Previous: Shared-memory parallelization, Up: Parallelization [Contents][Index]
The distributed-memory parallelization in FEMLISP is based on the
well-known standard MPI, for which a Common Lisp interface exists in the
form of the cl-mpi library developed by Alex Fukunaga and
Marco Heisig.
On top of cl-mpi, FEMLISP implements the higher-level
library ddo, which allows to make objects distributed between
processors, and keeps track of their relations when they are changed or
even deleted locally.
More about cl-mpi and ddo together with a report on
a serious practical application can be found in Heisig-Neuss
2017.
| • Distributed objects: | ||
| • Synchronization: | ||
| • Compiling the MPI worker: | ||
| • Starting the worker pool: | ||
| • Connecting to the worker pool: | ||
| • Communicating with the worker pool: |