The osc/rdma module did not wait for all pending atomics to complete
before tearing down. This could lead to weird issues as the target
location may no longer be registered or allocated.
This commit also fixes an offset calculation issue in
ompi_osc_get_data_blocking ().
Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
This commit is a large update to the osc/rdma component. Included in
this commit:
- Add support for using hardware atomics for fetch-and-op and single
count accumulate when using the accumulate lock. This will improve
the performance of these operations even when not setting the
single intrinsic info key.
- Rework how large accumulates are done. They now block on the get
operation to fix some bugs discovered by an IBM one-sided test. I
may roll back some of the changes if the underlying bug in the
original design is discovered. There appear to be no real
difference (on the hardware this was tested with) in performance so
its probably a non-issue. References #2530.
- Add support for an additional lock-all algorithm: on-demand. The
on-demand algorithm will attempt to acquire the peer lock when
starting an RMA operation. The lock algorithm default has not
changed. The algorithm can be selected by setting the
osc_rdma_locking_mode MCA variable. The valid values are two_level
and on_demand.
- Make use of the btl_flush function if available. This can improve
performance with some btls.
- When using btl_flush do not keep track of the number of put
operations. This reduces the number of atomic operations in the
critical path.
- Make the window buffers more friendly to multi-threaded
applications. This was done by dropping support for multiple
buffers per MPI window. I intend to re-add that support once the
underlying performance bug under the old buffering scheme is
fixed.
- Fix a bug in request completion in the accumulate, get, and put
paths. This also helps with #2530.
- General code cleanup and fixes.
Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
This commit renames the arithmetic atomic operations in opal to
indicate that they return the new value not the old value. This naming
differentiates these routines from new functions that return the old
value.
Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
This commit fixes the following bugs:
- Allow a btl to be used for communication if it can communicate with
all non-self peers and it supports global atomic visibility. In
this case CPU atomics can be used for self and the btl for any
other peer.
- It was possible to get into a state where different threads of an
MPI process could issue conflicting accumulate operations to a
remote peer. To eliminate this race we now update the peer flags
atomically.
- Queue up and re-issue put operations that failed during a BTL
callback. This can occur during an accumulate operation. This was
an unhandled error case.
Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
a buffer defined by (buf, count, dt)
will have data starting at buf+offset and ending len bytes later with
len = opal_datatype_span(&dt.super, count, &offset);
Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>
The osc_rdma_get_remote_segment() has the 3rd and 4th args as
* target_disp
* length
which it uses to determine if the rdma falls within the bounds of
the window or not (actually it only checks the upper bound, but I'm
okay with that).
Anyway the caller previously was passing in the length argument as
target_datatype->super.size * target_count
which which doesn't really represent the number of bytes after target_disp
for which data exists. In particular I could create a datatype as
{ disp -4, len 4 } and use target_disp 4
and that would be bytes 0-3 of the window where the original code
would think it was bytes 4-7 and could abort at the range check.
Ive changed it to use the opal_datatype_span() function.
Signed-off-by: Mark Allen <markalle@us.ibm.com>
since Open MPI now requires a C99, and ptrdiff_t type is part of C99,
there is no more need for the abstract OPAL_PTRDIFF_TYPE type.
Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>
Instead of ompi_datatype_get_extent(), use ompi_datatype_get_true_extent()
to get the local and remote lower bound. For derived types like
subarray, true_lb is the correct offset for RDMA operations.
This commit adds support for using network AMOs for MPI_Accumulate,
MPI_Fetch_and_op, and MPI_Compare_and_swap. This support is only
enabled if the ompi_single_intrinsic info key is specified or the
acc_single_interinsic MCA variable is set. This configuration
indicates to this implementation that no long accumulates will be
performed since these do not currently mix with the AMO
implementation.
This commit also cleans up the code somwhat. This includes removing
unnecessary struct keywords where the type is also typedef'd.
Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
This commit adds code to handle large unaligned gets. There are two
possible code paths for these transactions:
1) The remote region and local region have the same alignment. In
this case the get will be broken down into at most three get
transactions: 1 transaction to get the unaligned start of the region
(buffered), 1 transaction to get the aligned portion of the region,
and 1 transaction to get the end of the region.
