Chapter 1: Program Structure

What's in Chapter 1?

This chapter gives a basic overview of programming in C for an embedded system. We will introduce some basic terms so that you get a basic feel for the language. Since this is just the first of many chapters it is not important yet that you understand fully the example programs. The examples are included to illustrate particular features of the language.

Case Study : Microcomputer-Based Lock
      To illustrate the software development process, we will implement a simple digital lock. The lock system has 7 toggle switches and a solenoid as shown in the following figure. If the 7-bit binary pattern on Port A bit 6 to bit 0 becomes 0100011 for at least 10 ms, then the solenoid will activate. The 10 ms delay will compensate for the switch bounce. For information on switches and solenoids see Sections 4.2 and 8.6.3 of Embedded Systems: Introduction to ARM Cortex M Microcontrollers by Jonathan W. Valvano. For now what we need to understand is that Port A bits 6-0 are input signals to the computer and Port A bit 7 is an output signal.
      

Before we write C code, we need to develop a software plan. Software development is an iterative process. Even though we list steps the development process in a 1,2,3... order, in reality we iterative these steps over and over.
      1) We begin with a list of the inputs and outputs. We specify the range of values and their significance. In this example we will use PORTA. Bits 6-0 will be inputs. The 7 input signals represent an unsigned integer from 0 to 127. Port A bit 7 will be an output. If PA7 is 1 then the solenoid will activate and the door will be unlocked. In C, we use #define MACROS to assign a symbolic names to the corresponding addresses of the ports.
#define GPIO_PORTA_DATA_R       (*((volatile unsigned long *)0x400043FC))
#define GPIO_PORTA_DIR_R        (*((volatile unsigned long *)0x40004400))
#define GPIO_PORTA_DEN_R        (*((volatile unsigned long *)0x4000451C))
#define SYSCTL_PRGPIO_R         (*((volatile unsigned long *)0x400FEA08))

      2) Next, we make a list of the required data structures. Data structures are used to save information. If the data needs to be permanent, then it is allocates in global space. If the software will change its value then it will be allocated in RAM. In this example we need a 32-bit unsigned counter.
unsigned int cnt;

If data structure can be defined at compile time and will remain fixed, then it can be allocated in ROM. In this example we will define an 8-bit fixed constant to hold the key code, which the operator needs to set to unlock the door. The compiler will place these lines with the program so that they will be defined in Flash ROM memory.
const unsigned char key=0x23; // The key code 0100011 (binary) 

It is not real clear at this point exactly where in ROM this constant will be, but luckily for us, the compiler will calculate the exact address automatically. After the program is compiled, we can look in the listing file or in the map file to see where in memory each structure is allocated.
      3) Next we develop the software algorithm, which is a sequence of operations we wish to execute. There are many approaches to describing the plan. Experienced programmers can develop the algorithm directly in C language. On the other hand, most of us need an abstractive method to document the desired sequence of actions. Flowcharts and pseudo code are two common descriptive formats. There are no formal rules regarding pseudo code, rather it is a shorthand for describing what to do and when to do it. We can place our pseudo code as documentation into the comment fields of our program. The following shows a flowchart on the left and pseudo code and C code on the right for our digital lock example.

Normally we place the programs in Flash ROM. Typically, the compiler will initialize the stack pointer to the last location of RAM. On the ARM Cortex M, the stack is initialized to the 32-bit value located at ROM address 0. Next we write C code to implement the algorithm as illustrated in the above flowchart and pseudo code.

   4) The last stage is debugging. For information on debugging see  Sections 2.6, 4.6, 5.8 and 6.9 of Embedded Systems: Introduction to ARM Cortex M Microcontrollers by Jonathan W. Valvano. 

Free field language

In most programming languages the column position and line number affect the meaning. On the contrary, C is a free field language. Except for preprocessor lines (that begin with #, see Chapter 11), spaces, tabs and line breaks have the same meaning. The other situation where spaces, tabs and line breaks matter is string constants. We can not type tabs or line breaks within a string constant. For more information see the section on strings in the constants chapter. This means we can place more than one statement on a single line, or place a single statement across multiple lines. For example a function could have been written without any line breaks

"Since we rarely make hardcopy printouts of our software, it is not necessary to minimize the number of line breaks."

However, we could have added extra line breaks and comments to make it more readable.

