1
1
Граф коммитов

9725 Коммитов

Автор SHA1 Сообщение Дата
Brian Barrett
bc32a0f1bc Don't build all the network support code on systems without sockets (like
the Cray XT3)

This commit was SVN r14797.
2007-05-30 15:25:22 +00:00
Gleb Natapov
6b0d8c0858 TCP BTL ignores btl_tcp_bandwidth parameter. Fix it.
This commit was SVN r14795.
2007-05-30 14:12:05 +00:00
Rainer Keller
2cc4e6de85 - Allow detection of weak symbols support with Intel compiler
This commit was SVN r14794.
2007-05-30 13:01:36 +00:00
Donald Kerr
91c9b7b6f9 don't call dat_evd_resize if new value is less than or equal to current because ofed stack does not return DAT_INVALID_STATE
This commit was SVN r14792.
2007-05-29 20:08:16 +00:00
George Bosilca
6afbc02052 The idea behind this patch is to decrease the number of strcmp used in the replica
by using a small hash function before doing the strcmp. The hask key for each
registry entry is computed when it is added to the registry. When we're doing a
query, instead of comparing the 2 strings we first check if the hash key match,
and if they do match then we compare the 2 strings in order to make sure we
eliminate collisions from our answers.

There is some benefit in terms of performance. It's hardly visible for few
processes, but it start showing up when the number of processes increase. In fact
the number of strcmp in the trace file drastically decrease. The main reason it
works well, is because most of the keys start with basically the same chars
(such as orte-blahblah) which transform the strcmp on a loop over few chars.

This commit was SVN r14791.
2007-05-29 18:40:07 +00:00
Ralph Castain
a2964f429e Fix a compiler warning - strncmp returns an int, so you have to compare to 0 instead of NULL.
This commit was SVN r14790.
2007-05-29 18:02:10 +00:00
Anya Tatashina
de676d717b Ref Trac #1032; added suport for full path launching with TotalView
This commit was SVN r14789.
2007-05-29 17:39:11 +00:00
Tim Prins
f95442dec9 better error output
This commit was SVN r14788.
2007-05-29 16:53:40 +00:00
Tim Prins
b4e3ad8da0 Fixup tests for recent api changes
cleanup a ton of warnings, include proper files

fix orte_ring, it had a deadlock in it...

fix the abort test so it can be used with less than 4 processes

This commit was SVN r14787.
2007-05-29 16:22:50 +00:00
Rainer Keller
dd8bea2ea2 - Small comment fixes separately
This commit was SVN r14786.
2007-05-29 15:50:02 +00:00
Rainer Keller
43ce11d578 - Fortran 90 standard requires no name at:
end interface
   (ifort complains, pgf90 chokes on it).

This commit was SVN r14785.
2007-05-29 15:37:44 +00:00
Gleb Natapov
06bf5d74e7 Remove mca_pml_ob1_send_fin_btl function.
This commit was SVN r14784.
2007-05-28 06:51:12 +00:00
Gleb Natapov
f5078db0db Fix order of parameters to function.
This commit was SVN r14783.
2007-05-27 13:45:24 +00:00
Gleb Natapov
f191834e56 No need for MCA_BTL_FLAGS_NEED_ACK any more. As of commit r14768 this is the
default behaviour.

This commit was SVN r14782.

The following SVN revision numbers were found above:
  r14768 --> open-mpi/ompi@3401bd2b07
2007-05-27 11:25:39 +00:00
Gleb Natapov
444762456e Don't dereference NULL pointer. Fix bug introduced in r14768.
This commit was SVN r14781.

The following SVN revision numbers were found above:
  r14768 --> open-mpi/ompi@3401bd2b07
2007-05-27 09:24:56 +00:00
Gleb Natapov
ad69d3c6ac Fix out of resource handling for FIN packets broken by r14768.
This commit was SVN r14780.

The following SVN revision numbers were found above:
  r14768 --> open-mpi/ompi@3401bd2b07
2007-05-27 08:29:38 +00:00
Brian Barrett
80fa8eef6e Don't include malloc.h in mpool/base/base.h because it causes all kinds of problems
when the memory debugging stuff is enabled.  Push it down into the two .c files
that do use it.

This commit was SVN r14779.
2007-05-27 03:55:21 +00:00
Brian Barrett
eff77f7cbb Need sys/socket for these defines on some platforms (Solaris, and I"m sure others)
This commit was SVN r14778.
2007-05-27 03:52:56 +00:00
George Bosilca
eb43abf7ae Allow compilation when there is a progress thread.
This commit was SVN r14776.
2007-05-25 01:59:29 +00:00
George Bosilca
8b817e96fd Allow threaded compilation.
This commit was SVN r14775.
2007-05-25 01:53:29 +00:00
Josh Hursey
1e678c3f55 per conversation with Ralph and Jeff take out the opal_init_only logic.
This commit moves the initalization/finalization of opal_event and opal_progress
to opal_init/finalize. These were previously init/final in ORTE which is an
abstraction violation. After talking about it we concluded that there are no
ordering issues that require these to be init/final in ORTE instead of OPAL.

