ports/openmpi
ports/openmpi
1
1
Форкнуть 0
Код Релизы Активность
3 262 коммитов 3 Ветки 0 Теги
Граф коммитов

3262 Коммитов

This Branch
This Branch
Все ветки
Автор SHA1 Сообщение Дата
George Bosilca
57cd74c9db A small trick for count == 0 
This commit was SVN r3351.
 2004-10-26 20:12:26 +00:00
Jeff Squyres
93e09fbfa9 Updates for fortran bindings 
This commit was SVN r3350.
 2004-10-26 19:28:05 +00:00
Prabhanjan Kambadur
df3c0da4af Adding protection for c++ name mangling 
This commit was SVN r3349.
 2004-10-26 19:15:19 +00:00
Tim Woodall
7c25c4b499 correct the arch setting 
This commit was SVN r3348.
 2004-10-26 19:12:48 +00:00
Prabhanjan Kambadur
9421f0d154 adding windows linkage directive to this file. had previously left it untouched because george had commits then and forgot about it after he committed 
This commit was SVN r3347.
 2004-10-26 18:48:45 +00:00
Prabhanjan Kambadur
20242c3c66 until there is a better way of generating the version files. still dont have the code to generate these files on windows. trivial issue and probably does not matter 
This commit was SVN r3346.
 2004-10-26 18:30:34 +00:00
Rich Graham
7e55ce2d8c make sure that with the single PTL case, the number of bytes scheduled 
does not exceed the max size allowed.

This commit was SVN r3345.
 2004-10-26 18:08:51 +00:00
Edgar Gabriel
16abfe9ed7 group_null, and comm_null have to be accessible also in comm_dyn.c, therefore I had to add an 'extern' statement in the header file. 
This commit was SVN r3344.
 2004-10-26 17:25:49 +00:00
Prabhanjan Kambadur
05d6b30d66 adding protection. dont know how this was missed in the whole big prtection commit 
This commit was SVN r3343.
 2004-10-26 15:47:15 +00:00
Rich Graham
13b9c39958 2 fragment messages are now working. 
This commit was SVN r3342.
 2004-10-26 15:38:22 +00:00
Rich Graham
ae97a54070 set the number of bytes sent by the sender for use by the 
receive completion logic.

This commit was SVN r3341.
 2004-10-26 15:24:58 +00:00
Edgar Gabriel
17fd5308a6 moving all internal communication to negative tags 
This commit was SVN r3340.
 2004-10-26 15:06:51 +00:00
Tim Woodall
c01a17dad9 - fixed arg list in sendrecv 
- added brain dead sendrecv_replace implementation

This commit was SVN r3339.
 2004-10-26 14:55:23 +00:00
Edgar Gabriel
22acad7c5c adding the routine to handle the proper disconnect of all dynamic communicators in MPI_Finalize (in cases the user did not explicitly disconnect). 
This commit was SVN r3338.
 2004-10-26 14:54:23 +00:00
Prabhanjan Kambadur
10ff968ec7 adding protection for the flex files. also added ompi_config.h for them 
This commit was SVN r3337.
 2004-10-26 14:11:44 +00:00
Tim Woodall
78a18a4acd corrected error checking 
This commit was SVN r3336.
 2004-10-26 12:37:19 +00:00
Tim Woodall
90c300defb removed warning message 
This commit was SVN r3335.
 2004-10-26 12:34:51 +00:00
Jeff Squyres
17a10cee87 Ensure to set MPI_REQUEST_NULL's f_to_c index 
This commit was SVN r3334.
 2004-10-26 12:11:16 +00:00
Jeff Squyres
c7807685ec Ensure to set the proper value for fortran's MPI_REQUEST_FREE 
This commit was SVN r3333.
 2004-10-26 11:52:14 +00:00
Tim Woodall
58c0cdeb6e cleanup 
This commit was SVN r3332.
 2004-10-26 11:39:16 +00:00
Edgar Gabriel
ee8b9e1897 first cut on the new disconnect infrastructure. More to come, 
especially focusing on Finalize.

This commit was SVN r3331.
 2004-10-26 11:37:58 +00:00
Jeff Squyres
9f13808c1c Remove compiler warning 
This commit was SVN r3330.
 2004-10-26 11:18:10 +00:00
Jeff Squyres
9697a3977d Remove some compiler warnings 
This commit was SVN r3329.
 2004-10-26 10:56:08 +00:00
George Bosilca
317c894a26 Several minor & major bugfixes: 
- reorder the code to accept LB and UB with ZERO count (it's MPI legal)
- improuve the LB & UB computation, It's still not completly clear how to handle them in some specific cases.

This commit was SVN r3328.
 2004-10-26 06:15:22 +00:00
George Bosilca
2390f30522 Correct the indexed and vector constructor. Mainly take into account the ZERO value for some fields. 
Improuve the creation of those datatypes.

