The previous version of the protocol assumed that fi_trecv utilizes
direct address for matching. This does not hold, for example, for PSM
OFI provider (it does not set claim FI_DIRECTED_RECV capability). So,
the API1 set should work w/o directed recv assumption. Hence, we have
to encapsulate rank info in the tag in the RCD protocol in order to
avoid mismatch.

Change-Id: I7868f2f033cf3cdec29f7daded8b29c275b2632f
Signed-off-by: Charles J Archer <charles.j.archer@intel.com>

Change-Id: I25d9c8e80329dc185ee89da9fa1d8d9f54ea493e
Signed-off-by: Jithin Jose <jithin.jose@intel.com>
Signed-off-by: Charles J Archer <charles.j.archer@intel.com>

htmldir was not created before installing index.html. This caused
`make install` to fail on FreeBSD.  This patch fixes this issue.

Signed-off-by: Junchao Zhang <jczhang@mcs.anl.gov>

Some special options might be added by libtool, such as -Wl,-flat_namespace -Wl,-commons,use_dylibs,
and we do not want to drop them even when interlibrary deps is on.

Also reference [eb0e7712

]

Fixes #2271

Signed-off-by: Pavan Balaji <balaji@anl.gov>

Signed-off-by: Halim Amer <aamer@anl.gov>

Lena Oden authored Jun 02, 2015 and Pavan Balaji committed Jun 04, 2015

Add a test to configure.ac that tests, if type_tag_for_datatype is
agnostic for modifiers as const or volatile.  If the test fails, the
type_tag_for_datatype test is disabled in mpi.h by setting a flag to
avoid compiler warning.  In this case, there are no type tests for MPI
data types when compiling.

Fixes #2223.

Signed-off-by: Pavan Balaji <balaji@anl.gov>

In e50bec3a

, we add code in Nemesis
and TCP/Portals4/MXM netmods to issue extended packet header stored
in the request. In this patch, we did the similiar modification to
OFI netmod.

Signed-off-by: Pavan Balaji <balaji@anl.gov>

Halim Amer authored May 31, 2015 and Kenneth Raffenetti committed Jun 02, 2015

The implementation of Ibsend had several issues: 1) incorrect ref
counting when canceling a request, which leads to handle leaks; 2)
writing MPI_REQUEST_NULL directly in the handle field of a request
(MPID_Request) in the request pool because the handle field is passed
by reference to MPIR_Wait_impl; 3) the memory used in the attached
buffer is never freed after a successful cancellation

This patch fixes all the above mentioned issues. In addition it
fixes and closes ticket:287.

Signed-off-by: Ken Raffenetti <raffenet@mcs.anl.gov>

Pavan Balaji authored Jun 02, 2015 and Kenneth Raffenetti committed Jun 02, 2015

These tests are not passing yet, but the plan is to fix them for the
upcoming release.

Signed-off-by: Ken Raffenetti <raffenet@mcs.anl.gov>

Signed-off-by: Halim Amer <aamer@anl.gov>

No reviewer.

Cancel send is not implemented in all our tested netmods. xfail
this test for now.

No reviewer.

Xin Zhao authored May 31, 2015 and Kenneth Raffenetti committed May 31, 2015

Originally we add extended packet header for RMA packets and issue
it from Nemesis and netmods, unfortunately we forgot to modify
Sock channel. This patch fixes this problem.

Signed-off-by: Ken Raffenetti <raffenet@mcs.anl.gov>

No reviewer.

