THREAD_MULTIPLE. There's a new (hidden) MCA parameter to re-enable
these BTLs in the presence of THREAD_MULTIPLE:
btl_base_thread_multiple_override. This MCA parameter should ''only''
be used by developers who are working on make their BTLs thread safe;
it should ''not'' be used by end-users!
This commit was SVN r19826.
The following Trac tickets were found above:
Ticket 1588 --> https://svn.open-mpi.org/trac/ompi/ticket/1588
(this was missed in #1585). Also, fix a long-standing problem that
the F90 wrapper compilers were using the F77 wrapper compiler flags.
This commit was SVN r19819.
We had a public symbol named "already_opened". This commit changes
the name to mca_btl_base_already_opened.
I guess it's good we have the visibility stuff enabled by default!
:-)
This commit was SVN r19791.
1. Allow MX bonding via btl_mx_bonding MCA parameter. With this on, Open MPI
suppose that lib MX will do the bonding, and we will only return one BTL.
Otherwise, we return as many as devices.
2. Decrease the memory footprint, by cleaning up what we store about the
peers and how we store it.
3. Allow multiple MX routes that share the same mapper. In this particular
case we will link by their nic_id.
4. Allow multiple MX routes with multiple mappers. In this case we match
the NICs based on the last 6 digits of the mapper MAC.
5. Increase the size of the eager and rendez-vous eager limits in the
case where we are unable to register an unexpected callback with MX.
6. Increase the default max number of MX fragments.
7. Increase the max number of MX BTLs.
8. Only allow mx_if_include and mx_if_exclude if we have acess to the
mapper.
This commit was SVN r19788.
1. completely and cleanly separates responsibilities between the HNP, orted, and tool components.
2. removes all wireup messaging during launch and shutdown.
3. maintains flow control for stdin to avoid large-scale consumption of memory by orteds when large input files are forwarded. This is done using an xon/xoff protocol.
4. enables specification of stdin recipients on the mpirun cmd line. Allowed options include rank, "all", or "none". Default is rank 0.
5. creates a new MPI_Info key "ompi_stdin_target" that supports the above options for child jobs. Default is "none".
6. adds a new tool "orte-iof" that can connect to a running mpirun and display the output. Cmd line options allow selection of any combination of stdout, stderr, and stddiag. Default is stdout.
7. adds a new mpirun and orte-iof cmd line option "tag-output" that will tag each line of output with process name and stream ident. For example, "[1,0]<stdout>this is output"
This is not intended for the 1.3 release as it is a major change requiring considerable soak time.
This commit was SVN r19767.
There is still a problem with OpenIB and threads (external to C/R functionality). It has been reported in Ticket #1539
Additionally:
* Fix a file cleanup bug in CRS Base.
* Fix a possible deadlock in the TCP ft_event function
* Add a mca_base_param_deregister() function to MCA base
* Add whole process checkpoint timers
* Add support for BTL: OpenIB, MX, Shared Memory
* Add support Mpool: rdma, sm
* Sundry bounds checking an cleanup in some scattered functions
This commit was SVN r19756.
I run IMB exchange on two QS22 machines with r19674 and it got stucked after 256 or 512 bytes every time.
After applying r19717 the test passed, so I guess this is a essential patch.
This commit was SVN r19752.
The following SVN revision numbers were found above:
r19674 --> open-mpi/ompi@15c47a2473
r19717 --> open-mpi/ompi@0a765cd788
unconditionally, which can result in a flood of messages to the user
if all MPI processes invoke abort. Additionally, some users were
confused because they saw the MPI_ABORT opal_output() messages from
''some'' MPI processes, but not ''all'' of them (despite the fact that
every MPI process supposedly invoked MPI_ABORT). The reason is that
calling MPI_ABORT triggers ORTE to kill all MPI processes, so it's a
race condition as to whether a) all MPI processes actually invoke
MPI_ABORT, and/or b) whether every process is able to opal_output()
before they are killed.
