In C++11, you can initialize all members of a C-style array to zero.
For example, you can
int a[10] = {0}; // all elements 0
// or
int a[10] = {}; // all elements 0
But how the compilers implement it?
Assume we have this code:
void f() {
int a[10] = {};
}
With clang 11.0.1, we get this asm code:
f(): # @f()
push rbp
mov rbp, rsp
sub rsp, 48
xor esi, esi
lea rax, [rbp - 48]
mov rdi, rax
mov edx, 40
call memset
add rsp, 48
pop rbp
ret
As you see, clang use memset to initialize the array.
With gcc 10.3, we get this asm code:
f():
push rbp
mov rbp, rsp
mov QWORD PTR [rbp-48], 0
mov QWORD PTR [rbp-40], 0
mov QWORD PTR [rbp-32], 0
mov QWORD PTR [rbp-24], 0
mov QWORD PTR [rbp-16], 0
nop
pop rbp
ret
It looks like gcc expanded the function call. To get what gcc really do, we use int a[128], then get
f():
push rbp
mov rbp, rsp
sub rsp, 392
lea rdx, [rbp-512]
mov eax, 0
mov ecx, 64
mov rdi, rdx
rep stosq
nop
leave
ret
What rep stosq means?
From Intel’s documentation, the stosq instruction is described as follows:
Store RAX at address RDI or EDI.
And rep
Repeats a string instruction the number of times specified in the count register or until the indicated condition of the ZF flag is no longer met. The REP (repeat), REPE (repeat while equal), REPNE (repeat while not equal), REPZ (repeat while zero), and REPNZ (repeat while not zero) mnemonics are prefixes that can be added to one of the string instructions.
…
Repeat Prefix Termination Condition REP RCX or (E)CX = 0
In the above example, the CPU will fill 8 bytes of 0 each time and repeats 64 times.
An interesting point is when the array is too large, gcc would call memset rather than stosq instruction.