1
1
Граф коммитов

235 Коммитов

Автор SHA1 Сообщение Дата
Brian Barrett
ab73cc0487 * minor diagnostic printf that should have been in last commit (doh!)
This commit was SVN r6734.
2005-08-04 15:43:50 +00:00
Brian Barrett
a80b00ab5e * Don't change size of user frag - it's not needed, and causes the frag
to never be returned to the free list

This commit was SVN r6733.
2005-08-04 15:43:13 +00:00
Brian Barrett
9cfa6bba6a * If a message isn't successfully sent, reduce the pending sends counter, as
the message is no longer pending
* Try to push out new messages whenever we finish a send, whether it
   worked or not.  Means that in the case where the other side has too
  many sends pending, we'll constantly retry one (and only one, once the
  pending number is reached) message until goodness returns
* Make some warnings only happen in verbose case, as they are mainly
  diagnostics

This commit was SVN r6732.
2005-08-04 15:41:11 +00:00
Jeff Squyres
b2cfedf805 Add copyright headers, ompi_config.h, and stdio.h.
This commit was SVN r6729.
2005-08-04 12:06:08 +00:00
Brian Barrett
26adbfe713 checkpoint to move back to RS
* remove dead code
* add some debugging code

This commit was SVN r6725.
2005-08-03 20:21:23 +00:00
Brian Barrett
6c37ad4471 * more components to ignore on RS
* fix comment

This commit was SVN r6724.
2005-08-03 16:08:27 +00:00
Brian Barrett
67f96f7b46 * convert to new param registration code
* Fix RDMA book keeping

This commit was SVN r6723.
2005-08-03 16:02:02 +00:00
Brian Barrett
3867d59cf1 * Look at stdout instead of stdin for determining whether output is going
to a tty or not.  Now you can do something like:

     ompi_info -all | grep btl_portals

  and get the full line for each btl_portals parameter.
* For the case where stdout is a tty, we have my current nomination for
  Today's Useless OMPI Feature.  Autodetect the width of the terminal, so
  people with really wide terminals will get less wrapping

This commit was SVN r6722.
2005-08-03 14:48:48 +00:00
Jeff Squyres
11140e9cb8 We must eliminate and stamp out all forms of redundancy, however they
may appear.

(remove *error.h file from Makefile.am -- a cut-n-paste error that has
propagated to a surprising number of directories ;-) )

This commit was SVN r6721.
2005-08-03 14:47:04 +00:00
Brian Barrett
eb2748130b * don't build TCP component if we don't have IP sockets :)
This commit was SVN r6716.
2005-08-02 20:06:34 +00:00
Rainer Keller
f5f96f424d - Use for-loop
- Minor heck for stride != 1
- indenting

This commit was SVN r6715.
2005-08-02 19:56:18 +00:00
Brian Barrett
24116a3935 * fix up a bunch of threading issues when progress and/or mpi threads
are enabled.  Mostly just ADD32 -> ADD_SIZE_T issues and naming of
  variables in THREAD_{LOCK,UNLOCK}

This commit was SVN r6706.
2005-08-02 17:36:01 +00:00
Brian Barrett
4a748d62ae * add some missing header files for OS X
This commit was SVN r6702.
2005-08-02 14:59:50 +00:00
Tim Woodall
2214f0502d - first cut at tcp btl (working but not optimal)
- reworked btl error logging macros

This commit was SVN r6701.
2005-08-02 13:20:50 +00:00
Jeff Squyres
4c1dd716c7 Change and add new features to the MCA parameter system:
- new preferred API calls for registering MCA parameters are
  mca_base_param_reg_{int|string} and
  mca_base_param_reg_{int|string}_name.
- See opal/mca/base/mca_base_param.h for docs on new calls.  
- Can now register and lookup a value at the same time.
- Can now mark a parameter "read only" at registration time
- Can now mark a parameter "internal" at registration time
- Can now associate a help message with the parameter at registration
  time; displayed in the ompi_info output.

The old API calls are still available for backwards compatibility
(mca_base_param_register_{int|string}.  They will eventually be
removed -- all developers are encouraged to use the new APIs from here
on out and replace any old calls with the new API.

