1
1
Граф коммитов

657 Коммитов

Автор SHA1 Сообщение Дата
Brian Barrett
a070d9af63 * Update to make this work again with the changes in the mca changes
This commit was SVN r657.
2004-02-06 07:17:13 +00:00
Brian Barrett
6835324c4b * make patch not prompt when running with libtool 1.5.2. I'm going
to have to find a way to just not try to apply the patch, so that I
  don't get the error message ;)

This commit was SVN r656.
2004-02-06 05:42:33 +00:00
Jeff Squyres
2748878e9f Fix yet another linker error on OS X
This commit was SVN r655.
2004-02-06 04:20:13 +00:00
Tim Woodall
8947cebc8d - fixed issue w/ module close
This commit was SVN r654.
2004-02-05 17:45:54 +00:00
Tim Woodall
086fa5b7cb module init/close cleanup
This commit was SVN r653.
2004-02-05 17:12:59 +00:00
Jeff Squyres
056485c6da Move global variables into files with functions in them. POSIX says
that linkers don't have to pull in .o files unless they have functions
in them, so having global symbols in files by themselves does not
guarantee that a program trying to use them will be able to link
successfully.

This commit was SVN r652.
2004-02-05 01:52:56 +00:00
Tim Woodall
9d0b36c34d - reworked base headers
- when recv is posted check for matching frags

This commit was SVN r651.
2004-02-05 00:50:37 +00:00
Rich Graham
11fe4769aa add one more test - done for now.
This commit was SVN r650.
2004-02-05 00:05:18 +00:00
Rich Graham
d8c4b80918 list.h - remove prototype for unimplemented and unused function.
list.c - fix bug in lam_list_insert - loop had wrong ending parameter
list_test.c - add more tests

This commit was SVN r649.
2004-02-04 23:54:22 +00:00
Tim Woodall
b8d6c49fac - match fragments when a receive is posted
This commit was SVN r648.
2004-02-04 17:11:57 +00:00
Prabhanjan Kambadur
e195983847 Changing the F77 null condition. If not fortran compiler is found, then
F77 is set too be "none".

This commit was SVN r647.
2004-02-04 02:23:55 +00:00
Prabhanjan Kambadur
7e4da72a91 Forgot to add info as one of the subdirs to be traversed
This commit was SVN r646.
2004-02-04 01:34:06 +00:00
Prabhanjan Kambadur
2df9fad227 Adding some functionality to MPI_Info functions. Yet to add anything
relevant to Fortran functions. So, nothing has been changed in
MPI_Info_f2c and MPI_Info_c2f functions. Everything else has functioning
code now.

This commit was SVN r645.
2004-02-03 22:34:01 +00:00
Prabhanjan Kambadur
4cd799d590 Adding libmpi_info.la as one of the libraries to be built. Also adding
the necessary data structures to support MPI_Info* functions.

This commit was SVN r644.
2004-02-03 22:03:22 +00:00
Tim Woodall
a4738b19da allocate send request and first fragment together
This commit was SVN r643.
2004-02-03 21:37:17 +00:00
Tim Woodall
3836a7bf33 - factor common elements into fragment base class
- tcp cleanup

This commit was SVN r642.
2004-02-03 21:33:29 +00:00
Tim Woodall
103cb62480 cleanup pml/ptl init
This commit was SVN r641.
2004-02-02 22:58:24 +00:00
Tim Woodall
965c44204c fixed lam_ifnext based to correctly handle end of list semantics
This commit was SVN r640.
2004-02-02 20:21:26 +00:00
Jeff Squyres
5a13010974 Fix silly mistakes in figuring out the oob, pcm, and registry modules
This commit was SVN r639.
2004-02-02 19:16:21 +00:00
Tim Woodall
8908bf2cbc fixed a couple of porting (C++->C) issues
This commit was SVN r638.
2004-02-02 16:11:59 +00:00
Prabhanjan Kambadur
60be99b4cd Adding a few comments and description to the info* functions
This commit was SVN r637.
2004-02-01 22:28:20 +00:00
Prabhanjan Kambadur
fbc9be66a6 Adding structure lam_info_entry_t and member lam_info_list
This commit was SVN r636.
2004-02-01 22:27:22 +00:00
Jeff Squyres
96d8ab8db9 Just in case we ever go to subversion :-)
This commit was SVN r635.
2004-02-01 20:11:46 +00:00
Mark Taylor
e2f73b7edb added code to MPIO_Test, so it will also free the MPI request if
it is completed.  Also set requests to MPI_REQUEST_NULL after free'ing
them.  (at pressent, MPI-IO ROMIO component does not keep lists of
requests for re-use)

This commit was SVN r634.
2004-02-01 04:48:18 +00:00
Jeff Squyres
2ed7ef11c2 Display pcm, oob, and registry information
This commit was SVN r633.
2004-01-31 21:48:58 +00:00
Jeff Squyres
ea3855d1cb Take thread issues into account, but don't do much with the information
yet

This commit was SVN r632.
2004-01-31 21:48:41 +00:00
Jeff Squyres
7bafbd0043 Intialize and finalize the RTE MCA module types
This commit was SVN r631.
2004-01-31 21:47:59 +00:00
Jeff Squyres
cc05d5f606 First cut of the MPI thread leave determination stuff (i.e., actually
listen to what the modules say, but don't make any decisions based on that
yet)

This commit was SVN r630.
2004-01-31 21:45:32 +00:00
Jeff Squyres
c93e1bd291 Move finalize back up into the module struct. This the last time
I'll move finalize; I swear it.

