- 11 Sep, 2012 3 commits
-
-
James Dinan authored
An extra argument was added to the oputil macros to provide a name, which can be used to generate union types. This is needed for generating types to be included in the RMA packet headers. The RMA code is also modified to use this new functionality, replacing the old hand-coded unions. Reviewer: goodell@
-
James Dinan authored
The oputil.h code is useful to more than just the collectives code now, so it makes sense to move to the common include directory. Precursor to a refactor of oputil.h itself to add a third (type-name) parameter to the type macros. Reviewer: goodell
-
Darius Buntinas authored
-
- 07 Sep, 2012 6 commits
-
-
James Dinan authored
As of MPI 3.0, the tag argument in Intercomm_create can be ignored. Updated intercomm_create to use a fixed tag in the collectives tag space. Reviewer: balaji
-
James Dinan authored
-
James Dinan authored
Added builtin op implementations for RMA-only ops. MPI_REPLACE was not defined in the op_table prior to this commit. Also includes an alternate implementation of the linked list test that uses fetch-and-op with these operations to enable the use of shared locks. Reviewer: goodell
-
James Dinan authored
Adds implementation of atomic fetch-and-op operation and several test cases. Reviewer: buntinas
-
Darius Buntinas authored
-
Darius Buntinas authored
-
- 05 Sep, 2012 20 commits
-
-
Pavan Balaji authored
r10173. Dropped r10174. No reviewer.
-
Pavan Balaji authored
No reviewer.
-
Pavan Balaji authored
-
Pavan Balaji authored
No reviewer.
-
Pavan Balaji authored
through mpiexec. No reviewer.
-
Pavan Balaji authored
-
James Dinan authored
Reviewer: buntinas
-
James Dinan authored
Reviewer: buntinas
-
David Goodell authored
They were causing an unnecessary conflict with ROMIO's "lock.c", which libtool would handle for us with some extra internal hackery. This is a bit cleaner and faster. No reviewer.
-
David Goodell authored
No reviewer.
-
David Goodell authored
No reviewer.
-
David Goodell authored
This includes most remaining "Makefile.sm" files and the "simplemake.in" program itself. This commit also attempts to get the "test/mpi/maint/updatefiles" into better shape, although I have not yet tested most of these modifications. Partly reviewed by balaji@.
-
David Goodell authored
The code currently fails with assertions when I run any MPI program with hydra and pmi2: assert (*count * sizeof(struct HYD_pmcd_token)) failed No reviewer.
-
David Goodell authored
No reviewer.
-
David Goodell authored
It mostly contained stale, out of date information about p4 and mpirun. Reviewed by balaji@.
-
Pavan Balaji authored
-
Pavan Balaji authored
-
Pavan Balaji authored
-
William Gropp authored
-
William Gropp authored
-
- 04 Sep, 2012 4 commits
-
-
Pavan Balaji authored
Reviewed by buntinas.
-
Pavan Balaji authored
symbols for them. Reviewed by goodell.
-
Pavan Balaji authored
1. Move the new extension FT routines into the src/mpi/comm directory instead of the separate src/mpix directory, since it's clearly using the MPI internals. 2. Move the MPIX_mutex functionality to mpi.h.in instead of a separate mpix.h. This will allow us to use device-specific mpix_foo.h functionality that the user can use to buy-in to device-specific extensions. 3. Move the ARMCI and Mutex functionality to src/armci and src/mutex directories respectively. Reviewed by goodell.
-
Anthony Chan authored
[svn-r10186] added Fortran inplace tests for MPI_Alltoall and the v/w versions (added in mpi-2.2) and Fortran test of MPI_Reduce_scatter_block with or without inplace. This resolves part of req#1623.
-
- 03 Sep, 2012 3 commits
-
-
Pavan Balaji authored
RELEASE_NOTES.
-
Pavan Balaji authored
-
Pavan Balaji authored
-
- 01 Sep, 2012 1 commit
-
-
Pavan Balaji authored
-
- 30 Aug, 2012 3 commits
-
-
James Dinan authored
This is a workaround to allow "negative" target displacements (i.e. target addresses which are negative when interpreted as signed integers) in dynamic windows. Reviewer: none
-
James Dinan authored
The source_win_handle is used to inform the target that the RMA access epoch has ended. This was removed from the CAS packet to save space and was replaced with another packet type, MPIDI_CH3_PKT_CAS_UNLOCK, which conveys the same information.
-
James Dinan authored
Now that CAS is implemented, the type checking test should expect MPI_SUCCESS instead of ERR_NOTIMPL. Reviewer: buntinas
-