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.
point, the event engine has been shut down until btl finalization is
done, so opal_progress in the wait loop is not an option - we have
to drain from inside the btl.
Clean up the looping structure for the finalize routine
Update copyrights.
This commit was SVN r21620.
IPv4 and IPv6) is outside the legal boundaries. This fixes trac:1869.
This commit was SVN r21612.
The following Trac tickets were found above:
Ticket 1869 --> https://svn.open-mpi.org/trac/ompi/ticket/1869
btl_sm.c: In function ‘mca_btl_sm_sendi’:
btl_sm.c:734: warning: comparison between signed and unsigned
btl_sm.c: In function ‘mca_btl_sm_send’:
btl_sm.c:812: warning: comparison between signed and unsigned
This commit was SVN r21552.
The following SVN revision numbers were found above:
r21551 --> open-mpi/ompi@bd995d26b4
- poll FIFO occasionally even if just sending messages
- retry pending sends more often
- just before trying a new send
- as part of mca_btl_sm_component_progress
Maintain two new mca_btl_sm_component variables, num_outstanding_frags
and num_pending_sends, to keep overhead low.
Drain only one message fragment from the FIFO per btl_sm_component_progress
call (rather than drain until empty, which in retrospect everyone considers
to have been a mistake).
This commit was SVN r21551.
This commit was SVN r21533.
The following SVN revisions from the original message are invalid or
inconsistent and therefore were not cross-referenced:
r21524
(now works on both big and little endian machines)
* Be a little more flexible when looking for active devices in
btl_openib_component.c
* Add device name and port number to lots of verbose and help
messages
* Add a bunch of verbose messages to give insight into what is
occurring during all the CPC wireups
This commit was SVN r21418.
Yes, friends, our favorite PCIE BTL has resurfaced as mgmt vacillates over its existence. This is an updated version that actually mostly works, in its final stages of debugging.
Some generalization still remains to be done...
This commit was SVN r21358.
shm_fifos values are only partially updated, and this leads to wrong values
for the offset. Moving the write barrier at the right place, plus forcing
some read barriers might help.
In addition I get rid of the sm_offset array which is completely useless.
This commit was SVN r21253.
for printing size_t use "%lu" and cast to (unsigned long).
This commit was SVN r21238.
The following SVN revision numbers were found above:
r21234 --> open-mpi/ompi@22b6177fb9
I just found that we have 2 place where we call for XRC domain
creation. First one in init_one_device() and second one prepare_device_for_use().
They have absolutely identical code, but the call in init_one_device() is useless
because on this stage we don't know about QP configuration and we don't know if we need
XRC at all. So I removing the duplicated code from init_one_device().
This commit was SVN r21235.
for us already.
* Slightly clarify the error message strings; now they match the new
error strings for btl_openib_ipaddr_in|exclude
This commit was SVN r21197.
subnet specifications (in addition to interface names). These
parameters now take a comma-delimited list of interfaces names and/or
a.b.c.d/x specifications (only IPv4 currently supported for subnet
specifications). For example:
mpirun --mca btl_tcp_if_include 10.10.30.0/8,eth0
This commit was SVN r21189.
OMPI_* to OPAL_*. This allows opal layer to be used more independent
from the whole of ompi.
NOTE: 9 "svn mv" operations immediately follow this commit.
This commit was SVN r21180.
- 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.
several header files (previously included by header-files)
now have to be moved "upward".
This is mainly system headers such as string.h, stdio.h and for
networking, but also some orte headers.
This commit was SVN r21095.
and mo'bettah. Put in lengthy comments explaining what's going on.
We might still want to tweak this some more, but we can no longer get
IMB-EXT to hang with this new code anymore (e.g., even without eager
RDMA -- we discovered after the fact that the code in the v1.3.2
release will hang if eager RDMA is disabled).
Fixes trac:1890. Really.
This commit was SVN r21061.
The following Trac tickets were found above:
Ticket 1890 --> https://svn.open-mpi.org/trac/ompi/ticket/1890
_endpoint_post_send(), which could result in an infinite loop (see the
comment in the code).
This is part one of a proper fix; it's suitable for the v1.3 tree and
for an immediate release. Pasha and I plan to spend a little more
time and clean up this stuff properly, but it does not need to be
included in v1.3.2.
This commit was SVN r21047.
