Use C++ standard Library in a stm32 firmware

[Vergnani_ElEn]

Hi,

I'm currently developing firmware with an event-based architecture. My colleague and I thought about using the standard library, but this approach is risky because the standard library requires a lot of heap memory.

My question is: when is it possible or recommended to use standard library functions such as std::vector, std::map, and std::queue? And when is it mandatory to avoid using them?

What advice would you give me for implementing a good event-driven architecture?

[Andrew Neil]

---

@Vergnani_ElEn wrote:

What advice would you give me for implementing a good event-driven architecture?

---

It is certainly possible to do that without the C++ standard library.

In fact, without C++ at all.

State Machines, callbacks, etc can easily be done in plain ol' C ...

[TDK]

In general, dynamic memory allocation is frowned upon in embedded electronics applications. If you are designing safety critical software, the rules will generally disallow this. For example, see MISRA recommendations.

If your software crashing is not a big deal, just an inconvenience, then go for it. It's up to you what guidelines you want to follow in your code. There is no one size fits all rule.

Let the requirements of the application drive the design and implementation choices.

[Andrew Neil]

---

@Vergnani_ElEn wrote:

this approach is risky because the standard library requires a lot of heap memory.

---

The risk is not so much in the use of heap memory as in the re-use - ie, if you keep allocating & deallocating things.

If you only allocate - never deallocate - then there's not a problem.

If you are careful about de-allocation and garbage collections, then you may be able to avoid the problem.

The trouble with the C++ standard Library is knowing what it's actually doing behind the scenes...

[CTapp.1]

indeed, and the MISRA C++:2023 document has quite a good section on the issues related to dynamic memory and the techniques that can be used to work round them (such as using memory pools that are pre-allocated at startup).

As you say, best practice is to try to allocate all the memory that will be needed at startup so that the system can decide if it needs to operate normally, in a degrade mode (not enough memory available), or enter a fail-safe state.

[Vergnani_ElEn]

Where I could find a MISRA recommendations regarding heap and dinamic memory allocation.

[Andrew Neil]

From MISRA, of course!

https://misra.org.uk/publications/