GitLab

some libraries like HDF5 want to register their cleanup routines into
finalize.  these cleanup routines use MPI-IO, so they need to fire
before ROMIO cleans up.
Signed-off-by: Pavan Balaji <balaji@anl.gov>

Signed-off-by: Min Si <msi@il.is.s.u-tokyo.ac.jp>
Signed-off-by: Rob Latham <robl@mcs.anl.gov>

Signed-off-by: Sangmin Seo <sseo@anl.gov>

Signed-off-by: Sangmin Seo <sseo@anl.gov>

type promotions have resulted in a change to the device layer.

Ref: 1767
Signed-off-by: Pavan Balaji <balaji@anl.gov>
Signed-off-by: Sameh S Sharkawi <sssharka@us.ibm.com>

despite promoting types throughout the gather path, still had one case of
constructing structs with larger-than-int blocklens.  solution: borrow BigMPI
strategy and construct types-of-chunks to get around limitations.

Ref: #1767
Signed-off-by: Pavan Balaji <balaji@anl.gov>

The ongoing march towards 64-bit clean continues.  Address areas where large
product of two ints might have overflowed.

Ref: #1767
Signed-off-by: Pavan Balaji <balaji@anl.gov>

- preprocessor constants need parens
- which showed the "always fail" case wasn't big enough
- compiler warned about variables possibly being used uninitialized

Ref: #1767
Signed-off-by: Pavan Balaji <balaji@anl.gov>

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

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

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

No reviewer.

commit 83253a41

 triggerd a bunch of new warnings.  Take a different
approach.  For simplicity of implementation, do_accumulate_op is defined
as MPI_User_function.  We could split up internal routine and
user-provided routines, but that complicates the code for little
benefit:

Instead, keep do_accumlate_op with an int type, but check for overflow
before explicitly casting.  In many places the count is simply '1'.  In
stream processing there is an interal limit of 256k, so the assertion
should never fire.
Signed-off-by: Xin Zhao <xinzhao3@illinois.edu>

Encode request handles in unused tag bits, eliminating the need
for hash table.
Signed-off-by: Antonio Pena Monferrer <apenya@mcs.anl.gov>

rma/manyget was failing because we reached the maximum number of MEs
allowed: we were posting an ME for every remote get. This patch
implements a single persistent ME instead for the remote gets.

Fixes #2264
Signed-off-by: Ken Raffenetti <raffenet@mcs.anl.gov>

Signed-off-by: Antonio J. Pena <apenya@mcs.anl.gov>

MPID_Segment_pack/unpack should be all we need to manipulate noncontig
messages. Not sure why these were originally included.
Signed-off-by: Antonio J. Pena <apenya@mcs.anl.gov>

Because we don't drain the EQs in the event of flow control, we need
to use a dedicated EQ for messages related to pausing and unpausing
communication. These messages are all consumed by the Rportals layer,
the user will never see anything.
Signed-off-by: Antonio J. Pena <apenya@mcs.anl.gov>

Signed-off-by: Xin Zhao <xinzhao3@illinois.edu>

Loser of Head-to-Head connections are not necessarily
closed, if the sock set is destroyed. This patch
looks for all open connections, close the socket
and free the memory recourses. Fixes #2180
Signed-off-by: Ken Raffenetti <raffenet@mcs.anl.gov>

The loser of a head-to-head connection sometimes tries
to reconnect later, afer MPI_Finalize was called  This
can lead to several errors in the socket layer, depending
on the state of the disarded connection and the appereance
of the connection events. Refs #2180
This Patch has two ways to handle this:

1.)
Discarded connections are marked with CONN_STATE_DISCARD,
so they are hold from connection.  Furthermore, an error on
any discarded connection (because the remote side closed in
MPI_Finalize) is ignored and the connection is closed.

2.)
Add a finalize flag for process groups. If a process group is
closing and tries to close all VCs, a flag is set to mark this.
If the flag is set, a reconnection (in the socket state) is
refused and the connection is closed on both sided.

Both steps are necessary to catch all reconnection tries after
MPI_Finalize was called.
Signed-off-by: Ken Raffenetti <raffenet@mcs.anl.gov>

Ignore local completion events (SENDs) when counting outstanding
ops to remote targets.

No reviewer.

No reviewer.

Originally Request_load_recv_iov() function assumes that
the initial value of req->dev.segment_first is always zero,
which is not correct if we set it to a non-zero value for
streaming the RMA operations.

