1
1
Граф коммитов

3117 Коммитов

Автор SHA1 Сообщение Дата
Tim Prins
4033a40e4e Coding standards...
This commit was SVN r16118.
2007-09-13 14:00:59 +00:00
George Bosilca
617ff3a413 Add a MCA parameter for the ELAN MAP ID file.
Fix small memory bugs, and track the final segfault. Still some ork to do.

This commit was SVN r16117.
2007-09-12 21:25:35 +00:00
Aurelien Bouteiller
a1f5312afb Fixed two little warnings
This commit was SVN r16116.
2007-09-12 21:07:11 +00:00
Aurelien Bouteiller
ccb3f75e8f Make sure that the pml v parasite never get loaded when user did not requested FT. This does not break the ability to switch protocol on the fly.
This commit was SVN r16114.
2007-09-12 20:47:17 +00:00
George Bosilca
1e7a791349 Remove some of the problems identified by Coverty.
This commit was SVN r16112.
2007-09-12 20:13:26 +00:00
Aurelien Bouteiller
828af95be8 Major modification of the vprotocol framework build system. With a better integration in autogen.sh, it allows for generating static-components.h the usual way.
NOTE: This build system does not work with the current autogen.sh. Modified one is under heavy testing to make sure it does not have side effects 

This commit was SVN r16110.
2007-09-12 18:46:37 +00:00
George Bosilca
7b3dcff267 Coverty: Limit the strcpy to the maximum length of the destination.
This commit was SVN r16107.
2007-09-12 18:03:53 +00:00
George Bosilca
bfb4ddc3e2 Coverty: remove dead code.
This commit was SVN r16106.
2007-09-12 17:56:33 +00:00
George Bosilca
05ae27c68b Don't segfault if we receive a fragment for a non existing communicator.
Instead, drop it by now.

This commit was SVN r16105.
2007-09-12 17:52:02 +00:00
George Bosilca
c755938eb0 Coverty: release the temporary buffer on error.
This commit was SVN r16104.
2007-09-12 17:45:12 +00:00
George Bosilca
2b7ed6262b Update the communicator lowest_free when we rebuild the communicator list.
This commit was SVN r16102.
2007-09-12 16:41:14 +00:00
Shiqing Fan
b1ea3e0054 - add more lines for static import declaration on windows.
This commit was SVN r16101.
2007-09-12 15:32:54 +00:00
Shiqing Fan
a0660f4deb - Just some type casts.
This commit was SVN r16100.
2007-09-12 15:29:58 +00:00
Gleb Natapov
07c8fddeef Fix scheduling of pending send request. It should be scheduled req_lock times.
This commit was SVN r16096.
2007-09-12 07:08:38 +00:00
George Bosilca
d8fed2cfa1 Set a default value so that some compilers stop complaining about
uninitialized values.

This commit was SVN r16094.
2007-09-11 18:00:53 +00:00
George Bosilca
2e46809995 Only release the comm_reg is we have one.
This commit was SVN r16093.
2007-09-11 17:59:40 +00:00
Gleb Natapov
e82a6eec27 Restore check for lowest id. It prevents livelock situation if multiple threads
are inside the function and they failed to obtain new cid the first time around.

This commit was SVN r16090.
2007-09-11 15:32:46 +00:00
Gleb Natapov
58a018c16d The code tries to prevent itself from running for more then one communicator
simultaneously, but is doing it incorrectly. If the function is running already
for one communicator and it is called from another thread for other communicator
with lower cid the check comm->c_contextid != ompi_comm_lowest_cid()
will fail and the function will be executed for two different communicators by
two threads simultaneously. There is nothing in the algorithm that prevent it
from been running simultaneously for different communicators as far as I can see,
but ompi_comm_unregister_cid() assumes that it is always called for a communicator
with the lowest cid and this is not always the case. This patch removes bogus
lowest cid check and fix ompi_comm_register_cid() to properly remove cid from
the list.

This commit was SVN r16088.
2007-09-11 13:23:46 +00:00
George Bosilca
8659a864e9 This is the real fix for ticket 317 and ticket 1065 and ticket 278.
This commit was SVN r16084.
2007-09-10 22:27:59 +00:00
George Bosilca
8622beda54 This commit should fix the issues with ticket 1065. Now, we correctly
duplicate the MPI_UB and MPI_LB datatypes.

This commit was SVN r16083.
2007-09-10 22:13:42 +00:00
Pak Lui
e3fdfdbd9c Fix some typos here and there.
This commit was SVN r16058.
2007-09-06 14:56:08 +00:00
Tim Prins
f677ef5c12 Fix build failure on BigRed
This commit was SVN r16054.
2007-09-06 12:10:11 +00:00
Pak Lui
3d7b5b306f Fix a problem with OPAL_ALIGN that causes the upper bytes to get chopped
off and bogus addresses to show up for the requests, which in turns causes
message queues not showing up when debugging a 64 bit app on a 32 bit 
tvd and dll on only Solaris SPARC.

This commit was SVN r16052.
2007-09-05 23:52:36 +00:00
Pak Lui
99ae2c1c44 Nothing relevent (yet). Just making debugging more enjoyable.
This commit was SVN r16051.
2007-09-05 23:21:58 +00:00
Gleb Natapov
b0614931f4 Remove mpool_tree_item from the mpool_tree before unregistering/freeing memory.
Otherwise a race exists if another thread allocates already freed memory
which is not removed from the mpool_tree yet.

This commit was SVN r16038.
2007-09-03 10:56:55 +00:00
Rainer Keller
a3b30749b0 - Only lock/unlock when using threads.
Basically revert this part of r16015.

This commit was SVN r16029.

The following SVN revision numbers were found above:
  r16015 --> open-mpi/ompi@435e7d80e9
2007-08-31 12:34:48 +00:00
Rainer Keller
9c1c345c07 - head_lock is an opal_atomic_lock_t...
This commit was SVN r16028.
2007-08-31 12:20:21 +00:00
Pak Lui
bc34a46969 * complete the fix started in r15915, that is, to prevent negative tags
from showing up in the message queue graph. Tags are now casted to int
  before the negative checks, since tags by the spec are stored as 
  mqs_tword_t, an unsigned long long.

This commit was SVN r16027.

The following SVN revision numbers were found above:
  r15915 --> open-mpi/ompi@b9ea4c92e7
2007-08-31 03:02:24 +00:00
Shiqing Fan
b1250eba3a - Some more to be exported.
This commit was SVN r16023.
2007-08-30 15:13:08 +00:00
Shiqing Fan
efdcfa3807 - "extern 'C'" has been set twice. Remove one.
This commit was SVN r16022.
2007-08-30 15:03:59 +00:00
Shiqing Fan
80fdd5e2a4 - Need to be exported.
This commit was SVN r16021.
2007-08-30 14:16:03 +00:00
Gleb Natapov
79011279e5 Remove debug output.
This commit was SVN r16016.
2007-08-30 13:29:41 +00:00
Gleb Natapov
435e7d80e9 Remove rc parameter from MCA_BTL_SM_FIFO_WRITE() macro. It cannot fail in
current implementation.

This commit was SVN r16015.
2007-08-30 13:21:52 +00:00
Gleb Natapov
690fb95bda Cleanup send scheduling code.
This commit was SVN r16014.
2007-08-30 12:10:04 +00:00
Gleb Natapov
0b0f9d14aa Mark send request complete on PML level only when absolutely sure there is
no more work associated with this request. No more outstanding completions or
packets and send scheduling isn't running in another thread.

This commit was SVN r16013.
2007-08-30 12:08:33 +00:00
Gleb Natapov
fe414047bd registration may be freed inside mca_mpool_rdma_deregister().
This commit was SVN r16012.
2007-08-30 10:52:38 +00:00
Gleb Natapov
091862a25a Protect access to mca_mpool_base_tree by a lock.
This commit was SVN r16011.
2007-08-30 10:51:02 +00:00
Brian Barrett
f53b14bde5 George noted I had this logic completely backwards. Oops.
This commit was SVN r16005.
2007-08-29 16:18:04 +00:00
Gleb Natapov
eac2674f66 The inner voice tells me this is a typo.
This commit was SVN r16004.
2007-08-29 13:28:47 +00:00
George Bosilca
756eee571e Fix Coverty #24. This test didn't make sense in this branch of the if.
This commit was SVN r16001.
2007-08-29 02:02:19 +00:00
Jeff Squyres
466394a878 We only care about the value of ret in the
!OMPI_ENABLE_PROGRESS_THREADS case.  Reviewed by Brian.

This commit was SVN r16000.
2007-08-29 01:36:17 +00:00
Jeff Squyres
c4a38f47f6 Resolve Coverity CID 467: remove unused variable / dead code.
This commit was SVN r15997.
2007-08-29 01:23:18 +00:00
Jeff Squyres
f08cce16db Fix Coverity CID 468: remove unused variable.
This commit was SVN r15996.
2007-08-29 01:21:17 +00:00
Brian Barrett
59b22533f2 Enable RDMA for heterogeneous situations. Currently done by overloading
the ompi_convertor_need_buffers function to only return 0 if the convertor
is homogeneous (which it never does on the trunk, but does to on v1.2, but
that's a different issue).  Only enable the heterogeneous rdma code for
a btl if it supports it (via a flag), as some btls need some work for this
to work properly.  Currently only TCP and OpenIB extensively tested

This commit was SVN r15990.
2007-08-28 21:23:44 +00:00
Gleb Natapov
fa69c5cc10 If a memory on a sender's size is not registered don't register it on a receive
side too. Otherwise a content of the recvreq->req_rdma array is replaced later
without freeing previous content and refcount on registration in mpool become
wrong.

This commit was SVN r15978.
2007-08-28 07:43:06 +00:00
Rich Graham
bc97d22182 remove tabs. Remove old code that was commented out.
This commit was SVN r15975.
2007-08-28 03:08:36 +00:00
Rich Graham
4d58f9aed7 Add comments. Move temporary receive object from a free list object to
a stack object.

