supports a wide variety of operating systems and platforms; see the
opal/mca/paffinity/hwloc/hwloc/README file for details.
This component includes an embedded copy of hwloc, currently based on
hwloc-1.0rc6. But note that hwloc is properly SVN imported into the
/vendor branch, so it will be easy to update when 1.0 GA is released.
Note that the hwloc tree embedded in opal/mca/paffinity/hwloc/hwloc is
identical to a hwloc distribution tarball, except that much of the
documentation was rm -rf'ed (because we don't need it for the embedded
case).
Since the paffinity framework currently does not understand hardware
threads, the hwloc component compensates for this by identifying cores
by the "first" hardware thread on that core. Hopefully we'll update
paffinity someday to understand hardware threads. :-)
configure grew a --with-hwloc option, analogous to what we do for many
other external libraries that OMPI supports. However, there's a new
feature: due to the request of several distros, OMPI can be configured
to build with its internal copy of hwloc or with an external copy of
hwloc (e.g., a system-installed hwloc).
1. If --with-hwloc is not specified, Open MPI will try to use its
internal copy (but silently fail/ignore hwloc if that fails).
1. If --with-hwloc=<dir> is supplied, Open MPI looks for hwloc
support in <dir> (and --with-hwloc-libdir=<dir>, if specified).
1. If --with-hwloc=external is supplied, Open MPI will look for hwloc
in a compiler/linker default external location.
1. If --with-hwloc=internal is supplied, Open MPI will use its
internal copy of hwloc.
Some of OMPI's main configury had to be slightly re-arranged in the
bootstrapping phase to accomodate hwloc's configry needs.
This commit was SVN r23125.
It is okay to not have a paffinity module IF you aren't using paffinity anyway. So don't error out of MPI_Init because a paffinity module wasn't selected.
Cleanup error reporting in the odls default module to (once and for all!) eliminate messages originating in the fork'd process. Create some new error codes to allow us to pass enough info back to the parent process to provide useful error messages.
This commit was SVN r23106.
done this way a long time ago for the "gee whiz!" factor -- when in
reality, they really only need one-of-many-run-time priority
selection).
Changed run-time priorities to be as follows:
* darwin: 20
* linux: 20
* posix: 10
* solaris: 30
* test: 5
* windows: 20
I have a very dim (possibly untrue) recollection that Solaris needs to
have a higher priority than others just to ensure that no other is
chosen under Solaris. Make all other "native" components have a
priority of 20 (they shouldn't conflict with each other). Make the
posix fallback component have a priority of 10. Make the test
component priority 5, meaning someone can always select it, but you
can also make a "never select me" component that prioritizes itself
under test.
This commit was SVN r22997.
modify the OPAL_PAFFINITY_PROCESS_IS_BOUND macro to search the cpuset for
the maximum possible number of cpus rather than just the number of cpus
currently online. This corrects a problem where mpi_paffinity_alone was
not working properly on systems in which there can be cpu namespaces with
holes, such as on ppc64 with smt off (as discussed in #2365).
This commit was SVN r22927.
anything for non-MPI apps. Oops! (But before you freak out, gentle
reader, note that mpi_paffinity_alone for MPI apps still worked fine)
When we made the switchover somewhere in the 1.3 series to have the
orted's do processor binding, then stuff like:
mpirun --mca mpi_paffinity_alone 1 hostname
should have bound hostname to processor 0. But it didn't because of a
subtle startup ordering issue: the MCA param registration for
opal_paffinity_alone was in the paffinity base (vs. being in
opal/runtime/opal_params.c), but it didn't actually get registered
until after the global variable opal_paffinity_alone was checked to
see if we wanted old-style affinity bindings. Oops.
However, for MPI apps, even though the orted didn't do the binding,
ompi_mpi_init() would notice that opal_paffinity_alone was set, yet
the process didn't seem to be bound. So the MPI process would bind
itself (this was done to support the running-without-orteds
scenarios). Hence, MPI apps still obeyed mpi_paffinity_alone
semantics.
But note that the error described above caused the new mpirun switch
--report-bindings to not work with mpi_paffinity_alone=1, meaning that
the orted would not report the bindings when mpi_paffinity_alone was
set to 1 (it ''did'' correctly report bindings if you used
--bind-to-core or one of the other binding options).
This commit separates out the paffinity base MCA param registration
into a small function that can be called at the Right place during the
startup sequence.
This commit was SVN r22602.
party/"vendor" import, the changes are actually far smaller than the
size of this changeset implies. Here's a list of the changes:
* Update the AMD license header in plpa_map.c to be less restrictive
(see https://svn.open-mpi.org/trac/plpa/changeset/262 for details)
-- '''this is the most/only important change of this update.''' No
code is changed by this; only removing a clase from a license
header in plpa_map.c.
