2020 04 22

2020-04-22 meeting minutes

Attendees:

• Dan Holmes
• Martin Rüfenacht
• Puri Bangalore

Discussion of conversion of KINDs from arrays to classes

As a document author, it would be easier if all the Python code defining a KIND were in one place, i.e. like a class.

Discussion of future "editor pass"-like issues generated from this WG

For example,

Issue - go through all APIs and write down the valid range for all integer parameters. Use that to update "(integer)" vs "(non-negative integer)" designations, but also consider how to represent machine-readable "value range" information.

We should probably create a bunch of these issues (targeted at mpi-4.1 not mpi-4.0) and add them to a new github project (like Wesley's editor pass project). Initially, all of them would be in the "needs triage" state.

Progress for the Pythonisation itself (targeting MPI-4.0)

• Finish the Pythonization PR - DONE! All merged into mpi-4.x branch.
• Sessions has now been Pythonised (by Dan)
• Partitioned communication is being Pythonised (by Ryan)
• MPI_T_events is being Pythonised (by Marc-Andre)
• The Embiggening has been rebased onto the latest Pythonisation branch (by MartinR).
• Any other (big) proposals for MPI-4.0?

Relation to Semantics WG

Diagram in Miro shows logical relationships between some key semantic terms. Is this coverage sufficient? Do we need more terms included?

We have working textual definitions for some/most semantic terms. Is this coverage sufficient? Do we need more terms included?

We have little/no first-order-logic definitions for some/most semantic terms. Is this coverage sufficient? Do we need more terms included?

Next steps

Someone:

• Create a new github Pythonisation-Pass project.

All:

• Do any of these efforts suggest changes or improvements to the Python infrastructure? If so, create an issue assigned to the new github Pythonisation-Pass project.

Puri/Dan:

• Keep working with the semantic terms WG