This commit was SVN r15971.
2007-08-27 21:41:04 +00:00
Pak Lui
75c7d4e03b Temporary workaround for making Totalview be able to get those opal
symbols and load into the library when compiled with a Sun Studio C compiler

This commit was SVN r15970.
2007-08-27 19:04:56 +00:00
Gleb Natapov
e1a1d9d90e Receive request converter can be accessed in parallel by a thread that receives
data and a thread that run RDMA schedule function. Protect access to the
converter by a lock.

This commit was SVN r15967.
2007-08-27 11:41:42 +00:00
Gleb Natapov
33196d972b post_send() function is called without endpoint lock held from explicit credits
update function so eager_rdma_remote.head have to be updated in a thread safe
manner.

This commit was SVN r15966.
2007-08-27 11:37:01 +00:00
Gleb Natapov
32a61c3bf2 Credit fragment is not protected properly from concurrent access. There is a
race that can prevent further explicit credits update from been sent. Fix the
race.

This commit was SVN r15965.
2007-08-27 11:34:59 +00:00
Gleb Natapov
065d04dfde Do not free recvreq while schedule function is running in another thread.
This commit was SVN r15964.
2007-08-27 11:31:40 +00:00
Brad Benton
ccda5c9c74 Modified the MCA_BTL_TCP_CONNECTED case in mca_btl_tcp_endpoint_send_handler()
to always first check for a NULL frag pointer before trying to send the
fragment.  This avoids an issue in multi-threaded execution in which 
multiple threads working on the same endpoint can result in a thread 
finding itself here with nothing to send.

This commit was SVN r15963.
2007-08-26 23:40:02 +00:00
George Bosilca
475073c684 Be as user friendly as possible and provide more information. Now we make the
difference between the user specified length, and the one available from
Open MPI (this allow to se the truncated receives). Moreover, if the
data-type used is named we now print the count as well as the name of
the used data-type.

This commit was SVN r15962.
2007-08-26 23:07:14 +00:00
Jeff Squyres
18db56e270 Fix Coverity defect 675: possible NULL dereference in an error
condition.

This commit was SVN r15957.
2007-08-25 12:18:55 +00:00
Jeff Squyres
b69c7688a0 Fix Coverity defect 676: possible NULL dereference in an error
condition.

This commit was SVN r15956.
2007-08-25 12:17:02 +00:00
George Bosilca
a6723b34ea Cleanup the code. Remove all debugging messages.
This commit was SVN r15955.
2007-08-24 02:58:09 +00:00
Edgar Gabriel
a2f5cada1a convert the hiearch component to the new structure. More testing required before we remove the .ompi_ignore flag again.
This commit was SVN r15954.
2007-08-23 20:41:29 +00:00
George Bosilca
daaf5a9bf1 Correct the assert macro.
This commit was SVN r15953.
2007-08-23 19:48:04 +00:00
George Bosilca
db19f927e8 A lot of cleanups. Verbose is enabled right now as we're tracking down
an issue with the ompi_communicator_t structure.

This commit was SVN r15951.
2007-08-23 16:40:07 +00:00
Rainer Keller
b385f8a790 - ompi_comm_set(): PML add_comm may return something != OMPI_SUCCESS
Use OMPI_SUCCESS throughout. 
 - ompi_comm_allocate(): Initialize new_comm=NULL to get rid of
   warnings.

This commit was SVN r15948.
2007-08-23 07:40:40 +00:00
Rainer Keller
1b5fa48a29 - Add missing PERUSE_COMM_REQ_REMOVE_FROM_POSTED_Q when matching
from the posted generic_recv-queue.
 - Move the PERUSE_COMM_MSG_MATCH_POSTED_REQ from
   MCA_PML_OB1_RECV_REQUEST_MATCHED to
   mca_pml_ob1_recv_frag_match() as suggested by Terry Dontje
   Only post, if this is not a probe/iprobe request.
 - Do not post PERUSE_COMM_REQ_MATCH_UNEX for probes / iprobes and
   do in correct order before PERUSE_COMM_MSG_REMOVE_FROM_UNEX_Q

This commit was SVN r15947.
2007-08-23 07:09:43 +00:00
Rainer Keller
c175801f98 - Initialize in the order of mca_pml_ob1_comm_proc_t...
This commit was SVN r15946.
2007-08-23 05:56:22 +00:00
Rainer Keller
b0df55d53b - For MPI_Probe/MPI_Iprobe, we should not have a
PERUSE_COMM_REQ_ACTIVATE event.
   Therefore move the PERUSE_TRACE_COMM_EVENT for this event from
   MCA_PML_BASE_SEND_REQUEST_INIT / MCA_PML_BASE_RECV_REQUEST_INIT
   to the proper places into pml_ob1_isend.c / pml_ob1_irecv.c right
   after the MCA_PML_OB1_SEND_REQUEST_INIT /
   MCA_PML_OB1_RECV_REQUEST_INIT.

