local variable storage allocation in assembly

Question

the following snippet is often used when said to be allocate storage for local variables

addl $8,%esp  //allocate 8-byte storage

push %ebx     //store some value onto the stack

why not simply push the value onto the stack,but rather allocated some space in advance?

Answer 1

So that the allocated space is a fixed size, which is simplest for the compiler as the memory can be accessed with "DWORD PTR [ebp+(offset)]". The usual function prologue is something like:

(Intel syntax) push ebp add ebp, 8 mov ebp, esp

for a function with 8 bytes of local variables.

Answer 2

If you're asking why doesn't a compiler generate push instructions for local storage rather than direct manipulation of the stack pointer, it's a matter of efficiency.

Automatic variables (in C anyway) are uninitialized, so the code sequence would be (I'm going to use subl since I'm used to stacks growing down in memory):

C Code            Actual assembly        Your suggestion
void x(void) {
    int a;        subl 8,%esp            push 0
    int b;                               push 0
    : : :         blah blah              blah blah
}

My answer is that it's unnecessary and inefficient in cases such as:

C Code            Actual assembly        Your suggestion
void x(void) {
    int a[100];   subl 400,%esp          push 0
                                         push 0
                                         push 0
                                         : : :
                                         push 0
    : : :         blah blah              blah blah

What you propose may make sense for something like:

C Code            Your suggestion
void x(void) {
    int a = 7;    push 7
    int b = 9;    push 9
    int c[4];     subl 16,%esp
    : : :         blah blah
}