The following Trac tickets were found above:
Ticket 1890 --> https://svn.open-mpi.org/trac/ompi/ticket/1890
and https://svn.open-mpi.org/trac/ompi/ticket/1853, mallopt() hints do
not always work -- it is possible for memory to be returned to the OS
and therefore OMPI's registration cache becomes invalid.
This commit removes all use of mallopt() and uses a different way to
integrate ptmalloc2 than we have done in the past. In particular, we
use almost exactly the same technique as MX:
* Remove all uses of mallopt, to include the opal/memory mallopt
component.
* Name-shift all of OMPI's internal ptmalloc2 public symbols (e.g.,
malloc -> opal_memory_ptmalloc2_malloc).
* At run-time, use the existing glibc allocator malloc hook function
pointers to fully hijack the glibc allocator with our own
name-shifted ptmalloc2.
* Make the decision whether to hijack the glibc allocator ''at run
time'' (vs. at link time, as previous ptmalloc2 integration
attempts have done). Look at the OMPI_MCA_mpi_leave_pinned
and OMPI_MCA_mpi_leave_pinned_pipeline environment variables and
the existence of /sys/class/infiniband to determine if we should
install the hooks or not.
* As an added bonus, we can now tell if libopen-pal is linked
statically or dynamically, and if we're linked statically, we
assume that munmap intercept support doesn't work.
See the opal/mca/memory/ptmalloc2/README-open-mpi.txt file for all the
gory details about the implementation.
Fixes trac:1853.
This commit was SVN r20921.
The following Trac tickets were found above:
Ticket 1853 --> https://svn.open-mpi.org/trac/ompi/ticket/1853
after each process create it's FIFOs but before they access the
peer's FIFOs. Second, replace a one way synchronization by a real
barrier, so we know that every process is really where we expect
them to be.
This commit was SVN r20906.
* Ensure we don't try to do opal_list_get_next() on an item we just
deleted
* set myaddrs = NULL when we're done with it, just for good measure
Once this is ported to OMPI v1.3 branch, it fixes
https://bugs.openfabrics.org/show_bug.cgi?id=1579.
This commit was SVN r20896.
Add two new configure options that specify:
1. when to add padding to the openib control header - this *only* happens when the configure option is specified
2. when to use the dr-like checksum as opposed to the memcpy checksum. Not selectable at runtime - to eliminate performance impacts, this is a configure-only option
Also removed an unused checksum version from opal/util/crc.h.
The new component still needs a little cleanup and some sync with recent ob1 bug fixes. It was created as a separate module to avoid performance hits in ob1 itself, though most of the code is duplicative. The component is only selectable by either specifying it directly, or configuring with the dr-like checksum -and- setting -mca pml_csum_enable_checksum 1.
Modify the LANL platform files to take advantage of the new module.
This commit was SVN r20846.
- This patch solely _adds_ required headers and is rather localized
The next patch (after RFC) heavily removes headers (based on script)
- ompi/communicator/communicator.h: For sources that use
ompi_mpi_comm_world, don't require them to include "mpi.h"
- ompi/debuggers/ompi_common_dll.c: mca_topo_base_comm_1_0_0_t needs
#include "ompi/mca/topo/topo.h"
- ompi/errhandler/errhandler_predefined.h:
ompi/communicator/communicator.h depends on this header file!
To prevent recursion just have fwd declarations.
#include "ompi/types.h" for fwd declarations of the main structs.