* Changes to the generated {{{configure}}}, {{{config.guess}}}, and
{{{config.sub}}} scripts (which aren't used by OMPI).
* soname version tracking changes (which also aren't used by OMPI;
they're only used when PLPA is built/installed in "standalone"
mode).
* Update the "get version" m4 (which was stolen from OMPI's m4 to
begin with, and is only used during OMPI's autogen.sh step).
* Update various PLPA version numbers to 1.3.2.
* Bug fix in plpa-taskset (which is not built in the OMPI PLPA build).
This commit was SVN r22367.
opposite of MAKE_MEM_DEFINED. Also add in a call to NOACCESS to
(mostly) reverse the effects of MAKE_MEM_DEFINED (technically, page 0
was accessible before this, even though it's a Bad Idea to access it).
This commit was SVN r22056.
masks between the mask argument and a local PLPA mask. There were three
problems:
1) The "get" function computed the number of bits as sizeof(mask),
which is the size of the pointer to the mask rather than the mask
itself. So, only 4 bits were copied with m32 and 8 bits with m64.
There are actually 1024 bits.
2) The "get" and "set" functions both copied a number of bits computed
from the sizeof() mask, but sizeof() reports the number of bytes.
We have to multiply by 8 to get the number of bits.
3) These two functions check to make sure tha the mask argument is not
bigger than the PLPA mask. But, the set function copies a number
of bits in the PLPA mask, which is conceivably greater than the
number of bits in the mask argument. So, accesses to the mask
argument may overrun that argument.
Problems 1 and 2 meant that one would encounter errors when the number of
cores exceeded 4 (with -m32) or 8 (with -m64). Problem 3 probably caused
no errors.
This commit was SVN r21993.
The new options work by adding an ":if-avail" qualifier to the "bind-to-socket" and "bind-to-core" MCA params. If the system does not support this capability, the job will launch anyway. Without the qualifier, the job will abort with an error message indicating that the required functionality is not supported on this system.
This commit was SVN r21975.
#if defined (c_plusplus)
defined (__cplusplus)
followed by
extern "C" {
and the closing counterpart by BEGIN_C_DECLS and END_C_DECLS.
Notable exceptions are:
- opal/include/opal_config_bottom.h:
This is our generated code, that itself defines BEGIN_C_DECL and
END_C_DECL
- ompi/mpi/cxx/mpicxx.h:
Here we do not include opal_config_bottom.h:
- Belongs to external code:
opal/mca/backtrace/darwin/MoreBacktrace/MoreDebugging/MoreBacktrace.c
opal/mca/backtrace/darwin/MoreBacktrace/MoreDebugging/MoreBacktrace.h
- opal/include/opal/prefetch.h:
Has C++ specific macros that are protected:
- Had #if ... } #endif _and_ END_C_DECLS (aka end up with 2x
END_C_DECLS)
ompi/mca/btl/openib/btl_openib.h
- opal/event/event.h has #ifdef __cplusplus as BEGIN_C_DECLS...
- opal/win32/ompi_process.h: had extern "C"\n {...
opal/win32/ompi_process.h: dito
- ompi/mca/btl/pcie/btl_pcie_lex.l: needed to add *_C_DECLS
ompi/mpi/f90/test/align_c.c: dito
- ompi/debuggers/msgq_interface.h: used #ifdef __cplusplus
- ompi/mpi/f90/xml/common-C.xsl: Amend
Tested on linux using --with-openib and --with-mx
The following do not contain either opal_config.h, orte_config.h or
ompi_config.h
(but possibly other header files, that include one of the above):
ompi/mca/bml/r2/bml_r2_ft.h
ompi/mca/btl/gm/btl_gm_endpoint.h
ompi/mca/btl/gm/btl_gm_proc.h
ompi/mca/btl/mx/btl_mx_endpoint.h
ompi/mca/btl/ofud/btl_ofud_endpoint.h
ompi/mca/btl/ofud/btl_ofud_frag.h
ompi/mca/btl/ofud/btl_ofud_proc.h
ompi/mca/btl/openib/btl_openib_mca.h
ompi/mca/btl/portals/btl_portals_endpoint.h
ompi/mca/btl/portals/btl_portals_frag.h
ompi/mca/btl/sctp/btl_sctp_endpoint.h
ompi/mca/btl/sctp/btl_sctp_proc.h
ompi/mca/btl/tcp/btl_tcp_endpoint.h
ompi/mca/btl/tcp/btl_tcp_ft.h
ompi/mca/btl/tcp/btl_tcp_proc.h
ompi/mca/btl/template/btl_template_endpoint.h
ompi/mca/btl/template/btl_template_proc.h
ompi/mca/btl/udapl/btl_udapl_eager_rdma.h
ompi/mca/btl/udapl/btl_udapl_endpoint.h
ompi/mca/btl/udapl/btl_udapl_mca.h
ompi/mca/btl/udapl/btl_udapl_proc.h
ompi/mca/mtl/mx/mtl_mx_endpoint.h
