1
1
Граф коммитов

6317 Коммитов

Автор SHA1 Сообщение Дата
Brian Barrett
878676218e Rename opal/memory to opal/memoryhooks because XLC++ on Mac OS X is broken.
When compiling C++ code that includes something that looks for the C++
header file "memory" (stupid C++ headers not having .h extensions), it
goes through the header file search path, which includes $(topsrcdir)/opal,
so it finds the directory $(topsrcdir)/opal/memory/ and tries to load
that as the memory header file and all goes downhill.

This commit was SVN r8111.
2005-11-11 00:26:27 +00:00
Brian Barrett
660d2f61b6 Don't add external declarations for the PMPI_W{TICK,TIME} functions
if profiling isn't enabled.  It appers that some compilers (g95)
will try to resolve the symbols if they are prototyped.

This commit was SVN r8110.
2005-11-11 00:12:40 +00:00
Josh Hursey
5fa34df9ce Fix for orted / MPI_Abort problem reported from testers. They were seeing orteds
spining in orte_iof_base_flush() when running 
  intel_tests/src/MPI_Errhandler_fatal_c

When we close an endpoint by taking it out of the envent handler, we need to make
sure that it fits the criteria to pass through orte_iof_base_flush(), specificly
make sure we clean out the ep_frags list.
Note: This is more of a sanity check, since the endpoint should already be
      in this state at the point of closure.

Secondly in orte_iof_base_endpoint_read_handler(), if we determine that it is 
necessary to close the endpoint we have to "return" after doing so, otherwise
we add another frag to the endpoint which will cause it to hang in 
orte_iof_base_flush().

Bug go squish!

This commit was SVN r8109.
2005-11-11 00:09:07 +00:00
Jeff Squyres
bcd037315f Some Fortran compilers actually will return that a type exists even if
it doesn't support it -- the compiler will automatically convert the
unsupported type to a type that it *does* support.  For example, if
you try to use INTEGER*16 and the compiler doesn't support it, it may
well automatically convert it to INTEGER*8 for you (!).  So we have to
check the actual size of the type once we determine that the compiler
doesn't error if we try to use it (i.e,. the compiler *might* support
that type).  If the size doesn't match the expected size, then the
compiler doesn't really support it.

The F77 configure code actually handled this properly.  The F90 code
did not quite do it right.  This patch brings the F90 code up to the
same structure as the F77 code, albiet not m4-ized properly.  I also
added a comment to config/f77_check.m4 that explains *why* we do this
extra size check (because no explanation was given).

The impetus for this was that xlf* on OS X 10.3 was not recognizing
that INTEGER*16 was not supported, and mpi-f90-interfaces.h was being
assembled incorrectly.  This patch fixes this problem.

There is still one more problem, but waiting for some help from Craig
R on that (function pointers in F90 declarations).

This commit was SVN r8107.
2005-11-10 23:35:36 +00:00
Tim Woodall
654ba6d262 srq cleanup
This commit was SVN r8106.
2005-11-10 23:29:54 +00:00
Tim Woodall
2013104d1a SRQ cleanup
This commit was SVN r8104.
2005-11-10 20:51:56 +00:00
Tim Woodall
4a06e8463c port of flow control from mvapi
This commit was SVN r8102.
2005-11-10 20:15:02 +00:00
Tim Woodall
7f20198d49 Filter the set of data returned to the daemons during
startup using the new get_conditional command to improve
scalability during launch

This commit was SVN r8097.
2005-11-10 16:44:51 +00:00
Jeff Squyres
bacfb4fa2b Remove the generated F90 interfaces for all the "2 buffer" MPI API
functions (e.g., MPI_REDUCE).  We don't generate the back-end
subroutines for them (because it makes an expontential number of
subroutines, and compilers literally will segv), so we shouldn't
generate the f90 interfaces for them, either.  This allows user's MPI
F90 apps to automaitcally fall through to the F77 bindings for these
functions.

This commit was SVN r8094.
2005-11-10 16:04:39 +00:00
Tim Woodall
985c2ca943 cleanup
This commit was SVN r8093.
2005-11-10 15:40:27 +00:00
Tim Woodall
d62ea1835d correct typo
This commit was SVN r8090.
2005-11-10 15:29:52 +00:00
Brian Barrett
86e2adc43a * it appears that including event.h without calling opal_init annoys XLC on
OS X (you get an undefined symbol opal_event_lock).  Since the code is
  all #if 0'ed out, #if 0 out the header for now as well.