This commit does two simple things:
* Now use orte_show_help() for the MPI_ABORT message, so they are
aggregated.
* Add a note in the message that calling MPI_ABORT kills all
processes, so you might not see all output, yadda yadda yadda.
This commit was SVN r19735.
We still do an interreduce but it is now followed by an intrascatterv.
This fixes trac:1554.
This commit was SVN r19723.
The following Trac tickets were found above:
Ticket 1554 --> https://svn.open-mpi.org/trac/ompi/ticket/1554
following names are all new for v1.3, and therefore haven't been
officially released yet:
* btl_openib_of_cq_size
* btl_openib_of_max_inline_data
* btl_openib_of_pkey
* btl_openib_of_psn
* btl_openib_of_mtu
The "_of_" (for OpenFabrics) in there is redundant. It used to be
"_ib_", indicating that these values are pretty much passed directly
to the verbs stack. But I think the "openib" in the name implies this
already; having "_of_" in there just seems redundant, makes the name
longer, and seems redundant. It's also redundant.
So I took those "_of_"'s out of the MCA names. The old (v1.2) names
are still valid (but deprecated), such ash btl_openib_ib_cq_size.
This commit was SVN r19718.
* Change name: mca_btl_openib_of_pkey_value -> mca_btl_openib_of_pkey
(since now there's no index, the "_value" suffix is somewhat
superfluous)
* Put in a better help message for the _pkey MCA param (to agree with
the new help message in v1.2.8)
This commit was SVN r19716.
Commit from a long-standing Mercurial tree that ended up incorporating a lot of things:
* A few fixes for CPC interface changes in all the CPCs
* Attempts (but not yet finished) to fix shutdown problems in the IB CM CPC
* #1319: add CTS support (i.e., initiator guarantees to send first message; automatically activated for iWARP over the RDMA CM CPC)
* Some variable and function renamings to make this be generic (e.g., alloc_credit_frag became alloc_control_frag)
* CPCs no longer post receive buffers; they only post a single receive buffer for the CTS if they use CTS. Instead, the main BTL now posts the main sets of receive buffers.
* CPCs allocate a CTS buffer only if they're about to make a connection
* RDMA CM improvements:
* Use threaded mode openib fd monitoring to wait for for RDMA CM events
* Synchronize endpoint finalization and disconnection between main thread and service thread to avoid/fix some race conditions
* Converted several structs to be OBJs so that we can use reference counting to know when to invoke destructors
* Make some new OBJ's have opal_list_item_t's as their base, thereby eliminating the need for the local list_item_t type
* Renamed many variables to be internally consistent
* Centralize the decision in an inline function as to whether this process or the remote process is supposed to be the initiator
* Add oodles of OPAL_OUTPUT statements for debugging (hard-wired to output stream -1; to be activated by developers if they want/need them)
* Use rdma_create_qp() instead of ibv_create_qp()
* openib fd monitoring improvements:
* Renamed a bunch of functions and variables to be a little more obvious as to their true function
* Use pipes to communicate between main thread and service thread
* Add ability for main thread to invoke a function back on the service thread
* Ensure to set initiator_depth and responder_resources properly, but putting max_qp_rd_ataom and ma_qp_init_rd_atom in the modex (see rdma_connect(3))
* Ensure to set the source IP address in rdma_resolve() to ensure that we select the correct OpenFabrics source port
* Make new MCA param: openib_btl_connect_rdmacm_resolve_timeout
* Other improvements:
* btl_openib_device_type MCA param: can be "iw" or "ib" or "all" (or "infiniband" or "iwarp")
* Somewhat improved error handling
* Bunches of spelling fixes in comments, VERBOSE, and OUTPUT statements
* Oodles of little coding style fixes
* Changed shutdown ordering of btl; the device is now an OBJ with ref counting for destruction
* Added some more show_help error messages
* Change configury to only build IBCM / RDMACM if we have threads (because we need a progress thread)
This commit was SVN r19686.
