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.
Configure Option:
--enable-sysv
MCA Parameter:
mpi_common_sm
mpi_common_sm accepts a comma delimited list of: [sysv],mmap (order
dependent). The first component that is successfully selected is used. For
example, -mca mpi_common_sm sysv,mmap will first try sysv. If sysv is not
successfully selected, then mmap will be used. mmap will be used if
mpi_common_sm is not provided.
Notes:
Please make certain that your system's shmmax limit, or equivalent, is larger
than mpool_sm_min_size. Otherwise, shmget may fail.
This commit was SVN r23260.
This commit does a bunch of things:
* Address all remaining code review items from CMR #2023:
* Defer mmap setup to be lazy; only set it up the first time we
invoke a collective. In this way, we don't penalize apps that
make lots of communicators but don't invoke collectives on them
(per #2027).
* Remove the extra assignments of mca_coll_sm_one (fixing a
convertor count setup that was the real problem).
* Remove another extra/unnecessary assignment.
* Increase libevent polling frequency when using the RML to
bootstrap mmap'ed memory.
* Fix a minor procs-related memory leak in btl_sm.
* Commit a datatype fix that George and I discovered along the way to
fixing the coll sm.
* Improve error messages when mmap fails, potentially trying to
de-alloc any allocated memory when that happens.
* Fix a previously-unnoticed confusion between extent and true_extent
in coll sm reduce.
This commit was SVN r22049.
The following Trac tickets were found above:
Ticket 2023 --> https://svn.open-mpi.org/trac/ompi/ticket/2023
shmem progress (or the Windows equiv). Instead, poll hard on the
condition, but periocially call opal_progress(). This allows
badly-formed apps (e.g., the ibm test communicator/bsend_free) to
actually complete.
To be clear, there are far too many apps out there that assume that
MPI collectives will actually progress the rest of MPI. I don't like
putting in a feature to enable broken apps, but I have a dim
recollection of this issue coming up before (apps "hanging" when
testing the sm coll because they assumed that calling collectives
would trigger other MPI progress). Rather than have people claim that
OMPI is broken, I prefer to put in this "workaround". :-(
Indeed, the bsend_free test ''may'' be coded that way for exactly that
reason...? I don't remember offhand...
This commit was SVN r21984.
* Various cosmetic/style updates in the btl sm
* Clean up concept of mpool module (I think that code was written way
back when the concept of "modules" was fuzzy)
* Bring over some old fixes from the /tmp/timattox-sm-coll/ tree to
fix potential segv's when mmap'ed regions were at different
addresses in different processes (thanks Tim!).
* Change sm coll to no longer use mpool as its main source of shmem;
rather, just mmap its own segment (because it's fixed size --
there was nothing to be gained by using mpool; shedding the use of
mpool saved a lot of complexity in the sm coll setup). This
effectively made Tim's fixes moot (because now everything is an
offset into the mmap that is computed locally; there are no global
pointers). :-)
* Slightly updated common/sm to allow making mmap's for a specific
set of procs (vs. ''all'' procs in the process). This potentially
allows for same-host-inter-proc mmaps -- yay!
* Fixed many, many things in the coll sm (particularly in reduce):
* Fixed handling of MPI_IN_PLACE in reduce and allreduce
* Fixed handling of non-contiguous datatypes in reduce
* Changed the order of reductions to go from process (n-1)'s data
to process 0's data, because that's how all other OMPI coll
components work
* Fixed lots of usage of ddt functions
* When using a non-contiguous datatype, if the root process is not
(n-1), now we used a 2nd convertor to copy from shmem to the rbuf
(saves a memory copy vs. what was done before)
* Lots and lots of little cleanups, clarifications, and minor
optimizations (although still more could be done -- e.g., I think
the use of write memory barriers is fairly sub-optimal; they
could be ganged together at the root, for example)
I'm marking this as "fixes trac:1988" and closing the ticket; if something
is still broken, we can re-open the ticket.
This commit was SVN r21967.
The following Trac tickets were found above:
Ticket 1988 --> https://svn.open-mpi.org/trac/ompi/ticket/1988
OMPI
and a language agnostic part in OPAL. The convertor is completely
moved into OPAL. This offers several benefits as described in RFC
http://www.open-mpi.org/community/lists/devel/2009/07/6387.php
namely:
- Fewer basic types (int* and float* types, boolean and wchar
- Fixing naming scheme to ompi-nomenclature.
- Usability outside of the ompi-layer.
- Due to the fixed nature of simple opal types, their information is
completely
known at compile time and therefore constified
- With fewer datatypes (22), the actual sizes of bit-field types may be
reduced
from 64 to 32 bits, allowing reorganizing the opal_datatype
structure, eliminating holes and keeping data required in convertor
(upon send/recv) in one cacheline...
This has implications to the convertor-datastructure and other parts
of the code.
- Several performance tests have been run, the netpipe latency does not
change with
this patch on Linux/x86-64 on the smoky cluster.
- Extensive tests have been done to verify correctness (no new
regressions) using:
1. mpi_test_suite on linux/x86-64 using clean ompi-trunk and
ompi-ddt:
a. running both trunk and ompi-ddt resulted in no differences
(except for MPI_SHORT_INT and MPI_TYPE_MIX_LB_UB do now run
correctly).
b. with --enable-memchecker and running under valgrind (one buglet
when run with static found in test-suite, commited)
2. ibm testsuite on linux/x86-64 using clean ompi-trunk and ompi-ddt:
all passed (except for the dynamic/ tests failed!! as trunk/MTT)
3. compilation and usage of HDF5 tests on Jaguar using PGI and
PathScale compilers.