This commit was SVN r15945.
2007-08-23 05:52:33 +00:00
George Bosilca
b5af2ba6f2 Correctly retrieve the MPI_SOURCE field for receives.
This commit was SVN r15944.
2007-08-22 22:35:30 +00:00
Gleb Natapov
becf4aa9c9 ompi_pointer_array_get_size doesn't return how much elements are actually in an
array, so count them by ourselves.

This commit was SVN r15943.
2007-08-22 09:31:12 +00:00
Shiqing Fan
a497a3fcad - Fix some small bugs, copy-paste mistakes.
This commit was SVN r15941.
2007-08-21 19:57:28 +00:00
Sven Stork
3985a35c35 - export required symbol
This commit was SVN r15939.
2007-08-21 18:46:11 +00:00
Rainer Keller
08634e7e4a - Properly unlock the ompi_proc_lock in case of errors and otherwise.
This commit was SVN r15936.
2007-08-20 16:47:13 +00:00
Jeff Squyres
ad784a9ab0 Make "simultaneous" be a size_t; there's already a check to ensure
that it is >= 1, so making it a size_t makes it easier to interact
with all the other size_t variables and removes a compiler warning.

This commit was SVN r15935.
2007-08-20 13:22:46 +00:00
Gleb Natapov
d8f3063895 Create only one CQ for all BTLs on the same HCA. Many BTLs can be created for
one HCA. Multiple ports, LMC, multiple BTLs per one LID. Having only one CQ for
all of them substantially reduce polling time.

This commit was SVN r15933.
2007-08-20 12:28:25 +00:00
Gleb Natapov
5596aa5f53 The sizes of mca_pml_ob1_send_request_t and mca_pml_ob1_recv_request_t depend
on a parameter and are determined in runtime. r15346 removed calculation of
correct sizes for this structures. This patch adds it back and fixes trac:1116, #1114.

This commit was SVN r15932.

The following SVN revision numbers were found above:
  r15346 --> open-mpi/ompi@433f8a7694

The following Trac tickets were found above:
  Ticket 1116 --> https://svn.open-mpi.org/trac/ompi/ticket/1116
2007-08-20 12:06:27 +00:00
George Bosilca
7e8bd529dc A better management of the receive requests. If the request was matched
then we can report the real source (in both local and global rank) as
well as the real amount of data transfered.

This commit was SVN r15929.
2007-08-19 20:16:48 +00:00
George Bosilca
7ef49614fc Correctly read a bool from the application memory. Report back if the
receive request was matched or not.

This commit was SVN r15928.
2007-08-19 20:05:09 +00:00
George Bosilca
c7e0ab93ae Don't forget to include string.h for the strcmp function.
This commit was SVN r15927.
2007-08-19 19:59:15 +00:00
George Bosilca
5ac19c62a1 Translate between global and local ranks. A global rank is the rank
of the process in the MPI_COMM_WORLD, while a local rank is the rank
of the process in the communicator where the request was posted. In
order to get the message graph nicely, each request has to have the
global rank set correctly.

This commit was SVN r15926.
2007-08-19 19:50:26 +00:00
Brian Barrett
af4e86c25f Update collectives selection logic to allow for multiple components to be
used at nce (up to one unique collective module per collective function).
Matches r15795:15921 of the tmp/bwb-coll-select branch

This commit was SVN r15924.

The following SVN revisions from the original message are invalid or
inconsistent and therefore were not cross-referenced:
  r15795
  r15921
2007-08-19 03:37:49 +00:00
Brian Barrett
2b8af283de Add ability to completely turn off MPI one-sided support, so that users
can experiment with using ROMIO directly.

This commit was SVN r15922.
2007-08-18 21:35:51 +00:00
Josh Hursey
729c63cf9d Fix invalid MCA 'base' names so they appear in ompi_info.
A subset of this patch needs to be applied to v1.2

Refs trac:928

This commit was SVN r15918.

The following Trac tickets were found above:
  Ticket 928 --> https://svn.open-mpi.org/trac/ompi/ticket/928
2007-08-18 03:05:45 +00:00
Rolf vandeVaart
797078115d Fix the case where mpi_preconnect_oob=1 and
mpi_preconnect_oob_simultaneous > np.  Need to scale back
simultaneous to equal np in those cases.  Reviewed by Brian.

This commit fixes trac:1064.

This commit was SVN r15916.

The following Trac tickets were found above:
  Ticket 1064 --> https://svn.open-mpi.org/trac/ompi/ticket/1064
2007-08-17 20:18:42 +00:00
George Bosilca
b9ea4c92e7 Don't show the requests having a negative tag (they are internal
requests for Open MPI). Add a variable to allow Open MPI developers
to see all internal messages.

This commit was SVN r15915.
2007-08-17 18:59:57 +00:00
Edgar Gabriel
0684002812 fixes: 1127
fix some of the multi-threading problems for the cid allocation. Two bugs
specifically:
 - since we do not have a queue for incoming fragments of unknown cid, we need
 to synchronize all processes before exiting the communicator creation. This
 synchronization was/is located in comm_activate, which was however too late
 for the multi-threaded case. Thus, for multi-threaded scenarios we are now
 synchronizing 'before' we allow another thread to enter the cid-allocation
 loop.