This commit was SVN r3327.
 2004-10-26 06:13:35 +00:00
George Bosilca
62eaa080aa Correct the case MPI_COMBINER_HINDEXED. 
This commit was SVN r3326.
 2004-10-26 06:11:55 +00:00
George Bosilca
69f9781901 Remove unused macros. 
This commit was SVN r3325.
 2004-10-26 06:11:25 +00:00
George Bosilca
6f8a8b6cba Remove useless function. 
This commit was SVN r3324.
 2004-10-26 06:09:38 +00:00
George Bosilca
3598e00558 Correctly allocate the displ array. 
This commit was SVN r3323.
 2004-10-26 06:08:35 +00:00
Rich Graham
0648d1de85 fix bugs to get multi-fragment shared memory communications to work. 
This commit was SVN r3322.
 2004-10-26 03:33:02 +00:00
Tim Woodall
32f326a4d6 fix regression 
This commit was SVN r3321.
 2004-10-25 20:29:25 +00:00
Tim Woodall
978b7d8826 removed debug 
This commit was SVN r3320.
 2004-10-25 20:27:54 +00:00
Jeff Squyres
72f6cf2e4f Ensure that attributes that copy to fail will return an error out of 
MPI_COMM_DUP.

This commit was SVN r3319.
 2004-10-25 20:20:07 +00:00
Tim Woodall
9030610348 changed process list to hashtable 
This commit was SVN r3318.
 2004-10-25 19:55:04 +00:00
Tim Woodall
a0f95c028e error during startup/library init 
This commit was SVN r3317.
 2004-10-25 19:54:22 +00:00
Tim Woodall
ee20f3c6b0 added a hash table indexed by ompi_process_name_t 
This commit was SVN r3316.
 2004-10-25 19:53:44 +00:00
Tim Woodall
088c97b14e fix for deadlock in connect/accept 
This commit was SVN r3315.
 2004-10-25 19:52:37 +00:00
Rich Graham
de49778184 add a pointer to the sending ptl. Used to process acks. Can't assume 
that ptl address on send and receive side is the same.  This is set
once by the constructor.  Can't do this on the recv side, since a
fragment can be used for multiple destinations.

This commit was SVN r3314.
 2004-10-25 18:02:46 +00:00
Rich Graham
826b5ebdf6 fix numerous bugs in the shared memory implementation. Can pass a 
hello world test now.

This commit was SVN r3313.
 2004-10-25 17:22:47 +00:00
Ralph Castain
1b7a7bae23 Ensure that proc_info.h and sys_info.h are only gone through once. Minor cleanup in universe_setup_file_io.c. 
This commit was SVN r3312.
 2004-10-25 14:14:20 +00:00
Ralph Castain
3e19906b95 Begin to "leak" the changes to the registry and supporting subsystems to resolve the flood situation and support abnormal terminations. These changes just define a new message structure for returning all startup/shutdown information in a single broadcast-like transmission. Shouldn't have any impact on existing code as the message object isn't used yet. 
This commit was SVN r3311.
 2004-10-25 13:36:09 +00:00
Tim Woodall
e6066c59f2 init code 
This commit was SVN r3310.
 2004-10-25 13:33:30 +00:00
George Bosilca
6c21293555 I dont believe that I did such stupid mistake !!! I start thinking about my punishment ... 
This commit was SVN r3309.
 2004-10-25 05:27:12 +00:00
George Bosilca
542630ad68 Remove the fprintf statements. 
This commit was SVN r3308.
 2004-10-24 23:16:27 +00:00
Rich Graham
2d54712b37 bug fixes. 
This commit was SVN r3307.
 2004-10-24 00:26:03 +00:00
George Bosilca
941fec8f66 Correctly initialize the stack when we have an initial offset in the data. I still have to make the optimized version. 
This commit was SVN r3306.
 2004-10-23 22:06:10 +00:00
Jeff Squyres
1c4aafba31 Replace hard-coded OMPI_FORTRAN_HANDLE_MAX with a real configure test 
to determine min(INT_MAX, max fortran INTEGER value)

This commit was SVN r3305.
 2004-10-23 21:08:12 +00:00
Rich Graham
c7e6eb7e75 remove incorrect initialization. 
This commit was SVN r3304.
 2004-10-23 20:38:35 +00:00
Tim Woodall
725a1ae657 template for mx 
This commit was SVN r3303.
 2004-10-23 19:49:29 +00:00
Jeff Squyres
113ce723e1 Convert all ompi_bitmap_t and ompi_pointer_array_t indexing from 
size_t to int.  This is in recognition of the fact that these two
classes are primarily used for fortran<-->c convertsion of various
entities (attributes, MPI objects), and MPI defines that an int must
be used to hold MPI fortran handles.  Hence, why use size_t
internally?   See the comment in src/class/ompi_pointer_array.h for a
better description.

Mixed in a few attribute fixes in this commit as well.

This commit, therefore:
- converts the indexing from size_t to int
- changes all locations in the code that I could find that used size_t
  as the interface to ompi_bitmap_t or ompi_pointer_array_t
- convert to use OMPI_FINT_2_INT / OMPI_INT_2_FINT macros in various
  src/mpi/c/*f2c.c and src/mpi/c/*c2f.c files
- check the return code of ompi_comm_free() to ensure that everything
  worked properly before returning MPI_SUCCESS
- unified all src/mpi/c/*f2c.c functions (i.e., they all do the same
  thing)
- tie up some loose ends w.r.t. MPI_Request handling in fortran; set
  the req_f_to_c_index to MPI_UNDEFINED when it does not have a fortran
  index
- still need to add a configure test to find OMPI_FORTRAN_HANDLE_MAX
  for ompi_bitmap.c and ompi_pointer_array.c (hard-wired to INT_MAX
  right now)
- re-organized, consolidated, and unified some ompi_pointer_array.c
  code -- fixed a few minor bugs w.r.t. lowest_free (could have left
  some unintentional holes in the array)

This commit was SVN r3302.
 2004-10-23 19:24:00 +00:00