This commit updates the ugni btl to make use of multiple device
contexts to improve the multi-threaded RMA performance. This commit
contains the following:
- Cleanup the endpoint structure by removing unnecessary field. The
structure now also contains all the fields originally handled by the
common/ugni endpoint.
- Clean up the fragment allocation code to remove the need to
initialize the my_list member of the fragment structure. This
member is not initialized by the free list initializer function.
- Remove the (now unused) common/ugni component. btl/ugni no longer
need the component. common/ugni was originally split out of
btl/ugni to support bcol/ugni. As that component exists there is no
reason to keep this component.
- Create wrappers for the ugni functionality required by
btl/ugni. This was done to ease supporting multiple device
contexts. The wrappers are thread safe and currently use a spin
lock instead of a mutex. This produces better performance when
using multiple threads spread over multiple cores. In the future
this lock may be replaced by another serialization mechanism. The
wrappers are located in a new file: btl_ugni_device.h.
- Remove unnecessary device locking from serial parts of the ugni
btl. This includes the first add-procs and module finalize.
- Clean up fragment wait list code by moving enqueue into common
function.
- Expose the communication domain flags as an MCA variable. The
defaults have been updated to reflect the recommended setting for
knl and haswell.
- Avoid allocating fragments for communication with already
overloaded peers.
- Allocate RDMA endpoints dyncamically. This is needed to support
spreading RMA operations accross multiple contexts.
- Add support for spreading RMA communication over multiple ugni
device contexts. This should greatly improve the threading
performance when communicating with multiple peers. By default the
number of virtual devices depends on 1) whether
opal_using_threads() is set, 2) how many local processes are in the
job, and 3) how many bits are available in the pid. The last is
used to ensure that each CDM is created with a unique id.
Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
The send code in the ugni btl has an optimization that enables it to
return 1 (fragment gone) in some cases. This optimization involved
removing the btl ownership and callback flags to ensure the fragment
stuck around long enough for its completion flag to be checked. This
works fine for the single-threaded case but not in the multi-threaded
case. It is possible that a fragment will be completed by another
thread while a thread is in mca_btl_ugni_send. This competition can
lead to a leaked fragment, missed callback, or both. To fix the issue
without removing the optimization a reference count has been added to
the fragment. Callbacks and fragment release will not be made until
the fragment reference count has reach 0. The count is incremented
before sending the frag and decremented after the completion flag has
been checked. The fix has been verified to work using a multi-threaded
RMA benchmark with the osc/pt2pt component.
Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
This commit fixes a race condition that can cause an endpoint to be
added to the wait list multiple times. To fix the issue an additional
check has been added to ensure the endpoint is not on the wait list
after the wait list lock is held. The wait list processing code has
also been updated to keep the wait list lock until all wait listed
endpoints have been handled. This reduces the chance that an endpoint
that is being processed by the wait list code is not re-added to the
list by a competing send.
Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
The send inline optimization uses the btl_sendi function to achieve lower
latency and higher message rates. Before this commit BTLs were allowed to
assume the descriptor was non-NULL and were expected to return a valid
descriptor if the send could not be completed using btl_sendi. This
behavior was fine until the usage of btl_sendi was changed in ob1. This
commit allows the caller to specify NULL for the descriptor. The affected
btls have been updated to handle this case.
Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
value NULL for the descriptor
The send inline optimization uses the btl_sendi function to achieve
lower latency and higher message rates. The problem is the btl_sendi
function was allowed to return a descriptor to the caller. This is fine
for some paths but not ok for the send inline optimization. To fix
this the btl now must be able to handle descriptor = NULL.
The old BTL interface provided support for RDMA through the use of
the btl_prepare_src and btl_prepare_dst functions. These functions were
expected to prepare as much of the user buffer as possible for the RDMA
operation and return a descriptor. The descriptor contained segment
information on the prepared region. The btl user could then pass the
RDMA segment information to a remote peer. Once the peer received that
information it then packed it into a similar descriptor on the other
side that could then be passed into a single btl_put or btl_get
operation.
Changes:
- Removed the btl_prepare_dst function. This reflects the fact that
RDMA operations no longer depend on "prepared" descriptors.
- Removed the btl_seg_size member. There is no need to btl's to
subclass the mca_btl_base_segment_t class anymore.
...
Add more
We recognize that this means other users of OPAL will need to "wrap" the opal_process_name_t if they desire to abstract it in some fashion. This is regrettable, and we are looking at possible alternatives that might mitigate that requirement. Meantime, however, we have to put the needs of the OMPI community first, and are taking this step to restore hetero and SPARC support.
This commit adds initial ugni thread safety support.
With this commit, sun thread tests (excepting MPI-2 RMA)
pass with various process counts and threads/process.
Also osu_latency_mt passes.
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