- for synchronization, we used for the sake of simplicity allreduce
  operations. It turns out, that these operations interefered with the
 allreductions in the cid-allocation routine, which lead to non-sense results
  in the cid-allocation and potentially to endless loops.

Multi-threaded communicator creation seems to work now, is however still 'very
very' slow. I think, the busy wait of threads is killing the performance of
the active threads in the cid allocation. But this is another topic.

This commit was SVN r15910.
2007-08-17 16:15:26 +00:00
Brian Barrett
3b98b5f0a1 The reference implementation of Portals (which runs over TCP on Linux) is
only static libraries.  Previously, we were linking the libraries into 
directly into the common, btl, and mtl code.  This seemed to work fine
for me on my Opteron Fedora box, but caused Lisa some issues (PtlNIInit
would succeed, but the network handle would fail when used with
PtlEQAlloc).

Instead, link the portals libraries directly into libmpi and not at
all into the common, btl, or mtl components.  THen use some linker
tricks to force the linker to bring in the public interface for the
reference implementation (which thankfully is pretty small).

This commit was SVN r15902.
2007-08-17 03:56:49 +00:00
Brian Barrett
c9e3654a85 Allow OMPI components to modify the link options for libmpi.so. This
functionality used to exist, but I removed it like a year ago because
it wasn't being used.  Well, now I need it :).

This commit was SVN r15901.
2007-08-17 03:52:53 +00:00
George Bosilca
b1082c95ff Remove all output. This final commit should solve all [hopefully]
problems with the integration with parallel debuggers.

This commit was SVN r15898.
2007-08-17 02:35:14 +00:00
George Bosilca
4dacd163cc Don't grab the information from the MPI_Status for the send requests.
This commit was SVN r15897.
2007-08-17 02:31:50 +00:00
George Bosilca
2086b7b445 Optimize the group creation. Don't create a new group if there is
already one containing the same nodes (useful for MPI_Comm_dup).

This commit was SVN r15896.
2007-08-17 02:19:34 +00:00
George Bosilca
7efffdb1da Be smart about parsing the communicators list. Based on the values of
lowest_free and number_free detect if the communicator list has changed.
If not, there is no reason to rebuild it, just use the old one.

This commit was SVN r15895.
2007-08-16 22:51:55 +00:00
George Bosilca
9cce0eb4bc Deal with int/size_t differences. This introduce some small problems
with the reported length of the receive requests, but I'll fix
it soon.

This commit was SVN r15893.
2007-08-16 21:02:24 +00:00
Brad Benton
c254645383 Fixes trac:1134.
Fixed a condition test while checking that all segments are empty.
Without this fix, a NULL segment pointer could make it past the
test, resulting in a SegV when dereferenced.

This commit was SVN r15891.

The following Trac tickets were found above:
  Ticket 1134 --> https://svn.open-mpi.org/trac/ompi/ticket/1134
2007-08-16 19:39:52 +00:00
Brad Benton
1ddba9ec65 Lock the endpoint before doing endpoint_state processing. This ensures
that the subsequent unlock is valid.

This commit was SVN r15890.
2007-08-16 18:11:29 +00:00
Aurelien Bouteiller
3a83c61c40 Fixed a bug with available space in sender based.
This commit was SVN r15889.
2007-08-16 17:54:26 +00:00
George Bosilca
51e726ee8c Remove some old [and unused] code.
This commit was SVN r15887.
2007-08-16 17:06:17 +00:00
Mohamad Chaarawi
b18129c260 Fix to the invalid process lookup cause by group_intersection
This commit was SVN r15886.
2007-08-16 17:01:01 +00:00
George Bosilca
1ae49b7143 Don't use a fprintf. Instead a plain print will do the job.
This commit was SVN r15883.
2007-08-16 14:55:44 +00:00
Tim Prins
5a795128af Change it so that different components in orte use unique rml tags
This commit was SVN r15881.
2007-08-16 14:02:35 +00:00
Jeff Squyres
4b7a6cd922 Fix the following compile error:
{{{
ompi/debuggers/ompi_dll.c:102: error: initializer element is not
constant
}}}

The fix is stupid and I suspect that we'll want to ''not'' print out
all this debugging information all the time.  But I'll leave that to
George to fix...  :-)

This commit was SVN r15880.
2007-08-16 11:51:06 +00:00
George Bosilca
33a73a88c6 Add a lot more information about the requests (pending, matched,
completed). Correctly detect the tag is a receive was matched.

This commit was SVN r15879.
2007-08-16 07:11:56 +00:00
Aurelien Bouteiller
77565d60d9 Heavy modification of the pml_v framework.
* Code cleanup and rationalization
* Fixed: mca_pml_base_send/recv_request are now allocated before recreation by the PML-V
* Fixed: pointer arithmetic bug in sender based that crashed 
* Changed: directory structure. This is one step forward using autogen.sh to build static-components.h (it needs to have the directory structure of a mca framework for this). 

