[x] maybe be consistent about referencing WPM or SPM first time mentioned in a chapter
[x] drop Process from WPM and SPM and DPM
[x] once again we discuss tag for MPI_Comm_create_from_group. Perhaps str_tag.
[x] Jeff thinks we should add an advice to users about what to supply for tag. Jeff thinks we should reverse dot notation
[x] advice to users in Chapter 6.1.2 needs fixing.
[x] clarify errorhandler for MPI_Comm_create_from_group and MPI_Intercomm_create_from_groups and what happens when an error is encountered while creating the
[x] Jeff talks about attributes train wreck. Martin thinks we can leverage MPI_T functionality and maybe not need. So sounds like we'll drop sessions attributes and add pre-defined communicator attribute MPI_TAG_UB only. DELETE SESSIONS ATTRIBUTES
[x] discussion of MPI_TAG_UB and the MPI_COMM_WORLD discussion. Needs to be the same on all processes in the communicator. Update the Tag Values. DOUBLE CHECK THIS.
[x] discussion of Error handler section 8.3, in particular the sentence When using the world process model, MPI calls that are not related to any objects are considered to be attached to the communicator... . May be able to just remove and wait for Aurelian's ticket X.
[x] MPI_ERRORS_ABORT should be local for Sessions. Have a separate sentence for sessions.
[x] remove text for sessions in this error handling section.
[x] start of 13.7 I/O error handling is impacted Aurelian's ticket X, also check the rational at end of 13.7
[x] check language bindings if we remove attributes from sessions
[x] 18.2.7 remove change since we don't have create keyval for sessions
[x] rework the changelog a bit - Chapter 10, sections blah blah, appendix A
general discusion items
Discussion of semantic terms and usage of World Process Model etc.
Martin wonders whether we should call it process model, maybe something else? Dan points out the dynamic process model term. Model, vs. system, vs implementation.
Should we add a rank designator argument to MPI_GROUP_FROM_SESSION_PSET (color/key) + group split? Consensus seems to be to add this as a separate ticket. Todo item here for Guillaume Mercier.
External TODOS:
[x] Jeff will dig up problems about attributes API
specific todos
Processfrom WPM and SPM and DPMtagforMPI_Comm_create_from_group. Perhaps str_tag.tag. Jeff thinks we should reverse dot notationMPI_Comm_create_from_groupandMPI_Intercomm_create_from_groupsand what happens when an error is encountered while creating theMPI_TAG_UBonly. DELETE SESSIONS ATTRIBUTESTag Values. DOUBLE CHECK THIS.When using the world process model, MPI calls that are not related to any objects are considered to be attached to the communicator.... May be able to just remove and wait for Aurelian's ticket X.general discusion items
Discussion of semantic terms and usage of
World Process Modeletc.Martin wonders whether we should call it
process model, maybe something else? Dan points out thedynamic process modelterm. Model, vs. system, vs implementation.Should we add a rank designator argument to
MPI_GROUP_FROM_SESSION_PSET(color/key) + group split? Consensus seems to be to add this as a separate ticket. Todo item here for Guillaume Mercier.External TODOS: