Warning

This version of the documentation is NOT an official release. You are looking at ‘latest’, which is in active and ongoing development. You can change versions on the bottom left of the screen.

Recording backend mpi - Send data with MPI

Description

Availability

This stimulation backend is only available if NEST was compiled with support for MPI.

The mpi recording backend sends collected data to a remote process using MPI.

The name of the MPI port to send data to is read from a file for each device configured to use this backend. The file needs to be named according to the following pattern:

{data_path}/{data_prefix}{label}/{node_id}.txt

The data_path and data_prefix are global kernel properties, while label is a property of the device in question and node_id its node ID. This path can only be set outside of a Run context (i.e. after Prepare() has been called, but Cleanup() has not).

Communication Protocol

The following protocol is used to exchange information between both MPI processes. The protocol is described using the following format for the MPI messages: (value, number, type, source/destination, tag)

  1. Prepare : Connection of MPI port included in the port_file (see below)

  2. Run begin: Send at each beginning of the run (true, 1, CXX_BOOL, 0, 0)

  3. Run end : Receive at each ending of the run (true, 1, CXX_BOOL, 0, 0)

  4. Run end : Send shape of the data of the run (shape, 1,INT, 0, 0)

  5. Run end : Send data of the data of the run (data, shape, DOUBLE, 0, 0)

  6. Run end : Send at each ending of the run (true, 1, CXX_BOOL, 0, 1)

  7. Cleanup : Send at this en of the simulation (true, 1, CXX_BOOL, 0, 2)

Data format

The format of the data sent is an array consisting of (id device, id node, time is ms).