This commit was SVN r629.
2004-01-31 21:44:54 +00:00
Jeff Squyres
2df075a1ff - Update/revise pcm base functionality
- First cut of oob/registry base functionality

This commit was SVN r628.
2004-01-31 21:43:26 +00:00
Mark Taylor
3c09880e57 romio mca wrapper functions:
added code to each wrapper function to allocate or free lam_request_t
   when necessary.

   populate the romio action (function pointer table) in lam_file_t,
   when creating new lam_file_t.

This commit was SVN r627.
2004-01-31 07:44:05 +00:00
Mark Taylor
3c08f8faef minor updates to comments
This commit was SVN r626.
2004-01-31 06:52:47 +00:00
Mark Taylor
e05ae7ec7e Romio MCA wrapper functions - now all wrapper functions
are written & compiling.

This commit was SVN r625.
2004-01-31 06:45:24 +00:00
Mark Taylor
f79a364e39 more ROMIO MCA wrapper functions
This commit was SVN r624.
2004-01-31 06:36:51 +00:00
Mark Taylor
9af9b49b2c more romio MCA wrapper functions
This commit was SVN r623.
2004-01-31 06:05:28 +00:00
Mark Taylor
db6030ca87 more ROMIO MCA wrapper functions...
This commit was SVN r622.
2004-01-31 06:00:09 +00:00
Mark Taylor
161e021562 ROMIO MCA component:
completed all MPI-IO MCA v1 prototypes
    completed all MPI-IO MCA v1 typedefs
    add all functions to MCA v1 action struct

    completed about 50% of the wrapper functions.

This commit was SVN r621.
2004-01-30 23:14:19 +00:00
Jeff Squyres
cfff4bb79a No longer need to pass a lam_list_t to mca_ptl_base_select() since the
ptl base owns the list

This commit was SVN r620.
2004-01-30 23:10:06 +00:00
Jeff Squyres
2136489dda Return the real error, not LAM_ERROR
This commit was SVN r619.
2004-01-30 23:03:12 +00:00
Jeff Squyres
f3574207cc Close down the coll type
This commit was SVN r618.
2004-01-30 23:03:01 +00:00
Jeff Squyres
f28e15def9 - Pass a lam_list_t of ptl's to the pml function add_ptls()
- Use allow_multi_use_threads/have_hidden_threads instead of
  min_thread/max_thread
- rename fini -> finalize
- selection is based on ptl's that fail to initialize (they are weeded out)
- finalize all ptl's when shutting down

This commit was SVN r617.
2004-01-30 23:02:39 +00:00
Jeff Squyres
94a9aab796 - rename fini -> finalize
- use allow_multi_use_threads/have_hidden_threads instead of
  min_thread/max_thread

This commit was SVN r616.
2004-01-30 23:00:48 +00:00
Jeff Squyres
4debc92533 Ensure to only finalize if we did a selection
This commit was SVN r615.
2004-01-30 22:56:36 +00:00
Jeff Squyres
41b2f4e443 PTL list is now owned by the ptl base
This commit was SVN r614.
2004-01-30 22:55:48 +00:00
Jeff Squyres
34df3d80b8 Don't forget to finalize the selected module during shutdown.
This commit was SVN r613.
2004-01-30 22:28:32 +00:00
Jeff Squyres
dae32f6e58 - Fix bad memory referencing; added lam_list_item_t super to
opened_module_t
- Fix minor memory leak

This commit was SVN r612.
2004-01-30 19:42:31 +00:00
Tim Woodall
1b6464d815 STATIC_INIT requires an object - rather than a pointer - this
seems coutner-intuitive, as everything else requires a pointer....

This commit was SVN r611.
2004-01-30 14:12:35 +00:00
Jeff Squyres
97271026bc Do more stuff during startup; initialize various MCA types, etc.
This commit was SVN r610.
2004-01-30 03:59:39 +00:00
Jeff Squyres
fc555aaed7 Updates; show coll modules
This commit was SVN r609.
2004-01-30 03:59:17 +00:00
Jeff Squyres
e268e16358 Add some more globals per recent e-mails; keep the initially
requested MPI thread level as well as the provided MPI thread levels.
Also keep a bool indicating whether there may be multiple user
threads in MPI simultaneously (essentially whether we are
MPI_THREAD_MULTIPLE or not)

This commit was SVN r608.
2004-01-30 03:57:43 +00:00