No idea yet why the static builds fail...
This commit was SVN r7879.
2005-10-26 12:56:56 +00:00
Edgar Gabriel
d009d8de57 opening the hierarchical collective component to the public. I am at this
stage fairly confident that 
- it works in most scenarious (with symmetric hierarchies, with asymmetric
  hierarchies, wihout hierarchies - it just removes itself)
- it does not create too many problems (I am not aware of any at least)
- it does not slow down startup anymore dramatically (thanks to the fixes of
  Brian, Jeff, Tim and a significant reduction in the number of collective
  operations in the comm_query)
Any feedback is highly welcome.

This commit was SVN r7868.
2005-10-25 18:38:43 +00:00
Edgar Gabriel
00c04ab56a moving the hierarch collective component to the new parameter
registration interface.

This commit was SVN r7867.
2005-10-25 18:34:47 +00:00
Graham Fagg
382f05c7ad Infastructure changes.
started to add static (fixed if) statement based decision rules based on gigE numbers
added mca params so that a user can force a certain algorithm/segment/topo on a per collective basis
(this is not in the fixed call path but only in the dynamic (at com create) call path).
(these params can be used by test suites such as OCC to choice which algorithm they are using).

This commit was SVN r7854.
2005-10-25 03:55:58 +00:00
Graham Fagg
d8e32464cb ops. setting/reading mca option from the right varible would help.
This commit was SVN r7850.
2005-10-24 21:33:48 +00:00
Edgar Gabriel
3a7efaf4d9 fix for reduce and allreduce for an unsymmetric case
This commit was SVN r7802.
2005-10-18 19:20:48 +00:00
Edgar Gabriel
818b4af554 - reverting the logic in the hierarchy detection stuff. This can reduce the
number of collective operations and simplifies the logic significantly.
- introducing a special case if size of comm == 1, avoiding thus collective
 operations as well ( i.e. no need for hierarchies)
- fix for an unsymmetric case. Still to be tested.

This commit was SVN r7799.
2005-10-18 18:17:50 +00:00
Brian Barrett
1302cb4072 The next in a long line of crazed build system changes from Brian. This was
originally suggested by Ralf Wildenhues, to try to speed autogen, configure,
and make (and possibly even make install).  Use automake's include directive
to drastically reduce the number of Makefile files (although the number of
Makefile.am files is the same - most are just included in a top-level
Makefile.am).  Also use an Automake SUBDIRs feature to eliminate the
dynamic-mca tree, which was no longer really needed.  This makes adding
a framework easier (since you don't have to remember the dynamic-mca
tree) and makes building faster (as make doesn't have to recurse through
the dynamic-mca tree)

This commit was SVN r7777.
2005-10-17 00:21:10 +00:00
Edgar Gabriel
7e45f64065 reduce has now been tested quite extensively for all (predefined) operations
and for all root nodes and passed all tests.
First cut on barrier (which from my perspective does not make sense from the
performance point of view) and on allreduce (which might make sense),

This commit was SVN r7774.
2005-10-15 22:24:44 +00:00
Edgar Gabriel
3fab9c628c switching the root and creating (if necessary) the new local leader sub-communicators seems to work as well. Thoroughly tested with bcast, not yet that exhaustivly tested for the reduction.
This commit was SVN r7773.
2005-10-15 21:13:44 +00:00
Edgar Gabriel
7d34770456 further bugfixes. The hierarchy detection works now as far as I can see (even in unsymmetric sitations). Bcast and reduce work as well. Still to test: the code which generates new local leader communicators, in case the root of the operation is not yet part of the lleader comm.
This commit was SVN r7772.
2005-10-15 19:36:54 +00:00
Edgar Gabriel
63554d245f further bugfixes
This commit was SVN r7771.
2005-10-15 18:44:57 +00:00
Edgar Gabriel
92c7b77cbc minor bug fixes
This commit was SVN r7770.
2005-10-15 18:32:40 +00:00
Edgar Gabriel
ba163c611c checkpoint before moving to a real cluster. Most of the recoding should be
done. This version also doesn't break ompi (at least if its not chosen :-) ).
New features compared to the version from last Thursday (where bcast and
reduce seemed to work in most scenarios):
- clearer internal infrastructure
- ability to handle all root processes with a (hopefully) minimal number of
local leader communicators. 

This commit was SVN r7769.
2005-10-15 17:04:01 +00:00
Edgar Gabriel
84c070fc0f get rid of the different modes how to store the colorarray for now. Might be
reintroduced later as an optimization.

This commit was SVN r7762.
2005-10-14 18:11:21 +00:00
Edgar Gabriel
6d14440972 checkpoint for moving again to another machine. major rewrite to clean
up internal interfaces in progress.

