ports/openmpi
ports/openmpi
1
1
Форкнуть 0
Код Релизы Активность
12 163 коммитов 3 Ветки 0 Теги 130 MiB
9e0d64b9e5
Граф коммитов

1170 Коммитов

Автор SHA1 Сообщение Дата
Jeff Squyres
c40740947f Fix minor spelling error. 
This commit was SVN r18229.
 2008-04-22 13:11:50 +00:00
Galen Shipman
27c425b304 make portals level ack's optional (require ACK by default) 
This commit was SVN r18228.
 2008-04-21 22:22:18 +00:00
Ralph Castain
fa082cafa9 Shift the architecture calculation from the ompi/datatype engine to the opal/util area. This allows us to compute the architecture earlier in the launch and communicate it outside of the modex. 
Note: this is an early preliminary step in the movement of portions of the datatype engine to the opal layer.

This commit was SVN r18198.
 2008-04-17 20:43:56 +00:00
Adrian Knoth
d34dfbe12c fixed misleading comment. 
This commit was SVN r18170.
 2008-04-16 11:26:15 +00:00
Adrian Knoth
20473bfda2 on incoming connections, compare with every possible source address. 
Rational (taken from the code):

    /* This is PITA. We never know which source address an 
    * incoming/outgoing packet will have, so even with 
    * btl_tcp_if_include/exclude on the remote end, we 
    * might get a different source address. 
    * 
    * If this address isn't included in btl_proc->proc_addrs, 
    * we would erroneously drop the connection 
    */ 

merge -r18165:18167 to the trunk.

This commit was SVN r18169.

The following SVN revisions from the original message are invalid or
inconsistent and therefore were not cross-referenced:
  r18165
  r18167
 2008-04-16 11:24:09 +00:00
Adrian Knoth
e981a259bb btl_tcp_disable_family=4 and btl_tcp_disable_family=6 are mutually 
exclusive, so this should result in "unreachable" when set differently
between peers.

This commit was SVN r18168.
 2008-04-16 10:14:58 +00:00
Adrian Knoth
75c54616c7 renamed opal_sockaddr2str to opal_net_get_hostname for WANT_PEER_DUMP=1 
This commit was SVN r18154.
 2008-04-15 19:23:47 +00:00
Jeff Squyres
72af302360 Remove unused variable. 
This commit was SVN r18151.
 2008-04-15 14:58:32 +00:00
Aurelien Bouteiller
0f311ed824 Make sure the function returns NULL when no elan adapter is available instead of a random value. 
This commit was SVN r18136.
 2008-04-11 21:03:01 +00:00
Aurelien Bouteiller
20592cbcbf Fixes a warning about mallocing 0 bytes when no elan adapter is available. 
This commit was SVN r18135.
 2008-04-11 20:59:12 +00:00
Jon Mason
08ead87604 Potential double free of locks 
mca_btl_openib_endpoint_post_rr_nolock is freeing the endpoint lock on
the error case, but most/all of the functions calling this free the lock
regardless of its error case.  Thus resulting is a double free of the
lock.

This commit was SVN r18131.
 2008-04-10 21:15:01 +00:00
Donald Kerr
38e298cc9a report error message in all libs, not just debug 
This commit was SVN r18103.
 2008-04-08 22:58:28 +00:00
Gleb Natapov
713a27dc71 Counter of created RDMA channels should be incremented immediately after channel 
creation (not in control message completion) otherwise more than max_eager_rdma
channel may be created.

This commit was SVN r18082.
 2008-04-06 13:48:45 +00:00
Jeff Squyres
7072a32703 * Properly protect XRC stuff 
* A few minor style fixes

This commit was SVN r18076.
 2008-04-02 19:52:03 +00:00
George Bosilca
944453c4c1 Cleanups. 
This commit was SVN r18068.
 2008-04-02 06:37:42 +00:00
Jeff Squyres
d0f12f3df0 Make a better error message. 
This commit was SVN r18014.
 2008-03-29 12:54:24 +00:00
George Bosilca
be4b153f0d Another patch for thread safety in the TCP BTL (thanks to Pierre). 
This commit was SVN r17993.
 2008-03-27 18:36:08 +00:00
Jeff Squyres
5320c91ab3 Oops -- fix the constructor to also use opal_object_t instead of 
opal_list_item_t.

This commit was SVN r17945.
 2008-03-25 11:59:50 +00:00
Jeff Squyres
ebfdd133f5 AFACT, we never put endpoints on a list. 
This commit was SVN r17940.
 2008-03-24 18:32:55 +00:00
Ralph Castain
dc7f45dafd Remove the obsolete and largely unused orte_system_info structure. The only fields that were used in that struct were nodeid and nodename - these have been transferred to the orte_process_info structure. 
Only one place used the user name field - session_dir, when formulating the name of the top-level directory. Accordingly, the code for getting the user's id has been moved to the session_dir code.

This commit was SVN r17926.
 2008-03-23 23:10:15 +00:00
Galen Shipman
dcac824f59 Fix problem in releasing fragments during GET_END event (didn't check that 
portals btl has ownership and therefor didn't free the frag as it should) this
causes leakage and hangs in MPI_Finalize. 

Also added a bit more debugging. 

This commit was SVN r17900.
 2008-03-20 22:46:32 +00:00
George Bosilca
1d04ec4ded Correct the connection logic for TCP. Now we have not only a cleaner 
connection, but a more thread safe one. Thanks to Pierre for his
help on this.

