For a project to learn C, I decided for fun to put the code in Compiler Explorer and compare the assembly output with the C code. Here’s a minimal example of some code

unsigned char count[256][256];

int main()
{
    for (int i=0; i<256; ++i)
        for (int j=0; j<256; ++j)
            count[i][j] = 0;
}

-Os turned this into rep stosd which make sense.

        mov     edx, OFFSET FLAT:count
        xor     eax, eax
        mov     ecx, 16384
        mov     rdi, rdx
        rep stosd

But -O2 turned it into a memset call?

        mov     edx, 65536
        xor     esi, esi
        mov     edi, OFFSET FLAT:count
        call    memset

Isn’t calling memset slower than just inlining the assembly instructions? Or does it depend on memset which is optimized differently for different architectures?