1
1
Граф коммитов

644 Коммитов

Автор SHA1 Сообщение Дата
Jeff Squyres
5ab634555a Apparently, Cisco plans to be working on Open MPI for a veeeeery long time!
This commit was SVN r23433.
2010-07-19 19:31:59 +00:00
Jeff Squyres
57d89d1c0c Remove a lot of kruft from the hwloc paffinity directory that we're
not using in Open MPI (i.e., that stuff is only used in the standalone
builds of hwloc -- it's not compiled/installed/used by Open MPI).

This commit was SVN r23416.
2010-07-14 20:46:47 +00:00
Jeff Squyres
dc7d30b0ed We (Ralph and Jeff) discovered that if the OPAL_DESTDIR environment
variable was set, it was prefixed to ''all'' values in the wrapper
compiler data text files.  For example, if OPAL_DESTDIR was set to
/tmp/bogus and a wrapper compiler data file contained the line:

  preprocessor_flags=-pthread

The value would be exanded to:

  /tmp/bogus/-pthread

Which is clearly wrong.  After some back-and-forth with Ralph and
Brian, Brian submitted this patch that fixes the problem.  Now we
handle three cases properly (assume that configure was invoked with
--prefix=/opt/openmpi and no other directory specifications, and
$OPAL_DESTDIR is set to /tmp/buildroot):

1. Individual directories, such as libdir.  These need to be prepended
with DESTDIR.  I.e., return /tmp/buildroot/opt/openmpi/lib.

2. Compiler flags that have ${FIELD} values embedded in them.  For
example, consider if a wrapper compiler data file contains the
line:

  preprocessor_flags=-DMYFLAG="${prefix}/share/randomthingy/"

The value we should return is:

  -DMYFLAG="/tmp/buildroot/opt/openmpi/share/randomthingy/"

3. Compiler flags that do not have any ${FIELD} values.  For example,
consider if a wrapper compiler data file contains the line:

  preprocessor_flags=-pthread

The value we should return is:

  -pthread

Note, too, that this OPAL_DESTDIR futzing only needs to occur during
opal_init().  By the time opal_init() has completed, all values should
be substituted in that need substituting.  Hence, we take an extra
parameter (is_setup) to know whether we should do this futzing or
not.

This commit was SVN r23402.
2010-07-14 00:53:08 +00:00
Jeff Squyres
10185343a7 Ensure that we're actually checking for *linux*. Thanks to Aleksej
Saushev for the patch.

This commit was SVN r23336.
2010-07-01 23:26:49 +00:00
Jeff Squyres
6d07a1cc0b Per comments in this commit, hwloc isn't able to find cores on all
platforms (e.g., PPC64 running RHEL 5.4) -- sometimes it only finds
PUs.  So in that case, just run the same calculation, but with PUs
instead of cores.

This commit was SVN r23305.
2010-06-25 21:36:53 +00:00
Jeff Squyres
5cdd79ef13 Oops -- set the bits one at a time via _set. Using _cpu effectively
zeroed out the cpuset before setting the bit (i.e., we always had a
cpuset of 1).

This commit was SVN r23298.
2010-06-23 20:56:59 +00:00
Jeff Squyres
8ce59bb3e3 Use HWLOC_EMBEDDED_LIBS properly (new variable as of 1.0.2a12214).
Should fix some Solaris build issues.

This commit was SVN r23266.
2010-06-09 19:58:42 +00:00
Jeff Squyres
2887fe77c5 Refresh hwloc to an as-yet unreleased tarball from the hwloc 1.0
release branch in order to fix some Solaris bugs.