  I believe console and openmpi are to be removed from OMPI before 1.0
  release, so this doesn't need to go to the 1.0 branch

This commit was SVN r8089.
2005-11-10 15:24:57 +00:00
Jeff Squyres
6e08072113 Fix for the interface name for MPI_File_write_ordered_begin -- the
name was changed to shorten it too early (and then not restored), so
the "interface" name was not output correctly into
mpi-f90-interfaces.h.  Change to make it like the other long functions
-- temporarily change it to a shorter name while outputing the
subroutines, and then revert it when outputting the end interface.

This commit was SVN r8086.
2005-11-10 14:10:20 +00:00
Tim Woodall
3556757726 init callback from proxy
This commit was SVN r8085.
2005-11-10 05:27:11 +00:00
Tim Woodall
0b0d7f56c1 added support for callback on receipt of I/O
This commit was SVN r8084.
2005-11-10 04:49:51 +00:00
Tim Woodall
3699c924bd callback for init prior to launch - allow app to hookup stdout/stderr
prior to launch

This commit was SVN r8083.
2005-11-10 04:47:41 +00:00
Brian Barrett
d542e14120 * for some reason, some versions of linux didn't like that call to mmap. The
test isn't really needed to make sure the malloc intercept code was working,
  so just revert to not checking for now.  
  
  This should go to the 1.0 branch, as what is there now is causing issues.

This commit was SVN r8081.
2005-11-10 04:13:56 +00:00
George Bosilca
405d9794f8 Somehow I miss to remove one of the previous definition for the unavailable data.
This commit was SVN r8080.
2005-11-10 02:59:20 +00:00
Brian Barrett
5bf0a7bc62 * allow for the fact that svnversion might fail
This commit was SVN r8077.
2005-11-10 02:00:38 +00:00
George Bosilca
3507d5e9cd opal/util/output.h is required for optimized builds.
This commit was SVN r8076.
2005-11-10 01:19:27 +00:00
George Bosilca
8119c970db Improve the connection algorithm for MX. There are 2 problems here:
- first we setup the connections in the begining with all the peers
- MX does not handle well the case where several peers make connections to the same
  destination simultaneously.

So I change the order in which we connect. First we compute our rank in the array,
then in a round-robin fashion we setup connection starting with our left neighboard.

This commit was SVN r8075.
2005-11-10 01:15:49 +00:00
George Bosilca
dc1ad885d1 Move the output message outside the loop. We print an error message only once when we fail to
connect to a peer. Bonus, we print some additional informations like its MAC Address or name
if it's on our tables.

This commit was SVN r8074.
2005-11-10 01:13:18 +00:00
Tim Woodall
4c7c277b0a improve the scalability of MPI_Waitall ...
note that any code that sets a request to a completed state must
now increment a counter for every completed request

This commit was SVN r8073.
2005-11-10 00:45:27 +00:00
Tim Woodall
62fd74140b decrease socket buffers sizes to same as ptl code
This commit was SVN r8072.
2005-11-10 00:40:55 +00:00
Tim Woodall
2f6d50e0c6 init rdma count
This commit was SVN r8071.
2005-11-10 00:04:25 +00:00
Tim Woodall
b5ed723ea4 - check for null return
- disable debug

This commit was SVN r8070.
2005-11-10 00:02:18 +00:00
George Bosilca
55051b81c4 Activate the protection against unavailable datatypes. They get a flag DT_FLAG_UNAVAILABLE. We check now this flag in all the send/recv operations via the macros on mpi/c/bindings.h.
This flag is inherited by all datatypes create with unavailable datatypes. Basically, we let the user create the wrong datatype but we dont let him using it for any pt2pt communications or any pack/unpack.

This commit was SVN r8069.
2005-11-09 23:43:41 +00:00
Brian Barrett
5e6ab09424 apparently there are some platforms that do not like having munmap(NULL, 0)
called.  Instead, map and then munmap something.

This commit was SVN r8067.
2005-11-09 23:42:49 +00:00
Tim Woodall
78c98386d7 should reset the count (for persistent requests)
This commit was SVN r8064.
2005-11-09 22:02:48 +00:00
Tim Woodall
58b46d2da0 return mpool resources when request completes rather than in free
This commit was SVN r8063.
2005-11-09 21:59:01 +00:00
Graham Fagg
6b99301893 extra verbose in debug mode to help occ
This commit was SVN r8061.
2005-11-09 21:01:35 +00:00
Edgar Gabriel
b3d3552900 Fix for a problem Brian pointed out with cartesian communicators: in
comm_fill_rest there is no need for calling ompi_set_group_rank, since
we know already the rank of the process in the new comm. In case the
process was not part of the new communicator (rank = MPI_UNDEFINED)
calling this function caused a segfault on some platforms.

