These changes fix issue https://github.com/open-mpi/ompi/issues/1336
- improve abstractions: opal/memory/linux component should be single place that opeartes with
Memory Allocation Hooks.
- avoid collisions in case dynamic component open/close: it is safe because it is linked statically.
- does not change original behaivour.
converting an opal_process_name_t means the loss of one bit,
it was decided to restrict the local job id to 15 bits, so the
useful information of an opal_process_name_t can fit in 63 bits.
This commit fixes several bugs identified by @ggouaillardet and MTT:
- Fix SEGV in long send completion caused by missing update to the
request callback data.
- Add an MPI_Barrier to the fence short-cut. This fixes potential
semantic issues where messages may be received before fence is
reached.
- Ensure fragments are flushed when using request-based RMA. This
allows MPI_Test/MPI_Wait/etc to work as expected.
- Restore the tag space back to 16-bits. It was intended that the
space be expanded to 32-bits but the required change to the
fragment headers was not committed. The tag space may be expanded
in a later commit.
Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
This commit fixes several bugs identified by a new multi-threaded RMA
benchmarking suite. The following bugs have been identified and fixed:
- The code that signaled the actual start of an access epoch changed
the eager_send_active flag on a synchronization object without
holding the object's lock. This could cause another thread waiting
on eager sends to block indefinitely because the entirety of
ompi_osc_pt2pt_sync_expected could exectute between the check of
eager_send_active and the conditon wait of
ompi_osc_pt2pt_sync_wait.
- The bookkeeping of fragments could get screwed up when performing
long put/accumulate operations from different threads. This was
caused by the fragment flush code at the end of both put and
accumulate. This code was put in place to avoid sending a large
number of unexpected messages to a peer. To fix the bookkeeping
issue we now 1) wait for eager sends to be active before stating
any large isend's, and 2) keep track of the number of large isends
associated with a fragment. If the number of large isends reaches
32 the active fragment is flushed.
- Use atomics to update the large receive/send tag counters. This
prevents duplicate tags from being used. The tag space has also
been updated to use the entire 16-bits of the tag space.
These changes should also fixopen-mpi/ompi#1299.
Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
This commit adds code to handle large unaligned gets. There are two
possible code paths for these transactions:
1) The remote region and local region have the same alignment. In
this case the get will be broken down into at most three get
transactions: 1 transaction to get the unaligned start of the region
(buffered), 1 transaction to get the aligned portion of the region,
and 1 transaction to get the end of the region.
2) The remote and local regions do not have the same alignment. This
should be an uncommon case and is not optimized. In this case a
buffer is allocated and registered locally to hold the aligned data
from the remote region. There may be cases where this fails (low
memory, can't register memory). Those conditions are unlikely and
will be handled later.
Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
If the state of the request is not set to OMPI_REQUEST_ACTIVE
then MPI_Test would immediately signal such request completed
while hcoll may still be working on it.
Signed-off-by: Joshua Ladd <jladd.mlnx@gmail.com>
If atomics are not globally visible (cpu and nic atomics do not mix)
then a btl endpoint must be used to access local ranks. To avoid
issues that are caused by having the same region registered with
multiple handles osc/rdma was updated to always use the handle for
rank 0. There was a bug in the update that caused osc/rdma to continue
using the local endpoint for accessing the state even though the
pointer/handle are not valid for that endpoint. This commit fixes the
bug.
Fixesopen-mpi/ompi#1241.
Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
This commit makes ompi_datatype_get_pack_description thread safe. The
call is used by osc/pt2pt to send the packed description to remote
peers. Before this commit if MPI_THREAD_MULTIPLE is enabled and the
user uses MPI_Put, MPI_Get, etc we could hit a race where multiple
threads attempt to store the packed description on the datatype. Since
the code in question is not performance-critical the threading fix
uses opal_atomic_* calls instead of bothering with OPAL_THREAD_*.
Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
Optimizing put aggregation in the presence of threads will require a
redesign of the code. For now just ensure that put aggregation is
turned off when MPI_THREAD_MULTIPLE is enabled.
Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
A bus error occurs in sm OSC under the following conditions.
- sparc64 or any other architectures which need strict alignment.
- `MPI_WIN_POST` or `MPI_WIN_START` is called for a window created
by sm OSC.
- The communicator size is odd and greater than 3.
The lines 283-285 in current `ompi/mca/osc/sm/osc_sm_component.c` has
the following code.
```c
module->global_state = (ompi_osc_sm_global_state_t *) (module->segment_base);
module->node_states = (ompi_osc_sm_node_state_t *) (module->global_state + 1);
module->posts[0] = (uint64_t *) (module->node_states + comm_size);
```
The size of `ompi_osc_sm_node_state_t` is multiples of 4 but not
multiples of 8. So if `comm_size` is odd, `module->posts[0]` does
not aligned to 8. This causes a bus error when accessing
`module->posts[i][j]`.
This patch fixes the alignment of `module->posts[0]` by setting
`module->posts[0]` first.