This commit was SVN r7761.
2005-10-14 17:41:44 +00:00
Edgar Gabriel
770aeaf97b modifications towards adding new local-leader communicators.
This commit was SVN r7760.
2005-10-14 12:18:29 +00:00
Graham Fagg
636b42afff handle non existant recv buf in reduce for non root processes
(basic allreduce does this for mpi_in_place case)

This commit was SVN r7759.
2005-10-14 00:00:37 +00:00
Graham Fagg
61b8218d76 MPI_IN_PLACE fix for reduce.
(actually a work around for an optimisation in the reduce for not saving ops on the first recv of each segment)
Minor change in topo.

This commit was SVN r7758.
2005-10-13 23:38:21 +00:00
Edgar Gabriel
48f2563b4c checkpoint. Moving to another machine.
This commit was SVN r7757.
2005-10-13 20:04:26 +00:00
Edgar Gabriel
4b05359b16 minor fixes when freeing the component
This commit was SVN r7756.
2005-10-13 18:22:16 +00:00
Edgar Gabriel
0a5a346bbb first cut on the reduce operation.
This commit was SVN r7755.
2005-10-13 17:58:13 +00:00
Edgar Gabriel
30af775d40 further fixes. The first hierarchical MPI_Bcast works! Its just ~ 100 times slower then basic at the moment :-)
This commit was SVN r7754.
2005-10-13 17:34:42 +00:00
Edgar Gabriel
460b5cb840 further corrections to the hierarchy detection algorithms. It seems to work now as far as my tests show...
This commit was SVN r7753.
2005-10-13 16:21:13 +00:00
Edgar Gabriel
f5d16419b2 fix in the logic regarding protocol detection.
This commit was SVN r7749.
2005-10-13 15:07:35 +00:00
Edgar Gabriel
3e5ad3e681 Updates
This commit was SVN r7738.
2005-10-12 20:56:29 +00:00
Edgar Gabriel
25518b63c5 first version of coll_hierarch which does not crash the rest of the
library as long as its not selected :-)

This commit was SVN r7707.
2005-10-11 22:05:24 +00:00
Edgar Gabriel
0675c22dab updating with Jeff's help to the recent autogen/configure system
This commit was SVN r7705.
2005-10-11 21:50:16 +00:00
Edgar Gabriel
7b07dbc163 another round of fixes. Unfortunatly, I also have to provide a trivial
version of reduce and gather to make all this work....

This commit was SVN r7702.
2005-10-11 21:26:07 +00:00
Edgar Gabriel
c8adc2e65e coding around the collective operations
This commit was SVN r7698.
2005-10-11 20:34:17 +00:00
Edgar Gabriel
083d0b9630 Checkpoint: most of the coding should be done for the basic
infrastructure.

This commit was SVN r7696.
2005-10-11 19:45:21 +00:00
Graham Fagg
607bdf51b6 Last Cleanup BEFORE adding last two methods and final cross over points.
- new mca param calls
- move printfs to OPAL_OUTPUT

This commit was SVN r7692.
2005-10-11 18:51:03 +00:00
Edgar Gabriel
b42d4ac780 Checkpoint:
- update the hierarch stuff to use btl's instead of ptl's
- start the new logic regarding how to handle local leader communicators

This commit was SVN r7691.
2005-10-11 17:29:59 +00:00
Jeff Squyres
b22fab2826 Fix for a bug Galen noticed yesterday -- make the shared memory only
be allocated the first time a sm coll is selected for a communicator,
not before.

This commit was SVN r7647.
2005-10-06 13:17:27 +00:00
Jeff Squyres
83b5a675f9 Don't automatically take the first entry off the selected component
list; be sure to check its priority against the basic component and
take the one with the higher priority.

This commit was SVN r7621.
2005-10-04 17:09:45 +00:00
Jeff Squyres
b17c4334c4 - Remove all vestigates of using the built-in mcb_tree from the
reduce_inorder() function -- we don't use the tree at all.
- Add more relevant "volatile"'s for the control buffers in the
  fragment mpool (and associated casts where necessary)

This commit was SVN r7616.
2005-10-04 14:52:59 +00:00
Jeff Squyres
c7fe54ba44 - Remove some silly compiler warnings
- Move the "process 0" logic out of the main loop in reduce to make
  the code a bit less complex (at the price of slight code
  duplication, but it iss now significantly easier to read)
- Fix problem with uniquenes guarantee in the bootstrap mpool -- using
  the CID alone was not sufficient enough to guarantee uniquenes; now
  use (CID, rank 0 process name) tuple to check for uniqueness
- Made a few debugging help changes in coll_sm.h; especially helps
  debugging on uniprocessors

This commit was SVN r7599.
2005-10-03 21:34:58 +00:00
Jeff Squyres
2cedfeec53 - Eliminate some unused base globals
- Move one base global to the basic component and make it an MCA
  parameter 
- Convert the basic component to use the new MCA param API

This commit was SVN r7598.
2005-10-03 21:07:42 +00:00
Jeff Squyres
57fb96b018 Clarification of a help message
This commit was SVN r7597.
2005-10-03 21:06:13 +00:00
Jeff Squyres
ab099fa8cb Re-indent; real commit with some changes coming shortly.
This commit was SVN r7596.
2005-10-03 19:56:39 +00:00