Junchao Zhang authored May 29, 2015 and Kenneth Raffenetti committed May 30, 2015

Signed-off-by: Ken Raffenetti <raffenet@mcs.anl.gov>

Junchao Zhang authored May 29, 2015 and Kenneth Raffenetti committed May 30, 2015

Signed-off-by: Ken Raffenetti <raffenet@mcs.anl.gov>

Junchao Zhang authored May 29, 2015 and Kenneth Raffenetti committed May 30, 2015

Signed-off-by: Ken Raffenetti <raffenet@mcs.anl.gov>

Junchao Zhang authored May 29, 2015 and Kenneth Raffenetti committed May 30, 2015

Signed-off-by: Ken Raffenetti <raffenet@mcs.anl.gov>

Junchao Zhang authored May 30, 2015 and Kenneth Raffenetti committed May 30, 2015

Signed-off-by: Ken Raffenetti <raffenet@mcs.anl.gov>

Xin Zhao authored May 19, 2015 and Pavan Balaji committed May 30, 2015

Originally in CH3, we have two issuing functions to issue RMA messages:
issue_from_origin_buffer() for issuing non-streamed messages and
issue_from_origin_buffer_stream() for issuing streamed messages. Most
code in those two functions are the same, therefore here we merge them
into one function. The function requires stream_offset and stream_size
as input arguments, for non-streamed messages, we pass stream_offset
as 0 and stream_size as the size of the entire message.

Signed-off-by: Pavan Balaji <balaji@anl.gov>

Xin Zhao authored May 14, 2015 and Pavan Balaji committed May 30, 2015

Here we rename the request handlers for ACC/GACC as follows:

MPIDI_CH3_ReqHandler_AccumDerivedDTRecvComplete
        --> MPIDI_CH3_ReqHandler_AccumMetadataRecvComplete
MPIDI_CH3_ReqHandler_GaccumDerivedDTRecvComplete
        --> MPIDI_CH3_ReqHandler_GaccumMetadataRecvComplete

Those handlers are originally triggered when the
target side receives complete derived datatype info.
Now we extended those handlers to be triggered when
extended packet header or derived datatype info is
arrived, therefore we rename them here.

Signed-off-by: Pavan Balaji <balaji@anl.gov>

Xin Zhao authored May 14, 2015 and Pavan Balaji committed May 30, 2015

There are two APIs that are used to issue data with a request
created and passed:

iSendv() --- issue contiguous data;
sendNoncontig_fn() --- issue non-contiguous data;

In this patch, we modify the implementation of those two functions
in nemesis and netmod (tcp/mxm/ptl) to make them issue the extended
packet header stored in the request.

Signed-off-by: Pavan Balaji <balaji@anl.gov>

Xin Zhao authored May 18, 2015 and Pavan Balaji committed May 30, 2015

Here we added extended packet header in CH3 layer used to
transmit attributes that are only needed in RMA and are not
needed in two-sided communication. The key implementation
details are listed as follows:

Origin side:

(1) The extended packet header is stored in the request, and
the request is passed to the issuing function (iSendv() or
sendNoncontig_fn()) in the lower layer. The issuing function
checks if the extended packet header exists in the request,
if so, it will issue that header. (The modifications in lower
layer are in the next commit.)

(2) There is a fast path used when (origin data is contiguous &&
target data is predefined && extended packet header is not used).
In such case, we do not need to create a request beforehand
but can use iStartMsgv() issuing function which try to issue
the entire message as soon as possible.

Target side:

(1) There are two req handler being used when extended packet header
is used or target datatype is derived. The first req handler is
triggered when extended packet header / target datatype info is
arrived, and the second req handler is triggered when actual data
is arrived.

(2) When target side receives a stream unit which is piggybacked with
LOCK, it will drop the stream_offset in extended packet header, since
the stream unit must be the first one and stream_offset must be 0.

Signed-off-by: Pavan Balaji <balaji@anl.gov>

Xin Zhao authored May 13, 2015 and Pavan Balaji committed May 30, 2015

Signed-off-by: Pavan Balaji <balaji@anl.gov>

Xin Zhao authored Apr 30, 2015 and Pavan Balaji committed May 30, 2015

This reverts commit 19f29078

.