- ompi/mca/btl/btl.h: #include "opal/types.h" for ompi_ptr_t
- ompi/mca/mpool/base/mpool_base_tree.c: We use ompi_free_list_t and
ompi_rb_tree_t, so have the proper classes
- ompi/mca/op/op.h:
Op is pretty self-contained: Nobody up to now has done
#include "opal/class/opal_object.h"
- ompi/mca/osc/pt2pt/osc_pt2pt_replyreq.h:
#include "opal/types.h" for ompi_ptr_t
- ompi/mca/pml/base/base.h:
We use opal_lists
- ompi/mca/pml/dr/pml_dr_vfrag.h:
#include "opal/types.h" for ompi_ptr_t
- ompi/mca/pml/ob1/pml_ob1_hdr.h:
#include "ompi/mca/btl/btl.h" for mca_btl_base_segment_t
- opal/dss/dss_unpack.c:
#include "opal/types.h"
- opal/mca/base/base.h:
#include "opal/util/cmd_line.h" for opal_cmd_line_t
- orte/mca/oob/tcp/oob_tcp.c:
#include "opal/types.h" for opal_socklen_t
- orte/mca/oob/tcp/oob_tcp.h:
#include "opal/threads/threads.h" for opal_thread_t
- orte/mca/oob/tcp/oob_tcp_msg.c:
#include "opal/types.h"
- orte/mca/oob/tcp/oob_tcp_peer.c:
#include "opal/types.h" for opal_socklen_t
- orte/mca/oob/tcp/oob_tcp_send.c:
#include "opal/types.h"
- orte/mca/plm/base/plm_base_proxy.c:
#include "orte/util/name_fns.h" for ORTE_NAME_PRINT
- orte/mca/rml/base/rml_base_receive.c:
#include "opal/util/output.h" for OPAL_OUTPUT_VERBOSE
- orte/mca/rml/oob/rml_oob_recv.c:
#include "opal/types.h" for ompi_iov_base_ptr_t
- orte/mca/rml/oob/rml_oob_send.c:
#include "opal/types.h" for ompi_iov_base_ptr_t
- orte/runtime/orte_data_server.c
#include "opal/util/output.h" for OPAL_OUTPUT_VERBOSE
- orte/runtime/orte_globals.h:
#include "orte/util/name_fns.h" for ORTE_NAME_PRINT
Tested on Linux/x86-64
This commit was SVN r20817.
without mpi.h we have no notion of MPI_SUCCESS...
- ompi/mca/btl/sm/btl_sm.h: ptrdiff_t needs stddef.h
- ompi/mca/mpool/base/: If we use opal_pointer_array_t,
better include the class header.
This commit was SVN r20816.
Adapt orte_process_info to orte_proc_info, and
change orte_proc_info() to orte_proc_info_init().
- Compiled on linux-x86-64
- Discussed with Ralph
This commit was SVN r20739.
Anyway, this is blocking the move: do not include pml.h
if not really needed, aka none of the following used:
mca_pml
MCA_PML_CALL
OMPI_ANY_TAG
OMPI_ANY_SOURCE
OMPI_PROC_NULL
- Notable exceptions (deleting in one header->adding):
- ompi/mca/mtl/psm/
- ompi/mca/osc/rdma/
- ompi/mca/btl/openib/btl_openib_endpoint.c depended on
pml_base_sendreq.h
- Tested on Linux/x86-64, this time including make check
(thanks Jeff and Ralph)
This commit was SVN r20725.
get bitten by header depending on having already included
the corresponding [opal|orte|ompi]_config.h header.
When separating, things like [OPAL|ORTE|OMPI]_DECLSPEC
are missed.
Script to add the corresponding header in front of all following
(taking care of possible #ifdef HAVE_...)
- Including some minor cleanups to
- ompi/group/group.h -- include _after_ #ifndef OMPI_GROUP_H
- ompi/mca/btl/btl.h -- nclude _after_ #ifndef MCA_BTL_H
- ompi/mca/crcp/bkmrk/crcp_bkmrk_btl.c -- still no need for
orte/util/output.h
- ompi/mca/pml/dr/pml_dr_recvreq.c -- no need for mpool.h
- ompi/mca/btl/btl.h -- reorder to fit
- ompi/mca/bml/bml.h -- reorder to fit
- ompi/runtime/ompi_mpi_finalize.c -- reorder to fit
- ompi/request/request.h -- additionally need ompi/constants.h
- Tested on linux/x86-64
This commit was SVN r20720.
opal layer.
Add a check against a maximum (actually get rid of ifs internally to
opal_bitmap.c) -- the functionality to set the current maximum size
opal_bitmap_set_max_size() is currently only used in attribute.c
to set the maximum OMPI_FORTRAN_HANDLE_MAX...
Tested on linux/x86-64 with intel-tests with all_tests_no_perf_f
run with 6 procs.
Let's look into MTT as well...
This commit was SVN r20708.
a notifier module.
The Notifier framework was extended slightly to
convey more information about each event notice.
This works with the FTB v0.5 API.
To compile with FTB support, use --with-ftb=/path/to/ftb/install
CIFTS == Coordinated Infrastructure for Fault Tolerant Systems
FTB == Fault Tolerance Backplane
see http://wiki.mcs.anl.gov/cifts/index.php
This commit was SVN r20655.