This commit was SVN r17853.
 2008-03-18 02:42:16 +00:00
Gleb Natapov
9b6db25182 Fix compilation warning. 
This commit was SVN r17839.
 2008-03-17 13:37:57 +00:00
Pavel Shamis
54ad8d7446 The issue was reported/fixed by Jon Mason one month ago but the fix was not committed. So I'm commiting it now. 
This commit was SVN r17835.
 2008-03-17 11:13:06 +00:00
Brad Penoff
be13b86fc5 Clarifying and fixing SCTP btl_sctp_if_11 parameter 
This commit was SVN r17834.
 2008-03-17 09:18:31 +00:00
Gleb Natapov
f488b94899 More SM BTL initialization cleanups. 
This commit was SVN r17833.
 2008-03-16 10:01:56 +00:00
Jeff Squyres
6c77c995c2 Add missing dependencies in the static build case. 
This commit was SVN r17825.
 2008-03-15 12:11:36 +00:00
George Bosilca
5e229fe688 Thanks Ma for the patch. Correct the multi-rail support and 
rename some fields to something more clear.

This commit was SVN r17824.
 2008-03-14 19:17:28 +00:00
George Bosilca
ecebd5ae77 Update the Elan BTL to take in account multiple networks, and correctly deal 
with the node position in the network.

This commit was SVN r17822.
 2008-03-14 17:32:35 +00:00
Gleb Natapov
772772b944 Remove unneeded include. 
This commit was SVN r17813.
 2008-03-12 10:01:20 +00:00
Gleb Natapov
90c70e37b9 Clean up SM btl startup code. Remove no longer needed code leftovers from two 
BTL times. Remove old and no longer correct comment.

This commit was SVN r17805.
 2008-03-11 14:39:10 +00:00
Gleb Natapov
ffa09c44fd Pass correct pointer to mpool_base function. 
This commit was SVN r17795.
 2008-03-09 13:22:12 +00:00
Gleb Natapov
b0b21c68b4 Remove trailing spaces from SM BTL. 
This commit was SVN r17794.
 2008-03-09 13:17:13 +00:00
Tim Prins
5de3e1965e Remove the orte_proc_table. Migrate all users of it to the opal_hash_table and a new name hash function in orte. 
Everything should work, however I am unable to compile and test the sctp BTL.

This commit was SVN r17751.
 2008-03-05 22:44:35 +00:00
Donald Kerr
ef8f807c1c was not passing correct variable to dat_strerror 
This commit was SVN r17749.
 2008-03-05 21:45:16 +00:00
Jeff Squyres
ea5c0cb4a2 Now that the nightly tarball has safely been made, let's try this 
commit again.  Remove the svn:ignore from problematic directories and
try a merge from /tmp-public/plpa-merge-area2.

This commit was SVN r17718.
 2008-03-05 02:45:15 +00:00
Jeff Squyres
8189fcc7d5 Back out r17702; it went very badly. 
This commit was SVN r17704.

The following SVN revision numbers were found above:
  r17702 --> open-mpi/ompi@3df754ebd7
 2008-03-05 00:42:39 +00:00
Jeff Squyres
3df754ebd7 Bring over PLPA v1.1 from /tmp-public/plpa-v1.1 branch. 
This commit was SVN r17702.
 2008-03-05 00:16:49 +00:00
Christian Bell
c3d0a81cd3 Add new QLogic adapters to hca-params.init 
This commit was SVN r17699.
 2008-03-04 22:14:27 +00:00
Gleb Natapov
08abafdaa1 Initialize ib_pd to NULL. 
This commit was SVN r17674.
 2008-03-02 09:11:23 +00:00
Tim Prins
84b2099fe8 Remove the now-unused orte_value_array. As this is the last 'class' split between orte and ompi, remove the big comment about the split in ompi_bitmap. 
Also, update some properties (source files should not be executeable...), and remove a couple unneeded inclusions of orte_proc_table.h

This commit was SVN r17655.
 2008-02-28 21:39:42 +00:00
Ralph Castain
d70e2e8c2b Merge the ORTE devel branch into the main trunk. Details of what this means will be circulated separately. 
Remains to be tested to ensure everything came over cleanly, so please continue to withhold commits a little longer

This commit was SVN r17632.
 2008-02-28 01:57:57 +00:00
Galen Shipman
44003a41f2 Update common_portals to allow using portals interconnect with a modex rather 
than relying on cnos to get the nid/pid map. 

This commit was SVN r17588.
 2008-02-25 19:17:21 +00:00
Brian Barrett
bc8d863ce3 * Make Portals BTL compile again (looks like the frag ownership stuff didn't 
get copied well)
* Clean up a bunch of warnings

This commit was SVN r17562.
 2008-02-23 01:45:36 +00:00
Donald Kerr
437e280829 removing a few superfluous casts when the base or super is available 
This commit was SVN r17554.
 2008-02-22 20:10:55 +00:00
Donald Kerr
fe51084d8e fix compile warning by casting btl udapl module to base module before call to mca_btl_udapl_free 
This commit was SVN r17541.
 2008-02-21 16:19:06 +00:00
Pierre Lemarinier
2a99f89631 Modification of the mutex lock order to prevent races during connection stage. 
This commit was SVN r17535.
 2008-02-20 18:17:58 +00:00
Pavel Shamis
a0d12a9c92 Adding support for APM over different ports 
This commit was SVN r17521.
 2008-02-20 13:44:05 +00:00
Gleb Natapov
60c151608c Set flags inside fragment allocation function. 
This commit was SVN r17508.
 2008-02-19 12:26:45 +00:00
Nysal Jan
479f36adfc Fix a SEGV on ppc64. size_t is 8 bytes on a 64-bit build 
This commit was SVN r17507.
 2008-02-19 11:01:21 +00:00
Jeff Squyres
f22f62ef1f Fix typos. 
This commit was SVN r17502.
 2008-02-18 21:26:21 +00:00
Jeff Squyres
33a4aff18e Make openib btl a bit more resillient in the face of driver errors -- 
return OMPI_ERR_UNREACH if the port returns an invalid speed or
width.  OMPI_ERR_VALUE_OUT_OF_BOUNDS is reserved for when we exceed
the number of allowable BTLs.