This commit was SVN r15878.
2007-08-16 05:52:30 +00:00
Aurelien Bouteiller
ee708d702d Slight modification to register the name of the selected pml (from the pml framework) instead of the generic mca name. This might be a different name when enabling FT features. This name modification in the modex allows the PMLS to detect a FT protocol mismatch among hosts.
This commit was SVN r15877.
2007-08-16 05:46:11 +00:00
George Bosilca
de4813359a The message queue is now back online. It heavely depend on the
opal_list_t, ompi_free_list_t so every time there is a modification
in one of these files (such as changing the way we allocate the
elements in the free list) the debugger interface have to
reflect these changes.

This commit was SVN r15876.
2007-08-16 04:33:04 +00:00
Aurelien Bouteiller
fa7f6f6722 Improved error detection of request types
This commit was SVN r15857.
2007-08-14 17:24:46 +00:00
Aurelien Bouteiller
67399e7c31 Added a debug type checking for request types (to make sure request size is correctly computed).
This commit was SVN r15856.
2007-08-14 17:18:15 +00:00
Aurelien Bouteiller
1d97c183e7 Better argument checking for output function and added a routine for error printing.
This commit was SVN r15855.
2007-08-14 17:17:12 +00:00
Jeff Squyres
d7c5fea096 * Fix problem caused by r15848: the test parser was looking for
semicolons but the new specitifcation string used colons.  The text
   parser now looks for colons.
 * Changed all opal_output() error messages to
   much-more-helpful/descriptive opal_show_help() messages.
 * A few minor style/indenting fixes

This commit was SVN r15850.

The following SVN revision numbers were found above:
  r15848 --> open-mpi/ompi@dd30597f39
2007-08-14 14:46:13 +00:00
Jeff Squyres
dd30597f39 Change the default receive_queues value per
http://www.open-mpi.org/community/lists/devel/2007/08/2100.php.

This commit was SVN r15848.
2007-08-13 21:51:05 +00:00
Jelena Pjesivac-Grbovic
9bd9c92dbd Making sure that the decision function for scatter and gather correctly
computes everything for MPI_IN_PLACE case.

This commit was SVN r15841.
2007-08-13 17:35:50 +00:00
Brian Barrett
d166a2bb6d Change requested by Ralph -- Remove the dependency on GPR triggers for filling
in the OMPI proc structures.  For now, use an extension of the modex that is
keyed on strings.  Eventually, this should use the attribute put/get that is
part of the RSL interface.

This commit was SVN r15820.
2007-08-09 18:53:28 +00:00
Shiqing Fan
f84e919dc2 - one more to fix.
This commit was SVN r15814.
2007-08-09 13:53:10 +00:00
Shiqing Fan
7dc5dbd8ea - fix a few function export declarations.
This commit was SVN r15813.
2007-08-09 13:52:23 +00:00
Jelena Pjesivac-Grbovic
b558e820cb removing compiler wraning
This commit was SVN r15803.
2007-08-08 15:22:01 +00:00
Jelena Pjesivac-Grbovic
daa10b277e modifying scatter decision function to use binomial algorithm for
small message sizes.

This commit was SVN r15798.
2007-08-07 22:16:13 +00:00
Mohamad Chaarawi
8c458b0ee7 removing unused variables, that cause warnings..
This commit was SVN r15791.
2007-08-07 15:13:46 +00:00
Pak Lui
0790c4cc40 * Update the comment for the previous fix. Thanks Gleb for pointing out.
This commit was SVN r15790.
2007-08-07 14:40:13 +00:00
Jeff Squyres
50bae9c603 Bring in the modular-wireup stuff for the openib BTL (from
/tmp/jms-modular-wireup branch):

 * This commit moves all the openib BTL connection code out of
   btl_openib_endpoint.c and into a connect "pseudo-component" area,
   meaning that different schemes for doing OFA connection schemes can
   be chosen via function pointer (i.e., MCA parameter) at run-time.
 * The connect/connect.h file includes comments describing the
   specific interface for the connect pseudo-component.
 * Two pseudo-components are in this commit (more can certainly be
   added).
   * oob: use the same old oob/rml scheme for creating OFA connections
     that we've had forever; this now just puts the logic into this
     self-contained pseudo-component.
   * rdma_cm: a currently-empty set of functions (that currently
     return NOT_IMPLEMENTED) that will someday use the RDMA connection
     manager to make OFA connections.

This commit was SVN r15786.
2007-08-06 23:40:35 +00:00
Mohamad Chaarawi
96e132b11d typo in comment of the group struct...
This commit was SVN r15785.
2007-08-06 23:09:37 +00:00
Aurelien Bouteiller
ca69915b1e Code cleanup
This commit was SVN r15783.
2007-08-06 22:20:44 +00:00
Brian Barrett
69952d9603 Fix abort caused by calling PtlEQGet on an invalid eq, which could occur
if add_procs was never called.

