1
1
Граф коммитов

748 Коммитов

Автор SHA1 Сообщение Дата
George Bosilca
0376dce258 Keep track of the ompi_proc in the comm_proc. This avoid a lookup for the processor and
simplify the execution path. The peer proc (ompi_proc_t) is set at the matching stage.

This commit was SVN r8962.
2006-02-10 18:55:43 +00:00
Galen Shipman
44fe6c3896 allow pml pipeline to cache memory registrations
to enable this (off by default) use: 
-mca pml_ob1_leave_pinned_pipeline 1 
!!AND!!!
-mca mpool_use_mem_hooks 1 

This commit was SVN r8949.
2006-02-09 15:49:51 +00:00
George Bosilca
269fc0c13a Cleanup:
1. remove all useless macros from the proc header file
2. merge 2 of the match macros (they share the same logic except one list)

This commit was SVN r8946.
2006-02-09 06:59:54 +00:00
George Bosilca
83f83e5730 Specialize the MCA_PML_OB1_FREE macro. When we call this macro we already know what kind
of request we are playing with (send or receive). Therefore, it's useless to have another
switch inside this macro and make the code bigger. Now, we have 2 versions
MCA_PML_OB1_SEND_REQUEST_FREE and MCA_PML_OB1_RECV_REQUEST_FREE.

This commit was SVN r8945.
2006-02-08 22:42:00 +00:00
Galen Shipman
1c71ab73d1 Move ompi_pointer_array initialization and destruction outside of the memory
callback, note that ompi_pointer_array_remove_all does not call free. 
This needs to go to the release branch 

This commit was SVN r8944.
2006-02-08 22:40:40 +00:00
George Bosilca
600f664db2 No comment !!! But this time I really remove the second lock ...
This commit was SVN r8943.
2006-02-08 21:43:46 +00:00
George Bosilca
89db0be4a8 Remove the second lock.
This commit was SVN r8940.
2006-02-08 19:19:57 +00:00
Galen Shipman
18bbb049d1 fix bsend bug, need to check that data length before packing.
This needs to go to the release branch

This commit was SVN r8937.
2006-02-08 17:39:33 +00:00
George Bosilca
0775c88a9b Remove all unused variables.--This line, and those below, will be ignored--
M    ob1/pml_ob1_recvreq.c

This commit was SVN r8934.
2006-02-08 07:20:48 +00:00
George Bosilca
e9706e6db0 Remove the macro to compute the length of the segments from the send header
and add a new macro that can be used for both sends and receives.
Move to atomic operations to manage the length of the sended or received
status. There is one instance where the atomic operation is not required
as the code can cannot be executed in same time by 2 differents threads.

This commit was SVN r8933.
2006-02-08 06:03:54 +00:00
George Bosilca
9f1357fb89 Remove all the useless includes. Most of the endpoint do not depend on the
orte includes.

This commit was SVN r8932.
2006-02-08 05:10:48 +00:00
Brian Barrett
ec7b60d501 * Fix some error checking code for Lock / Unlock at the MPI layer
* Implement win_lock and win_unlock in the pt2pt component.  Not well
  tested, but appears to move bits if properly motivated...

This commit was SVN r8922.
2006-02-07 18:45:18 +00:00
Galen Shipman
c8045bf397 Fixup for ORTE datatype checkin,
- use appropriate header files 
- change calls from orte_dps to orte_dss 

This commit was SVN r8920.
2006-02-07 15:20:44 +00:00
Brian Barrett
340bf14191 * support non-predefined datatypes for all three communication mechanisms
* rework the thread locking so that it at least makes sense to me.  Still
  need to do a bunch of testing before I'm happy with it, but it's a tad
  bit closer...

