OpenVMS RTL Parallel Processing (PPL$) Manual

This manual provides detailed usage and reference information for the OpenVMS Run-Time Library's Parallel Processing (PPL$) facility. It is intended for system and application programmers developing concurrent programs on OpenVMS systems, whether single-CPU or multiprocessor.

The PPL$ facility enables parallel processing by dividing applications into multiple tasks that execute simultaneously. It helps simplify complex tasks related to process management, shared memory, and synchronization. Key functionalities include:

• Process Management: Routines to create and delete parallel applications and their subordinate processes, manage participant information (e.g., indices, PIDs), and generate application-unique names.
• Shared Memory Operations: Mechanisms for creating, flushing, and deleting global sections (shared memory) and virtual memory zones to facilitate interprocess communication.
• Synchronization Operations: A suite of tools including barriers, events, semaphores, spin locks, and work queues to control the ordering of events and manage access to critical sections and shared resources, ensuring data integrity in parallel execution.

The document describes various software models for parallel processing, such as master/slave, pipelining, and work queue processing. It also covers important programming considerations like granularity and decomposition, data dependence (true, antidependence, output, control), deadlock prevention, and performance measurement using models like the geometric model. System requirements, including necessary privileges (PRMGBL, SYSGBL, SYSLCK) and process quotas (subprocess, AST, enqueue, global section limits), are outlined. The manual discourages the use of SYS$HIBER and disabling ASTs in conjunction with PPL$ routines due to potential unpredictable results and performance issues. Practical examples in BLISS, DEC FORTRAN, and DEC C are provided.