This commit was SVN r17500.
 2008-02-18 20:28:06 +00:00
George Bosilca
7a21d77b29 Remove some compilation warnings. 
This commit was SVN r17498.
 2008-02-18 18:55:32 +00:00
George Bosilca
fa31ec81d0 Add the ownership flags to the PML/BTL interface. The layer 
owning the descriptor is responsible for releasing it once
the descriptor is not in use anymore.

This commit was SVN r17497.
 2008-02-18 17:39:30 +00:00
George Bosilca
be2579467a With the new ompi_free_list this is not needed anymore. 
This commit was SVN r17465.
 2008-02-15 03:22:16 +00:00
Donald Kerr
58bf7f5a1d add uintptr_t to prevent the possibility of a signed extension occuring 
This commit was SVN r17456.
 2008-02-14 19:16:34 +00:00
Jeff Squyres
6420db7088 Add missing header file that caused compilation errors in the 
rhc-step2b branch last night.

This commit was SVN r17453.
 2008-02-14 14:10:27 +00:00
George Bosilca
255cd2186b Improve the performance of the MX BTL. Correct the fake PUT 
protocol.

This commit was SVN r17452.
 2008-02-14 04:38:55 +00:00
Adrian Knoth
f1648f08df Advanced address selection code from Thomas Peiselt. Re #1207, #1027 
This commit was SVN r17450.
 2008-02-13 21:53:00 +00:00
Sharon Melamed
5b2dab2439 Reverted commit # r17443 
This commit was SVN r17446.

The following SVN revision numbers were found above:
  r17443 --> open-mpi/ompi@88ce5a2b73
 2008-02-13 14:07:12 +00:00
Sharon Melamed
88ce5a2b73 Replaced PLPA to the latest PLPA (plpa-1.1a3r123) 
This commit was SVN r17443.
 2008-02-13 13:09:11 +00:00
Rainer Keller
7621800477 - Fix and add comments -- output full name for pd 
- Protect argument in macro...

This commit was SVN r17434.
 2008-02-12 16:59:59 +00:00
Gleb Natapov
cf801edfe5 Use carto topology framework to choose which HCAs to use. 
This commit was SVN r17414.
 2008-02-11 10:34:11 +00:00
George Bosilca
ee321748a6 The lost space. 
This commit was SVN r17413.
 2008-02-10 22:08:49 +00:00
Pavel Shamis
df787bbeab Fixing compilation issue on machines with ofed under 1.3. 
Also finx in apm migration flow.

This commit was SVN r17383.
 2008-02-06 13:54:58 +00:00
Pavel Shamis
3ba3f70624 Adding apm support for xrc. 
This commit was SVN r17382.
 2008-02-06 10:19:51 +00:00
Gleb Natapov
03c80bdfe3 Fix old libiverbs case. 
This commit was SVN r17370.
 2008-02-04 14:05:01 +00:00
Pavel Shamis
f0c478e7e0 XRC - replacing the new old API with new one. 
This commit was SVN r17369.
 2008-02-04 14:03:38 +00:00
Gleb Natapov
67f752dd50 Add compatibility function between old libibverbs and current libibverbs 
way of detecting HCAs.

This commit was SVN r17365.
 2008-02-03 15:16:24 +00:00
George Bosilca
3a6d2e3894 The latest and greatest Elan improvements. 
This commit was SVN r17361.
 2008-02-01 21:29:57 +00:00
Gleb Natapov
f73adf69c0 Fix compiler warnings on 32bit systems. 
This commit was SVN r17346.
 2008-01-31 09:05:25 +00:00
Adrian Knoth
8ae4a10b4c Reverted r17331, r17332. Still broken. I'm in a bad hurry. :-( Re #1206 
This commit was SVN r17333.

The following SVN revision numbers were found above:
  r17331 --> open-mpi/ompi@3846e2a797
  r17332 --> open-mpi/ompi@c03de08c55
 2008-01-30 16:51:55 +00:00
Adrian Knoth
c03de08c55 Logic is wrong. I'm going to revert it again. Re #1206 
This commit was SVN r17332.
 2008-01-30 16:48:50 +00:00
Adrian Knoth
3846e2a797 When checking incoming connections, also care about aliased interfaces. 
Re #1206

This commit was SVN r17331.
 2008-01-30 16:45:41 +00:00
Adrian Knoth
7f79c68930 Reverted r17307 and r17308. It broke parallel TCP connections. Re #1206 
This commit was SVN r17329.

The following SVN revision numbers were found above:
  r17307 --> open-mpi/ompi@7a59b3f58c
  r17308 --> open-mpi/ompi@72b29bc21f
 2008-01-30 14:31:47 +00:00
Adrian Knoth
72b29bc21f Cosmetic patch. Use IN6_ARE_ADDR_EQUAL instead of memcmp(). Re #1206. 
This commit was SVN r17308.
 2008-01-29 16:02:24 +00:00
Adrian Knoth
7a59b3f58c accept incoming connections from hosts with multiple addresses. 
We loop over all peer addresses and accept when one of them matches.
Note that this might break functionality: mca_btl_tcp_proc_insert now
always inserts the same endpoint. (is the lack of endpoints the problem?
should there be one for every remote address?)