This commit was SVN r8060.
2005-11-09 21:00:58 +00:00
Brian Barrett
586a9be557 * make it easier to compare free timings with / without memory hooks enabled
This commit was SVN r8059.
2005-11-09 18:20:08 +00:00
Brian Barrett
05f8785e4f * OBJ_RELEASE is probably going to call free() in the way we're using it in
this function, and the lock will still be held, so we'd have deadlock.
  Instead, unlock, then release.  We don't need the lock for calling the
  release - we're not modifying any shared variables my that point.

  This should probbaly go onto the 1.0 branch, although it would be nearly
  impossible to trigger the deadlock with the way we use the free handlers
  at present.

This commit was SVN r8057.
2005-11-09 17:58:41 +00:00
George Bosilca
a6fdc2b2b4 Turn off the missing data-type message on MPI_Init.
This commit was SVN r8056.
2005-11-09 17:34:44 +00:00
Galen Shipman
3079fc2da1 use correct lock for threaded build..
This commit was SVN r8055.
2005-11-09 16:09:05 +00:00
George Bosilca
de0676a3dd Do not do any local copy into the storage if this convertor is finished. This is usefull in the Pack/Unpack case when there is more data in the packed buffer than the one we try to extract.
This commit was SVN r8054.
2005-11-09 07:46:12 +00:00
George Bosilca
025a8a04c5 More optimization of the data-type description are now possibles. Some corner cases are corrected. As a result we discover more accurately the contiguous part of the data memory layout.
This commit was SVN r8051.
2005-11-09 00:02:39 +00:00
Tim Woodall
78522ed454 send credits on correct qp
This commit was SVN r8050.
2005-11-08 22:59:44 +00:00
George Bosilca
63ba3bde11 Allow the convertor to remember the last trucated unpack. If the same convertor is used for the next unpack it will put the data back correctly. However, if the BTL/PTL create a new convertor, even if it clone the last one this magic will not happens !
This commit was SVN r8048.
2005-11-08 21:48:48 +00:00
George Bosilca
cdfe5e71fd By default there is no pending length on the convertor.
This commit was SVN r8047.
2005-11-08 21:45:45 +00:00
Tim Woodall
b4ca28da4b removed debug
This commit was SVN r8046.
2005-11-08 21:41:02 +00:00
Jeff Squyres
6de5c208f2 Fix propblem with prototypes for wtick and wtime in prototypes_pmpi.h.
This commit was SVN r8043.
2005-11-08 19:45:51 +00:00
George Bosilca
7582ae3ef1 A simpler way to get output about the packing/unpacking. Now there are 2 MCA parameters datatype_pack_debug and datatype_unpack_debug. When they are set to 1 the ddt engine will dump a lot of messages. Dont turn them to one by default. But if you notice any problems in the ddt you can turn them to one and send me the output.
First step toward adding memory to the convertor. It will be able to keep partial basic datatype between calls ...

This commit was SVN r8042.
2005-11-08 17:49:51 +00:00
George Bosilca
2b9b5500b9 Change some variable's names.
This commit was SVN r8041.
2005-11-08 17:44:56 +00:00
George Bosilca
c63e4dcef9 When we finish one of the loops take care of the index of the begining of the loop. If it's -1 then we just complete the full datatype ... therefore we have to do something special.
This commit was SVN r8040.
2005-11-08 16:53:31 +00:00
Tim Woodall
2d9c509add flow control
This commit was SVN r8039.
2005-11-08 16:50:07 +00:00
Graham Fagg
bcf8744bf6 valgrind saved me from a nasty order of eval error... i.e. derefing slected_data before setting it.
Anyway fixed and no memory leaks in coll tuned so far.

This commit was SVN r8037.
2005-11-08 04:52:30 +00:00
Graham Fagg
5b3ba944a8 Enabled, and running...
todos. turn the debug messages  into ompi ignorables and inot do some ops in ompi_bug mode

This commit was SVN r8036.
2005-11-08 04:43:17 +00:00
Graham Fagg
833b558046 Full configuration file based control of tuned collectives.
(verbose on bad config file and even cleans up after itself enought to make valgrind happy).

This commit was SVN r8035.
2005-11-08 03:36:38 +00:00