At this point I will warn the reader, just because C allows many different forms of syntax,  proper syntax will have a profound impact on the quality of our code. After much experience you will develop a programming style that is easy to understand. Although spaces, tabs, and line breaks are syntactically equivalent, their proper usage will have a profound impact on the readability of your software. For more information on programming style see Section 5.6 of Embedded Systems: Introduction to ARM Cortex M Microcontrollers by Jonathan W. Valvano.

A token in C can be a user defined name (e.g., the variable Info and function Lock_Init) or a predefined operation (e.g., *, unsigned, while). Each token must be contained on a single line. We see in the above example that tokens can be separated by white spaces (space, tab, line break) or by the special characters, which we can subdivide into punctuation marks (Table 1-1) and operations (Table 1-2). Punctuation marks (semicolons, colons, commas, apostrophes, quotation marks, braces, brackets, and parentheses) are very important in C. It is one of the most frequent sources of errors for both the beginning and experienced programmers.

punctuation	Meaning
;	End of statement
:	Defines a label
,	Separates elements of a list
( )	Start and end of a parameter list
{ }	Start and stop of a compound statement
[ ]	Start and stop of a array index
" "	Start and stop of a string
' '	Start and stop of a character constant

Table 1-1: Special characters can be punctuation marks

The next table shows the single character operators. For a description of these operations, see Chapter 5.

operation	Meaning
=	assignment statement
@	address of
?	selection
<	less than
>	greater than
!	logical not (true to false, false to true)
~	1's complement
+	addition
-	subtraction
*	multiply or pointer reference
/	divide
%	modulo, division remainder
|	logical or
&	logical and, or address of
^	logical exclusive or
.	used to access parts of a structure

Table 1-2: Special characters can be operators

The next table shows the operators formed with multiple characters. For a description of these operations, see Chapter 5.

operation	Meaning
==	equal to comparison
<=	less than or equal to
>=	greater than or equal to
!=	not equal to
<<	shift left
>>	shift right
++	increment
--	decrement
&&	Boolean and
||	Boolean or
+=	add value to
-=	subtract value to
*=	multiply value to
/=	divide value to
|=	or value to
&=	and value to
^=	exclusive or value to
<<=	shift value left
>>=	shift value right
%=	modulo divide value to
->	pointer to a structure

Table 1-3: Multiple special characters also can be operators

Although the operators will be covered in detail in Chapter 9, the following section illustrates some of the common operators. We begin with the assignment operator. Notice that in the line x=1; x is on the left hand side of the = . This specifies the address of x is the destination of assignment. On the other hand, in the line z=x; x is on the right hand side of the = . This specifies the value of x will be assigned into the variable z. Also remember that the line z=x; creates two copies of the data. The original value remains in x, while z also contains this value.

short x,y,z; /* Three variables */
void Example(void){
   x = 1;          /* set the value of x to 1 */
   y = 2;          /* set the value of y to 2 */
   z = x;          /* set the value of z to the value of x (both are 1) */  
   x = y = z = 0;  /* all all three to zero   */   
}

Next we will introduce the arithmetic operations addition, subtraction, multiplication and division. The standard arithmetic precedence apply. For a detailed description of these operations, see Chapter 5.

short x,y,z; /* Three variables */
void Example(void){
   x=1; y=2;   /* set the values of x and y */
   z = x+4*y;  /* arithmetic operation */  
   x++;        /* same as x=x+1;   */   
   y--;        /* same as y=y-1;   */   
   x = y<<2;   /* left shift same as x=4*y;   */    
   z = y>>2;   /* right shift same as x=y/4;  */    
   y += 2;     /* same as y=y+2;   */   
}

Next we will introduce a simple conditional control structure. Assume PORTB is configured as an output port, and PORTE as an input port. For more information on input/output ports see Chapter 4 of Embedded Systems: Introduction to ARM Cortex M Microcontrollers by Jonathan W. Valvano. The expression PORTE&0x04 will return 0 if PORTE bit 2 is 0 and will return a 4 if PORTE bit 2 is 1. The expression (PORTE&0x04)==0 will return TRUE if PORTE bit 2 is 0 and will return a FALSE if PORTE bit 2 is 1. The statement immediately following the if will be executed if the condition is TRUE. The else statement is optional.

#define PORTB (*((volatile unsigned long *)0x400053FC))
#define PORTE (*((volatile unsigned long *)0x400243FC))
void Example(void){
  if((PORTE&0x04)==0){ /* test bit 2 of PORTE */
    PORTB = 0;       /* if PORTE bit 2 is 0, then make PORTB=0 */  
  }else{
    PORTB = 100;     /* if PORTE bit 0 is not 0, then make PORTB=100 */
  }
}

In the next example lets assume that PORTA bit 3 is configured as an output pin on the TM4C123. Like the if statement, the while statement has a conditional test (i.e., returns a TRUE/FALSE). The statement immediately following the while will be executed over and over until the conditional test becomes FALSE.

