Sequential and pipeline processing modes

The data logger has two processing modes: sequential mode and pipeline mode. In sequential mode, data logger tasks run more or less in sequence. In pipeline mode, data logger tasks run more or less in parallel. Mode information is included in a message returned by the data logger, which is displayed by software when the program is sent and compiled, and it is found in the Status Table, CompileResults field. The CRBasic Editor pre-compiler returns a similar message.

The default mode of operation is pipeline mode. However, when the data logger program is compiled, the data logger analyzes the program instructions and automatically determines which mode to use. The data logger can be forced to run in either mode by placing the PipeLineMode or SequentialMode instruction at the beginning of the program (before the BeginProg instruction).

For additional information, visit the Campbell Scientific blog article, "Understanding CRBasic Program Compile Modes: Sequential and Pipeline ." Or watch an instructional video at: http://www.campbellsci.com/videos/pipeline-sequential .

Sequential mode

Sequential mode executes instructions in the sequence in which they are written in the program. After a measurement is made, the result is converted to a value determined by processing arguments that are included in the measurement instruction, and then program execution proceeds to the next instruction. This line-by-line execution allows writing conditional measurements into the program.

Pipeline mode

Pipeline mode handles measurement, most digital, and processing tasks separately, and, in many cases, simultaneously. Measurements are scheduled to execute at exact times and with the highest priority, resulting in more precise timing of measurements, and usually more efficient processing and power consumption.

In pipeline mode, it will take less time for the data logger to execute each scan of the program. However, because processing can lag behind measurements, there could be instances, such as when turning on a sensor using the SW12() instruction, that the sensor might not be on at the correct time to make the measurement.

Pipeline scheduling requires that the program be written such that measurements are executed every scan. Because multiple tasks are taking place at the same time, the sequence in which the instructions are executed may not be in the order in which they appear in the program. Therefore, conditional measurements are not allowed in pipeline mode. Because of the precise execution of measurement instructions, processing in the current scan (including updating public variables and data storage) is delayed until all measurements are complete. Some processing, such as transferring variables to control instructions, like PortSet() and ExciteV(), may not be completed until the next scan.

When a condition is true for a task to start, it is put in a queue. Because all tasks are given the same priority, the task is put at the back of the queue. Every 1 ms (or faster if a new task is triggered) the task currently running is paused and put at the back of the queue, and the next task in the queue begins running. In this way, all tasks are given equal processing time by the data logger.

Slow Sequences

Priority of a slow sequence (SlowSequence) in the data logger will vary, depending upon whether the data logger is executing its program in pipeline mode or sequential mode. With the important exception of measurements, when running in pipeline mode all sequences in the program have the same priority. When running in sequential mode, the main scan has the highest priority for measurements, followed by background calibration (which is automatically run in a slow sequence), then the first slow sequence, the second slow sequence, and so on. The effects of this priority are negligible; however, since, once the tasks begin running, each task is allotted a 1 msec time slice, after which, the next task in the queue runs for 1 msec. The data logger cycles through the queue until all instructions for all sequences are complete.