2) The remote and local regions do not have the same alignment. This
should be an uncommon case and is not optimized. In this case a
buffer is allocated and registered locally to hold the aligned data
from the remote region. There may be cases where this fails (low
memory, can't register memory). Those conditions are unlikely and
will be handled later.
Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
Optimizing put aggregation in the presence of threads will require a
redesign of the code. For now just ensure that put aggregation is
turned off when MPI_THREAD_MULTIPLE is enabled.
Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
This commit fixes several bugs in the osc/rdma component:
- Complete aggregated requests immediately. Completion of RMA
requests indicates local completion anyway. This fixes a hang in
the c_reqops test.
- Correctly mark Rget_accumulate requests.
- Set the local base flag correctly on the local peer.
- Clear or set the no locks flag on the window if the value is
changed by MPI_Win_set_info.
- Actually update the target when using MPI_OP_REPLACE.
Fixesopen-mpi/ompi#1010
Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
This commit adds support for performing one-sided operations over
supported hardware (currently Infiniband and Cray Gemini/Aries). This
component is still undergoing active development.
Current features:
- Use network atomic operations (fadd, cswap) for implementing
locking and PSCW synchronization.
- Aggregate small contiguous puts.
- Reduced memory footprint by storing window data (pointer, keys,
etc) at the lowest rank on each node. The data is fetched as each
process needs to communicate with a new peer. This is a trade-off
between the performance of the first operation on a peer and the
memory utilization of a window.
TODO:
- Add support for the accumulate_ops info key. If it is known that
the same op or same op/no op is used it may be possible to use
hardware atomics for fetch-and-op and compare-and-swap.
Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
osc/rdma uses counters to determine if all messages have been received
before exiting synchronization calls. The problem is that the active
target counter is always increasing (never zeroed). If over 2^31-1
messages are sent this causes the counter to overflow (in itself this
isn't an error). This causes test/wait to return before the communication
is complete. There is an additional error in the use of the fragment
flush function. If PSCW synchronization is in use this function CAN NOT
be called unless a post message has arrived.
Relevant mailing list thread: http://www.open-mpi.org/community/lists/devel/2014/10/16016.php
This commit fixes both issues. Tested against MTT and issue reproducer.
Closes#224.
This commit adds a check to see if the target is in an access epoch. If
not we return OMPI_ERR_RMA_SYNC. This fixes test_start3 in the onesided
test suite. The cost of this extra check is 1 byte/peer for the boolean
flag indicating that the peer is in an access epoch.
I also fixed a problem where mupliple unexpected post messages are not
correctly handled.
cmr=v1.8.2:reviewer=jsquyres
This commit was SVN r32160.
The wrong type was used when calculating the amount of space needed
for an accumulate fragment. Fixed the calculation and took the
opportunity to eliminate the get_acc header as it is identical to the
acc header.
This fixes trac:4719 and #4718
Tracking these fixes for 1.8.2 in this CMR.
Throwing this to Brad for review as he is the one who ran into the issue.
cmr=v1.8.2:reviewer=bbenton
This commit was SVN r32015.
The following Trac tickets were found above:
Ticket 4719 --> https://svn.open-mpi.org/trac/ompi/ticket/4719
in self optimization
This addresses an issue found with the MPICH pscw_ordering test. Eager sends
were not yet active (which is ok for the standard path) but not ok for the
self optimization. Fixed by waiting for all post messages before checking
the sync state.
Fixes trac:4724
Tracking the 1.8.2 issue in this CMR.
cmr=v1.8.2:reviewer=bbenton
This commit was SVN r32012.
The following Trac tickets were found above:
Ticket 4724 --> https://svn.open-mpi.org/trac/ompi/ticket/4724
When sending PUT_LONG, the data is sent before headers, and sometimes
the header is not flushed immediately. This creates a lot of unexpected
receives in the peer, since it would posts a receive only when gets the
header, which makes it run out of receive buffers. When the sender
eventually flushes the window, the receiver already has no buffers to
receive the header, which causes a deadlock.
The fix is to always flush the headers when doing put_long.
cmr=v1.8.1:reviewer=hjelmn
This commit was SVN r31378.