This commit was SVN r23265.
2010-06-09 19:56:18 +00:00
Jeff Squyres
f1a7b5cc33 Make "processor affinity not supported" error message a little better:
* Remove OPAL_ERR_PAFFINITY_NOT_SUPPORTED; fit it into the generic
   OPAL_ERR_NOT_SUPPORTED case.
 * When odls_default detects that processor affinity is not supported,
   it prints a specific message about it, and then it suppressed a
   generic HNP help message that would normally follow it (i.e., it's
   easier to have the "processor affinity is not supported" show_help
   message last).
 * Use some symbolic names in odls_default instead of fixed int's,
   just for slight readability improvements in the code.
 * Introduce orte_show_help_suppress(), which gives the ability to
   suppress any future showings of any arbitrary show_help() message.
   This is useful if you display message X and want to suppress
   message Y.  This suppression *only* works in environments where
   orte_show_help() does coalescing.

This commit was SVN r23249.
2010-06-08 20:16:07 +00:00
Jeff Squyres
61f5528ec4 Update to hwloc 1.0.1rc1:
* Should fix the issues with 32 bit builds on 64 bit platforms
 * A few windows fixes
 * A few other minor / misc fixes

This commit was SVN r23226.
2010-06-01 14:51:25 +00:00
Jeff Squyres
e41603fb64 Add files into 3 directories that would not otherwise exist in a
distribution tarball, and would therefore cause automake to fail (in
case someone invokes autogen.sh on a distribution tarball).

This commit was SVN r23218.
2010-05-28 19:33:22 +00:00
Jeff Squyres
fec7918eea Some paffinity functions had their return status overloaded:
* If < 0, it's an OPAL_ERR_* value
 * If >= 0, it's the actual output value of the function

This is problematic for the OPAL_SOS stuff.  This commit changes those
functions to always return OPAL_* statuses and send the output value
back through output parameters (like 95% of the rest of the code
base).  This avoids the confusion with OPAL_SOS stuff and makes
paffinity work again (e.g., mpirun --bind-to-core ...).

I updated all paffinitiy modules for the new function signatures, and
bumped the paffinity API version up to 2.0.1.  I don't think the
version change will matter, though, because we'll be introducing
support for hardware threads soon, which will either bump the
paffinity version again or we'll replace paffinity with 
a new framework.

This commit was SVN r23197.
2010-05-21 16:55:28 +00:00
Jeff Squyres
e597c4f9cd Add --with-libltdl option to allow building Open MPI with an external installation of libltdl. Fixes trac:2407
This commit was SVN r23189.

The following Trac tickets were found above:
  Ticket 2407 --> https://svn.open-mpi.org/trac/ompi/ticket/2407
2010-05-20 22:42:02 +00:00
Abhishek Kulkarni
c63c4d6892 Fix bugs where (OMPI_ERROR == *) checks cannot be converted to (OMPI_SUCCESS != *) since the return codes are overloaded to return an "index" on success.
The fix is to just check if the return value is positive or not, since all the SOS encoded errors are *always* negative.

The real fix (as Ralph points out) is to change these functions (opal_pointer_array_add and mca_base_param*) to return the index as a pointer.

This commit was SVN r23173.
2010-05-18 20:54:11 +00:00
Jeff Squyres
32417b9802 Bump up to hwloc v1.0.
This commit was SVN r23171.
2010-05-18 17:11:45 +00:00
Abhishek Kulkarni
afbe3e99c6 * Wrap all the direct error-code checks of the form (OMPI_ERR_* == ret) with
(OMPI_ERR_* = OPAL_SOS_GET_ERR_CODE(ret)), since the return value could be a
 SOS-encoded error. The OPAL_SOS_GET_ERR_CODE() takes in a SOS error and returns
 back the native error code.

* Since OPAL_SUCCESS is preserved by SOS, also change all calls of the form
  (OPAL_ERROR == ret) to (OPAL_SUCCESS != ret). We thus avoid having to
  decode 'ret' to get the native error code.

This commit was SVN r23162.
2010-05-17 23:08:56 +00:00
Jeff Squyres
b43288f01e Add missing header file
This commit was SVN r23154.
2010-05-17 21:31:24 +00:00
Jeff Squyres
b0cfe91eca Re-enable hwloc component; it should be working now.
I forgot to mention one more thing in the r23152 commit message:

 * Copy the fix for hwloc's m4 to disable the configure flag
   --enable-debug when building in embedding mode, because it can be
   hijacked by the outter-level application.  In this case, if you
   configured OMPI with --enable-debug (or have --enable-debug in a
   platform file), you'd see all of hwloc's debug output.  Ick.  hwloc
   1.0 will include this fix.