The following Trac tickets were found above:
Ticket 1210 --> https://svn.open-mpi.org/trac/ompi/ticket/1210
got a whole lot smaller, decreasing the memory footprint of the
running application. How much it's a good question. Here is a
breakdown:
- in mca_bml_base_endpoint_t: 3 *size_t + 1 * uint32_t
- in mca_bml_base_btl_t: 1 * int + 1 * double - 1 * float
+ 6 * size_t + 9 * (void*)
The decrease in mca_bml_base_endpoint_t is for each peer and the
decrease in mca_bml_base_btl_t is for each BTL for each peer.
So, if we consider the most convenient case where there is only
one network between all peers, this decrease the memory foot print
per peer by
9*size_t + 9*(void*) + 2 * int32_t + 1 * double - 1 * float.
On a 64 bits machine this will be 156 bytes per peer.
Now we access all these fields directly from the underlying BTL
structure, and as this structure is common to multiple BML endpoint,
we are a lot more cache friendly. Even if this do not improve the
latency, it makes the SM performance graph a lot smoother.
This commit was SVN r19659.
There was an argument that was barely used, and on return at the PML
level it contained nothing usable. It has been removed, so now we're
using less memory ...
This commit was SVN r19657.
(related to the presence of posix threads and ptmalloc2) is now a
little outdated: since we don't build ptmalloc2 as part of libopal
anymore, the openib BTL's requirements are not directly tied to
ptmalloc2's anymore. Specifically, I altered the test to:
1. At compile time, if no threads are found, the ptmalloc2 component
is going to be built, '''and the ptmalloc2 component is going to be
inside libopal,''' then refuse to build the openib BTL.
1. At run time, if no threads were available at compile time and the
ptmalloc2 component is part of the process, then refuse to use the
openib BTL.
Fixes trac:1537.
This commit was SVN r19652.
The following Trac tickets were found above:
Ticket 1537 --> https://svn.open-mpi.org/trac/ompi/ticket/1537
always in a heterogeneous way in order to be able to support extern32. It
doesn't really matter as it is outside the critical path.
This commit was SVN r19651.
Rationale:
1. This value has already changed since v1.2 (v1.2 MPI_MAX_PORT_NAME
== 36). Hence, this commit simply increases the value from a
previous change.
1. The changes does increase OMPI's memory footprint slightly, but
only when using MPI-2 dynamics. So it is expected that the change
will have minimal impact on the overall footprint.
1. The change is helpful for nodes that have 4 or more IP networks
(e.g., regular ethernet and multiple IP-over-<pick your favorite
high-speed network> networks). Without this change, invoking
MPI_COMM_SPAWN on hosts with 4 or more IP networks will fail
because we'll exceed 256 bytes for the port name. Some OMPI
developer test clusters already have this kind of configuration
(e.g., Cisco); it is expected that this is not too common in the
real world yet, but with "manycore" coming, having multiple
IP-based networks in a single server will likely become more
common.
This commit was SVN r19638.
we already have them in orte_process_info. Refs trac:1523.
This commit was SVN r19615.
The following Trac tickets were found above:
Ticket 1523 --> https://svn.open-mpi.org/trac/ompi/ticket/1523
help messages so that users only see the message once instead of N
times when their MPI app crashes.
Note that there is a tradeoff here -- we now call malloc in this
particular "show the error" code path. This shouldn't usually be a
problem, because the errors typically displayed through this mechanism
are MPI API argument problems (e.g., sending a negative count to
MPI_SEND), and not memory errors. But such API argument errors could
be a consequence of of a prior memory error, so there's a nonzero
chance that the error failure will fail to print because malloc
failed. In this case, the user can disable help message aggregation
(via the orte_base_want_aggregate MCA parameter) and we'll fall back
to the no-malloc code path (but without aggregation).
