GitLab

Reviewer: buntinas

Implementation of MPI-3 window info query routines.

Reviewer: balaji

This commit implements MPI-3 RMA's flush and flush_all operations.

Reviewer: buntinas

When enabled, this mode of operation immediately requests the lock when
MPI_Win_lock is called.  Currently, this is enabled by setting the
MPICH_RMA_LOCK_IMMED environment variable.  In the future, we can also make
this mode of operation available though an info/assert.  This capability is
needed to implement MPI-3's flush operations.

Reviewer: buntinas

Reviewer: goodell

By setting "indent-tabs-mode:nil" we should hopefully begin to slowly
squeeze out hard tabs from the source without a disruptive (to
downstream projects) whitespace-fixing change.

No reviewer.

Updated the fetch_and_op implementation to have two data transfer paths; one
where data can be embedded in the packet header and one where it is sent
separately.  With this change, the header size is back to 40 bytes.

Reviewer: buntinas

Adds implementation of atomic fetch-and-op operation and several test cases.

Reviewer: buntinas

Reviewed by buntinas.

This patch adds support for the new MPI-3 compare-and-swap routine.  The
implementation embeds the data into the packet header to reduce the number of
messages.

Reviewer: buntinas

Added the missing disp_unit argument (was added in a later revision of the MPI
3.0 spec) and fixed a bug in base pointer calculations when processes pass a
size of 0.  Added a test case to test MPI-2 ops on shared memory windows.

Reviewer: buntinas

Moved RMA implementation constants from mpidimpl.h to the RMA implementation
header.  Also updated constants to use enumeration types and removed an old
fixme note, which indicated that this should be done.

Reviewer: buntinas

This patch was a quick fix, which moved shared memory segment support from
Nemesis up to ch3 in order to support shared memory windows.  This
functionality has been pushed down into Nemesis, so the r10057 fix is no longer
needed.

Reviewer: buntinas

This adds the win_fns table to ch3, which allows the channel to override the
default implementation of window creation routines provided by ch3.  This also
pushes the implementation of shared memory windows down into Nemesis, includes
window functions for sock, and contains multiple improvements to the window
creation functions code.

Reviewer: buntinas

Removed the old RMA virtual function infrastructure from CH3 -- this code was
all already dead.  Function overrides are already provided per-window in the
MPID_Win structure.  Overrides for non-window-specific (window creation)
operations will be added shortly.

Reviewer: buntinas

Reviewer: balaji

Nemesis.  In MPI-3, ch3 will require shared memory in order to support
MPI_Win_allocate_shared().

yet implemented below the ADI level.

disabled for the public interfaces, but all internal functions use
const.

No reviewer.

RMA operations.

No reviewer.

No reviewer.

selectability feature is now added into nemesis for its netmods.  See
tt#984.

No reviewer.

and defined inside the ch3 device.

No reviewer.

Convert to using it in CH3's MPID_Irecv

Reviewed by buntinas@.

This adds the appropriate code in the device to support networks
that natively do tag matching, such as mx and newmad.  The appropriate
new functions in those netmods throw a new "**nomprobe" error message
and MPI_ERR_INTERN error class.

Due to lack of time and easy availability of MX+NewMad platforms, this
code has not been tested.  (Sorry Guillaume!)

No reviewer.

This includes MPIX_Mprobe, MPIX_Improbe, MPIX_Mrecv, and MPIX_Imrecv.

!!!NOTE!!! This change adds four more functions to the ADI.

The current implementation of MPIX_Mprobe is much less efficient than it
could be.  Right now it polls the unexpected queue (UQ), much like
MPI_Probe does.  A better implementation would enqueue the "message"
object on the PQ so that the matching is handled by the progress engine
in the various packet handlers.  Unfortunately, such an implementation
is surprisingly complicated at this point.

Includes only a simple test for now (test/mpi/pt2pt/mprobe), but it at
least shows that our Mprobe functionality basically all works.

Reviewed by buntinas@.

[svn-r9424] Added MPIX_Comm_reenable_anysource functionality.  This is a manual merge of the dev/reenableas branch.  Reviewed by goodell@

[svn-r9394] Added query functions for failed processes.  This is a manual merge of the groupfailed branch, because svn choked.  Reviewed by goodell@

Most of the problems associated with this implementation have to do with
initialization bootstrapping/teardown issues.  Using the new interface
to profile MPI_Init/MPI_Finalize certainly has at least a few bugs right
now.  Use entirely within an MPI_Init/MPI_Finalize region should
basically work fine.

Two performance variables are exposed right now, the expected and
unexpected queue lengths.  All environment variables from the parameter
interface are exposed via the new "control variable" API.

No reviewer.

[svn-r9034] Added parameter to allow netmods to force ready sends to use rendezvous for large messages.  Reviewed by goodell@

This change moves the greq functions and state into a subordinate
struct that is only created for generalized requests.  This saves ~48 bytes
per request on many platforms in the common pt2pt request case.

Reviewed by buntinas@.

This change consists of four main parts:

 1. Updating all of the MPIR_Ifoo_impl routines to assume that all
    nonblocking collective functions are set in the coll_fns field.

 2. Defining the MPID_Sched_ portion of the ADI in mpir_nbc.h.

 3. Implementing that interface in src/mpid/common/sched

 4. Modifications to ch3:nemesis and ch3:sock in order to call the
    progress engine hook for the scheduling mechanism, as well as
    correctly including the various scheduling headers in CH3 headers.

Reviewed by balaji@.

[svn-r7728] Added attribute to COMM_WORLD to list failed processes.  The attribute is called MPICH_ATTR_FAILED_PROCESSES, and should be declared 'extern' in the application. (I'm not sure about the implications of adding a symbol to mpi.h on the ABI.)

[svn-r7671] merged in error-return branch.  This branch completes with an error recvs on failed VCs or anysource recvs on comms with failed VCs.  See log for r7669 for details.

    - Clean up some error checking code.
    - Report communication errors in request rather than function
      returns.  This fixes cases where an error from a nonblocking
      communication operation is reported in the return value of a
      subsequent unrelated blocking operation.
    - When a VC is moved to the MORIBUND state, all pending send
      requests and posted receive requests are completed with errors.
Nontrivial changes reviewed by goodell@

[svn-r7416] Major improvement to RMA performance for long lists of operations, an immediate mode accumulate for single ints, store the MPID_Comm within the window, and added a basic performance instrumentation interface that was extensively used to improve the RMA performance (enabled with --enable-g=instr).  With these fixes, MPICH2 can run the one-sided version of the Graph500 benchmark at a respectable if not great rate

[svn-r7312] Correct use of MPI_ERR_NO_MEM (should be MPI_ERR_OTHER) for out-of-memory that does not pertain to MPI_Alloc_mem and add detail to error messages about out-of-memory