Myths

As I started working in the embedded environment I was astonished that there was so much prejudice against the usage of C++ in the embedded programming. Most are based on a wrong understanding of the programming language C++.

The Myths

First, I have to say a few words about the myths around C++ I often heard. Of course, this post reflects my perception. Examples?

• Templates inflate the code.
• Objects must live on the heap
• Exceptions are expensive.
• C++ is too slow and needs too much memory.
• C++ is too dangerous in safety-critical systems.
• You have to program object-oriented in C++.
• C++ can only be used for applications.
• The iostream library is too big; the Standard Template Library to slow.

Or to say it in one statement.

=> C++ is a nice toy but we are dealing with serious challenges.

The list of (pre)judice is long consisting partially of half-truth and untruth statements often stated by experienced C programmer. I will only refer to the untruth statements. The half-truth statements are to a large extent questions due to the right usage of C++ and to small extent questions of the implementation of the core and the libraries of C++. 

• Objects must live on the heap.
  • Objects can be created on the stack or at an arbitrary position with the help of placement new.
• C++ is too dangerous in safety-critical systems.
  
  • Of course, it depends on the experience of the developer. But whoever uses C strings instead of C++ strings; uses C arrays instead of C++ arrays; uses macros instead of constant expressions or templates, can not argue that C++ is not well suited for safety-critical systems. Honestly, the contrary holds. C++ has a lot to offer in safety-critical systems. 
• You have to program object-oriented in C++.
  
  • C++  is a multi-paradigm language. Therefore, you can solve your problem in an object-oriented, structured, functional, generic, or generative style.
• C++ can only be used for applications.
  • C++ is for example used for a fire extinguisher, defibrillators, and cars. ARM maintains with  ARM GCC  the current GCC collection together with the gnu toolchain. Therefore, the current g++ compiler is available. The packages that are very frequently used is maintained by ARM for their processors that are the default architecture for the embedded world.

What's the reason for the half-truth? I think, there are more reasons.

• Old C++ compilers
  • The knowledge is based on old C++ compilers of the last millennium. They implement the C++98 standard but they have a large potential for optimization.
• Training deficit
  
  • On one hand, many embedded programmers learned only C, on the other hand, there is no time to experiment with new technologies.
• Loss of expert state
  
  • You have to be brave to leave your area as a C expert and continue the next day as a C++ novice.
• Legacy codebase in C
  • The existing code base is in C therefore is the obvious decision to fix a bug or implement a feature in the first place in C.
• Many C experts
  
  • There are many C experts. They are training the novices and becoming the leaders.
• Curse of the monoculture
  
  • I perceive the embedded world often as a monoculture. I worked 15 years as a consultant in the automobile area and used about 10 languages. On the contrary, I used only 3 languages in the embedded area. 
• The pressure of the standard
  
  • There are a lot of standards that you have to fulfil. The courage to use new technologies seems to be inversely proportional to the pressure of the standards. 
• Insufficient knowledge about C++
  • Many developers have not sufficient knowledge of classical C++ and no knowledge of modern C++.

   

Maybe, I will polarize with this post. But If it helps to make the great features of modern C++ better known in the embedded world then I want to do it voluntarily. In the next post, I compare the myths with the facts. In particular, I will write about the Technical Report on C++ Performance.