Resource Sharing in Parallel Computing

The Wolfram Language's symbolic parallel computation architecture provides a uniquely convenient mechanism for communicating and sharing resources between parallel processes. Its foundation is a virtual shared memory model, implemented on top of WSTP-based message passing, running seamlessly on arbitrary clusters or networks of processors.

Shared Variables and Functions

SetSharedVariable set symbols to share values across subkernels

SetSharedFunction set functions to share downvalues across subkernels

$SharedVariables, $SharedFunctions lists of shared variables, functions

UnsetShared stop sharing variables or functions


CriticalSection synchronize concurrent access to a shared resource

Broadcasting Definitions

DistributeDefinitions distribute definitions to all parallel subkernels

ParallelNeeds load the same package into all parallel subkernels

ParallelEvaluate evaluate an expression on all parallel subkernels