This commit was SVN r8918.
2006-02-07 12:16:23 +00:00
George Bosilca
20fd358327 A BTL cannot depend on the orte data-types.
This commit was SVN r8915.
2006-02-07 06:05:36 +00:00
Ralph Castain
4b9f015c0b Merge in the new data support subsystem for ORTE. MPI folks should not notice a difference. Longer explanation will be sent to developers mailing list.
This commit was SVN r8912.
2006-02-07 03:32:36 +00:00
Galen Shipman
8f8ed186f1 When leave pinned is set, only enable mem hooks if disable sbrk is not set.
This commit was SVN r8908.
2006-02-06 21:22:20 +00:00
Tim Woodall
e920ec67b9 - add new send flag (MCA_PML_OB1_SEND_COMPLETE) to force local completion prior to signalling mpi completion
This commit was SVN r8907.
2006-02-06 19:51:49 +00:00
Tim Woodall
c0d18c9579 mem hooks must be enabled if leave_pinned is set
This commit was SVN r8906.
2006-02-06 19:44:21 +00:00
George Bosilca
e20265bd2b Dont let any external to the data-type code check directly for the predefined
data-types. Instead, use the newly provided data-type function ompi_ddt_is_predefined..

This commit was SVN r8903.
2006-02-06 18:01:45 +00:00
George Bosilca
eb1d2dd290 Working down the latency (0.2 micro-sec on a Xeon 2Ghz) by removing the
second instance of the ompi_proc from the send and receive request. This
information is already available on the base request, so there is no
need for duplication. The drawback is that now (in order to avoid a
second lookup in the communicator array of procs) we have to set the base
proc in the PML's _ALLOC macro.

This commit was SVN r8900.
2006-02-05 06:13:07 +00:00
Galen Shipman
6a6d86f963 Port mvapi free list for registrations to openib
This commit was SVN r8886.
2006-02-02 19:44:39 +00:00
Rainer Keller
7ac0ffc349 - Instead of doing the unlock inside the if, just move the if-statement
later after the mandatory unlock.

This commit was SVN r8885.
2006-02-02 17:32:22 +00:00
Galen Shipman
a0d3ddd267 Use free list for registrations, otherwise we may OBJ_RELEASE in the memory
callbacks. 

This commit was SVN r8884.
2006-02-02 17:14:49 +00:00
Graham Fagg
232bb9534a Start moving stuff out of modules that should be in the component.
This commit was SVN r8874.
2006-02-01 20:50:14 +00:00
Graham Fagg
5f2d82347f a couple of changes to make barrier synchronous.. means last communication to any possible peer must
be locally completing. for now using synchronous calls until the new functionality is available. then will change
the code to use the new PML send flags.

This commit was SVN r8867.
2006-01-31 23:21:46 +00:00
Brian Barrett
44a516d966 * Clean up error checking in the MPI interface for MPI-2 onesided
* Implement fortran handle -> c handle tracking
* Remove some unneeded locking around free lists (the free list
  macros do their own locking)
* Try to be a bit more memory friendly with the w_mode setting /
  checking

This commit was SVN r8865.
2006-01-31 21:40:12 +00:00
Tim Woodall
a2fde48f2f changes from release branch
This commit was SVN r8858.
2006-01-31 16:17:18 +00:00
Tim Woodall
bcd6c525f8 removed duplicate locks
This commit was SVN r8857.
2006-01-31 16:12:37 +00:00
Brian Barrett
762fe70bd1 * Fix uninitialized variable warning. Technically, start's group should be
a subset of win's group, so this should never happen.  But users have
  been known to screw up before, so return a reasonable error.

This commit was SVN r8855.
2006-01-31 15:32:40 +00:00
Tim Woodall
9d484916db remove locks already held
This commit was SVN r8853.
2006-01-31 14:23:08 +00:00
George Bosilca
42516d0785 Look like the name changed from OSC_PT2PT_DATA to P2P_MODULE.
This commit was SVN r8852.
2006-01-31 06:34:36 +00:00
Brian Barrett
0d507ad970 * Fix a bunch of checks for proper access / exposure epochs
* Add implementation of Post/Wait/Start/Complete synchronization

This commit was SVN r8848.
2006-01-31 02:44:08 +00:00
Brian Barrett
58f074816d * declare the variable names the same way they were used... At least at
some point, I realized having a variable name "new" might cause some
  problems for those people using a C++ compiler to compile C code ;).

This commit was SVN r8846.
2006-01-30 12:16:52 +00:00
George Bosilca
418f1b8986 The module name is ompi_osc_pt2pt_module_t not ompi_osc_p2p_module_t.
This commit was SVN r8843.
2006-01-29 18:00:30 +00:00
George Bosilca
9d9fc78d0e Like in the PML the path to the statically included file have to start
with the sub-project name.

