1
1

1284 Коммитов

Автор SHA1 Сообщение Дата
George Bosilca
93aa0d87bf Start to cleanup the header dependencies
Allow the TCP driver to release the memory allocated by the convertor. For that I add one integer (keeping the state of the actual allocation for each send_fragment) and a saved iovec (just to be able to remember the correct starting point for each iovec).

This commit was SVN r3122.
2004-10-14 07:09:17 +00:00
George Bosilca
da1c28b966 Remove some useless includes
Use the struct to avoid additional dependencies.

This commit was SVN r3121.
2004-10-14 07:07:05 +00:00
George Bosilca
1fafed88c6 Move the function mca_ptl_tcp_recv_frag_init in the header file and make it static and inline.
Solve the byte swap error for the ack. (on the receiver side the ack header was converted as a frag header).
Some indentation.

This commit was SVN r3120.
2004-10-14 07:05:46 +00:00
Brian Barrett
54dea835ed * be more verbose on errors in the rsh pcm. No functionality change, really
This commit was SVN r3098.
2004-10-13 23:11:19 +00:00
George Bosilca
236f44df72 Add the libmosal.so to the libvapi check. Now normally IB should work.
This commit was SVN r3089.
2004-10-13 21:51:20 +00:00
George Bosilca
fd5cd005a0 Update to the new datatype engine. Jeff will compile the code.
This commit was SVN r3088.
2004-10-13 21:13:50 +00:00
Tim Woodall
ce03565c4f - exchange procs arch/byte ordering at startup
- added tcp ptl support for heterogenous architectures

This commit was SVN r3075.
2004-10-12 21:50:25 +00:00
Tim Woodall
3c352a3ca4 restore prior version
This commit was SVN r3073.
2004-10-12 21:14:25 +00:00
Tim Woodall
0d58bace41 cleanup compiler warnings
This commit was SVN r3072.
2004-10-12 20:29:31 +00:00
Tim Woodall
7219f2d4f2 handle requests not completing initially due to resource exhaustion
This commit was SVN r3071.
2004-10-12 19:34:31 +00:00
Tim Woodall
dceae77546 cleanup valgrind warning
This commit was SVN r3070.
2004-10-12 18:11:26 +00:00
Tim Woodall
97752f1b06 - added ompi_request_wait which is optimized for a single request
- added georges comments regarding MPI_REQUEST_NULL handling to ompi_request_wait_any

This commit was SVN r3069.
2004-10-12 18:01:39 +00:00
Tim Woodall
a97d6dc4d9 restored code to disable event processing during cleanup
This commit was SVN r3068.
2004-10-12 16:31:06 +00:00
Tim Woodall
a86dc99b02 call the appropriate constructors
This commit was SVN r3067.
2004-10-12 16:14:12 +00:00
Tim Woodall
0df8ca90ea small re-org of request completion - moved test/wait out of pml
This commit was SVN r3066.
2004-10-12 15:50:01 +00:00
Tim Woodall
aa07811da5 intialize variables
This commit was SVN r3065.
2004-10-12 15:18:43 +00:00
Jeff Squyres
fcc40ec717 Don't let anyone else compile this
This commit was SVN r3064.
2004-10-12 14:29:30 +00:00
Jeff Squyres
a49fabdbd9 First cut of a "demo" component. Has hilighted a few errors in 3rd
party components, but I'll fix those after the tutorial.  :-)

This commit was SVN r3060.
2004-10-12 14:19:29 +00:00
Jeff Squyres
5d6174a348 Ensure basic coll component registers its priority properly (necessary
to show demo coll component in tutorial).

This commit was SVN r3059.
2004-10-12 12:38:56 +00:00
George Bosilca
0385f4b87d Adapt to the new datatype engine
Remove some of the compiler warnings

This commit was SVN r3057.
2004-10-12 04:17:29 +00:00
George Bosilca
37419ffc5f Rmove a lot of warning and make it compiling.
This commit was SVN r3056.
2004-10-12 03:57:08 +00:00
Rich Graham
cc46f30e6c add comment to mca_ptl_base_match_in_order_network_delivery.
This commit was SVN r3050.
2004-10-12 00:11:08 +00:00
Rich Graham
e83db5af32 save work in progress.
This commit was SVN r3049.
2004-10-11 22:56:46 +00:00
Graham Fagg
12c5be481d Don't allow tcp to register for Self as there is a Self module.
This fixes a bug in the PML logic that switches the PTL per message not
just per fragment.

This commit was SVN r3048.
2004-10-11 15:46:01 +00:00
George Bosilca
acca9f2f5e Improuve the logi of the wait function. I add a label simply because it was more convenient and faster than using additional variable to get out of the double loops.
This commit was SVN r3047.
2004-10-11 15:42:22 +00:00
George Bosilca
22418fb788 Change the PML logic a little bit. If there are several path between 2 nodes, I dont need message multiplexing, just fragments. So when we create the request we always pick up the first PTL attached to the destination. Now a communication between 2 nodes with small messages will always take the same path.
This commit was SVN r3046.
2004-10-11 15:40:36 +00:00
George Bosilca
1d2a03a0b2 In case of error dump some more informations just to make it more clear.
This commit was SVN r3045.
2004-10-11 15:37:53 +00:00
George Bosilca
da97ee8eba Typo ...
This commit was SVN r3044.
2004-10-11 15:36:01 +00:00
Weikuan Yu
58b4ec4d3e -- Leave the thread to pcm to cleanup
-- Update DTT engines, now it should be OK for basic showcase

This commit was SVN r3037.
2004-10-11 06:49:55 +00:00
Weikuan Yu
ad60bfdb6b -- Update ptl/elan with the latest tree
This commit was SVN r3034.
2004-10-11 02:08:39 +00:00
Weikuan Yu
508dbc4453 -- Commit some code from experimentation
-- Will update the code with the tree soon

BTW, Jeff, it seems to me that adding three NULL entries to an dummy module
does not matter. But let them hold the spots can be a little more
informative
though.