4. compilation and usage on Scicortex.
- Please note, that for the heterogeneous case, (-m32 compiled
binaries/ompi), neither
ompi-trunk, nor ompi-ddt branch would successfully launch.
This commit was SVN r21641.
* add "register" function to mca_base_component_t
* converted coll:basic and paffinity:linux and paffinity:solaris to
use this function
* we'll convert the rest over time (I'll file a ticket once all
this is committed)
* add 32 bytes of "reserved" space to the end of mca_base_component_t
and mca_base_component_data_2_0_0_t to make future upgrades
[slightly] easier
* new mca_base_component_t size: 196 bytes
* new mca_base_component_data_2_0_0_t size: 36 bytes
* MCA base version bumped to v2.0
* '''We now refuse to load components that are not MCA v2.0.x'''
* all MCA frameworks versions bumped to v2.0
* be a little more explicit about version numbers in the MCA base
* add big comment in mca.h about versioning philosophy
This commit was SVN r19073.
The following Trac tickets were found above:
Ticket 1392 --> https://svn.open-mpi.org/trac/ompi/ticket/1392
used at nce (up to one unique collective module per collective function).
Matches r15795:15921 of the tmp/bwb-coll-select branch
This commit was SVN r15924.
The following SVN revisions from the original message are invalid or
inconsistent and therefore were not cross-referenced:
r15795
r15921
This merge adds Checkpoint/Restart support to Open MPI. The initial
frameworks and components support a LAM/MPI-like implementation.
This commit follows the risk assessment presented to the Open MPI core
development group on Feb. 22, 2007.
This commit closes trac:158
More details to follow.
This commit was SVN r14051.
The following SVN revisions from the original message are invalid or
inconsistent and therefore were not cross-referenced:
r13912
The following Trac tickets were found above:
Ticket 158 --> https://svn.open-mpi.org/trac/ompi/ticket/158
allocation logic is completely done outside the data-type engine (in the PML) there is
no need for any special case inside the data-type engine. There is less arguments for
the ompi_convertor_pack and ompi_convertor_unpack as well (the last field free_after is
not required anymore as there is no memory allocated in the engine itself). This change
affect all components using datatypes. I test most of them, but it might happens that I
miss some ... If it's the case please let me know (don't shoot the pianist!!).
This commit was SVN r12331.
reduce_inorder() function -- we don't use the tree at all.
- Add more relevant "volatile"'s for the control buffers in the
fragment mpool (and associated casts where necessary)
This commit was SVN r7616.
- Move the "process 0" logic out of the main loop in reduce to make
the code a bit less complex (at the price of slight code
duplication, but it iss now significantly easier to read)
- Fix problem with uniquenes guarantee in the bootstrap mpool -- using
the CID alone was not sufficient enough to guarantee uniquenes; now
use (CID, rank 0 process name) tuple to check for uniqueness
- Made a few debugging help changes in coll_sm.h; especially helps
debugging on uniprocessors
This commit was SVN r7599.
- remove redundant OBJ_CONSTRUCT in bcast
- fix up some macros in coll_sm.h
- check to ensure that if there are too many processes in the
communicator (i.e., if we couldn't fit a flag for each of them in
the control segment), then fail selection
- setup the in_use flags properly
- adapt to new mpool API
- first working copy of reduce -- not tree-baed (but still
NUMA-aware), and only processes in order from process 0 to process
N-1 -- do not have a tree-based and/or commutative version yet
(i.e., process the results in whatever order they arrive)
Reduce now passes the new ibm reduce_big.c test. Woo hoo! Time to
declare success for the evening (and run the intel test tomorrow).
This commit was SVN r7379.
- finally added "in use" flags -- one flag protects a set of segments
- these flags now used in bcast to protect (for example) when a
message is so long that the root loops around the segments and has
to re-use old segments -- now it knows that it has to wait until the
non-root processes have finished with that set of segments before it
can start using them
- implement allreduce as a reduce followed by a bcast (per discussion
with rich)
- removed some redundant data on various data structures
- implemented query MCA param ("coll_sm_shared_mem_used_data") that
tells you how much shared memory will be used for a given set of MCA
params (e.g., number of segments, etc.). For example:
ompi_info --mca coll_sm_info_num_procs 4 --param coll sm | \
grep shared_mem_used_data
tells you that for the default MCA param values (as of r7172), for 4
processes, sm will use 548864 bytes of shared memory for its data
transfer section
- remove a bunch of .c files from the Makefile.am that aren't
implemented yet (i.e., all they do is return ERR_NOT_IMPLEMENTED)
Now on to the big Altix to test that this stuff really works...
This commit was SVN r7205.
The following SVN revision numbers were found above:
r7172 --> open-mpi/ompi@bc72a7722b
- bcast now works properly for root!=0 and multi-fragment messages
- destroy mpool when communicator is destroyed
Still need to implement:
- "in use" flags for groups of fragments so that "wrapping around" in
the data segment doesn't overwrite not-yet-read data
- ensure that shared memory isn't removed before all processes have
finished with it (e.g., during COMM_FREE)
This commit was SVN r7172.
much time) and somewhat-lame implementation of barrier (need to
precompute some more stuff rather than calculate it every time).
Checkpointing so I can try this on another machine...
This commit was SVN r6985.
multiple components to share a single mpool module (e.g., the
ptl/btl and coll sm components).
- Re-tool the ptl, btl, and coll sm components to first look for the
target mpool module, and if they don't find it, to create it.
- coll sm component now correctly identifies when it is supposed to
run or not (i.e., if all the processes in the communicator are on
the same host). Now we just need to fill in some algorithms. :-)
This commit was SVN r6530.