Signed-off-by: Pavan Balaji <balaji@anl.gov>

Xin Zhao authored Apr 29, 2015 and Pavan Balaji committed May 30, 2015

Signed-off-by: Pavan Balaji <balaji@anl.gov>

Xin Zhao authored May 18, 2015 and Pavan Balaji committed May 30, 2015

Signed-off-by: Pavan Balaji <balaji@anl.gov>

Xin Zhao authored May 18, 2015 and Pavan Balaji committed May 30, 2015

In RMA communication, when data to be sent is derived datatype but
contiguous, we can avoid copy/packing but directly issue it from
the user buffer. However, the data may have non-zero lb and we should
add the true lb when calculating the starting address. This patch
fixes this issue.

Signed-off-by: Pavan Balaji <balaji@anl.gov>

Xin Zhao authored May 28, 2015 and Pavan Balaji committed May 30, 2015

Originally the implementation of sendNoncontig() in Nemesis
assume that req->dev.segment_first is 0 and req->dev.segment_size
is the size of data, which potentially asssumes that the data passed
from upper layer must be the entire data, prohibiting the possibility
of streaming that data in the upper layer. The general way to use them
should be that req->dev.segment_first specifies the current starting
location and req->dev.segment_size specifies the current ending location.

We fixed this issue in commit 5132e070

, but
there are some places missed in that fixing. Here we fixed those places.

Signed-off-by: Pavan Balaji <balaji@anl.gov>

Xin Zhao authored May 28, 2015 and Pavan Balaji committed May 30, 2015

Note that here we move the definition of streaming unit size to
mpidimpl.h in CH3, so that Nemesis can see it and do assert check on it.

Signed-off-by: Pavan Balaji <balaji@anl.gov>

Xin Zhao authored May 19, 2015 and Pavan Balaji committed May 30, 2015

Signed-off-by: Pavan Balaji <balaji@anl.gov>

Min Si authored May 19, 2015 and Pavan Balaji committed May 30, 2015

The original mtest routine only generated tests for noncontiguous
derived datatypes. This patch added two tests for every derived
datatype: (1) contiguous ddt (stride=block length) without lower-bound;
(2) contiguous ddt with lower-bound.

Signed-off-by: Pavan Balaji <balaji@anl.gov>

Min Si authored May 19, 2015 and Pavan Balaji committed May 30, 2015

The original mtest routine only generated datatype tests with zero
lower-bound. This patch added a test for every derived datatype that
checks datatype structure with non-zero lower-bound.

Signed-off-by: Pavan Balaji <balaji@anl.gov>

Min Si authored May 19, 2015 and Pavan Balaji committed May 30, 2015

The original mtest routine used extent as the space required by
derived datatypes (i.e., to calculate the offset per datatype in a
data buffer). However, it is incorrect if the lower-bound is not zero.
The space must be lower-bound + extent. This patch fixed it.

Signed-off-by: Pavan Balaji <balaji@anl.gov>

No reviewer.

Ensures this test runs only when ROMIO is enabled.

Signed-off-by: Rob Latham <robl@mcs.anl.gov>

Removes unnecessary thread yielding in threaded nonblocking context id
allocation algorithm. The error was introduced by "copy-pasting" from
the blocking context id allocation algorithm
(MPIR_Get_contextid_sparse_group) when implementing the nonblocking
algorithm. Note the subtle difference on thread handling between the
two. In the blocking algorithm, yield is needed to allow another thread
to make progress. In nonblocking algorithm, there is no need to yield to
another thread because this thread will not block the progress. On the
contrary, unnecessary yield will allow other threads to execute and
insert wrong order of entries to the nonblocking schedule and cause
errors.

Fixes #2183

Signed-off-by: Junchao Zhang <jczhang@mcs.anl.gov>

Refs #2088

Signed-off-by: Rob Latham <robl@mcs.anl.gov>

Use PAC macro to AC_SUBST and export WRAPPER_XXX variables. Ensures
that they are available to the testsuite configure script.

Closes #2190

Signed-off-by: Pavan Balaji <balaji@anl.gov>