-

The PCJ library was created with some principles.

Tasks (PCJ threads)

Each task executes its own set of instructions. Variables and instructions are private to the task. PCJ offers methods to synchronize tasks.

Local variables

Variables are accessed locally within each tasks and are stored in the local memory.

Shared variables

There is dedicated class called Storage which represents shared memory. Each task can access other tasks variables that are stored in a shared memory. Shareable variable has to have a special annotation @Shared.

There is distinction between nodes and tasks (PCJ threads). One instance of JVM is understood as node. In principle it can run on a single multicore node. One node can hold many tasks (PCJ threads) – separated instances of threads that run calculations. This design is aligned with novel computer architectures containing hundreds or thousands of nodes, each of them built of several or even more cores. This forces us to use different communication mechanism for inter- and intranode communication.

In the PCJ there is one node called Manager. It is responsible for setting unique identifiers to the tasks, sending messages to other tasks to start calculations, creating groups and synchronizing all tasks in calculations. In contrast to our previous version of the PCJ library, the Manager node has its own tasks and can execute parallel programs.

 Execution in multinode multicore environment

The application using PCJ library is run as typical Java application using Java Virtual Machine (JVM). In the multinode environment one (or more) JVM has to be started on each node. PCJ library takes care on this process and allows user to start execution on multiple nodes, running multiple threads on each node. The number of nodes and threads can be easily configured, however the most resonable choice is to limit on each node number of threads to the number of available cores. Typically, single Java Virtual machine is run on each physical node although PCJ allows for multiple JVM scenario.

Since PCJ application is not running within single JVM, the communication between different threads has to be realized in different manners. If communicating threads run within the same JVM, the Java concurrency mechanisms can be used to synchronize and exchange information. If data exchange has to be realized between different JVM's the network communication using for example sockets has to be used.

The PCJ library handles both situations hiding details from the user. It distinguishes between inter- and intranode communication and pick up proper data exchange mechanism. Moreover, nodes are organized in the graph which allows to optimize global communication.