Note that we won't aggregate before MPI_INIT or after MPI_FINALIZE.
So if you call an MPI function before MPI_INIT / after MPI_FINALIZE,
you'll still see the error message N times. Nothing we can do about
that; we need ORTE to do the aggregation properly (which is obviously
unavailable before MPI_INIT / after MPI_FINALIZE).
This commit was SVN r19611.
Terry and George in the non-sparse-groups scenarios. Fixes trac:1464.
Will file a new ticket to actually resolve IDs when sparse groups are
used.
This commit was SVN r19610.
The following Trac tickets were found above:
Ticket 1464 --> https://svn.open-mpi.org/trac/ompi/ticket/1464
Thanks to George and Jeff for pointing out a better way to do this.
This commit was SVN r19573.
The following SVN revision numbers were found above:
r19566 --> open-mpi/ompi@351c3a3a86
figure it out at runtime (really meaning: we'll still default to "0"
unless something explicitly overrides to 1, such as the openib BTL).
This way, ompi_info doesn't confusingly report mpi_leave_pinned==0 for
mpi_leave_pinned, but we end up running with mpi_leave_pinned==1.
Fixes trac:1502.
This commit was SVN r19571.
The following Trac tickets were found above:
Ticket 1502 --> https://svn.open-mpi.org/trac/ompi/ticket/1502
remove the unconditional opal_output's when mmap() fails, and instead,
conditionally output the failure message via btl_base_verbose settings.
This commit was SVN r19547.
file. breaks windows compilation. see r19502
This commit was SVN r19544.
The following SVN revision numbers were found above:
r19502 --> open-mpi/ompi@ce42e749a0
The new component fixes a number of problems with the old component. The core algorithm is the same, but by changing the data strucutres a bit we have improved performance and memory utilization.
There are still a couple corner cases that still need some work. However, I did not want to delay bringing this into the trunk (and v1.3 branch) for too much longer.
This commit was SVN r19537.
This fixes trac:1477.
Help provided by Jeff and Terry.
This commit was SVN r19533.
The following Trac tickets were found above:
Ticket 1477 --> https://svn.open-mpi.org/trac/ompi/ticket/1477
checking for contiguous datatypes by using our native DDT engine
(rather than several MPI_* calls). The majority of the work is in the
IO ROMIO module.c file, but there's a small part in
adio/common/iscontig.c that we're also submitting upstream.
This commit was SVN r19509.
mostly don't use this mechanism, as we have to be thread safe in order to
really take full advantage of it. The unexpected handler is called by one
of th MX threads, and we do not have control on the moment. In a non threaded
case, this will completely destroy our recv requests queues, so the safest
approach is to don't the unexpected handler.
This commit was SVN r19496.
MPI::SEEK_SET and friends, suggested by Doug Gregor. This way allows
users to utilize SEEK_SET in a case statement, which they could not do
with our previous method.
This commit was SVN r19494.
way".
Don't modify coords in the top-level API function because coords is an
IN variable. Instead, as Nysal noted, the real cause of the problem
was a missing ! down in topo_base_cart_rank.c. Put a comment down in
topo_base_cart_rank.c explaining what's going on so that the code is
not so cryptic.
Refs trac:1363.
This commit was SVN r19487.
The following Trac tickets were found above:
Ticket 1363 --> https://svn.open-mpi.org/trac/ompi/ticket/1363
correctly in case of MPI_PROC_NULL with persistent requests.
As of r12339, the req_status of the request is initialized
correclty and being used in {{{ompi_request_test}}} and
{{{ompi_request_wait}}}.
So, just set the status, getting rid of the if-statement.
Checked by Jeff using mtt.
This commit was SVN r19485.
The following SVN revision numbers were found above:
r12339 --> open-mpi/ompi@020efdf1f9
need to clean them up during ompi_mpi_finalize(), though...
This commit was SVN r19462.