This commit was SVN r15779.
2007-08-06 17:28:11 +00:00
Brian Barrett
1fb78a35f9 Back out part of r15756. The common_portals_utcp.c file is only used with
the Sandia reference implementation of Portals, and doesn't have the cnos
functions.  This file should never be compiled (and wasn't being compiled)
on the Cray machines, so doesn't need to be updated to support CNL.

This commit was SVN r15778.

The following SVN revision numbers were found above:
  r15756 --> open-mpi/ompi@755658694e
2007-08-06 17:21:00 +00:00
Sven Stork
9e2263f29f - fix a small memory leak
This commit was SVN r15768.
2007-08-06 13:35:32 +00:00
Mohamad Chaarawi
59a7bf8a9f Merging in the Sparse Groups..
This commit includes config changes..

This commit was SVN r15764.
2007-08-04 00:41:26 +00:00
George Bosilca
8baeadb761 The PTLs are now long gone ...
This commit was SVN r15763.
2007-08-04 00:37:52 +00:00
George Bosilca
78e2d3523b Remove some old and unused code. Update some of the comments.
This commit was SVN r15761.
2007-08-04 00:34:42 +00:00
George Bosilca
e41ee17ca5 Add a small comment that hopefully will enforce the correct ordering of
the fields between CM and the other PML in the requests structure.

This commit was SVN r15760.
2007-08-03 23:59:29 +00:00
Josh Hursey
755658694e Bring in changes to support Cray's Compute Node Linux (CNL) and
Application Level Placement Scheduler (ALPS).

This commit was tested under two Cray machines at ORNL: Jaguar (Catamount)
and Rizzo (CNL Test cage). Both machines performed as they should across
the commit.

It is likely that mor changes will follow this the work and environment
stabilizes.

Most of the infrastructure works the same for Catamount and CNL
except for a few bits. Below are the highlights:

Default IFACE Change:
 On Catamount we can use PTL_IFACE_DEFAULT, but on the CNL system we have access
 to will fail on this interface, and should be set to:
    IFACE_FROM_BRIDGE_AND_NALID(PTL_BRIDGE_UK,PTL_IFACE_SS).
 So if we detect that we are running with YOD then use the former interface
 and if we detect that we are running with ALPS then use the latter.
 We will want to pursue a more elegant solution if this interface continues to 
 change across machines.

PtlGetId and cnos_register_ptlid:
 The header suggests that these should never be called when launching with YOD.
 But in the ALPS environment the cnos_barrier() will hang forever if these 
 functions are not called after PtlNIInit(). Since these functions only need to
 be called once, and the orte rmgr/cnos component is loaded before the ompi 
 common/portals componet then just call these functions once in the rmgr/cnos
 component.

cnos_barrier_init():
 This is a noop for YOD, but critical for ALPS. So be sure to call it before
 calling the first barrier in the rmgr/cnos component.

cnos_barrier vs cnos_pm_barrier:
 It is suggested the cnos_pm_barrier only be used during finalization 
 as it will indicate to the launcher (yod or aprun) that the app is about
 to complete. It was suggested that we use the regular cnos_barrier() instead.
 I want to look into this a bit more to make sure there are not adverse
 side effects. A note has been placed in the code to indicate this reasoning.

This commit was SVN r15756.
2007-08-03 19:46:38 +00:00
Pak Lui
010d216db9 * restrict the user with 32 bit app to specify a sm_size to be between
2GB to 4GB-1 by using long instead of size_t for the sm size.
   * it is done to prevent user from running into the ftruncate() in 
	 common sm component (and possibly others) problem that ftruncate 
	 takes an off_t which is a signed long integer. If we use an 
	 unsigned long, it'll run into an invalid argument errno=22.
   * See trac #1117

This commit was SVN r15752.
2007-08-03 15:43:02 +00:00
Aurelien Bouteiller
1d160ca583 Needed change for vampir pml to work
This commit was SVN r15750.
2007-08-03 02:23:24 +00:00
Jeff Squyres
d3f008492f Introduce a new debugging MCA parameter:
mpi_show_mpi_alloc_mem_leaks

When activated, MPI_FINALIZE displays a list of memory allocations
from MPI_ALLOC_MEM that were not freed by MPI_FREE_MEM (in each MPI
process).

 * If set to a positive integer, display only that many leaks.
 * If set to a negative integer, display all leaks.
 * If set to 0, do not show any leaks.

This commit was SVN r15736.
2007-08-01 21:33:25 +00:00
Jeff Squyres
0fb8cf65a8 If you have an HCA with no active ports, we still create an mpool.
This mpool will have no btl module owner there was no btl created for
the HCA with no ports, but it will still be tracked in the mpool
framework (i.e., it's available).