[yuw@p3 src]$ diff ptl_elan.c src/ptl_elan.c
46,49c46
<         mca_ptl_elan_req_fini,
<         NULL,
<         NULL,
<         NULL
---
>         mca_ptl_elan_req_fini

This commit was SVN r3033.
2004-10-11 00:14:39 +00:00
Jeff Squyres
31f6c00f3f Initialize a previously-uninitialized variable
This commit was SVN r3032.
2004-10-10 15:18:48 +00:00
George Bosilca
a62dafec84 Behave as expected by the MPI layer.
Use the defines from the mpi.h
Release the correct request.
Set the correct status
And return MPI_UNDEFINED if all requests are MPI_REQUEST_NULL.

This commit was SVN r3028.
2004-10-10 01:12:42 +00:00
George Bosilca
356cc37c13 Handle the 0 count communications through the self module.
This commit was SVN r3017.
2004-10-09 21:38:42 +00:00
George Bosilca
1777c361e9 If the rc variable is not initialized then there is a execution path who return whatever is in rc. Of course everything fail afterward as rc NEVER contain what we really expect by default.
This commit was SVN r3014.
2004-10-09 21:20:04 +00:00
Rich Graham
eaae910b5b add matching logic that assumes in-order delivery with no
dropped packets (shared memory, for example).  Reviewed by Tim W.

This commit was SVN r2999.
2004-10-08 22:24:57 +00:00
Tim Woodall
dded02fdcb add appropriate cast
This commit was SVN r2997.
2004-10-08 18:18:07 +00:00
Brian Barrett
96b3c503f9 * Add useful error messages to bootproxy for when it fails
* don't push pcmclient env variables to clients, since they don't have useful
  information in them
* add check to comm_spawn code to make sure we can spawn

This commit was SVN r2993.
2004-10-08 16:22:35 +00:00
Prabhanjan Kambadur
4c9f883a9b Changing the logic in ompi_comm_split to pass on the topo information only when required. Also had to change other files to go along wit this. Tested the intel test suites (relevant tests ... MPI_Cart_sub_c MPI_Comm_split1_c MPI_Comm_split2_c ... everything seems to work fine
This commit was SVN r2982.
2004-10-07 19:14:27 +00:00
Rich Graham
145ce7d68c set pointer to fragment matching logic routine, if this
not already set by the ptl - reviewed by tim.

This commit was SVN r2964.
2004-10-06 22:10:43 +00:00
Rich Graham
725afde500 just initialize some of the function pointers to null, so that when
mca_pml_teg_add_ptls is called, the value of NULL can be used to
determine if this needs to be set, or a previously set value needs
to be used.

This commit was SVN r2963.
2004-10-06 22:06:52 +00:00
Rich Graham
7fef6135d9 just initialize some of the function pointers to null, so that when
mca_pml_teg_add_ptls is called, the value of NULL can be used to
determine if this needs to be set, or a previously set value needs
to be used.

This commit was SVN r2962.
2004-10-06 22:06:02 +00:00
Rich Graham
281411f698 just initialize some of the function pointers to null, so that when
mca_pml_teg_add_ptls is called, the value of NULL can be used to
determine if this needs to be set, or a previously set value needs
to be used.

This commit was SVN r2961.
2004-10-06 22:05:15 +00:00
Tim Woodall
93a54d248f init pml callback function pointers to null
This commit was SVN r2960.
2004-10-06 21:08:54 +00:00
Prabhanjan Kambadur
a8764a7155 Changing graph create
This commit was SVN r2941.
2004-10-05 23:25:59 +00:00
Prabhanjan Kambadur
ffbe23e01b This file really should not be here
This commit was SVN r2940.
2004-10-05 22:27:54 +00:00
Prabhanjan Kambadur
baf370d0b8 The silliest of all cut and paste errors :-( .... had forgotten to change from dest to source
This commit was SVN r2939.
2004-10-05 22:26:35 +00:00
Tim Woodall
52a9776ab1 updating docs
This commit was SVN r2937.
2004-10-05 20:58:01 +00:00
Prabhanjan Kambadur
4655fe3943 Correcting code which allows topo to function properly ... also adding code to make comm_split and comm_dup work properly
This commit was SVN r2933.
2004-10-05 16:32:52 +00:00
Brian Barrett
b01bfc26a4 * Fix bug #1000 - I had the test wrong for strcmp. We want to use the
local hostname when we *FIND* "localhost" or "127.0.0.1", not when
  we don't find it.  Duh.

This commit was SVN r2925.
2004-10-04 20:00:36 +00:00