Some params were also renamed -- the previous convention of using
"base_" as a prefix for any param that was not associated with a
component is henceforth deprecated.  Instead, use one of the following
prefixes:

mca: for anything in the MCA base itself
opal: for anything in OPAL
orte: for anything in ORTE
mpi: for anything in OMPI

This commit was SVN r6698.
2005-08-01 22:38:17 +00:00
Tim Woodall
2324f36065 call ob1 progress rather than endpoint - as it may not have one
This commit was SVN r6696.
2005-08-01 22:30:25 +00:00
George Bosilca
ea0cf322f3 Taking care of the source pointer.
This commit was SVN r6691.
2005-08-01 17:09:46 +00:00
George Bosilca
f8ccce7503 One step further.
This commit was SVN r6690.
2005-08-01 17:08:59 +00:00
Brian Barrett
41f7bb3a2a * undo removing of .ompi_unignore file (there is actually a .ompi_ignore
in this directory)

This commit was SVN r6688.
2005-07-29 23:56:48 +00:00
Ralph Castain
4e79a51395 Add a job_info segment to the system that holds a container for each job. Within each container is a keyval indicating the job state (i.e., all procs at stage1, finalized, etc.). This provides a rough state-of-health for the job.
This required a little fiddling with a number of areas. Biggest problem was that it uncovered a potential for an infinite loop to be created in the registry. If a callback function modified the registry, the registry checked the triggers to see if anything had fired. Well, if the original callback was due to a trigger firing, that condition hadn't changed - so the trigger fired again....which caused the callback to be called, which modified the registry, which checked the triggers, etc. etc.

Triggers are now checked and then "flagged" as being "in process" so that the registry will NOT recheck that trigger until all callbacks have been processed. Tried doing this with subscriptions as well, but that caused a problem - when we release processes from a stagegate, they (at the moment) immediately place data on the registry that should cause a subscription to fire. Unfortunately, the system will just hang if that subscription doesn't get processed. So, I have left the subscription system alone - any callback function that modifies the registry in a fashion that will fire a subscription will indeed fire that subscription. We'll have to see if this causes problems - it shouldn't, but a careless user could lock things up if the callback generates a callback to itself.

Also fixed the code that placed a process' RML contact info on the registry to eliminate the leading '/' from the string.

This commit was SVN r6684.
2005-07-29 14:11:19 +00:00
Brian Barrett
6528ee4692 * remove some useless printfs
This commit was SVN r6683.
2005-07-29 00:23:28 +00:00
George Bosilca
8e1f169ffd When we copy a convertor we should copy it's flags too. Otherwise the informations inside will
not match the flags (uninitialized).

This commit was SVN r6681.
2005-07-29 00:17:02 +00:00
George Bosilca
8e0d8a0e99 The datatype should be initialized as soon as possible. Inside we detect the local architecture and
create the local convertor. They will get used on the ompi_proc_init.

This commit was SVN r6680.
2005-07-29 00:15:26 +00:00
George Bosilca
dc2a3d7917 We dont have a .ompi_ignore so I dont see why we have a .ompi_unignore.
This commit was SVN r6679.
2005-07-29 00:14:18 +00:00
George Bosilca
10d95144b6 Add a global local architecture. It is required by the proc.c file as all new procs by default
have the same architecture as the current one.

This commit was SVN r6678.
2005-07-29 00:12:33 +00:00
Brian Barrett
170777d399 * exit on error, not success...
This commit was SVN r6677.
2005-07-29 00:07:52 +00:00
Brian Barrett
cbf04e3d3f * if the frag isn't going to go, reduce the pending frags count, don't
increase it

This commit was SVN r6675.
2005-07-28 22:29:05 +00:00
Jeff Squyres
6e1110188b Allow coll v1 component authors to be lazy in their module_init
functions  (i.e., just return comm->c_coll_selected_module).

