in this code and it ended up changing the logic that is used to set up eager RDMA.
Rather than setting up eager RDMA with a high priority message, it did it the other
way around. For some reason, CUDA-aware support did not like this. So, basically,
restore the logic to the way it was prior to the refactoring. The refactoring did not
intend to change this. Lightly reviewed by hjelmn.
There was a bug in the openib btl handling this valid sequence of
calls:
desc = btl_alloc ();
btl_free (desc);
When triggered the bug would cause either fragment loss or undefined
behavior (SEGV, etc). The problem occured because btl_alloc contained
the logic to modify the pending fragment (length, etc) and these
changes were not corrected if the fragment was freed instead of sent.
To fix this issue I 1) moved some of the coalescing logic to the
btl_send function, and 2) retry the coalesced fragment on btl_free
if it was never sent. This appears to completely address the issue.
For systems with OFED's lacking XRC support, commit b3617e73
broke the build of the openib btl. This commit addresses
the issues introduced by this commit.
- by default allow to register maximum possible (i.e 2 * total_memory)
memory. This beheviour can be turned off using mca parameter
"btl_openib_allow_max_memory_registration"
- In fallback case, use device specific parameters to calulate
memory limit.
structure
This structure member was originally used to specify the remote segment
for an RDMA operation. Since the new btl interface no longer uses
desriptors for RDMA this member no longer has a purpose. In addition
to removing these members the local segment information has been
renamed to des_segments/des_segment_count.
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 enables IBM's software iWARP provider. With this driver you can
run iWARP/RDMA over any ethernet NIC. Useful for testing OMPI RDMA
logic without requiring an expensive RDMA adapter/infrastructure.
The Soft iWARP code is at: https://www.gitorious.org/softiwarp
These two macros set the prefix for the OPAL and ORTE libraries,
respectively. Specifically, the OPAL library will be named
libPREFIXopen-pal.la and the ORTE library will be named
libPREFIXopen-rte.la.
These macros must be called, even if the prefix argument is empty.
The intent is that Open MPI will call these macros with an empty
prefix, but other projects (such as ORCM) will call these macros with
a non-empty prefix. For example, ORCM libraries can be named
liborcm-open-pal.la and liborcm-open-rte.la.
This scheme is necessary to allow running Open MPI applications under
systems that use their own versions of ORTE and OPAL. For example,
when running MPI applications under ORTE, if the ORTE and OPAL
libraries between OMPI and ORCM are not identical (which, because they
are released at different times, are likely to be different), we need
to ensure that the OMPI applications link against their ORTE and OPAL
libraries, but the ORCM executables link against their ORTE and OPAL
libraries.
Properly setup the opal_process_info structure early in the initialization procedure. Define the local hostname right at the beginning of opal_init so all parts of opal can use it. Overlay that during orte_init as the user may choose to remove fqdn and strip prefixes during that time. Setup the job_session_dir and other such info immediately when it becomes available during orte_init.
The question of whether or not the openib BTL supports loopback is a separate question. It may be more appropriate to make the modex be PMIX_GLOBAL for cases where openib can support loopback so someone can run without a shared memory component. I'll leave that decision to the IB vendors.
This commit was SVN r32702.
- do not have mca_btl_openib_component.default_recv_qps point to the stack
- do not reset mca_btl_openib_component.default_recv_qps in btl_openib_component_open
cmr=v1.8.3:reviewer=miked
This commit was SVN r32642.
WHAT: Merge the PMIx branch into the devel repo, creating a new
OPAL “lmix” framework to abstract PMI support for all RTEs.
Replace the ORTE daemon-level collectives with a new PMIx
server and update the ORTE grpcomm framework to support
server-to-server collectives
WHY: We’ve had problems dealing with variations in PMI implementations,
and need to extend the existing PMI definitions to meet exascale
requirements.
WHEN: Mon, Aug 25
WHERE: https://github.com/rhc54/ompi-svn-mirror.git
Several community members have been working on a refactoring of the current PMI support within OMPI. Although the APIs are common, Slurm and Cray implement a different range of capabilities, and package them differently. For example, Cray provides an integrated PMI-1/2 library, while Slurm separates the two and requires the user to specify the one to be used at runtime. In addition, several bugs in the Slurm implementations have caused problems requiring extra coding.
All this has led to a slew of #if’s in the PMI code and bugs when the corner-case logic for one implementation accidentally traps the other. Extending this support to other implementations would have increased this complexity to an unacceptable level.
Accordingly, we have:
* created a new OPAL “pmix” framework to abstract the PMI support, with separate components for Cray, Slurm PMI-1, and Slurm PMI-2 implementations.
* Replaced the current ORTE grpcomm daemon-based collective operation with an integrated PMIx server, and updated the grpcomm APIs to provide more flexible, multi-algorithm support for collective operations. At this time, only the xcast and allgather operations are supported.
* Replaced the current global collective id with a signature based on the names of the participating procs. The allows an unlimited number of collectives to be executed by any group of processes, subject to the requirement that only one collective can be active at a time for a unique combination of procs. Note that a proc can be involved in any number of simultaneous collectives - it is the specific combination of procs that is subject to the constraint
* removed the prior OMPI/OPAL modex code
* added new macros for executing modex send/recv to simplify use of the new APIs. The send macros allow the caller to specify whether or not the BTL supports async modex operations - if so, then the non-blocking “fence” operation is used, if the active PMIx component supports it. Otherwise, the default is a full blocking modex exchange as we currently perform.
* retained the current flag that directs us to use a blocking fence operation, but only to retrieve data upon demand
This commit was SVN r32570.
