This is not GCC, but a general programming question.
What do you use for huge finite state machines? I've wrote this huge state machine with about 20 states (it will probably grow as I add new features), with about 500 lines long (including comments) but I don't quite like how huge it is.
This is not GCC, but a general programming question.What do you use for huge finite state machines? I've wrote this huge state machine with about 20 states (it will probably grow as I add new features), with about 500 lines long (including comments) but I don't quite like how huge it is.
May I ask what the code does?
20 states isn't THAT much, but I guess you can always find a way to rewrite some of the states as variables and/or flags, thus reducing the number of "apparent" states. Or you can make two parallel FSM's.
Not sure it would always make things easier though. It depends on the overall logic of the application.
clawson 
A common solution is to use a table processor and state tables rather than doing all the state transitions programmatically.
hierarchical state machine?
May I ask what the code does?20 states isn't THAT much, but I guess you can always find a way to rewrite some of the states as variables and/or flags, thus reducing the number of "apparent" states. Or you can make two parallel FSM's.
Not sure it would always make things easier though. It depends on the overall logic of the application.
The state machine parses commands received by uart (char by char) of my VFD display/boost converter circuit. I may be able to divide it in two, one for each section (VFD or boost), but I don't think it will help that much.
One thing I'm thinking about is creating a inline function for each big state and calling it instead of leaving the code inside the switch statement. As the function is inline, it won't have any penalty in execution.
hierarchical state machine?
This is kind of dividing the state machine, right?
Do you use a big switch statement for the state machine?
When I desigin a state machine I usually give each state of the state machine it's own function.
The current state is remembered via a function pointer, the switch statement is completely eliminated. Just execute the function to which the function pointer points.
The context of the state machine can be put in some global variables.
Or you can put the whole context in a structure and pass a pointer to that stucture to the current state.
This gives you the flexibility of running multiple state machines at the same time (concurrently?) with different contexts.
You can find a complete example with source code and some documentation on my website
http://www.hoevendesign.com/#Net...
20 states is not such a big state machine. I've been working on an SDI-12 interface with about that many. Yes, it gets awkward and difficult to manage, but that the case with any project that gets big.
Jim
clawson
Is it something like this?
Another good example is the menu processing in Atmel's Butterfly application.
curtvm If you are using gcc, labels as values can be handy-
https://www.avrfreaks.net/index.p...
but I don't quite like how huge it is.
What is it that you don't like? The number of lines of code, the size of the code produced, to slow, maintaining/understanding it without a proper state diagram and design, the abstraction of FSM's?
about 500 lines long (including comments)
What is it that you don't like? The number of lines of code, the size of the code produced, to slow, maintaining/understanding it without a proper state diagram and design, the abstraction of FSM's?
Too many lines of code inside a single function makes it hard to follow. I like small simple functions, but looks like I won't have that with a switch statement FSM. I just wanted to check if there were any other way of doing it...
Quote:about 500 lines long (including comments)
As the comments are probably a third of the number of lines of code I would call the size of it very small.
Actually most of the comments are in the same line of the code, so i'm guessing about 400 lines of code.
A major factor of course is the number of inputs to the state machine and how much you are able to to simplify by
- having multiple state machines each with less inputs
- elimination of redundant/duplicated states states
Too many lines of code inside a single function makes it hard to follow.
stevech State machine drawings - Anyone know of a free drawing software tool, or free add-on to MS office/Visio? I tried Visio, but too awkward.
buffi If you really want to dig into it, try LaTeX and the tikz package. LaTeX is a typesetting system using its own kind of markup language, you actually write source code for your document, which is then compiled to dvi, ps or pdf.
tikz is a drawing package for LaTeX, which can be used to draw state diagrams with all kinds of fancy additions.
The approach is totally different from any wysiwyg tool like word or visio, and the learning curve is very steep - but it's free and I'd always recommend it. It's free, customizable and extendable. You can use your favorite editor for coding. The source code can be split into several files, just as you would split up c code.
JohanEkdahl 
I use a lead pencil on A4 & an eraser for editing. Quick cheap & effective. Final product can be documented electronically using http://qfsm.sourceforge.net/scre... or just scanned into PDF for for subsequent filing if required.
Most packages just do not have a nice bezier curve drawing tool which will go from a state to previous state.
:idea: GUI FSM design interface for AVRStudio 5 :idea:
Edit.
I have just had a fiddle with Qfsm (after many years) and it now has a syntax checker & simulator. The bezier curves are easy to use.
gchapman State machine drawings - Anyone know of a free drawing software tool, or free add-on to MS office/Visio? I tried Visio, but too awkward.
Visio -
Quantum Leaps has Visio stencils to ease use of Visio
and has a link to an open UML tool:
http://www.state-machine.com/resources/goodies.php
stevech Qsfm requires, (my win 7 PC) an install of QT and all its baggage. Is that onerous?
Hi Steve, I had Qt installed on my Windows 7 box anyway , but I never saw or have seen, a problem with it. I reckon that the next best thing to XP-SP3 is Windows7
Qfsm runs nicely on my box, but I still reckon I will do the design on paper and do the final drawing with Qfsm(if required)
I really think that the OP is probably getting confused because he is trying to understand whats going on by looking at code. Really hard to with more that 5 ot 6 states.
stevech Yes, simple pencil and paper state diagrams is what I use to ensure the overall design makes sense - and so that I can see the big picture. With computer documents, I always feel like the view is via a soda-straw.
Even when the top level design gets to a B sized drawing, it's still better than a dozen little computer files on the LCD.
I have a professional endeavor now that is 10 or so independent state machines, each running under the supervision of a cooperative scheduler (single stack, to save RAM). This is a run-to-completion scheduler. Single stack means the state machines either use static variables (in C) or get some RAM, cast it to a struct of variables, and keep the pointer to that in the task's TCB - making the task's code reentrant.
Using a single stack eliminates the worry "is my stack big enough for the worst case nesting which may occur after I think I'm done testing?" and the "I don't have enough RAM - dare I make some stacks smaller?.
The scheduler for this is a later version of "OPEX", for which the source code is in the projects section on this forum. Lots of changes made, and a port to an ARM7 done.