Update the configure logic for the new pmix120 component
ckpt
Get the pmix120 component to work - still not really registering or handling notifications, but infrastructure now operates
Cleanup some of the symbol scopes, and provide a more comprehensive rename.h file. Will pretty it up later - let's see how this works
Cleanup the rename files to use the pretty macros
When mtl-portals4 is configured for logical mapping, coll-portals4
must disqualify because it does not yet support logical mapping.
coll-portals4 looks for the endpoint pid to be zero which tells it
that mtl-portals4 is configured for logical mapping. This commit
initializes the endpoint nid/pid to zero for logical mapping.
Since OS X 10.11 (aka El Capitan) DYLD_LIBRARY_PATH is no more
propagated to children, so try to dlopen libmpi with the full path
using the directory of libmpi_java
Fixesopen-mpi/ompi#1220
Thanks Alexander Daryin for reporting this
NOTE: Building with external pmix *requires* that you also build with external libevent and hwloc libraries. Detect this at configure and error out with large message if this requirement is violated.
Closes#1204 (replaces it)
Fixes#1064
A previous commit updated the one-sided code to register the state
region only once. This created an issue when using the scratch lock
with fetching atomics. In this case on any rank that isn't local rank
0 the module->state_handle is NULL. This commit fixes the issue by
removing the scratch lock and using a fragment pointer instead.
Fixesopen-mpi/ompi#1290
Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
accelerate message queue lookups if the lookups have
the proper tag&mask layout. OpenMPI should follow
PSM2's preferred tag&mask spec, so that PSM2 can provide
a performance benefit.
Remove send of the extra message. This bug hase triggered on
MPICH/coll/nbicbarrier test. In this test a series of communicators
are created.
This extre-message was reseived after original communicator was destroyed
and queued into non_existing_communicator_pending. When new completely
unrelated communicator with the same id as original was created this message
was pushed into the frags_cant_match queue and caused seq numbers skew and hang.
Remove excessive parameter check to avoid premature exit from the collective.
MPI standard says:
The type signature associated with sendcount, sendtype, at a process must be equal to
the type signature associated with recvcount, recvtype at any other process. This implies
that the amount of data sent must be equal to the amount of data received, pairwise between
every pair of processes.
In case of inter-communicator we have 2 group of processes and "left" group may call
MPI_Alltoall(NULL, 0, MPI_INT, buf, 10, MPI_INT, comm, ...);
and the right one:
MPI_Alltoall(buf,10,MPI_INT, NULL, 0, MPI_INT, comm, ...);
And it would be legal though one of the group will receive 0 bytes from others.
This was triggered by MPICH/coll test called icalltoall.
`MPI_WIN_TEST` must update the `flag` parameter to 0 when not all
origin processes called `MPI_WIN_COMPLETE`. But sm OSC doesn't.
If the caller initialize the `flag` argument to a non-0 value,
the caller will receive the non-0 `flag` value.
Previous commit f2794740 reverts Nathans changes. However it turns out
that I was unable to trace his logic until I started investigation of
icsplit hang. Bug was triggered when splitting Intercom was giving a group
where on side of the communicator was empty (icsplit, intercom create #2).
in this case remote_size == 0 and there is no way to distinguish between
inter- and intra-communicator.
Conclusion: We do need to distinguish between intra- and inter-communicators.
So we should use ompi_mpi_group_null.group.
This commit adds a call to ompi_request_wait_completion for buffered
sends. Without this line it is possible to get into a state where the
data is never sent.
Fixesopen-mpi/ompi#1185
Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
Instead of solely relying on the out value definitions in
MPI_Waitsome.3, explicitly copy this text here.
Note that the original text in this man page was copied verbatim from
the MPI spec; we've now added a bit more text (copied from
MPI_Waitsome.3in) that explains the out values so that users don't
have to cross-reference to another man page.
Thanks to Eric Schnetter for the suggestion.
Fixesopen-mpi/ompi#1153
some of the collective modules. Added a new function
opan_datatype_span, to compute the memory span of
count number of datatype, excluding the gaps in the
beginning and at the end. If a memory allocation is
made using the returned value, the gap (also returned)
should be removed from the allocated pointer.
This commit adds a helper function for creating groups from proc
lists. The function is used by ompi_comm_fill_rest to create the local
and remote groups.
Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
This commit changes the way ompi_proc_t's are retained/released by
ompi_group_t's. Before this change ompi_proc_t's were retained once
for the group and then once for each retain of a group. This method
adds unnecessary overhead (need to traverse the group list each time
the group is retained) and causes problems when using an async
add_procs.
Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
This commit changes the OPAL_THREAD_LOCK/OPAL_THREAD_UNLOCK calls in
ompi/proc to opal_mutex_lock/opal_mutex_unlock. This will allow
multi-threaded BTLs the ability to creat ompi_proc_t's without having
to set opal_using_threads. There should be no performance hits as none
of the lock points are in the critical path.
Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
This commit removes two pieces of unneeded code from gather. First
it removes destroy_tree() calls from linear_top(), because the
linear algorithm does not create a tree, so there is no need to
destroy it. Second it removes unpack_bytes from the gather request
because it was calculated but never used.
This commit allows to control output during abnormal oshmem/ompi application
termination.
Fixed issue in backtrace output. HAVE_BACKTRACE was never set so user was limited
in control of this variable.
Two related mca variables are moved to opal layer. Corresponding aliases are
added for ompi and oshmem.