Added a few comments and changed the return code after the FIFO write to be SUCCESS,
even if the FIFO write indicated an error. Such an error would only mean that the
FIFO was full, but the FIFO-write operation would still be queued. Therefore, the
PML should think of this as successful.
This commit was SVN r20644.
anyhow -- if oob functionality is neededm then orte/mca/oob/oob.h
Nevertheless compiles fine with -Wimplicit-function-declaration
This commit was SVN r20641.
Only proc_info.h-internal include file is opal/dss/dss_types.h
- In one case (orte/util/hnp_contact.c) had to add proc_info.h again.
- Local compilation (Linux/x86_64) w/ -Wimplicit-function-declaration
works fine, no errors.
Again, let's have MTT the last word.
This commit was SVN r20631.
* compute mmap-file size more wisely and pass requested size to allocator
* change MCA parameters:
- get rid of mpool_sm_per_peer_size
- get rid of mpool_sm_max_size
- set default mpool_sm_min_size to 0
* no longer pad sm allocations to page boundaries
* have sm_btl_first_time_init check return codes on free-list creations
Have mca_btl_sm_prepare_src() check to see if it can allocate an EAGER fragment
rather than a MAX fragment if the smaller size works.
Remove ompi/class/ompi_[circular_buffer_]fifo.h and references thereto.
Remove opal/util/pow2.[c|h] and references thereto.
This commit was SVN r20614.
* The main thing done here is to convert from multiple FIFOs/queues per
receiver (each receiver has one FIFO for each sender) to a single FIFO/queue
per receiver (all senders sharing the same FIFO for a given receiver).
* This requires rewriting the FIFO support, so that
ompi/class/ompi_[circular_buffer_]fifo.h is no longer used and FIFO
support is instead in btl_sm.h.
* The number of FIFOs per receiver is actually an MCA tunable parameter,
but it appears that 1 or possibly 2 FIFOs (even for 112 local processes)
per receiver is sufficient.
This commit was SVN r20578.
Often, orte/util/show_help.h is included, although no functionality
is required -- instead, most often opal_output.h, or
orte/mca/rml/rml_types.h
Please see orte_show_help_replacement.sh commited next.
- Local compilation (Linux/x86_64) w/ -Wimplicit-function-declaration
actually showed two *missing* #include "orte/util/show_help.h"
in orte/mca/odls/base/odls_base_default_fns.c and
in orte/tools/orte-top/orte-top.c
Manually added these.
Let's have MTT the last word.
This commit was SVN r20557.
by r20496 for the sm BTL, openib BTL on iWarp, and the sm & sm2 coll modules.
This commit was SVN r20515.
The following SVN revision numbers were found above:
r20496 --> open-mpi/ompi@4cdf91a8d4
The prior ompi_proc_t structure had a uint8_t flag field in it, where only one
bit was used to flag that a proc was "local". In that context, "local" was
constrained to mean "local to this node".
This commit provides a greater degree of granularity on the term "local", to include tests
to see if the proc is on the same socket, PC board, node, switch, CU (computing
unit), and cluster.
Add #define's to designate which bits stand for which local condition. This
was added to the OPAL layer to avoid conflicting with the proposed movement of
the BTLs. To make it easier to use, a set of macros have been defined - e.g.,
OPAL_PROC_ON_LOCAL_SOCKET - that test the specific bit. These can be used in
the code base to clearly indicate which sense of locality is being considered.
All locations in the code base that looked at the current proc_t field have
been changed to use the new macros.
Also modify the orte_ess modules so that each returns a uint8_t (to match the
ompi_proc_t field) that contains a complete description of the locality of this
proc. Obviously, not all environments will be capable of providing such detailed
info. Thus, getting a "false" from a test for "on_local_socket" may simply
indicate a lack of knowledge.
This commit was SVN r20496.
we remembered to use strcasecmp() every time I see this entry in the
file... (we did, but I just don't want to have to keep remembering
that ;-) )
This commit was SVN r20461.
btl_openib_connect_rdmacm_reject_causes_connect_error (yes, it's
still long -- on purpose :-) )
* Add INI file parameter rdmacm_reject_causes_connect_error
* Now only treat CONNECT_ERROR events as a REJECT if:
* It's on a connection where we were expecting a REJECT, ''and''
* The MCA parameter is true ''or'' the INI parameter for this
device is true
* Set the INI parameter for true for the NE020
This commit was SVN r20459.
versions of the NE driver will report the OUI while others will report
the PCI ID. We'll put in the Intel values when we get them (may not
be for a few more weeks).