Re #1206

This commit was SVN r17307.
 2008-01-29 15:55:56 +00:00
Pavel Shamis
7b59f8ae0b Fixing warning in apm code. 
This commit was SVN r17306.
 2008-01-29 15:45:18 +00:00
Gleb Natapov
bb03e07ec4 Move eager RDMA channels accounting into completion callback. Otherwise it 
can go wrong with XRC as endpoint may be not yet connected at the time
eager rdma channel is created.

This commit was SVN r17302.
 2008-01-29 14:35:33 +00:00
Pavel Shamis
92ef832472 Making sure that XRC will not overrun ib_dev_attr.max_qp_wr 
This commit was SVN r17300.
 2008-01-29 13:15:21 +00:00
Pavel Shamis
7d83f34eb0 Protecting the apm code with OMPI_HAVE_THREADS. 
This commit was SVN r17284.
 2008-01-28 16:10:18 +00:00
Jeff Squyres
6a49c97368 Remove erroneous #if 
This commit was SVN r17282.
 2008-01-28 14:38:03 +00:00
Pavel Shamis
28a3917306 Adding APM support (over different lids). 
This commit was SVN r17280.
 2008-01-28 10:38:08 +00:00
George Bosilca
3418485085 Replace the tport by a queue. 
This commit was SVN r17221.
 2008-01-25 01:15:18 +00:00
Donald Kerr
66acac8ff3 the value for invalid idx was just plain wrong, a more appropriate value is now used 
This commit was SVN r17201.
 2008-01-24 15:01:26 +00:00
Jeff Squyres
2227d5ec4a Add configure check for struct ibv_device.transport type, which was added in OFED v1.2. Still need to fix up oob and rdma_cm cpc's to do something better with this information... 
This commit was SVN r17198.
 2008-01-24 12:14:21 +00:00
Gleb Natapov
52c94fa7ea Fix compilation warnings. 
This commit was SVN r17169.
 2008-01-21 15:07:39 +00:00
Gleb Natapov
c9a1b06771 Remove trailing whitespaces. No code changes in this commit. 
This commit was SVN r17167.
 2008-01-21 12:11:18 +00:00
George Bosilca
170416797d This commit was SVN r17162. 2008-01-18 20:10:57 +00:00
George Bosilca
0081202195 Mark the receives as ELAN_TPORT_RXBUF | ELAN_TPORT_RXANY ... 
This commit was SVN r17161.
 2008-01-18 20:00:44 +00:00
George Bosilca
bf299bb833 Keep most of the functions as static. Improve the progress function. Get rid 
of all internal quues that are not really useful.

This commit was SVN r17160.
 2008-01-18 19:28:50 +00:00
Donald Kerr
5f884b1ca4 fix for #1130 - adds support for multi-rail configurations 
This commit was SVN r17152.
 2008-01-17 17:30:50 +00:00
Donald Kerr
908b514ac5 update use of internal tag values to accommodate the active message change found in r17140 
This commit was SVN r17148.

The following SVN revision numbers were found above:
  r17140 --> open-mpi/ompi@6310ce955c
 2008-01-16 21:17:25 +00:00
Pavel Shamis
add4d9df8a XRC fixes for MPI2 dynamics. 
This commit was SVN r17144.
 2008-01-15 21:14:48 +00:00
Jeff Squyres
251842ff6a Remove this AS_IF -- it breaks "make dist". 
This commit was SVN r17143.
 2008-01-15 12:33:08 +00:00
George Bosilca
e8ac5ff04d Typos. 
This commit was SVN r17141.
 2008-01-15 05:37:42 +00:00
George Bosilca
6310ce955c The first patch related to the Active Message stuff. So far, here is what we have: 
- the registration array is now global instead of one by BTL.
- each framework have to declare the entries in the registration array reserved. Then
  it have to define the internal way of sharing (or not) these entries between all
  components. As an example, the PML will not share as there is only one active PML
  at any moment, while the BTLs will have to. The tag is 8 bits long, the first 3
  are reserved for the framework while the remaining 5 are use internally by each
  framework.
- The registration function is optional. If a BTL do not provide such function,
  nothing happens. However, in the case where such function is provided in the BTL
  structure, it will be called by the BML, when a tag is registered.

Now, it's time for the second step... Converting OB1 from a switch based PML to an
active message one.

This commit was SVN r17140.
 2008-01-15 05:32:53 +00:00
Jon Mason
a0d4122606 The new cpc selection framework is now in place. The patch below allows 
for dynamic selection of cpc methods based on what is available.  It
also allows for inclusion/exclusions of methods.  It even futher allows
for modifying the priorities of certain cpc methods to better determine
the optimal cpc method.

This patch also contains XRC compile time disablement (per Jeff's
patch).

At a high level, the cpc selections works by walking through each cpc
and allowing it to test to see if it is permissable to run on this
mpirun.  It returns a priority if it is permissable or a -1 if not.  All
of the cpc names and priorities are rolled into a string.  This string
is then encapsulated in a message and passed around all the ompi
processes.  Once received and unpacked, the list received is compared
to a local copy of the list.  The connection method is chosen by
comparing the lists passed around to all nodes via modex with the list
generated locally.  Any non-negative number is a potentially valid
connection method.  The method below of determining the optimal
connection method is to take the cross-section of the two lists.  The
highest single value (and the other side being non-negative) is selected
as the cpc method.

svn merge -r 16948:17128 https://svn.open-mpi.org/svn/ompi/tmp-public/openib-cpc/ .