The following SVN revision numbers were found above:
r19458 --> open-mpi/ompi@697dc524c1
* Various changes to enable 0-dimensional cartesian communicators:
* Set various mtc_* members to NULL when there are 0 dimensions (and
don't bother trying to memcpy these arrays when duplicating the
communicator -- because they're NULL)
* adjust topo_base_cart_sub to correctly handle 0 dimensions
(simplified it a bit)
* adjust a few error codes to return ERR_OUT_OF_RESOURCE
* adjust error checking of CART_CREATE, CART_RANK
* Allow MPI_GRAPH_CREATE to accept 0 == nnodes.
* Bump reported MPI version in mpi.h to 2.1
This commit was SVN r19461.
The following Trac tickets were found above:
Ticket 1236 --> https://svn.open-mpi.org/trac/ompi/ticket/1236
for the F90 type create functions to the requirements of MPI 2.1 standard.
Advice to implementors. An application may often repeat a call to
MPI_TYPE_CREATE_F90_xxxx with the same combination of (xxxx,p,r).
The application is not allowed to free the returned predefined, unnamed
datatype handles. To prevent the creation of a potentially huge amount of
handles, the MPI implementation should return the same datatype handle for
the same (REAL/COMPLEX/INTEGER,p,r) combination. Checking for the
combination (p,r) in the preceding call to MPI_TYPE_CREATE_F90_xxxx and
using a hash-table to find formerly generated handles should limit the
overhead of finding a previously generated datatype with same combination
of (xxxx,p,r). (End of advice to implementors.)
This commit fixes trac:1239, and #712.
This commit was SVN r19458.
The following Trac tickets were found above:
Ticket 1239 --> https://svn.open-mpi.org/trac/ompi/ticket/1239
* Protect an orte variable used in the orte debugger stuff
* Initialize the datatype code in the Catamount code, as we need it
for intercommunicators (the proc code needs it to pack the remote
name)
* Turn on a bunch of the orte datatype code so that ORTE_NAME is available.
This commit was SVN r19457.
guarantee local progress, my opinion is we don't need to do anything
special for complete sends. Perhaps I'm wrong on that one, in which
case this implementation of send_complete is not quite right.
* Give a slightly better error message when a catchall is hit, so that
we can at least know *which* catchall is hit.
This commit was SVN r19456.
See opal/mca/paffinity/paffinity.h for explanation as to the physical vs logical nature of the params used in the API.
Fixes trac:1435
This commit was SVN r19391.
The following Trac tickets were found above:
Ticket 1435 --> https://svn.open-mpi.org/trac/ompi/ticket/1435
2. fix a bug in the initial communicator creation of llcomm
3. fix a bug which showed up as the result of fixing issue number 2: we have
to check now whether llcomm has really be created before freeing the
according llcomm in hierarch_destruct.
This commit was SVN r19361.
generate an overlap between the different layers. Why four versions? Because
there is right now always the trade-off between using non-blocking operations
on a layer with a trivial, linear algorithm and using the more sophisticaed
algorithms in a blocking manner.
- bcast_intra_seg used the bcast of lcomm and llcomm, similarly
to original algorithm in hierarch. However, it can segment
the message, such that we might get an overlap between the two
layers. This overlap is based on the assumption, that a process
might be done early with a bcast and can start the next one.
- bcast_intra_seg1: replaces the llcomm->bcast by isend/irecvs
to increase the overlap, keeps the lcomm->bcast however
- bcast_intra_seg2: replaced lcomm->bcast by isend/irecvs
to increase the overlap, keeps however llcomm->bcast
- bcast_intra_seg3: replaced both lcomm->bcast and llcomm->bcast
by isend/irecvs
The code is lightly tested, more testing to follow right now.
This commit was SVN r19358.
Roland noticed that the QLogic HCA driver was using the PCIe vendor ID
for the ibv_query_device so the IEEE OUI value is now used. This means
the config file should recognize the vendor ID value 0x1175 too.
This commit was SVN r19277.