This commit adds large datatype description support to the osc/rdma
component. Support is provided by an additional send/recv of the datatype
description if the description does not fit in an eager buffer. The
code is designed to require minimal new code and not for speed. We
consider this code path to be a slow path.
Refs trac:1905
cmr=v1.8:reviewer=jsquyres
This commit was SVN r31197.
The following Trac tickets were found above:
Ticket 1905 --> https://svn.open-mpi.org/trac/ompi/ticket/1905
Dave Goodell correctly pointed out that it is unusual to return MPI
error classes from internal ompi functions. Correct this in the RMA
case by adding an internal error code to match MPI_ERR_RMA_SYNC.
This does change OMPI_ERR_MAX. I don't think this will cause any
problems with ABI.
cmr=v1.7.5:reviewer=jsquyres
This commit was SVN r31012.
Roll in the ORTE state machine. Remove last traces of opal_sos. Remove UTK epoch code.
Please see the various emails about the state machine change for details. I'll send something out later with more info on the new arch.
This commit was SVN r26242.
operations, not ints.
Sorry for the mid-day configure.ac change, folks...
This commit was SVN r23449.
The following Trac tickets were found above:
Ticket 2472 --> https://svn.open-mpi.org/trac/ompi/ticket/2472
(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.
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.
to happen
* Properly error out (rather than cause buffer overflow) in case where
the datatype packed description is larger than our control fragments.
This still isn't standards conforming, but at least we know what
happened.
* Expose win_set_name to external libraries (like the osc modules)
* Set default window name to the CID of the communcator it's using
for communication
Refs trac:1905
This commit was SVN r21134.
The following Trac tickets were found above:
Ticket 1905 --> https://svn.open-mpi.org/trac/ompi/ticket/1905
* If the accumulate is local, make it short-circuit the request path. Accumulate requires local
ops due to its window rules, so this is likely to help a bunch (on the codes I"m messing
with at least)
* Due a better job at flushing everything that can go out on the wire in a resource constrained problem
* Move some debugging values around to make large problems somewhat easier to deal with
This commit was SVN r20277.
After much work by Jeff and myself, and quite a lot of discussion, it has become clear that we simply cannot resolve the infinite loops caused by RML-involved subsystems calling orte_output. The original rationale for the change to orte_output has also been reduced by shifting the output of XML-formatted vs human readable messages to an alternative approach.
I have globally replaced the orte_output/ORTE_OUTPUT calls in the code base, as well as the corresponding .h file name. I have test compiled and run this on the various environments within my reach, so hopefully this will prove minimally disruptive.
This commit was SVN r18619.
such, the commit message back to the master SVN repository is fairly
long.
= ORTE Job-Level Output Messages =
Add two new interfaces that should be used for all new code throughout
the ORTE and OMPI layers (we already make the search-and-replace on
the existing ORTE / OMPI layers):
* orte_output(): (and corresponding friends ORTE_OUTPUT,
orte_output_verbose, etc.) This function sends the output directly
to the HNP for processing as part of a job-specific output
channel. It supports all the same outputs as opal_output()
(syslog, file, stdout, stderr), but for stdout/stderr, the output
is sent to the HNP for processing and output. More on this below.
* orte_show_help(): This function is a drop-in-replacement for
opal_show_help(), with two differences in functionality:
1. the rendered text help message output is sent to the HNP for
display (rather than outputting directly into the process' stderr
stream)
1. the HNP detects duplicate help messages and does not display them
(so that you don't see the same error message N times, once from
each of your N MPI processes); instead, it counts "new" instances
of the help message and displays a message every ~5 seconds when
there are new ones ("I got X new copies of the help message...")
opal_show_help and opal_output still exist, but they only output in
the current process. The intent for the new orte_* functions is that
they can apply job-level intelligence to the output. As such, we
recommend that all new ORTE and OMPI code use the new orte_*
functions, not thei opal_* functions.
=== New code ===
For ORTE and OMPI programmers, here's what you need to do differently
in new code:
* Do not include opal/util/show_help.h or opal/util/output.h.
Instead, include orte/util/output.h (this one header file has
declarations for both the orte_output() series of functions and
orte_show_help()).
* Effectively s/opal_output/orte_output/gi throughout your code.