This commit was SVN r6668.
2005-07-28 20:31:26 +00:00
George Bosilca
c8bc529df4 The second cut of MX ... still not working yet
This commit was SVN r6666.
2005-07-28 19:53:27 +00:00
Tim Woodall
b2ec71a609 correct for case where eager limit is greater than min_rdma_size
This commit was SVN r6663.
2005-07-28 19:35:47 +00:00
Brian Barrett
5e75cb2495 * properly set unlink thresholds - START/END combined are 1 event
This commit was SVN r6662.
2005-07-28 19:28:04 +00:00
Brian Barrett
7441dfc4c3 fix some printfs
This commit was SVN r6660.
2005-07-28 19:15:07 +00:00
Brian Barrett
8a56cd567f * make poll time 0 so that our latency isn't way too high
* learn to spell...

This commit was SVN r6659.
2005-07-28 18:48:30 +00:00
Brian Barrett
93671ab150 * remove ompi_{ignore,unignire} files
This commit was SVN r6658.
2005-07-28 18:46:46 +00:00
Brian Barrett
05720c099f * use catamount header file
* fix some printfs

This commit was SVN r6654.
2005-07-28 17:09:23 +00:00
Brian Barrett
6cf88caeb4 * remove some unneeded printfs in Portals btl
* add some svn:ignores

This commit was SVN r6653.
2005-07-28 17:04:52 +00:00
Brian Barrett
3f09d5f2a4 * make btl open be safe to call multiple times (btl close already was)
* add btl back into ompi_info.  Since it now directly calls the
  open/close, the missing symbol problems Ralph was seeing when ob1 is
  ignored will not occur.

This commit was SVN r6652.
2005-07-28 16:31:29 +00:00
Rainer Keller
42f23932e0 In part revert 6647, btl_sm_fifo was in repos,
but not in Makefile.am

This commit was SVN r6651.
2005-07-28 16:25:09 +00:00
Brian Barrett
b0b6ddd078 * add --enable-heterogeneous (default: enabled) to enable heterogeneous
support in OMPI.  Currently only enables/disables the architecture
  sharing modex in ob1 pml.
* Add sds framework to ompi_info
* Figure out table ids to use for Portals BTL at configure time, since
  we should use 30 & 31 on Red Storm, but the reference implementation
  only supports 0-8.
* Some bug fixes in Portals UTCP sds

This commit was SVN r6650.
2005-07-28 16:16:13 +00:00
Rainer Keller
29465f0f28 There is no file btl_sm_fifo.h
This commit was SVN r6647.
2005-07-28 15:47:46 +00:00
Galen Shipman
168274db45 Call progress on send schedule so that prepare_src can run if the rdma ack
lands between sending fragments up to rdma_offset. 

This commit was SVN r6646.
2005-07-28 15:38:31 +00:00
Rainer Keller
6b7eb3b2d9 Add btl_sm_endpoint.h
This commit was SVN r6644.
2005-07-28 15:05:16 +00:00
Brian Barrett
052b4d4da4 * only give warning about removing -pedantic and -Wall if we are actually
going to build the component

This commit was SVN r6641.
2005-07-28 06:05:27 +00:00
Jeff Squyres
a7a9196350 There is no such file btl_gm_error.h.
This commit was SVN r6636.
2005-07-28 00:08:36 +00:00
George Bosilca
801037ed6c Migrating MX PTL to the new configure sub-system (shared by both MX BTL and MX PTL).
This commit was SVN r6634.
2005-07-27 23:39:56 +00:00
George Bosilca
94fe5e6ac8 MX BTL use the new configure sub-system.
This commit was SVN r6633.
2005-07-27 23:38:31 +00:00
Brian Barrett
6aa464b67e More changes from Red Storm port
- only call sched_yield if it exists
  - don't fail out if modex doens't work in ob1
  - bunch of fixes for Portals BTL
  - add cnos rml component
  - add NULL gpr component (should only be used if replica AND proxy
    fail to load)  

This commit was SVN r6629.
2005-07-27 23:07:14 +00:00
George Bosilca
9603996a31 Self BTL is working just fine.
This commit was SVN r6624.
2005-07-27 21:46:44 +00:00
Rainer Keller
b9e092a9db There is no btl_mx_error.h, its not included anywhere.
This commit was SVN r6623.
2005-07-27 20:48:19 +00:00
George Bosilca
e1b3758fa5 The first cut for he MX BTL.
This commit was SVN r6621.
2005-07-27 19:46:36 +00:00