#define PORTA (*((volatile unsigned long *)0x400043FC))
#define PORTB (*((volatile unsigned long *)0x400053FC))
void Example(void){ /* loop until PORTB equals 200 */
  PORTB = 0;
  while(PORTB != 200){
    PORTA = PORTA^0x08;}  /* toggle PORTA bit 3 output */
    PORTB++;}             /* increment PORTB output */  
}

The for control structure has three parts and a body. for(part1;part2;part3){body;} The first part PORTB=0 is executed once at the beginning. Then the body PORTA = PORTA^0x08; is executed, followed by the third part PORTB++. The second part PORTB!=200 is a conditional. The body and third part are repeated until the conditional is FALSE. For a more detailed description of the control structures, see Chapter 6.

#define PORTB (*((volatile unsigned long *)0x400053FC))
void Example(void){       /* loop until PORTB equals 200 */
  for(PORTB=0; PORTB != 200; PORTB++){
    PORTA = PORTA^0x08;}  /* toggle PORTA bit 3 output */
  }    
}

Precedence

As with all programming languages the order of the tokens is important. There are two issues to consider when evaluating complex statements. The precedence of the operator determines which operations are performed first. In the following example, the 2*x is performed first because * has higher precedence than + and =. The addition is performed second because + has higher precedence than =. The assignment = is performed last. Sometimes we use parentheses to clarify the meaning of the expression, even when they are not needed. Therefore, the line z=y+2*x; could also have been written z=2*x+y; or z=y+(2*x); or z=(2*x)+y;.

The second issue is the associativity. Associativity determines the left to right or right to left order of evaluation when multiple operations of the precedence are combined. For example + and - have the same precedence, so how do we evaluate the following?

We know that + and - associate the left to right, this function is the same as z=(y-2)+x;. Meaning the subtraction is performed first because it is more to the left than the addition. Most operations associate left to right, but the following table illustrates that some operators associate right to left.

Precedence	Operators	Associativity
highest	()  []  .  ->  ++(postfix)   --(postfix)	left to right
	++(prefix)  --(prefix)   !~ sizeof(type) +(unary)  -(unary) &(address) *(dereference)	right to left
	*   /   %	left to right
	+   -	left to right
	<<   >>	left to right
	<    <=   >   >=	left to right
	==  !=	left to right
	&	left to right
	^	left to right
	|	left to right
	&&	left to right
	||	left to right
	? :	right to left
	=   +=   -=  *=  /=  %=  <<=  >>=  |=  &=  ^=	right to left
lowest	,	left to right

Table 1-4: Precedence and associativity determine the order of operation

"When confused about precedence (and aren't we all) add parentheses to clarify the expression."

Comments

There are two types of comments. The first type explains how to use the software. These comments are usually placed at the top of the file, within the header file, or at the start of a function. The reader of these comments will be writing software that uses or calls these routines. The second type of comments assists a future programmer (ourselves included) in changing, debugging or extending these routines. We usually place these comments within the body of the functions. The comments on the right of each line are examples of the second type. For more information on writing good comments see Section 5.6 of Embedded Systems: Introduction to ARM Cortex M Microcontrollers by Jonathan W. Valvano.

Comments begin with the /* sequence and end with the */ sequence. They may extend over multiple lines as well as exist in the middle of statements. The following is the same as PORTA=0x80;

Keil uVision does allow the use of C++ style comments. The start comment sequence is // and the comment ends at the next line break or end of file. Thus, the following two lines are equivalent:

C does allow the comment start and stop sequences within character constants and string constants. For example the following string contains all 7 characters, not just the ac

Most compilers unfortunately do not support comment nesting. This makes it difficult to comment out sections of logic that are themselves commented. For example, the following attempt to comment-out the PORTA = 0x00; will result in a compiler error.

The conditional compilation feature can be used to temporarily remove and restore blocks of code.

Preprocessor Directives

Preprocessor directives begin with # in the first column. As the name implies preprocessor commands are processed first. I.e., the compiler passes through the program handling the preprocessor directives. Although there are many possibilities (assembly language, conditional compilation, interrupt service routines), I thought I'd mention the two most important ones early in this document. We have already seen the macro definition (#define) used to define I/O ports and bit fields. A second important directive is the #include, which allows you to include another entire file at that position within the program. The following directive will define all the TM4C123 I/O port names.

