static double m = 30000;
double foo(double x, double y) {
return x/m + y;
}
This doesn't win you anything. A copy of m has to be made to do the computation.
Also if you do:
double bar( double x, double y) {
m += x + y;
return m;
}
Then all calls to bar will change m.
Static variables outside of functions (or classes) are really global variables with file scope. Other files can't get at them by extern
Static variables inside a function are still like global variables, except that even other functions in the same file can't see them directly.
const double m = 30000;
This is better and in many cases best. If the compiler sees this global const and then sees a reference to m then it knows that rather than generate code to load the value from where ever it is (which likely requires loading a literal address into a register first) to a register or stack position to do computations it can just make a register be 30000 or sometimes generate an instruction with 30000 encoded right in there.
The down side to this is that the compiler has to assume that other souce files will want to read m and has to actually store a copy as a variable (but a constant variable) in the object file.
I'm not sure if it is standard but you can sometimes do extern const double m = 30000;
and the compiler will use 30000 to optimize and assume that another file actually has a copy of m that will be stored in the executable. You can also do static const double m = 30000;
and the compiler can assume that no one else will expect that a copy of m is stored in the object code generated from this source file.
Doing
#define m 30000
is more risky. You will not get a warning or error if previously there was another m declared as a variable, constant, or function. Also, for preprocessor macros like this it is easy to mess up.
For example:
#define BASE_ADDRESS 48
#define MY_OFFSET 9
#define MY_ADDRESS BASE_ADDRESS+MY_OFFSET
...
return MY_ADDRESS*4;
Yes, this is a stupid example, but what this looks like after the preprocessor gets done with it is
...
return 48+9*4;
Which is
return 48+(9*4);
And that's not what you probably wanted.
Another place where macros are bad is when you have large constants, such as strings. Strings require that they be addressable by pointer and are more difficult to optimize away than integers and floating point literal or constant numbers. You could easily make a very large program if you had lots of stuff like:
#define JIM "Jim"
#define JOHN "John"
and then used JIM and JOHN all over your programs because the compiler might not be able to see that you really only needed the strings "Jom" and "John" once in the program.
That being said, it is not uncommon to see constants being declared like that, and often they are properly done that way by people who know what they are doing.