ompi/mca/mtl/mx/mtl_mx.h
ompi/mca/mtl/psm/mtl_psm_endpoint.h
ompi/mca/mtl/psm/mtl_psm.h
ompi/mca/pml/cm/pml_cm_component.h
ompi/mca/pml/csum/pml_csum_comm.h
ompi/mca/pml/dr/pml_dr_comm.h
ompi/mca/pml/dr/pml_dr_component.h
ompi/mca/pml/dr/pml_dr_endpoint.h
ompi/mca/pml/dr/pml_dr_recvfrag.h
ompi/mca/pml/example/pml_example.h
ompi/mca/pml/ob1/pml_ob1_comm.h
ompi/mca/pml/ob1/pml_ob1_component.h
ompi/mca/pml/ob1/pml_ob1_endpoint.h
ompi/mca/pml/ob1/pml_ob1_rdmafrag.h
ompi/mca/pml/ob1/pml_ob1_recvfrag.h
ompi/mca/pml/v/pml_v_output.h
opal/include/opal/prefetch.h
opal/mca/timer/aix/timer_aix.h
opal/util/qsort.h
test/support/components.h
This commit was SVN r21855.
The following SVN revision numbers were found above:
r2 --> open-mpi/ompi@58fdc18855
Adds several new mpirun options:
* -bysocket - assign ranks on a node by socket. Effectively load balances the procs assigned to a node across the available sockets. Note that ranks can still be bound to a specific core within the socket, or to the entire socket - the mapping is independent of the binding.
* -bind-to-socket - bind each rank to all the cores on the socket to which they are assigned.
* -bind-to-core - currently the default behavior (maintained from prior default)
* -npersocket N - launch N procs for every socket on a node. Note that this implies we know how many sockets are on a node. Mpirun will determine its local values. These can be overridden by provided values, either via MCA param or in a hostfile
Similar features/options are provided at the board level for multi-board nodes.
Documentation to follow...
This commit was SVN r21791.
Revamp the affinity detection/set procedure in mpi_init to correctly detect when we have already been bound to processors, given the revised understanding of paffinity_get. Add a new paffinity macro to make checking for already bound a little nicer.
This commit was SVN r21402.
1. replacing mpi_paffinity_alone with opal_paffinity_alone - for back-compatibility, I have aliased mpi_paffinity_alone to the new param name. This caus
es a mild abstraction break in the opal/mca/paffinity framework - per the devel discussion...live with it. :-) I also moved the ompi_xxx global variable
that tracked maffinity setup so it could be properly closed in MPI_Finalize to the opal/mca/maffinity framework to avoid an abstraction break.
2. Added code to the odls/default module to perform paffinity binding and maffinity init between process fork and exec. This has been tested on IU's odi
n cluster and works for both MPI and non-MPI apps.
3. Revise MPI_Init to detect if affinity has already been set, and to attempt to set it if not already done. I have *not* tested this as I haven't yet f
igured out a way to do so - I couldn't get slurm to perform cpu bindings, even though it supposedly does do so.
This has only been lightly tested and would definitely benefit from a wider range of evaluation...
This commit was SVN r21209.
- Delete unnecessary header files using
contrib/check_unnecessary_headers.sh after applying
patches, that include headers, being "lost" due to
inclusion in one of the now deleted headers...
In total 817 files are touched.
In ompi/mpi/c/ header files are moved up into the actual c-file,
where necessary (these are the only additional #include),
otherwise it is only deletions of #include (apart from the above
additions required due to notifier...)
- To get different MCAs (OpenIB, TM, ALPS), an earlier version was
successfully compiled (yesterday) on:
Linux locally using intel-11, gcc-4.3.2 and gcc-SVN + warnings enabled
Smoky cluster (x86-64 running Linux) using PGI-8.0.2 + warnings enabled
Lens cluster (x86-64 running Linux) using Pathscale-3.2 + warnings enabled
This commit was SVN r21096.
forgotten to save before the commit was sent.
This comment explains why we're doing a cache check here rather than a
real check.
This commit was SVN r20975.
component (which we probably don't test regularly because we probably
only test environments where the other paffinity components are used)
was not getting built because it had a bad configure test.
This commit was SVN r20974.