This commit was SVN r17138.
 2008-01-14 23:22:03 +00:00
Pavel Shamis
6e50fca2dd Fixing permissions for XRC domain file. 
This commit was SVN r17127.
 2008-01-13 19:23:11 +00:00
Jon Mason
626e0814a2 Style clean-up 
This commit was SVN r17126.
 2008-01-12 18:47:17 +00:00
Jon Mason
3970c3ff6c Add Chelsio T3 to ompi/mca/btl/openib/mca-btl-openib-hca-params.ini 
This commit was SVN r17101.
 2008-01-09 22:14:18 +00:00
Jon Mason
597c7e68f1 Minor cleanups 
This commit was SVN r17100.
 2008-01-09 21:54:11 +00:00
Rolf vandeVaart
870fa8b1f1 Pad the sm btl header to double-word alignment. Preserves PML 
header as double-word aligned and prevents bus errors on SPARC
based servers.  This is part of fix for #1148.

Refs trac:1148

This commit was SVN r17090.

The following Trac tickets were found above:
  Ticket 1148 --> https://svn.open-mpi.org/trac/ompi/ticket/1148
 2008-01-09 18:50:51 +00:00
Gleb Natapov
25ce70bb92 Call mca_btl_openib_endpoint_post_send() holding endpoint lock and not holding 
qp lock since this is what the function assumes.

This commit was SVN r17086.
 2008-01-09 14:46:41 +00:00
Pavel Shamis
99f51482e3 Fixing openib finalization flow. 
This commit was SVN r17085.
 2008-01-09 12:36:30 +00:00
Gleb Natapov
51d6ca0cb6 Provide no lock version of mca_btl_openib_endpoint_post_rr(). On connection 
creation we call it with endpoint lock already held.

This commit was SVN r17084.
 2008-01-09 10:39:35 +00:00
Gleb Natapov
50af6b9e78 Rearrange functions order so that functions are defined before they are used. No 
code changes here.

This commit was SVN r17083.
 2008-01-09 10:27:15 +00:00
Gleb Natapov
621fa223c5 Create free lists of fragments per HCA, not per BTL. Saves memory in case of 
multiple LMCs.

This commit was SVN r17082.
 2008-01-09 10:26:21 +00:00
Gleb Natapov
5ce3213158 Rearrange functions order so that functions are defined before they are used. No 
code changes here.

This commit was SVN r17081.
 2008-01-09 10:05:41 +00:00
Pavel Shamis
fbf7bcd9a9 We need to prepost on srq/xrc before reply with ENDPOINT_XOOB_CONNECT_XRC_RESPONSE. 
This commit was SVN r17066.
 2008-01-08 10:30:16 +00:00
Rolf vandeVaart
0f0fde3490 Partial fix for #1148. Enable this for 32-bit sparc as well as 64-bit sparc. 
This commit was SVN r17059.
 2008-01-07 15:43:44 +00:00
Gleb Natapov
c3bbf69356 Set send_flags correctly in btl_openib_put. Otherwise we may reuse flags from 
previous use of the buffer and they may be incorrect.

This commit was SVN r17058.
 2008-01-07 10:19:07 +00:00
George Bosilca
48f5a26e8c Cast to keep VC happy (quiet). 
This commit was SVN r17054.
 2008-01-04 23:13:32 +00:00
Jeff Squyres
a234ba198a Remove superflous / unused -D from Makefile.am. 
This commit was SVN r17030.
 2008-01-02 18:00:20 +00:00
Jeff Squyres
c9bea80f8f Fix unbalanced parenthesees noticed by Paul Hargove. 
This commit was SVN r17029.
 2008-01-02 13:34:07 +00:00
Gleb Natapov
2fb6947f88 Destroy endpoints that use eager rdma communication before destroying SRQ. Do't 
skip async event thread destruction if SRQ was not destroyed, or it will segfault
on module removal.

This commit was SVN r17025.
 2007-12-23 13:58:31 +00:00
Gleb Natapov
b06d92bdab OpenIB BTL has three channels through which data can be received (eager rdma, 
high prio QPs and low prio QPs) and because not all of them are polled each time
progrgess() is called (to save on latency) starvation is possible. The commit
fixes this. Now each channel is polled, but higher priority channels are polled
more often. Three new parameters are introduced that control polling ratios 
between different channels.

This commit was SVN r17024.
 2007-12-23 12:29:34 +00:00
Brad Penoff
4c2571b54c fixed more 64 bit SCTP BTL warnings 
This commit was SVN r17022.
 2007-12-21 21:50:00 +00:00
Brad Penoff
195faa37b6 fixed send side of 64 bit compilation warnings 
This commit was SVN r17019.
 2007-12-21 19:11:50 +00:00
Jeff Squyres
558d179e2e Fix typo. 
This commit was SVN r17012.
 2007-12-21 14:25:48 +00:00
George Bosilca
906e8bf1d1 Replace the ompi_pointer_array with opal_pointer_array. The next step 
(sometimes after the merge with the ORTE branch), the opal_pointer_array
will became the only pointer_array implementation (the orte_pointer_array
will be removed).

This commit was SVN r17007.
 2007-12-21 06:02:00 +00:00
Tim Mattox
bbeef5b84b Change the MX BTL's exclusivity to MCA_BTL_EXCLUSIVITY_DEFAULT, 
so that it is higher than the new TCP BTL exclusivity as of r16942.

The portals BTL maintainer may want to do the same...

This commit was SVN r16995.

