Allocation, initialization and scope

Question

I realized I don't fully understand how Arduino code handles variable lifetime. My tendency is to keep variables in the narrowest scope (which I realize isn't as important in Arduino code, but, habits...). But how inefficient is this?

Given these possible variations:

 const int c1 = 2, c2 = 7, c3 = 9;

 void functy();

 int foo[3] = {c1 | c2, c3, c2 | c3};

 static int foo[3] = {c1 | c2, c3, c2 | c3};

 const int foo[3] = {c1 | c2, c3, c2 | c3};

  ....

In case 1, foo is both allocated and initialized each call.

In case 2, foo is allocated only once, but is it initialized on each call?

In case 3 I assume that allocation and initialization are done only once? Is that true?

Answer 1

Here's a pretty definitive list for you:

// Allocated once in the global area and assigned during 
// program startup
int foo = 1;

// May not even be allocated - the compiler will most likely
// optimise it out.
const int foo = 1;

// Allocated once and assigned during program startup. Cannot be
// accessed from outside the file.
static int foo = 1;

// The static here has no real effect.
static const int foo = 1;

void function() {
    // Allocated in the stack each call and initialised each call.
    // Only accessible from inside this function.
    int foo = 1;

    // Allocated in the global area and initialized on the first call
    // of the function. Only accessible from inside this function.
    static int foo = 1;

    // May not be allocated at all. Only accessible from inside
    // this function.
    const int foo = 1;
}

Note that non-simple const values, things like arrays and strings, depending on the architecture (i.e., if it's a Harvard architecture or not) may be copied into a block of global RAM for simpler access. See PROGMEM in the Arduino Reference Manual for more information.