I ran the IBM test suite against this commit and it didn't turn up any new
failures so I think it is good to go.

Let us know if this causes problems.

This commit was SVN r14773.
2007-05-24 21:54:58 +00:00
Jeff Squyres
d627918678 Add item about ompi_info fix
This commit was SVN r14772.
2007-05-24 21:37:47 +00:00
Jeff Squyres
4bf964eb3f Fix issue found by Josh England: ompi_info deliberately clears
environment variables corresopnding to framework MCA parameters so
that the opal MCA base loads '''all''' components (not just the ones
specified in the environment variables).  This has the side-effect of
not showing the user's value when displaying the framework MCA
parameters via --param output.  For example:

{{{
shell% setenv OMPI_MCA_btl foo
shell% ompi_info --param btl base
}}}

The above sequence would show a "<none>" value for the "btl" parameter
instead of "foo".

This commit restores the environment after we munge it to make the
loader load all components.  Hence, the above command sequence will
show "foo" for the "btl" parameter value, not "<none>".

This commit was SVN r14771.
2007-05-24 21:30:38 +00:00
Josh Hursey
e8b85faf28 Fix for the invalid arguments case. we were not finalizing cleanly.
This commit was SVN r14770.
2007-05-24 21:27:06 +00:00
Josh Hursey
a296ef5487 Checkpoint/restart fix:
Still recovering from interface changes.

This commit was SVN r14769.
2007-05-24 21:12:34 +00:00
Galen Shipman
3401bd2b07 Add optional ordering to the BTL interface.
This is required to tighten up the BTL semantics. Ordering is not guaranteed,
but, if the BTL returns a order tag in a descriptor (other than
MCA_BTL_NO_ORDER) then we may request another descriptor that will obey
ordering w.r.t. to the other descriptor.


This will allow sane behavior for RDMA networks, where local completion of an
RDMA operation on the active side does not imply remote completion on the
passive side. If we send a FIN message after local completion and the FIN is
not ordered w.r.t. the RDMA operation then badness may occur as the passive
side may now try to deregister the memory and the RDMA operation may still be
pending on the passive side. 

Note that this has no impact on networks that don't suffer from this
limitation as the ORDER tag can simply always be specified as
MCA_BTL_NO_ORDER.

This commit was SVN r14768.
2007-05-24 19:51:26 +00:00
Rainer Keller
a665b7a20d - Getting rid of "missing initializer" warnings
This commit was SVN r14766.
2007-05-24 19:19:52 +00:00
Jeff Squyres
81df632e29 Clarification to MCA parameter help messages
This commit was SVN r14765.
2007-05-24 19:18:29 +00:00
Rainer Keller
7d84de8510 - now the formatting (just getting rid of spaces at the end)....
This commit was SVN r14764.
2007-05-24 19:10:32 +00:00
Rainer Keller
ff3cfc0011 - Get rid of "set but never used" warning
This commit was SVN r14763.
2007-05-24 19:07:45 +00:00
Rainer Keller
c8668ef83f - Get rid of unused variables / set but never used warnings.
This commit was SVN r14762.
2007-05-24 18:57:51 +00:00
Rainer Keller
0e2a335297 - When not debugging, do not initialize a unused mutexattr.
This commit was SVN r14761.
2007-05-24 18:54:22 +00:00
George Bosilca
146989fee7 Allow for datatype with more than 2^16-1 entries. The new limit is 2^32-1 and it
is enforced at the data-type creation.

This commit was SVN r14758.
2007-05-24 17:24:57 +00:00
Brian Barrett
5ec421e1b0 Create a new queue (to simplify locking) for requests that are started but
can not be started by the BTL.

This commit was SVN r14757.
2007-05-24 17:21:56 +00:00
George Bosilca
bd5be6ed79 Decrease the dependencies on the rest of the Open MPI code base.
This commit was SVN r14756.
2007-05-24 16:59:00 +00:00
George Bosilca
7459ab45f1 This is the complete commit for the TCP header issue. Jeff commit a partial
fix (r14749) and then backed it out (r14753).

As we are unable to send more than a 32 bits length over TCP in one go, there
is no reason to have an uint64 length in the header. This reduce the size
of the TCP header.

This commit was SVN r14755.