The following SVN revision numbers were found above:
  r16942 --> open-mpi/ompi@80e9730100
 2007-12-19 21:24:45 +00:00
Pavel Shamis
fcbca510d8 The ib_inline_max should be updated only when SEND qp is created. 
This commit was SVN r16973.
 2007-12-17 10:30:30 +00:00
Gleb Natapov
f79e344ea4 Fix bug in debug build. 
This commit was SVN r16972.
 2007-12-17 10:26:18 +00:00
Gleb Natapov
64a95f63cd Fix error reporting in openib if parameter value is out of range. 
This commit was SVN r16971.
 2007-12-16 14:04:36 +00:00
Gleb Natapov
8b511b969d Introduce a new BTL parameter btl_rndv_eager_limit which determines size of a 
first fragment of rendezvous protocol. Remove no longer used btl_min_send_size
parameter.

This commit was SVN r16969.
 2007-12-16 08:35:17 +00:00
Jeff Squyres
213b5d5c6e Per long threads on the mailing list and much confusion discussion 
about linkers, have all OPAL, ORTE, and OMPI components '''not'' link
against the OPAL, ORTE, or OMPI libraries.

See ttp://www.open-mpi.org/community/lists/users/2007/10/4220.php for
details (or https://svn.open-mpi.org/trac/ompi/wiki/Linkers for a
better-formatted version of the same info).

This commit was SVN r16968.
 2007-12-15 13:32:02 +00:00
Brad Penoff
540d483dd3 64 bit fix and initial Solaris support 
This commit was SVN r16967.
 2007-12-15 03:28:10 +00:00
Donald Kerr
d05d3afaed clean up and make consistent the reporting out from the udapl btl; report out readeable event string instead of just a number 
This commit was SVN r16954.
 2007-12-13 15:32:26 +00:00
Brad Penoff
ecd563b0fa reduced noise for SCTP BTL on RHEL4U4 
This commit was SVN r16951.
 2007-12-13 03:15:29 +00:00
Jeff Squyres
80e9730100 Per http://www.open-mpi.org/community/lists/devel/2007/12/2698.php and 
this thread:
http://www.open-mpi.org/community/lists/devel/2007/12/2807.php, set
TCP's exclusivity to LOW+100 and SCTP's exclusivity to LOW.

This commit was SVN r16942.
 2007-12-12 15:55:37 +00:00
Jon Mason
e05cd7b0e4 To modify the default connection method, a "btl_openib_connect <arg>" 
should be passed via commandline.  However, there is a slight coding
bug in the openib connect code.  When registering the name of the
option, mca_base_param_reg_string will prepend the relevant info
("btl_openib_" in this case).  The existing code will require
"btl_openib_btl_openib_connect" instead of "btl_openib_connect".
This patch corrects this.

This commit was SVN r16937.
 2007-12-11 20:36:36 +00:00
Galen Shipman
a04d21b459 Make CNL compile again.. 
This commit was SVN r16929.
 2007-12-11 16:14:30 +00:00
Gleb Natapov
2a59b2a68f 1. Set segments length in prepare_src() after packing because actual size may be 
smaller then allocated size.

2. If reserve zero don't allocate coalesced frag since it will be RDMAed, not
send.  The logic was other way around.

This commit was SVN r16928.
 2007-12-11 13:10:52 +00:00
Jon Mason
df82fcb917 Slight word usage and grammar error in the openib btl help test. I 
believe the change below is the intended meaning.

This commit was SVN r16921.
 2007-12-10 21:50:48 +00:00
Donald Kerr
a604fca52c follow on change to r16901 and r16898; the interface change mca_btl_udapl_alloc() was not applied to two locations in this file 
This commit was SVN r16918.

The following SVN revision numbers were found above:
  r16898 --> open-mpi/ompi@7364b7cf47
  r16901 --> open-mpi/ompi@e2e211f23b
 2007-12-10 18:10:52 +00:00
Gleb Natapov
17611dafbe Fix pointer casting on 32bit machines. 
This commit was SVN r16907.
 2007-12-09 14:15:35 +00:00
Gleb Natapov
2f9c5b46cf Return OMPI_ERR_RESOURCE_BUSY from openib_btl_send() if fragment is not on wire. 
This commit was SVN r16906.
 2007-12-09 14:14:11 +00:00
Gleb Natapov
493951e09d Add heterogeneous support to message coalescing. 
This commit was SVN r16903.
 2007-12-09 14:10:25 +00:00
Gleb Natapov
b4698dc6df Use flags provided during allocation to coalesce to correct priority queue. 
This commit was SVN r16902.
 2007-12-09 14:08:55 +00:00
Gleb Natapov
e2e211f23b Add flags parameter to btl_alloc() and btl_prepare_src() functions. If BTL 
knows at the time of allocation priority of a descriptor it may do some
optimizations.

This commit was SVN r16901.
 2007-12-09 14:08:01 +00:00
Gleb Natapov
5313a2baa7 Message coalescing for openib BTL. If fragment is waiting to be transmitted in 
a pending queue pack another message into it if there is enough space there.

This commit was SVN r16900.
 2007-12-09 14:05:13 +00:00
Gleb Natapov
7302cd24eb Call btl_alloc() from btl_prepare_src() to have one point of frag allocation. 
This commit was SVN r16899.
 2007-12-09 14:02:32 +00:00
Gleb Natapov
7364b7cf47 Add endpoint parameter to btl_alloc() function. Enables various optimizations 
inside BTL.

This commit was SVN r16898.
 2007-12-09 14:00:42 +00:00
Gleb Natapov
de3761208a Send cm_seen by eager rdma channel. Encode qp index into credits filed. If 
cm_seen is not send here non symmetric eager rdma connection may hang.

This commit was SVN r16896.
 2007-12-09 13:56:13 +00:00
Tim Mattox
d188642715 Apparently the SCTP BTL has a btl_sctp_component.h file that needs to be 
part of the "sources" list.  Hopefully this will clear of the nightly
tarball creation for the trunk.

