libmuscle.peer_manager module

class libmuscle.peer_manager.PeerManager(kernel: Reference, index: List[int], conduits: List[Conduit], peer_dims: Dict[Reference, List[int]], peer_locations: Dict[Reference, List[str]])[source]

Bases: object

Manages information about peers for a Communicator

Create a PeerManager.

Peers here are instances, and peer_dims and peer_locations are indexed by a Reference to an instance. Instance sets are multi-dimensional arrays with sizes given by peer_dims.

Parameters

  • kernel – The kernel for the instance whose peers we’re managing.

  • index – The index of the instance whose peers we’re managing.

  • conduits – A list of conduits attached to this compute element, as received from the manager.

  • peer_dims – For each peer we share a conduit with, the dimensions of the instance set.

  • peer_locations – A list of locations for each peer instance we share a conduit with.

get_peer_dims(peer_kernel: Reference) → List[int][source]

Get the dimensions of a peer kernel.

Parameters

peer_kernel – The peer kernel whose dimensions to get.

get_peer_endpoint(port: Identifier, slot: List[int]) → Endpoint[source]

Determine the peer endpoint for the given port and slot.

Parameters

  • port – The port on our side to send or receive on.

  • slot – The slot to send or receive on.

Returns

The peer endpoint.

get_peer_locations(peer_instance: Reference) → List[str][source]

Get the locations of a peer instance.

There may be multiple, if the peer supports more than one

protocol.

Parameters

peer_instance – The instance whose locations to get.

get_peer_port(port: Identifier) → Reference[source]

Get a reference for the peer port.

Parameters

port – Name of the port on this side.

is_connected(port: Identifier) → bool[source]

Determine whether the given port is connected.

Parameters

port – The port to check.