The following SVN revision numbers were found above:
  r14749 --> open-mpi/ompi@48c026ce6b
  r14753 --> open-mpi/ompi@28ed850b4c
2007-05-24 16:40:49 +00:00
George Bosilca
f744e09462 The hopefully final correction for the ticket #919. Make sure we are always aligned
to the max width (MPI_Aint) when we pack the description of a data-type.

This commit was SVN r14754.
2007-05-24 16:08:23 +00:00
Jeff Squyres
28ed850b4c Back out r14749; it wasn't quite ready for prime time yet...
This commit was SVN r14753.

The following SVN revision numbers were found above:
  r14749 --> open-mpi/ompi@48c026ce6b
2007-05-24 15:46:15 +00:00
Brian Barrett
1b025798d2 remove some now unneeded volatiles
This commit was SVN r14752.
2007-05-24 15:42:06 +00:00
Brian Barrett
1a9f48c89d Some much needed cleanup of the rdma one-sided component, similar to
r14703 for the point-to-point component.

  * Associate the list of long message requests to poll with the 
    component, not the individual modules
  * add progress thread that sits on the OMPI request structure
    and wakes up at the appropriate time to poll the message
    list to move long messages asynchronously.
  * Instead of calling opal_progress() all over the place, move
    to using the condition variables like the rest of the project.
    Has the advantage of moving it slightly further along in the
    becoming thread safe thing.
  * Fix a problem with the passive side of unlock where it could 
    go recursive and cause all kinds of problems, especially 
    when progress threads are used. Instead, have two parts of 
    passive unlock -- one to start the unlock, and another to 
    complete the lock and send the ack back. The data moving 
    code trips the second at the right time. 

This commit was SVN r14751.

The following SVN revision numbers were found above:
  r14703 --> open-mpi/ompi@2b4b754925
2007-05-24 15:41:24 +00:00
Jeff Squyres
48c026ce6b Commit a patch from George (reviewed by Brian): reduce the size of the
mca_btl_tcp_hdr_t struct and remove the need for the heterogeneous
padding by changing the type of the "size" member to be uint32_t
(vs. uint64_t).  The value would never be greater than 32 bits anyway,
so having the type be uint64_t was wasteful.

This commit was SVN r14749.
2007-05-24 15:08:57 +00:00
Jeff Squyres
379a4ec5e2 While we're editing MCA params in the oob tcp component, ditch the use
of the deprecated MCA param API for registering MCA parameters and
update to the current API.

This commit was SVN r14747.
2007-05-24 13:01:55 +00:00
Jeff Squyres
839c1db95c Fix something that has been bugging me for a while:
Rename the oob_tcp_include and oob_tcp_exclude MCA parameters to be
oob_tcp_if_include and oob_tcp_if_exclude (to match the convention
with btl_tcp_if_[in|ex]clude).  Keep "hidden" synonyms oob_tcp_include
and oob_tcp_exclude in case anyone is actually using them (and some
users undoubtedly are), but do not have them show up in ompi_info
--param output.  Instead, the new "oob_tcp_if_*" names will show up in
ompi_info output.

This commit was SVN r14746.
2007-05-24 12:52:26 +00:00
Gleb Natapov
be71b78f6a Initialize btl_send_limit before use.
This commit was SVN r14745.
2007-05-24 08:40:26 +00:00
Josh Hursey
caea631032 Fix a compiler error when not building with ipv6 support.
This commit was SVN r14743.
2007-05-23 21:53:00 +00:00
Brian Barrett
5f15becf4e Allow multiple connections to be started simultaneously when doing the OOB
wireup.  For small clusters or clusters with decent ARP lookup and
connect times, this will have marginal impact.  For systems with either
bad ARP lookup times or long connect times, increasing this number
to something much closer to SOMAXCONN (128 on most modern machines) will
result in a faster OOB wireup.  Don't set higher than SOMAXCONN or you
can end up with lots of connect() retries and we'll end up slower.

This commit was SVN r14742.
2007-05-23 21:35:44 +00:00
Brian Barrett
075389f67d fix some printf warnings
This commit was SVN r14740.
2007-05-23 21:19:26 +00:00
Brian Barrett
38b0d22243 Some cleanups to the pt2pt component
* Remove unused declaration
  * remove unused variable warning when not using progress threads
  * If we're using progress threads, we want to lock, not trylock
    when in progress, since it was called from the wakeup thread
    and not the progress function

This commit was SVN r14739.
2007-05-23 20:31:25 +00:00
Jeff Squyres
28a102feab Add .libs to svn:ignore
This commit was SVN r14738.
2007-05-23 20:07:02 +00:00
George Bosilca
7485c920c4 Remove the useless comment as it does not reflect the reality anymore.
This commit was SVN r14737.
2007-05-23 18:55:02 +00:00