This commit was SVN r16895.
 2007-12-08 04:05:59 +00:00
Karl Mroz
71b54d8e4e Removed .ompi_ignore and .ompi_unignore from SCTP BTL. 
This commit was SVN r16893.
 2007-12-07 17:02:32 +00:00
Jon Mason
20294e7800 There is a double call to ompi_btl_openib_connect_base_open in 
mca_btl_openib_mca_setup_qps().  It looks like someone just forgot to
clean-up the previous call when they added the check for the return
code.

I ran a quick IMB test over IB to verify everything is still working.

This commit was SVN r16870.
 2007-12-06 17:25:38 +00:00
Pavel Shamis
e8aeadb11e XRC fixes: 
- create separate xrc domain file for each hca
- return error if we failed to create xrc file.

This commit was SVN r16853.
 2007-12-05 14:32:44 +00:00
Pavel Shamis
f60ca0e4e5 Removing unused mca_btl_openib_ib_address_status 
This commit was SVN r16835.
 2007-12-04 13:16:26 +00:00
Pavel Shamis
57728986f8 Fixing XRC multiport/multisubnet support. 
This commit was SVN r16819.
 2007-12-03 09:49:53 +00:00
Gleb Natapov
b2858236fb Use new free list interface. 
This commit was SVN r16818.
 2007-12-02 15:13:11 +00:00
Gleb Natapov
a774cd98f8 Put send completions to low prio CQ. Receive is more important. 
This commit was SVN r16817.
 2007-12-02 14:46:37 +00:00
Gleb Natapov
b17f5b7480 Change how default receive queues parameters are calculated. Current default 
parameters don't make any sense. Credits are never piggybacked. Also make
default queue sizes to be calculated from eager_limit and max_send_size values.

This commit was SVN r16816.
 2007-12-02 14:43:28 +00:00
Rich Graham
6e77414a68 changes to the ompi_free_list_ex - called ompi_free_list_ex_new, for now. 
This commit was SVN r16803.
 2007-11-29 21:18:37 +00:00
Jeff Squyres
8c0060701c Stub out the ibcm CPC. 
This commit was SVN r16800.
 2007-11-29 13:23:17 +00:00
Pavel Shamis
8aca6eb31b OFED 1.3 doesn't implement ibv_resize_cq for connectX. 
On error exit from ibv_resize_cq we should to check if the function
is implemented.

This commit was SVN r16799.
 2007-11-28 15:23:19 +00:00
Gleb Natapov
5f242c77f2 Post each recv wr not separately but in one call to ibv_post_recv(). 
This commit was SVN r16798.
 2007-11-28 14:57:15 +00:00
Gleb Natapov
14cffee726 Uninline mca_btl_openib_post_srr() function. 
This commit was SVN r16797.
 2007-11-28 14:52:31 +00:00
Pavel Shamis
1c314ef4c3 If XRC qp was specified in btl_openib_receive_queues we automatically should 
choose xoob connection module.

This commit was SVN r16796.
 2007-11-28 10:33:32 +00:00
Pavel Shamis
488a508732 Removing comments from help file. 
This commit was SVN r16795.
 2007-11-28 10:16:08 +00:00
Pavel Shamis
3e2e4f6d2a Removing unused lid. 
This commit was SVN r16794.
 2007-11-28 10:06:57 +00:00
Pavel Shamis
aa79bdabc8 Removing port_touse - we don't really need it 
This commit was SVN r16793.
 2007-11-28 09:57:48 +00:00
Pavel Shamis
2ffbe8776a Fixing compilation problems in openib 
This commit was SVN r16792.
 2007-11-28 09:38:49 +00:00
Gleb Natapov
218adb2a96 Account for eager rdma credit fragments when creating send queue. Create XRC 
receive QP with zero receive and send queue length. We don't going to use this
QP for send and receives a posted to SRQs.

This commit was SVN r16791.
 2007-11-28 07:22:01 +00:00
Gleb Natapov
601952a952 Don't shared endpoint->qps array, only pointer to actual QP. Calculate send 
queue size for shared QP based on all endpoints that want to use it.

This commit was SVN r16790.
 2007-11-28 07:21:07 +00:00
Gleb Natapov
b46c9cc7bc Make xrc use srq_qp unions instead of the xrc_qp which is exactly like srq_qp. 
This commit was SVN r16789.
 2007-11-28 07:20:26 +00:00
Gleb Natapov
be0981fc07 Change a type of xrc_recv_qp to "struct ibv_qp". 
This commit was SVN r16788.
 2007-11-28 07:19:36 +00:00
Gleb Natapov
bd47da4699 Initial XRC support by Mellanox. 
This commit was SVN r16787.
 2007-11-28 07:18:59 +00:00
Gleb Natapov
b49788c499 Receive queue is not used in case of SRQ QP, so don't create one. 
This commit was SVN r16786.
 2007-11-28 07:17:22 +00:00
Gleb Natapov
923666b75c Process pending put/get frags on endpoint connection establishment. 
This commit was SVN r16785.
 2007-11-28 07:16:52 +00:00
Gleb Natapov
e502402470 Fix endpoint destructor to not skip closed endpoints. 
This commit was SVN r16784.
 2007-11-28 07:15:54 +00:00
Gleb Natapov
5a4e953aaa Allow share the same qp for different buffer sizes. Needed for XRC support. 
This commit was SVN r16783.
 2007-11-28 07:15:20 +00:00
Gleb Natapov
b123696d57 Fix async thread creation and destruction. Create async thread only when it is 
needed instead of creating it and then canceling if it is not needed. Change
error handling during finalize so that it will not skip async thread
destruction. Otherwise async thread may segfault during openib module unloading.

