RT inter-process communication

Summary

• Inter-process communication refers to data exchanges between two process loops running in parallel on a given target
• Some inter-process communication techniques facilitate data exchange between a deterministic process loopA special type of while-loop structure that executes with a precisely-defined time per loop iteration. and a nondeterministic process loopA conventional while-loop structure whose execution time may vary unexpectedly from one loop iteration to the next. to ensure that the deterministic loop operates properly
• Communication techniques between deterministic and nondeterministic process loops include:
  • tagTransfer only the current (most recent) data value between two process loops or between two targets communication:
    • Single-process shared variable with RT FIFO enabled (single element)
  • streamingTransfer every data value through a first-in first-out (FIFO) queue and messageTransfer every value of intermittent data with low latency. communication:
    • Single-process shared variable with RT FIFO enabled (multi-element)
    • RT FIFO functions
• Communication techniques between nondeterministic process loops include:
  • tagTransfer only the current (most recent) data value between two process loops or between two targets communication:
    • Local variable
    • Single-process shared variable
    • Functional global variable
  • streamingTransfer every data value through a first-in first-out (FIFO) queue and messageTransfer every value of intermittent data with low latency. communication:
    • Queues
• The channel wireVisual representation of asynchronous data flow between parallel loops (new feature in LabVIEW 2016) (introduced with LabVIEW 2016) implements all of the above communication techniques, often with much less programming effort!

Communication between deterministic and nondeterministic process loops

• The Timed Loop implements a deterministic process loopA special type of while-loop structure that executes with a precisely-defined time per loop iteration. that reliably operates at a desired rate with very low jitterUnpredictable variation in process loop execution time from one iteration to the next.

• Deterministic process loops represent one of the most important constructs offered by a real-time operating system (RTOS) such as the NI Linux Real-Time operating system that runs on the RT target: the RTOS provides a wide range of high-level programming features similar to those available on the PC host such as parallel execution threads, file I/O and network access, all the while ensuring that the code running inside a timed loop strictly meets proscribed timing requirements required for process control systems and other time-sensitive tasks

• Place non-deterministic code in adjacent conventional loops and then use appropriate inter-process data communication techniques to exchange values; in this way the timed loop process concentrates on the time-critical task (e.g., managing a high-speed PID process control loop) while the conventional loop deals with the non-time-critical tasks (e.g., logging measurements to a data file or communicating with the PC host through the network)

Code examples

• Single-process shared variable (SPSV) with RT FIFO enabled (single element) – Use the SPSV like a global variable to link deterministic and nondeterministic process loops; this method is suitable for tagTransfer only the current (most recent) data value between two process loops or between two targets / latest valueTransfer only the current (most recent) data value between two process loops or between two targets communication
• Channel wires – this method supports all communication types

Procedures

• Create a network-published shared variable (NPSV) – follow this general procedure, but select “Single Process” as the desired variable type

Communication between nondeterministic process loops

• The conventional while-loop structure is a nondeterministic process loopA conventional while-loop structure whose execution time may vary unexpectedly from one loop iteration to the next. ; its actual execution time can be quite accurate on average, but the execution time for any given loop iteration is not well controlled; this variation in loop time known as jitterUnpredictable variation in process loop execution time from one iteration to the next. can be quite significant, e.g., tens of percent or more, and is generally unacceptable for time-critical tasks such as process loop control

• Sources of jitter include use of subVIs that occasionally require more time to complete a task (e.g., file system and network I/O), shared resource contention, and operating system scheduling for parallel processes and routine housekeeping tasks.

• Use nondeterministic loops for non-time-critical tasks.

• See this application example to compare loop jitter of pulsed waveforms generated by the FPGA, a deterministic loop, and a non-deterministic loop

Code examples

• Local variable – this method is suitable for tagTransfer only the current (most recent) data value between two process loops or between two targets / latest valueTransfer only the current (most recent) data value between two process loops or between two targets communication
• Global variable – this method is suitable for tagTransfer only the current (most recent) data value between two process loops or between two targets / latest valueTransfer only the current (most recent) data value between two process loops or between two targets communication
• Single-process shared variable (SPSV) – use the SPSV like a global variable to link nondeterministic process loops – do not enable the “RT FIFO” option; this method is suitable for tagTransfer only the current (most recent) data value between two process loops or between two targets / latest valueTransfer only the current (most recent) data value between two process loops or between two targets communication
• Functional global variable (FGV) – use the FGV like a global variable with extended functionality; this method is suitable for tagTransfer only the current (most recent) data value between two process loops or between two targets / latest valueTransfer only the current (most recent) data value between two process loops or between two targets communication
• Queue – this method is suitable for streamingTransfer every data value through a first-in first-out (FIFO) queue / messageTransfer every value of intermittent data with low latency. communication
• Channel wires – this method supports all communication types

For more information

1. Using the LabVIEW Shared Variable (http://www.ni.com/white-paper/4679/en)
   Introduces the single-process shared variable and discusses features and performance.
   
2. Channel Wires (http://www.ni.com/white-paper/53423/en/)
   From the overview: "The channel wire, a new, feature in LabVIEW 2016, simplifies the programming involved in data transfer between parallel loops, such as the common producer/consumer pattern. This white paper provides an introduction to channel wires and highlights specific use cases".