When a model's behavior has been validated, the code generator has to reliably transfer it to the target processor, whose resources are usually designed for the greatest possible cost-efficiency.

In other words, the final production ECU generally has less memory and processing power than the RCP system on which the algorithm was developed and tested.

In 1999, dSPACE introduced the production code generator TargetLink,[8] which is integrated into Simulink, the environment for model-based development.

In addition to performing the actual autocoding, including code generation for AUTOSAR software components, TargetLink also makes it possible for developers to compare the behavior of the generated code with that of the original Simulink model (by means of software-in-the-loop (SIL) and processor-in-the-loop (PIL) simulation).

This method provides a way to test new functions reproducibly in a safe environment, before a prototype of the product has even been produced.