This commit was SVN r16782.
 2007-11-28 07:14:34 +00:00
Gleb Natapov
5463eb892c Send all explicit credits for PP QPs of all orders over smallest PP qp. 
This commit was SVN r16781.
 2007-11-28 07:13:34 +00:00
Gleb Natapov
a9f864d15c If there is an eager rdma credit, but there is no WQE to send a packet we add it 
to a pending queue of eager rdma QP instead of correct pending list. This patch
fixes this by getting reed of "eager rdma qp" notion. Packet is always send
over its order QP. The patch also adds two pending queues for high and low prio
packets. Only high prio packets are sent over eager RDMA channel.

This commit was SVN r16780.
 2007-11-28 07:12:44 +00:00
Gleb Natapov
6a2d210b7d Use OMPI object system to make fragment hierarchy more object oriented. The 
main idea (except of cleanup) is to save on initialisation of unneeded fields
and to use C type checking system to catch obvious errors.

This commit was SVN r16779.
 2007-11-28 07:11:14 +00:00
Gleb Natapov
267cd2342a Cleanup. Remove unused functions. 
This commit was SVN r16778.
 2007-11-28 07:08:56 +00:00
Ron Brightwell
924414f92f Added support for Accelerated Portals for the btl. 
This commit was SVN r16771.
 2007-11-21 21:34:17 +00:00
Brad Penoff
fb5536f11d conforming SCTP BTL to Open MPI naming conventions and IP requirements 
This commit was SVN r16764.
 2007-11-21 10:13:41 +00:00
Andrew Friedley
c50f2aa74c fix warning 
This commit was SVN r16759.
 2007-11-20 16:55:12 +00:00
Brad Penoff
ede8a6a7a1 adjusting for Linux when sctp_recvmsg returns 0 for remote close 
This commit was SVN r16742.
 2007-11-20 06:02:08 +00:00
Tim Prins
f42fcd36db make the mx btl compile again after the free list changes 
This commit was SVN r16735.
 2007-11-19 19:41:22 +00:00
Brad Penoff
f34ddfef80 for SCTP BTL, added Mac OS X support for systems using SCTP NKE (Network Kernel Extension) 
This commit was SVN r16729.
 2007-11-17 02:56:27 +00:00
Brad Penoff
5abd2d8064 initial SCTP BTL commit 
This commit was SVN r16723.
 2007-11-13 23:39:16 +00:00
Jeff Squyres
a4d571f8ad Fix typo that broke the build. 
This commit was SVN r16635.
 2007-11-02 09:19:55 +00:00
Rich Graham
27a748e7eb change all instances of ompi_free_list_init to ompi_free_list_init_new. Header 
and payload data are specified separately at this stage.

This commit was SVN r16633.
 2007-11-01 23:38:50 +00:00
Andrew Friedley
46516d98e1 Update MCA params -- sd_num_peer is no longer used, change rd_num_init to rd_num 
This commit was SVN r16601.
 2007-10-29 22:56:30 +00:00
Andrew Friedley
8273b61471 Bugfix for hangs in certain communication patterns, particularly alltoall. 
This commit was SVN r16600.
 2007-10-29 21:51:28 +00:00
George Bosilca
d67c0eefb4 Remove a compilation warning about using uninitialized variables. 
This commit was SVN r16589.
 2007-10-26 20:15:28 +00:00
George Bosilca
b1b5cb6453 Looks like SO_REUSEPORT it's not defined on some platforms. Switch 
to the conventional SO_REUSEADDR instead.

This commit was SVN r16588.
 2007-10-26 19:56:21 +00:00
George Bosilca
337f78a4a8 Restrict the port range for the OOB and the BTL. Each protocols (v4 and v6) 
has his own range which is defined by a min value and a range. By default
there is no limitation on the port range, which is exactly the same
behavior as before.

This commit was SVN r16584.
 2007-10-26 16:36:51 +00:00
Gleb Natapov
3a63eb6c17 Cleanup macro definitions. 
This commit was SVN r16554.
 2007-10-23 13:33:19 +00:00
Gleb Natapov
d836f3dbbe Remove unused macro. 
This commit was SVN r16552.
 2007-10-23 13:18:10 +00:00
Gleb Natapov
18ed60edeb Revert previous commit. There was no memory leak, the pointer is saved inside 
free list for future use. This patch moves BTL initialization into separate
function too.

This commit was SVN r16551.
 2007-10-23 12:57:45 +00:00
Gleb Natapov
657e544e02 Fix memory leak. Define init_data on a stack instead of allocation it each time. 
This commit was SVN r16550.
 2007-10-23 11:10:52 +00:00
Gleb Natapov
9e2d5acf8e Remove unused filed from openib fragment structure. 
This commit was SVN r16549.
 2007-10-23 07:38:29 +00:00
Gleb Natapov
63dde87076 If SM BTL cannot send fragment because the cyclic buffer is full put the 
fragment on the pending list and send it later instead of spinning on
opal_progress().

This commit was SVN r16537.
 2007-10-22 12:07:22 +00:00
Jeff Squyres
b7eeae0a74 Remove the mvapi BTL. Woo hoo! 
This commit was SVN r16483.
 2007-10-17 14:08:03 +00:00
Jeff Squyres
94b1e9cff9 Update to use BTL_VERBOSE and BTL_ERROR instead of opal_output'ing to 
the mca_btl_base_output stream directly (and relying on it to be -1 if
we didn't want any output).

This commit was SVN r16449.
 2007-10-15 17:53:02 +00:00