The way Request_load_recv_iov() works is that, it is triggered
multiple times for the same receiving request until all data is
received. During this process, req->dev.segment_first is rewritten
to the current offset value. When the initial value of
req->dev.segment_first is non-zero, we need another variable
to store that value until the receiving process for this request
is finished. Here we use a static variable in this function to
reach the purpose.
Signed-off-by: Pavan Balaji <balaji@anl.gov>

In this patch, we fix the mistakes in calculating the streaming
size in GetAccumulate pkt handler on the target side. The original
code has two mistakes here:

1. The original code use the size and extent of the target datatype,
   which is wrong. Here we should use the size / extent of the basic
   type in the target datatype.

2. The original code always use the total data size to calculate
   the current streaming size, which is wrong. Here we should use
   the current rest data size to calculate.

This patch fixes these two issues.
Signed-off-by: Pavan Balaji <balaji@anl.gov>

Here we assign req->dev.segment_first to orig_segment_first.
Since req->dev.segment_first is a MPIDI_msg_sz_t type, we should
use the same type for orig_segment_first.
Signed-off-by: Pavan Balaji <balaji@anl.gov>

This test occasionally run more than 3 min (default time limit)
on OFI platform. This patch increases the time limit to 5 min.
Signed-off-by: Pavan Balaji <balaji@anl.gov>

This test uses irecv and isend to transfer data in
an alltoall manner between multiple processes.
The idea of this test is testing, if MPI can handle
multiple processes trying to connect to each other from
both sides at the same time.
Signed-off-by: Antonio J. Pena <apenya@mcs.anl.gov>

This patch includes three changes:
(1) Added netmod API get_ordering to allow netmod to expose the network
ordering. A netmod may issue some packets via multiple connections in
parallel if those packets (such as RMA) do not require ordering, and
thus the packets may be unordered. This patch sets the network ordering
in every existing netmod (tcp|mxm|ofi|portals|llc) to true, since all
packets are sent orderly via one connection.
(2) Nemesis exposes the window packet orderings such as AM flush
ordering at init time. It supports ordered packets only when netmod
supports ordered network.
(3) If AM flush is ordered (flush must be finished after all previous
operations), then CH3 RMA only requests FLUSH ACK on the last operation.
Otherwise, CH3 must request per-OP FLUSH ACK to ensure all operations
are remotely completed.
Signed-off-by: Xin Zhao <xinzhao3@illinois.edu>
Signed-off-by: Pavan Balaji <balaji@anl.gov>

When win resource is used up, the current code frees OPs before
completion only if flush_remote is ordered. However, we can always free
them even on out-of-order network. Because remote completion is waited
by ack counter, and local completion (flush_local) is translated to
remote completion (flush).
Signed-off-by: Xin Zhao <xinzhao3@illinois.edu>
Signed-off-by: Pavan Balaji <balaji@anl.gov>

The outstanding_acks counter was increased at each sync call (such as
fence and flush). However, the counter had to be decreased again if
flush ack is not required. It is more straightforward if increasing it
only when the flush packet is issued (FLUSH flag piggyback or a separate
flush message).
Signed-off-by: Xin Zhao <xinzhao3@illinois.edu>
Signed-off-by: Pavan Balaji <balaji@anl.gov>

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

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

Here we make check_and_switch_target/window_state to return a flag
indicating if the current window/target states are OK for issuing
operations.
Signed-off-by: Pavan Balaji <balaji@anl.gov>

check_window_state ---> check_and_switch_window_state
check_target_state ---> check_and_switch_target_state

Both of those two functions are used to check and switch
(if possible) RMA state. Here we change their name to
proper ones.
Signed-off-by: Pavan Balaji <balaji@anl.gov>

When GACC/FOP is used with MPI_NO_OP, the operation is essentially
an atomic GET. Originally MPICH implemented this by converting
GACC/FOP to GET, which lost the atomicity of that operation.

In this patch, we modify the implementation of GACC/FOP to support
atomic GET. Main modifications are listed below:

(1) When streaming GACC operation, originally we use origin data
    size to calculate the stream unit size. Since origin data is
    zero in atomic GET, here we use target data size instead
    to calculate the stream unit size.

(2) On the origin side, if it is atomic GET, CH3 just issues packet
    header and metadata for derived datatypes (if needed) and does
    not try to issue from origin buffer; on the target side, after
    packet header and metadata for derived datatypes (if needed) are
    received, the final request handler is triggered, CH3 does not
    try to receive any data from origin.
Signed-off-by: Pavan Balaji <balaji@anl.gov>

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

Originally MPICH check datatype of FOP by judging if it is a
BUILTIN type, this prohibits all pair types. This patch fix this
issue.
Signed-off-by: Pavan Balaji <balaji@anl.gov>

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