- 15 May, 2013 4 commits
-
-
Rob Latham authored
maddeningly we cannot dup MPI_INFO_NULL, so let's check for that case
-
Pavan Balaji authored
Reviewed by Mike Blocksome @ IBM.
-
Pavan Balaji authored
We now provide two types of function pointers: _sched and _req. The _sched functions just add instructions to the schedule, but no request is created. The _req functions return an MPID request. Reviewed by Mike Blocksome @ IBM.
-
The following collectives are updated: MPIX_Ibcast MPIX_Iallgather MPIX_Iallgatherv MPIX_Iallreduce MPIX_Ialltoall MPIX_Ialltoallv MPIX_Ialltoallw MPIX_Iexscan MPIX_Igather MPIX_Igatherv MPIX_Ireduce_scatter_block MPIX_Ireduce_scatter MPIX_Ireduce MPIX_Iscan MPIX_Iscatter MPIX_Iscatterv These collectives are updated similar to how the optimized MPIX_Ibarrier is implemented: | The previous MPIR_Ibarrier_impl() function forced all adi implementations | to create a MPID_Sched_t opaque object which was then passed in to the | specific ibarrier implementation via a function pointer table. | | The MPID_Sched_t object represents a completely new state machine that | must be advanced whenever mpi progress is made. | | The required construction of the MPID_Sched_t object and the required | advance of the schedule state machine would be extremely detrimental to | pamid performance. (ibm) 47b497e15cd87df666c6031a28f5f8a5ae53cece (ibm) aee872eb91f4300066957349e5a3d8103da33f40 Signed-off-by:
Charles Archer <archerc@us.ibm.com>
-
- 14 May, 2013 4 commits
-
-
Pavan Balaji authored
No reviewer.
-
Pavan Balaji authored
This is an important warning flag, but given the number of places we are breaking this in mpich, it is causing us to miss other warnings. This is meant to be a temporary disabling. No reviewer.
-
(ibm) D190161 (ibm) 55566af07e7d69df17005e50c783ff3ae26cc325 Signed-off-by:
Michael Blocksome <blocksom@us.ibm.com>
-
Rob Latham authored
In response to IBM's integration ticket 1822 (but reworked), turn MPIO_CHECK_INFO into a collective macro; exchange result of info inspection with all procs. Now a bogus info on one proc won't cause a process hang.
-
- 13 May, 2013 4 commits
-
-
(ibm) D186334 (ibm) 7Z6 (ibm) 63716b02ccab7cd99116e927a210eca43b03d52e Signed-off-by:
Michael Blocksome <blocksom@us.ibm.com>
-
(ibm) D185675 (ibm) 7693f538360130f4cca71f0975c0bb345d2f6239
-
For some reason these changes from the original commit did not migrate into the 'mpich/master' branch. (ibm) D185675 (ibm) 3728f432f99fc499955a2c9d830a5026e5f87948
-
Signed-off-by:
Michael Blocksome <blocksom@us.ibm.com>
-
- 10 May, 2013 2 commits
-
-
(ibm) D190081 (ibm) 25939d9ffd5267261f70e714aaec434d84e07093 Signed-off-by:
Michael Blocksome <blocksom@us.ibm.com>
-
Also remove MPIR_Comm_group_ext() function declaration and reference. (ibm) D189966 (ibm) 5681895bcb45798680b249dbf993b9a634954b80 Signed-off-by:
Michael Blocksome <blocksom@us.ibm.com>
-
- 09 May, 2013 12 commits
-
-
Pavan Balaji authored
No reviewer.
-
Pavan Balaji authored
Typo in mpl's configure which was causing MPL headers to be installed, which should not be needed. No reviewer.