This commit was SVN r23153.

The following SVN revision numbers were found above:
  r23152 --> open-mpi/ompi@ca3362021e
2010-05-17 21:07:57 +00:00
Jeff Squyres
ca3362021e Fix some problems noted by Ralph:
* Fix disabling hwloc build (i.e., put the AM_CONDITIONALs where they
   belong in the configure.m4 file)
 * Update some svn:ignores
 * r23142 removed some extraneous code, but forgot to remove the
   variables used only by that code

This commit was SVN r23152.

The following SVN revision numbers were found above:
  r23142 --> open-mpi/ompi@610fc67d12
2010-05-17 21:05:27 +00:00
Ralph Castain
cc8ebe7dd5 Protect against NULL when looking for an MCA param in an environment
This commit was SVN r23151.
2010-05-17 02:50:39 +00:00
Ralph Castain
12590202d8 Cleanup warnings
This commit was SVN r23148.
2010-05-16 20:22:00 +00:00
Ralph Castain
da170a7ab9 Turn off the blasted hwloc component as it generates a ton of garbage. Note that this means linux-based systems will -not- have paffinity for now since the good old plpa module was removed.
Clean up some missing ignores

This commit was SVN r23147.
2010-05-16 20:06:14 +00:00
Jeff Squyres
e2ab4f2baf Should be working now...
This commit was SVN r23143.
2010-05-14 15:20:47 +00:00
Jeff Squyres
610fc67d12 Oops -- don't convert to a processor ID here; just return the OS index
of the core.

This commit was SVN r23142.
2010-05-14 15:14:28 +00:00
Jeff Squyres
a27da2473a Ensure the whole directory is built.
This commit was SVN r23140.
2010-05-14 13:21:09 +00:00
Jeff Squyres
3ba4086b0f Remove another debugging message.
This commit was SVN r23139.
2010-05-14 13:20:46 +00:00
Jeff Squyres
a1848ef8d5 Arf. Ignore this component while I fix vpath builds...
This commit was SVN r23138.
2010-05-14 13:03:02 +00:00
Jeff Squyres
2d01a67516 Remove these generates files from SVN.
This commit was SVN r23137.
2010-05-14 11:58:17 +00:00
Jeff Squyres
8c8efa9bf3 Remove debugging message.
This commit was SVN r23136.
2010-05-14 11:57:43 +00:00
Jeff Squyres
21178f9379 Remove the "linux" paffinity component (i.e., the one that was based
on the now-defunct PLPA) -- the new hwloc component supersedes it.  

So long, PLPA -- we loved ya!

This commit was SVN r23126.
2010-05-13 23:59:21 +00:00
Jeff Squyres
3129ccd9ec Make the hwloc paffinity component available for everyone. hwloc
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.
2010-05-13 23:56:05 +00:00
Jeff Squyres
ca6d95a9c8 Clean up some comments; make paffinity/base/base.h comments agree with
paffinity/paffinity.h. 

This commit was SVN r23124.
2010-05-13 23:43:28 +00:00
Jeff Squyres
bf7954c1de Bump up to 1.0rc6 from the vendor branch.
This commit was SVN r23117.
2010-05-12 17:04:48 +00:00
Jeff Squyres
c7c3de87f5 Add ummunotify support to Open MPI. See
http://marc.info/?l=linux-mm-commits&m=127352503417787&w=2 for more
details.

