Use an OpenMP parallel region to implement a WorkerGroup that manages threads

Unit testing all the components of the code.

Extend virtual context infrastructure for to support indexable collections of virtual entities.

Virtual context collections will need performance data to feed into Charm++ load balancing suite. We need per-index timers/counters.

Implement Virtual Context proxy/ID, manager, message routing to virtual contexts

Every will be moved to the "topos" package

Implement a Worker that uses OpenMP for threading

Cleanup RDMA code now that method renaming has been complete

Everything is identifiable with a 64-bit ID EXCEPT for a collection and a given index in the collection which is two 64-bit IDs. Would be better to unify IDs into a single thing that identifies collection + index.

Extract from boost fcontext (without a dependency) the library that enables very efficient ULTs. This will become useful for some worker interactions.

Make sequencing work with virtual contexts

VT added groupings (could potentially match to MPI_Groups -limiting scope of collectives). Only incast and bcast across the entire machine are otherwise allowed on collective communication.

Discuss classification of components: core, virtualization, etc. and the current state (done, partially complete, or design phase).

Also discuss current unit/integration testing state (whether tests exist, how comprehensive they are, etc.), examples, performance testing.

Map virtual contexts to workers and allow them to execute off the communication thread.

Allow the MPI communicator to be split to support interoperability.

Ensure termination detector implementation is sufficient for the PIC use case.

Implement sequencer component on top of active messaging layer

As a optional add-on component to vt, implement the standard C++ template detection idiom to statically check at compile-time for adherence to a pre-defined interface.

This may be applied to virtual contexts, messages, indices, etc.

A new format for test filenames.