This commit was SVN r20457.
properly at all. NetEffect's current driver (OFED 1.4.0) will return
a CONNECT_ERROR event to the initiator rather than the REJECTED event.
Doh! Additionally -- unfortunately -- NetEffect's vendor_id and
vendor_part_id are reported as 0 in OFED 1.4.0, so we can't
automatically detect these cards and work around the problem. So all
we can do is add a new MCA parameter
(btl_openib_connect_rdmacm_ignore_connect_errors -- yes, it's long on
purpose ;-) ) that says that if we get a CONNECT_ERROR, bascially
treat it exactly as a REJECT for the WRONG_DIRECTION reason (which is
a "good" reject). This allows OMPI to function with NetEffect/Intel
cards on OFED 1.4.0.
Note that NetEffect has been bought by Intel; I'm waiting for
information from them to update the ini file for their new OUI/PCI
ID's and/or new vendor_part_id values.
This commit was SVN r20454.
problem seems to come from the free list, but due to lack of time to
understand it completely, I provide this fix. Basically, there is no
waiting in the MX BTL anymore, if we cannot allocate a fragment we
rely on the PML to take the corrective actions.
This commit was SVN r20124.
1. minor modification to include two new opal MCA params:
(a) opal_profile: outputs what components were selected by each framework
currently enabled for most, but not all, frameworks
(b) opal_profile_file: name of file that contains profile info required
for modex
2. introduction of two new tools:
(a) ompi-probe: MPI process that simply calls MPI_Init/Finalize with
opal_profile set. Also reports back the rml IP address for all
interfaces on the node
(b) ompi-profiler: uses ompi-probe to create the profile_file, also
reports out a summary of what framework components are actually
being used to help with configuration options
3. modification of the grpcomm basic component to utilize the
profile file in place of the modex where possible
4. modification of orterun so it properly sees opal mca params and
handles opal_profile correctly to ensure we don't get its profile
5. similar mod to orted as for orterun
6. addition of new test that calls orte_init followed by calls to
grpcomm.barrier
This is all completely benign unless actively selected. At the moment, it only supports modex-less launch for openib-based systems. Minor mod to the TCP btl would be required to enable it as well, if people are interested. Similarly, anyone interested in enabling other BTL's for modex-less operation should let me know and I'll give you the magic details.
This seems to significantly improve scalability provided the file can be locally located on the nodes. I'm looking at an alternative means of disseminating the info (perhaps in launch message) as an option for removing that constraint.
This commit was SVN r20098.
of more than one of the btl_openib_if_include, btl_openib_if_exclude,
btl_openib_ipaddr_include, or btl_openib_ipaddr_exclude MCA parameters.
This commit was SVN r20053.
determining which IP address to use when transmitting data. Also it adds logic
to prevent usage of more than one of the btl_openib_if_include,
btl_openib_if_exclude, btl_openib_ipaddr_include, or btl_openib_ipaddr_exclude
MCA parameters.
This should complete the code modifications needed for ticket 1665.
This commit was SVN r20052.
determing of the IP subnet. The netmask was being used improperly when
determining which subnet each connection is on. Part two is the ability to
include/exclude specific subnets.
This patch fixes ticket #1665
This commit was SVN r20016.
* If max_inline_data == -1 perform runtime detection
* If max_inline_data >=0 use the value provided
* If the user does not explicitly set this via command line, use the value from INI file
This commit fixes trac:1662
This commit was SVN r19995.
The following Trac tickets were found above:
Ticket 1662 --> https://svn.open-mpi.org/trac/ompi/ticket/1662
This will sit in trunk for a few days - would like to actually see some errors reported to syslog before moving the code to 1.3
This commit was SVN r19986.
another bug. This also causes 0-byte requests to be treated as a buffer
error, causing the base request to be requeued. On Cray XT, it may be
temporarily impossible to make allocations for buffer requests, as the
default stack size is small (8 MB) and there is no true swap device.
Even with the stack size increased, there will be cases in which this
condition recurs.
One possibility is to make the buffer allocations off of the heap; but,
this does not change the fact that eventually an out-of-memory condition
will occur and we need to support multiple receives in transit, a
condition for which the available buffer space may change. On the other
hand, if we switch to allocating the buffer space from the heap, we will
need to return an error when the allocation fails and there are no other
buffers in transit.
This commit was SVN r19981.