Examples of #include are shown below, and more in Chapter 11.

Global Declarations

An object may be a data structure or a function. Objects that are not defined within functions are global. Objects that may be declared in Keil uVision include:

Keil uVision supports 64-bit long long integers and floating point. In this document we will focus on 8-bit 16-bit and 32-bit objects. Object code generated with most ARM compilers is often more efficient using 32-bit parameters rather than 8-bit or 16-bit parameters.

Declarations and Definitions

It is important for the C programmer to distinguish the two terms declaration and definition. A function declaration specifies its name, its input parameters and its output parameter. Another name for a function declaration is prototype. A data structure declaration specifies its type and format. On the other hand, a function definition specifies the exact sequence of operations to execute when it is called. A function definition will generate object code (machine instructions to be loaded into memory that perform the intended operations). A data structure definition will reserve space in memory for it. The confusing part is that the definition will repeat the declaration specifications. We can declare something without defining it, but we cannot define it without declaring it. For example the declaration for the function UART_OutChar could be written as

We can see that the declaration shows us how to use the function, not how the function works. Because the C compilation is a one-pass process, an object must be declared or defined before it can be used in a statement. (Actually the preprocess performs a pass through the program that handles the preprocessor directives.)  One approach is to first declare a functions, use it, and lastly defines the functions:

       long add (long, long);

    int main(void) {
      long x,y,z;
      x = 2; y = 3;
      z = add(x,y);
      return 1;
    }

    long add(long a, long b){
      return (a+b);
    }

An object may be said to exist in the file in which it is defined, since compiling the file yields a module containing the object. On the other hand, an object may be declared within a file in which it does not exist. Declarations of data structures are preceded by the keyword extern. Thus,

defines 32-bit signed integer called RunFlag; whereas,

only declares the RunFlag to exist in another, separately compiled, module. Thus the line

declares the function name and type just like a regular function declaration. The extern tells the compiler that the actual function exists in another module and the linker will combine the modules so that the proper action occurs at run time. The compiler knows everything about extern objects except where they are. The linker is responsible for resolving that discrepancy. The compiler simply tells the assembler that the objects are in fact external. And the assembler, in turn, makes this known to the linker.

Functions

A function is a sequence of operations that can be invoked from other places within the software. We can pass 0 or more parameters into a function. The code generated by the Keil uVision C compiler passes the first four input parameters in Register R0, R1, R2 and R3 and the remaining parameters are passed on the stack. A function can have 0 or 1 output parameter. The return parameter is placed in Register R0 (8-bit or 16-bit return parameters are promoted to 32 bits.) The add function below (an improvement that checks for overflow) has two 16-bit signed input parameters, and one 16-bit output parameter. Again the numbers in the first column are not part of the software, but added to simplify our discussion.

Listing 1-8: Example of a function call

The interesting part is that after the operations within the function are performed, control returns to the place right after where the function was called. In C, execution begins with the main program. The execution sequence is shown below:

Notice that the return from the first call goes to line 8, while all the other returns go to line 11. The execution sequence repeats lines 9,10,1,2,3,4,5,11 indefinitely.

The programming language Pascal distinguishes between functions and procedures. In Pascal a function returns a parameter while a procedure does not. C eliminates the distinction by accepting a bare or void expression as its return parameter.

C does not allow for the nesting of procedural declarations. In other words you can not define a function within another function. In particular all function declarations must occur at the global level.

A function declaration consists of two parts: a declarator and a body. The declarator states the name of the function and the names of arguments passed to it. The names of the argument are only used inside the function. In the add function above, the declarator is (short x, short y) meaning it has two 16-bit input parameters. 

I prefer to include the void because it is a positive statement that there are no input parameters. However, one must specify the output parameter, even if there is none. For more information on functions see Chapter 10.

The body of a function consists of a statement that performs the work. Normally the body is a compound statement between a {} pair. If the function has a return parameter, then all exit points must specify what to return. In the following median filter function shown in Listing 1-4, there are six possible exit paths that all specify a return parameter.

On a reset or a power initialization, the 32-bit value in ROM location 0x000000004 is loaded into the program counter, PC. The programs created using Keil uVision actually begin execution at a place called Reset_Handler, which can be found in the start.s file. After a power on or hardware reset, the embedded system will initialize the stack, initialize the heap, and clear all RAM-based global variables. After this brief initialization sequence the function named main() is called. Consequently, there must be a main() function somewhere in the program. If you are curious about what really happens, look in the assembly file start.s For programs not in an embedded environment (e.g., running on your PC) a return from main() transfers control back to the operating system. As we saw earlier, software for an embedded system usually does not quit. 