If MPI_ALLOC_MEM is called by the app, one of two things will happen:

 1. if there's an HCA on the host with some active ports, the openib
    btl component will still be in the process space, and therefore
    the "mpool with no btl" (MWNB) module will still be able to call
    the reg/dereg functions, and all will be fine.  However, if
    MPI_FREE_MEM is never invoked to free the memory, bad things will
    happen during MPI_FINALIZE.  The pml is finalized, which finalizes
    all the btls.  The btls finalize all their mpools and all is fine.
    But later we close down the mpool framework which then finalizes
    any left over mpool modules, such as MWNB.  However, the openib
    BTL module functions that the MWNB was registered with are no
    longer in the process space, and it segv's while trying deregister
    the memory.
 2. if there are *no* HCA's on the host with active ports, then the
    openib btl will have been unloaded, and when the MWNM tries to
    register the memory, the functions it tries to call (in the openib
    btl) are no longer there, and we segv.

This commit was SVN r15735.
2007-08-01 20:53:34 +00:00
Gleb Natapov
627d9bc8ed Delay freeing of a send request if scheduling function is running by other
thread.

This commit was SVN r15722.
2007-08-01 12:19:16 +00:00
Gleb Natapov
758f932aa6 Handle credit in a thread safe manner. I am sure more work will have to be done
in this are.

This commit was SVN r15721.
2007-08-01 12:15:43 +00:00
Gleb Natapov
9c20d67301 1) Return IB header to it's previous size by using char for cm_seen field.
2) Allow to specify rd_win/rd_rsv parameters by user, but make them optional.

This commit was SVN r15719.
2007-08-01 12:10:56 +00:00
Aurelien Bouteiller
a403fed18a More checkings (assert) on the output system so that malformed format string does not crash the application at a later random time.
Changed various debug messages to retain most usefull messages

This commit was SVN r15715.
2007-07-31 19:33:39 +00:00
Aurelien Bouteiller
cec9ce8106 Fixed: various warnings with printf(%x, uint64_t) on 32 bit architectures + some left (long) cast for size_t printf.
This commit was SVN r15706.
2007-07-31 17:12:21 +00:00
Shiqing Fan
0f468f3668 - Remove the solution and project files, will commit them later.
This commit was SVN r15705.
2007-07-31 17:07:02 +00:00
Aurelien Bouteiller
a5d0e53bb3 Moved replay macros to functions. The performance improvement in process recovery does not worth the debugging hassle.
This commit was SVN r15703.
2007-07-31 16:01:32 +00:00
Aurelien Bouteiller
5a792a3fad (hopefully) fixed various pedantic warning about casts on 32bit machines. Not tried only have 64bits available.
This commit was SVN r15702.
2007-07-31 15:58:19 +00:00
Sven Stork
fd778a5539 - put the label to the right place
This commit was SVN r15699.
2007-07-31 09:34:41 +00:00
Sven Stork
a13d2dcb96 - fix possible memory leak found by coverity
This commit was SVN r15698.
2007-07-31 09:32:49 +00:00
Sven Stork
27422e05ac - add parameter check for NULL pointer
This commit was SVN r15697.
2007-07-31 09:01:39 +00:00
George Bosilca
d6a676b29e Remove unused variable.
This commit was SVN r15693.
2007-07-30 19:38:26 +00:00
George Bosilca
cf9bccf2e6 prefer snprintf to sprintf.
This commit was SVN r15692.
2007-07-30 19:37:34 +00:00
Aurelien Bouteiller
3559fd5d1a Fixed issues with "verbose" output being too silent.
This commit was SVN r15691.
2007-07-30 19:11:15 +00:00
Sven Stork
4c5836c2ee - add missing va_end found by coverity
This commit was SVN r15689.
2007-07-30 16:08:18 +00:00
Sven Stork
d830318bbb - fix typo
This commit was SVN r15687.
2007-07-30 15:47:37 +00:00
Sven Stork
6c8d921a76 - coverity found dead code, but it's a typo
This commit was SVN r15686.
2007-07-30 15:41:41 +00:00
Sven Stork
80cdafb8f4 - remove dead code found by coverity
This commit was SVN r15685.
2007-07-30 15:36:00 +00:00
Sven Stork
855434de59 - fixes several coverty issues
- add missing initialisation for variables
  - use strncpy instead of strcpy

This commit was SVN r15683.
2007-07-30 14:44:37 +00:00
Jeff Squyres
327576b2a3 Fix incorrect behavior noted by Lisandro Dalcini: when MPI_COMM_SELF is
passed to MPI_COMM_FREE, it invoked the error handler on
MPI_COMM_WORLD, not on MPI_COMM_FREE.  This commit changes the
behavior: if MPI_COMM_SELF is passed to MPI_COMM_FREE, we invoke the
error handler on MPI_COMM_SELF (not MPI_COMM_WORLD).  Fixes trac:1109.

This commit was SVN r15682.

The following Trac tickets were found above:
  Ticket 1109 --> https://svn.open-mpi.org/trac/ompi/ticket/1109
2007-07-30 13:01:33 +00:00
Gleb Natapov
afac5eb93f Guard recv request with lock against simultaneous access from different
threads.

This commit was SVN r15681.
2007-07-30 12:50:38 +00:00