Note that orte_output_open() takes a slightly different argument
list (as a way to pass data to the filtering stream -- see below),
so you if explicitly call opal_output_open(), you'll need to
slightly adapt to the new signature of orte_output_open().
* Literally s/opal_show_help/orte_show_help/. The function signature
is identical.
=== Notes ===
* orte_output'ing to stream 0 will do similar to what
opal_output'ing did, so leaving a hard-coded "0" as the first
argument is safe.
* For systems that do not use ORTE's RML or the HNP, the effect of
orte_output_* and orte_show_help will be identical to their opal
counterparts (the additional information passed to
orte_output_open() will be lost!). Indeed, the orte_* functions
simply become trivial wrappers to their opal_* counterparts. Note
that we have not tested this; the code is simple but it is quite
possible that we mucked something up.
= Filter Framework =
Messages sent view the new orte_* functions described above and
messages output via the IOF on the HNP will now optionally be passed
through a new "filter" framework before being output to
stdout/stderr. The "filter" OPAL MCA framework is intended to allow
preprocessing to messages before they are sent to their final
destinations. The first component that was written in the filter
framework was to create an XML stream, segregating all the messages
into different XML tags, etc. This will allow 3rd party tools to read
the stdout/stderr from the HNP and be able to know exactly what each
text message is (e.g., a help message, another OMPI infrastructure
message, stdout from the user process, stderr from the user process,
etc.).
Filtering is not active by default. Filter components must be
specifically requested, such as:
{{{
$ mpirun --mca filter xml ...
}}}
There can only be one filter component active.
= New MCA Parameters =
The new functionality described above introduces two new MCA
parameters:
* '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that
help messages will be aggregated, as described above. If set to 0,
all help messages will be displayed, even if they are duplicates
(i.e., the original behavior).
* '''orte_base_show_output_recursions''': An MCA parameter to help
debug one of the known issues, described below. It is likely that
this MCA parameter will disappear before v1.3 final.
= Known Issues =
* The XML filter component is not complete. The current output from
this component is preliminary and not real XML. A bit more work
needs to be done to configure.m4 search for an appropriate XML
library/link it in/use it at run time.
* There are possible recursion loops in the orte_output() and
orte_show_help() functions -- e.g., if RML send calls orte_output()
or orte_show_help(). We have some ideas how to fix these, but
figured that it was ok to commit before feature freeze with known
issues. The code currently contains sub-optimal workarounds so
that this will not be a problem, but it would be good to actually
solve the problem rather than have hackish workarounds before v1.3 final.
This commit was SVN r18434.
r14703 for the point-to-point component.
* Associate the list of long message requests to poll with the
component, not the individual modules
* add progress thread that sits on the OMPI request structure
and wakes up at the appropriate time to poll the message
list to move long messages asynchronously.
* Instead of calling opal_progress() all over the place, move
to using the condition variables like the rest of the project.
Has the advantage of moving it slightly further along in the
becoming thread safe thing.
* Fix a problem with the passive side of unlock where it could
go recursive and cause all kinds of problems, especially
when progress threads are used. Instead, have two parts of
passive unlock -- one to start the unlock, and another to
complete the lock and send the ack back. The data moving
code trips the second at the right time.
This commit was SVN r14751.
The following SVN revision numbers were found above:
r14703 --> open-mpi/ompi@2b4b754925
tell if the remote proc should be in an exposure epoch or not.
Refs trac:325
This commit was SVN r11746.
The following Trac tickets were found above:
Ticket 325 --> https://svn.open-mpi.org/trac/ompi/ticket/325
implemented entirely on top of the PML. This allows us to have a
one-sided interface even when we are using the CM PML and MTLs for
point-to-point transport (and therefore not using the BML/BTLs)
* Old pt2pt component was renamed "rdma", as it will soon be having
real RDMA support added to it.
Work was done in a temporary branch. Commit is the result of the
merge command:
svn merge -r10862:11099 https://svn.open-mpi.org/svn/ompi/tmp/bwb-osc-pt2pt
This commit was SVN r11100.
The following SVN revisions from the original message are invalid or
inconsistent and therefore were not cross-referenced:
r10862
r11099
