Cpp vs CCpp targets

[stustd]

The documentation says:

NOTE: If you intend to use C++ code or import C++ libraries in the C target, we provide a special CCpp target that automatically uses a C++ compiler by default. Alternatively, you might want to look at the native Cpp target, built for C++ code if features such as federated execution are not needed.

I'm interested in the combination that CCpp provides: application code in C++ and federation. However, what's the (recommended) advantage of using the C++ scheduler instead of the C version (where the latter provides both, federation and clock synchronization)?

[edwardalee]

In the Cpp target, reactions are written using a C++ style, whereas in the CCpp target, they are written with a C style. Specifically, accessing inputs, state variables, and parameters is via method calls in the Cpp target. Also, the type system is more flexible in the Cpp target, supporting generics and handling allocation/deallocation in a more intuitive way. Also, the code generated by the Cpp target is more elegant and is object oriented.

I think (hope) the Cpp target will eventually support federated execution, at which point the CCpp target may become less interesting, but I'm not sure about that.

[stustd]

Thanks for the offer Christian, after I've informed myself in comparing the C and C++ scheduler implementations I'll reach out to you for help.

[stustd]

My interest in Lingua Franca (LF) can be summarized by the motto that "A person's non-deterministic decision is not exclusively his/her decision." IMO it's also the core issue the real-time community is grappling with (e.g. ROS2, a viable real-time solution?). Can we say that LF's approach conclusively addresses this problem?

[lhstrh]

@stustd: could you elaborate a bit on this motto? Is this about control decisions being affected by external factors that are beyond the control of the system engineer? If so, then I think, yes, LF confronts this problem. But, interestingly, it is not a black-and-white issue; there are trade-offs. These trade-offs become more apparent in distributed applications (see Quantifying and Generalizing the CAP Theorem (preprint).

[stustd]

@lhstrh: indeed, that was the intent of my question, and I'm glad to hear it confirmed. In the terminology of the CA{P,L} paper, indeed, when Availability decreases (i.e. user latency increases), synchronization will break as dependent reactions at some point in time will have to resume without that information; then we're in a (partially) asynchronous, non-deterministic communication world. These effects will have to be accounted for in mission-critical systems, Under (controlled) lab conditions, however, the LF paradigm (with milestone violations traced as errors), provides a rock-solid testing framework.

[lhstrh]

What is key here is that Lingua Franca provides guarantees under quantifiable assumptions. In centralized coordination, violations of assumptions about observable latency yield a reduction of availability. In decentralized coordination, violations of assumptions about observable latency yield a reduction of consistency.