Compound Statements

A compound statement (or block) is a sequence of statements, enclosed by braces, that stands in place of a single statement. Simple and compound statements are completely interchangeable as far as the syntax of the C language is concerned. Therefore, the statements that comprise a compound statement may themselves be compound; that is, blocks can be nested. Thus, it is legal to write

Listing 1-9: Example of nested compound statements.

Although C is a free-field language, notice how the indenting has been added to the above example. The purpose of this indenting is to make the program easier to read. On the other hand since C is a free-field language, the following two statements are quite different

In both cases n2=100; is executed if n1>100. In the first case the statement n3=0; is always executed, while in the second case n3=0; is executed only if n1>100.

Global Variables

Variables declared outside of a function, like Count in the following example, are properly called external variables because they are defined outside of any function. While this is the standard term for these variables, it is confusing because there is another class of external variable, one that exists in a separately compiled source file. In this document we will refer to variables in the present source file as globals, and we will refer to variables defined in another file as externals.

There are two reasons to employ global variables. The first reason is data permanence. The other reason is information sharing. Normally we pass information from one module to another explicitly using input and output parameters, but there are applications like interrupt programming where this method is unavailable. For these situations, one module can store data into a global while another module can view it. 

In the following example, we wish to maintain a counter of the number of times PA3 is toggled. This data must exist for the entire life of the program. This example also illustrates that with an embedded system it is important to initialize RAM-based globals at run time. Most C compilers (including uVision) will automatically initialize globals to zero at startup.

Listing 1-10: A global variable contains permanent information

Although the following two examples are equivalent, I like the second case because its operation is more self-evident. In both cases the global is allocated in RAM, and initialized at the start of the program to 1.

Listing 1-11: A global variable initialized at run time by the compiler

Listing 1-12: A global variable initialized at run time by the compiler

From a programmer's point of view, we usually treat the I/O ports in the same category as global variables because they exist permanently and support shared access.

Local Variables

Local variables are very important in C programming. They contain temporary information that is accessible only within a narrow scope. We can define local variables at the start of a compound statement. We call these local variables since they are known only to the block in which they appear, and to subordinate blocks. The following statement adjusts x and y such that x contains the smaller number and y contains the larger one. If a swap is required then the local variable z is used.

Notice that the local variable z is declared within the compound statement. Unlike globals, which are said to be static, locals are created dynamically when their block is entered, and they cease to exist when control leaves the block. Furthermore, local names supersede the names of globals and other locals declared at higher levels of nesting. Therefore, locals may be used freely without regard to the names of other variables. Although two global variables can not use the same name, a local variable of one block can use the same name as a local variable in another block. Programming errors and confusion can be avoided by understanding these conventions.

Source Files

Our programs may consist of source code located in more than one file. The simplest method of combining the parts together is to use the #include preprocessor directive. Another method is to compile the source files separately, then combine the separate object files as the program is being linked with library modules. The linker/library method should be used when the programs are large, and only small pieces are changed at a time. On the other hand, most embedded system applications are small enough to use the simple method. In this way we will compile the entire system whenever changes are made. Remember that a function or variable must be defined or declared before it can be used. The following example is one method of dividing our simple example into multiple files.

Listing 1-13: Header file for Port A I/O ports

Listing 1-14: Header file for the Port A functions

Listing 1-15: Implementation file for the Port A interface

Listing 1-16: Main program file for this system

With Keil uVision, we do not need the #include "Lock.c" because Lock.c will be included in the project. I make the following general statement about good programming style.

While the main focus of this document is on C syntax, it would be improper to neglect all style issues. This system was divided using the following principles:

Breaking a software system into files has a lot of advantages. The first reason is code reuse. Consider the code in this example. If a Lock output function is needed in another application, then it would be a simple matter to reuse the lock.h and lock.c files. The next advantage is clarity. Because the details have been removed, the overall approach is easier to understand. The next reason to break software into files is parallel development. As the software system grows it will be easier to divide up a software project into subtasks, and to recombine the modules into a complete system if the subtasks have separate files. The last reason is upgrades. Consider an upgrade in our simple example where the Port A is replaced with Port B. For this kind of upgrade we implement the Port B functions in the Lock.c file with the new version. If we plan appropriately, we should be able to make this upgrade without changes to the files lock.h and main.c.

Go to Chapter 2 on Tokens Return to Table of Contents