Parallel, non-blocking access to an inference module

This sample shows how an instance of FB_MllPrediction can be accessed from two tasks running concurrently.

The sample can be downloaded here: concurrent sample.

The instance fbpredict is declared in the GVL_ML. All programs in the PLC thus have access to the instance. The following are created as programs:

• P_InitML: The step sequence for initializing/loading an ML model is described here.
• P_Predict_Task1: The Predict method of the fbpredict is called, wherein PRG is executed on Core 1.
• P_Predict_Task2: The Predict method of the fbpredict is called, wherein PRG is executed on Core 2.

The essential components for the concurrent execution of 2 Predict calls are:

• The maximum number of concurrent accesses must be specified with the Configure method: GVL_ML.fbpredict.stPredictionParameter.MaxConcurrency := nMaxConcurrency; with nMaxConcurrency = 2.
  The instance then keeps this number of independent inference machines available.
• A unique ID of the calling context must be specified when calling the Predict method. These are declared as constants in P_Predict_Task1 and P_Predict_Task2, see nConcurrencyId.
  The user must ensure that each calling context transfers a unique ID with the Predict call.
• The remainder of the source code is largely identical to the Quick start sample.