-
(ibm) D189618 (ibm) f30c0544fcc37fc18cb0f9b27f414373d5d5308b Signed-off-by:
Bob Cernohous <bobc@us.ibm.com>
-
(ibm) D189618 (ibm) 73870926a4a9598deacbc5c0b91fa6150ef8f7ed Signed-off-by:
Bob Cernohous <bobc@us.ibm.com>
-
(ibm) 36f5ac777f738068690184193b557fab53b3472f Signed-off-by:
sssharka <sssharka@us.ibm.com>
-
(ibm) Issue 7967 (ibm) 6b7ba8eb718d8323c8dd638c27cca6a05baa202b Signed-off-by:
sssharka <sssharka@us.ibm.com>
-
(ibm) Trac #659 (ibm) 9d36aed8f988859eba1a284292f0a1a6fc75d01d Signed-off-by:
sssharka <sssharka@us.ibm.com>
-
Do not tie scatterv to scatter selection. (ibm) Issue 9416 (ibm) 4901e679071e489e38ee1a6925da36eab9c3ffac Signed-off-by:
sssharka <sssharka@us.ibm.com>
-
(ibm) Issue 9416 (ibm) 723fb4ce39b5e06ef63cc37231acdaeb6d1c08cf Signed-off-by:
sssharka <sssharka@us.ibm.com>
-
(ibm) 26edfa903df4881e1ab0190bc1859f5dc30e868e Signed-off-by:
sssharka <sssharka@us.ibm.com>
-
(ibm) 88338876fc003a9fe936298097b42df1e54ead32 Signed-off-by:
sssharka <sssharka@us.ibm.com>
-
(ibm) CPS 95PHUZ (ibm) 1b4e2da6d5fb0990d530be6f896f539dc7571620 Signed-off-by:
sssharka <sssharka@us.ibm.com>
-
- 08 May, 2013 4 commits
-
-
Pavan Balaji authored
PAC_VAR_PUSHFLAG and PAC_VAR_POPFLAG are duplicates of PAC_PUSH_FLAG and PAC_POP_FLAG. Also, unlike the PAC_VAR_ macro variants, the flag is not updated by the PAC_PUSH/PAC_POP macros, so the older value is retained. Also cleaned up macros that do additional logic to replace empty action-if-true/action-if-false statements with ":". This is not required. The macros work fine without this replacement. Reviewed by gropp.
-
Pavan Balaji authored
Reviewed by gropp.
-
Pavan Balaji authored
We no longer need PAS_VAR_COPY since AS_VAR_COPY is present in newer autoconf versions. Reviewed by gropp.
-
Fix seg fault in _mpi_world_exiting_handler, fix hang at MPI_Finalize, removing calling PAMI_Context_advance in dispatch. (ibm) D189340 (ibm) 7c57f7ed7a8aa1f0156ba5112664a7c2cd35f227 Signed-off-by:
Bob Cernohous <bobc@us.ibm.com>
-
- 07 May, 2013 10 commits
-
-
James Dinan authored
Myrank was caching win_ptr->comm_ptr->rank, so we now use that directly rather than caching it in the MPID_Win object. Reviewer: balaji
-
James Dinan authored
Several fields of MPID_Win were not used above the ADI. These have been moved to CH3 to reduce clutter. Downstream devices should redefine these fields in MPID_DEV_WIN_DECL if they are needed. Reviewer: balaji
-
James Dinan authored
-
William Gropp authored
-
(ibm) D189463 (ibm) e7be2d5b3ae75a32383c32eb203d864d1bbbc386 Signed-off-by:
Bob Cernohous <bobc@us.ibm.com> Signed-off-by:
Michael Blocksome <blocksom@us.ibm.com>
-
(ibm) D189345 (ibm) 7Z8 (ibm) 9e1fffe10532d58f09acdbbd81ee582b370bf804 Signed-off-by:
Bob Cernohous <bobc@us.ibm.com> Signed-off-by:
Michael Blocksome <blocksom@us.ibm.com>
-
(ibm) D189342 (ibm) 7Z8 (ibm) c2e13cd5979dbba4019cb95c01688a629b7b0f9e Signed-off-by:
Michael Blocksome <blocksom@us.ibm.com>
-
This commit includes a fix for all rooted collectives where type info is local. The fix makes sure that non-contig data is converted and packed on all tasks and then calling PAMI. Another change in this commit is related to performance. MPICH2 queries the collsel extension for best algorithm before attempting any conversions. If algorithm is mpich, then it is called directly instead of going through the collsel extension path in PAMI. (ibm) D187892 (ibm) 7Z8 (ibm) dad6ba8f7883508cd7f83433232b8dadb6ccddcb Signed-off-by:
Bob Cernohous <bobc@us.ibm.com> Signed-off-by:
Michael Blocksome <blocksom@us.ibm.com>
-
(ibm) F183421 (ibm) 6463405fd65ec7a45b137334c7608d30f8c29055 Signed-off-by:
Bob Cernohous <bobc@us.ibm.com>
-
Pavan Balaji authored
No reviewer. Signed-off-by:
Michael Blocksome <blocksom@us.ibm.com>
-