jsquyres [Wed, 27 Jul 2005 23:44:37 +0000] rev 18044
Submitted by: Rainer Keller
Reviewed by: Jeff Squyres and George Bosilca
(from r6611 and r6620 on the trunk)
r6611:
Any non-blocking P2P function should set the request to something else
then MPI_REQUEST_NULL, when passed a source/dest of MPI_PROC_NULL.
Introduce a new ompi_request_empty, which returns the
MPI_SOURCE=MPI_PROC_NULL, MPI_TAG=MPI_ANY_TAG, count=0 as specified by
the MPI-standard.
r6620:
See comment in file (src/request/request.c).
We need to reset the empty-request to the MPI_REQUEST_NULL.
jsquyres [Fri, 15 Jul 2005 14:04:30 +0000] rev 18043
Submitted by: Edgar Gabriel
Reviewed by: Jeff Squyres
minor fixes to make the code compile, if ompi has been configured
with the --enable-mpi2-one-sided option
Reviewer's note: this is, strictly speaking, not necessary for the
beta, but Edgar raised a compelling point that some people (including
him) need MPI 1-sided stubs to compile their benchmark and testing
apps (i.e., they only need the stubs, not the functionality). So
since these definitely were bugs on the beta, they were worth fixing.
jsquyres [Fri, 15 Jul 2005 12:19:12 +0000] rev 18042
Submitted by: Edgar Gabriel
Reviewed by: Jeff Squyres
make sure, that the value for TYPE_ATTR is identical, independent of
the definition of OMPI_WANT_MPI2_ONE_SIDED
Reviewer note: this change was made on the trunk. It is completely
harmless, and although the beta code is "done", seems like a good idea
to keep this tiny piece of functionality in sync between the trunk and
the branch.
brbarret [Tue, 05 Jul 2005 22:14:33 +0000] rev 18041
* initialize size_count
jsquyres [Sun, 03 Jul 2005 16:14:07 +0000] rev 18040
Needs no review
Updates to README.
jsquyres [Sat, 02 Jul 2005 16:03:34 +0000] rev 18039
Submitted by: Jeff Squyres
Reviewed by: Brian Barrett
Need to case sizeof() to (long) so that we can printf() it.
jsquyres [Fri, 01 Jul 2005 20:40:27 +0000] rev 18038
Submitted by: Jeff Squyres
Approved by: Brian Barrett
Nasty typo that was causing nondeterministic failures when running
MPI-2 dynamic tests that did not disconnect from the parent before
invoking MPI_FINALIZE.
jsquyres [Fri, 01 Jul 2005 17:50:34 +0000] rev 18037
Submitted by: Jeff Squyres
Reviewed by: Edgar Gabriel
Found another size_t<-->int issue.
jsquyres [Fri, 01 Jul 2005 16:51:18 +0000] rev 18036
Submitted by: Jeff Squyres
Reviewed by: Edgar Gabriel
Similar to the changes yesterday, this is a clash of the APIs -- MPI
needs int's, and ORTE needs size_t's. So do the "safe" downcast to an
int when required.
jsquyres [Thu, 30 Jun 2005 22:00:42 +0000] rev 18035
Submitted by: Jeff Squyres
Reviewed by: Josh Hursey
Arrgh -- checked functionality, but forgot to check compiler warnings
on 32 bit. Separate out the "very little to do" case from the "must
do lots of work" case with a single #if.