RTOS programming Strategy needed

[LOSTISLAND]

Hi all ,
It’s my first experience with RTOS ,
I’m completely familiar with RTOS (Tasks , priorities , semaphor , mutex , priority inversion , message boxs , Interrupt handling …) but as far as I haven’t actually had any experience with that I need your help .
I don’t want to make another mistake like I’ve made so far in non OS programming strategy .

Concerning non OS programming strategy :
As a matter of fact it’s just 6 months I’ve became familiar with State Machine strategy . by the time ,
I had never heard of it . I usually used some flags that got invoked when needed but when the program grew larger these flags got somewhat confusing . when I had to implement ISO15765 protocol I found out I need something more than these flags and decided to have some states .

Now , in my latest project , I have some main states and from these states , some have some sub states , _S indicates State and _SS indicates sub state . As a result with a quick look you can trace the program and Main loop isn’t more than 30 lines while the program code in more than 3000 lines .
Unfortunately It was too late to reach to this simple point after 6 years of programming .
if 6 years ago someone had taught that simple matter to me , I woud have had another 6 years .

Now I’m on the eve of RTOS . As I said I don’t want to make another funny mistake like that .
I’m completely familiar with RTX/OS and somewhat with Uc/OS .
Now , I have a project which has Ethernet , CAN , TFT , MMC/SD , Digital and Analog IO’s , … .
Just Ethernet program I’m going to embed in my RTX program has got several states and is somewhat a huge program by itself .
I’ve finished Ethernet , CAN , TFT , MMC , … Program separately (Without RTOS) and want to combine all of these into one RTX program using UVision4 .
The microcontroller is from LPC17xx family .
Any idea , strategy , help , … ?
Thanks in advanced

 

[sky_123]

With large pieces of software, it's not really an RTOS specific question, but more a general question, and best answered by CompSci people
(hopefully there are some on the forum).
There are recommended practices to developing complex software, but usually a framework and architecture is employed.
(For example, if you were running on Linux, then ACE (**broken link removed**) could be used. It provides a library of
functions that support real-time software). Here is an example of what I mean: Say you wanted to pass information from one process to another.
You could use the (effectively) low-level features found in the RTOS, such as mutexes. However, if you have a framework and supporting
software, then it will provide the functions to pass information without needing to use the low-level features, and you'd be able to
wait and act on the information in a more scalable and easier way for larger software, e.g. pass any type of variable type, arrays etc,
and invoke a process just to handle the data, and and it would probably log and provide debug information too, so that you can troubleshoot
your larger code easier.
When you're talking about a microcontroller, then the chances are that the code, while large, is probably still small enough that you
may not need such sophisticated features, but nevertheless it's a decision you need to make - should you spend time creating functions
that will provide supporting features for your main software, like good logging functions, debugging functions etc? Probably yes.
Programming language is also important. There are people who will suggest "you can do anything in C" or "you can do anything in assembler"
but if you've got the space, then C++ will make your life simpler.
Also, there are some well-established methods for doing certain common things. As you've spotted, a state machine is an established method.
Similarly, there are established ways to handle data structures, and established ways to do a whole host of "things". You'd probably need
to spend a few days reading up a book on design patterns. So much to read..
If you're coding for a uni project, then you won't have time to do everything listed above. But then again, you possibly don't need to -
it is possible to implement a lot of software even without an RTOS for example, or using very few processes and try to keep it as simple
as possible - the more processes and complexity then the more troubleshooting that is needed when you get conditions that
were unexpected because of process execution speeds and synchronization, and then the more the need for better logging and
debugging features..
In short, spend some time figuring out what bits of code need to share data or synchronize, can you collapse to a few processes,
and do you have some good debugging facilities.

 

[LOSTISLAND]

Thanks alot