 * Remove the ptmalloc memory component; replace it with a new "linux"
   memory component.
 * The linux memory component will conditionally compile in support
   for ummunotify.  At run-time, if it has ummunotify support and
   finds run-time support for ummunotify (i.e., /dev/ummunotify), it
   uses it.  If not, it tries to use ptmalloc via the glibc memory
   hooks. 
 * Add some more API functions to the memory framework to accomodate
   the ummunotify model (i.e., poll to see if memory has "changed").
 * Add appropriate calls in the rcache to the new memory APIs to see
   if memory has changed, and to react accordingly.
 * Add a few comments in the openib BTL to indicate why we don't need
   to notify the OPAL memory framework about specific instances of
   registered memory.
 * Add dummy API calls in the solaris malloc component (since it
   doesn't have polling/"did memory change" support).

This commit was SVN r23113.
2010-05-11 21:43:19 +00:00
Ralph Castain
d6a1d7a082 Little more cleanup on paffinity. Provide a specific error code for affinity not supported so we can better report the problem. Move the error reporting to orterun so we only get one error message. Update the darwin paffinity module to return the correct new error codes.
This commit was SVN r23107.
2010-05-07 14:04:55 +00:00
Ralph Castain
d4f56cff61 More cleanup on paffinity....groan
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.
2010-05-06 20:57:17 +00:00
Jeff Squyres
71cbe1a69f Bump up to hwloc v1.0rc3
This commit was SVN r23070.
2010-04-29 15:59:01 +00:00
Jeff Squyres
f064056a07 We don't need all this stuff in OMPI.
This commit was SVN r23056.
2010-04-28 00:31:15 +00:00
Jeff Squyres
2fe1bc043d Bump up to hwloc 1.0rc2
This commit was SVN r23042.
2010-04-26 21:57:51 +00:00
Ralph Castain
13a7338289 Ensure we get past the '=' in the parameter
This commit was SVN r23039.
2010-04-26 20:46:50 +00:00
Ralph Castain
e1b9f400ba Add some new utilities that support searching an environ string list (not just our own environ) for specific MCA params and returning their value. Helpful when a daemon needs to check an app_context's environ for params that can impact how the daemon launches and/or interacts with it, but don't pertain to the daemon's own environ.
This commit was SVN r23034.
2010-04-26 03:35:09 +00:00
Jeff Squyres
ea8b0ea569 Add a new function in the paffinity base:
opal_paffinity_base_cset2str().  This function basically makes a
prettyprint string out of an opal_paffinity_base_cset_t.

This commit was SVN r23017.
2010-04-21 17:26:36 +00:00
Jeff Squyres
53ab6600e6 Minor update to comments.
This commit was SVN r23013.
2010-04-20 20:59:42 +00:00
Jeff Squyres
f1d4a748eb Minor fix: pass by pointer to the new function so that the caller
can see the results.

This commit was SVN r23012.
2010-04-20 19:52:47 +00:00
Ralph Castain
7717c970a3 Ahem...it requires 2 hex chars to describe each byte of a bitmask...
This commit was SVN r23001.
2010-04-20 05:11:16 +00:00
Ralph Castain
86228aee38 Provide two new opal paffinity utilities for printing a hex representation of the cpu set and parsing that string back into a cpu set on the other end. Also add a new MCA param for passing the cpu set applied to a process during launch down to that process so it can know what we attempted to do.
All to be used in some new MPI extensions provided by Jeff so that users can easily query their binding situation.

This commit was SVN r22998.
2010-04-19 22:16:35 +00:00
Jeff Squyres
338920656f Remove the compile-time proiorities for paffinity modules (they were
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.
2010-04-19 22:14:06 +00:00
Jeff Squyres
9f5ddbcc6e 3rd party import hwloc 1.0rc1 into the SVN trunk
This commit was SVN r22996.
2010-04-19 19:48:58 +00:00
Jeff Squyres
8b163ccd70 Add dummy hwloc directory for staged import into svn
This commit was SVN r22994.
2010-04-19 19:43:43 +00:00
Ralph Castain
4d06125a33 Establish a method by which a process knows if it has been bound by mpirun. This helps resolve a problem where a process gets "bound" to all available resources, which looks to the opal paffinity system as "not bound". This can cause mpi_init to attempt to "bind" the process itself, causing unintended behavior.
This commit was SVN r22985.
2010-04-17 01:58:26 +00:00