This commit was SVN r8842.
2006-01-29 17:33:45 +00:00
Brian Barrett
d7056e6e2a remove some unneeded debugging output
This commit was SVN r8841.
2006-01-28 19:59:54 +00:00
Brian Barrett
05c9de058c * add missing header include
This commit was SVN r8837.
2006-01-28 16:09:35 +00:00
Brian Barrett
b1d2424013 Merge in present work on the MPI-2 onesided chapter. The current code is not
complete, but stable enough that it will have no impact on general development,
so into the trunk it goes.  Changes in this commit include:

 - Remove the --with option for disabling MPI-2 onesided support.  It
   complicated code, and has no real reason for existing
 - add a framework osc (OneSided Communication) for encapsulating
   all the MPI-2 onesided functionality
 - Modify the MPI interface functions for the MPI-2 onesided chapter
   to properly call the underlying framework and do the required
   error checking
 - Created an osc component pt2pt, which is layered over the BML/BTL
   for communication (although it also uses the PML for long message
   transfers).  Currently, all support functions, all communication
   functions (Put, Get, Accumulate), and the Fence synchronization
   function are implemented.  The PWSC active synchronization
   functions and Lock/Unlock passive synchronization functions are
   still not implemented

This commit was SVN r8836.
2006-01-28 15:38:37 +00:00
Rainer Keller
60c2ae768b - Change the spacing preventing finding the struct from script.
This commit was SVN r8819.
2006-01-26 11:55:00 +00:00
George Bosilca
a2f31b2bf6 Use the MCA_PML_BASE_REQUEST_MPI_COMPLETE macro to set the completed state to true.
Add back the mutex protection around the size computations as they are not yet atomic
operations.

This commit was SVN r8812.
2006-01-25 23:17:17 +00:00
George Bosilca
58c9c82dab Add a macro to mark the MPI request as completed (MCA_PML_BASE_REQUEST_MPI_COMPLETE) and
broadcast the request condition if required. This macro should be called with the request's
mutex locked.

This commit was SVN r8811.
2006-01-25 23:15:36 +00:00
George Bosilca
0f1c6d79e8 Make the MVAPI BTL thread safe again. The problem was a double locking on the endpoint mutex.
It's still not very clean as we still lock the mvapi_btl mutex inside a critical section
protected by the endpoint mutex ...

This commit was SVN r8810.
2006-01-25 23:14:06 +00:00
Galen Shipman
ddc22d8c7e Use endpoint_lock, not ib_lock (copy paste error from openib btl)
This commit was SVN r8806.
2006-01-25 15:04:37 +00:00
Andrew Friedley
0ff8e47075 Checkpoint for the night:
- Updated copyrights
 - First pass at the uDAPL mpool component looks complete - is it really this easy?

This commit was SVN r8803.
2006-01-25 04:08:08 +00:00
Andrew Friedley
f5f8a1b874 Initial commit of uDAPL mpool component. Just a stripped copy of the GM mpool component right now.
This commit was SVN r8802.
2006-01-25 02:22:50 +00:00
Andrew Friedley
ec995160e6 Checkpoint for switch to mpool work:
- Remove printing of CFLAGS in configure.m4
 - Set MCA_BTL_FLAGS_SEND flag
 - Improved error handling during module initialization
 - Extract the address of each interface with dat_ia_query
 - Start playing around with fragment stuff - probably wrong
 - Misc code cleanup (removal of GM-specific code)

This commit was SVN r8801.
2006-01-25 02:21:34 +00:00
Tim Woodall
51ec050647 port of revised flow control from openib
This commit was SVN r8799.
2006-01-24 23:44:30 +00:00
Tim Woodall
e861158fcd - removed debug code
- removed extraneous memset

This commit was SVN r8798.
2006-01-24 23:38:41 +00:00
Galen Shipman
1e0ea9dd6d Major fixes for the RDMA registration cache (leave_pinned).
This commit fixes issues with HPL runs on node counts > 4. 

This commit was SVN r8793.
2006-01-23 22:51:50 +00:00