Assembly Language for Beginners

3.12 C99 restrict.

$LN6@memcmp_123:
sbb eax, eax
or eax, 1
pop esi
ret 0
_memcmp_1235 ENDP

strcat()

This is inlined strcat() as it has been generated by MSVC 6.0. There are 3 parts visible: 1) getting source
string length (firstscasb); 2) getting destination string length (secondscasb); 3) copying source string
into the end of destination string (movsd/movsbpair).

Listing 3.52: strcat()

lea edi, [src]

or ecx, 0FFFFFFFFh

repne scasb

not ecx

sub edi, ecx

mov esi, edi

mov edi, [dst]

mov edx, ecx

or ecx, 0FFFFFFFFh

repne scasb

mov ecx, edx

dec edi

shr ecx, 2

rep movsd

mov ecx, edx

and ecx, 3

rep movsb

IDA script

There is also a smallIDAscript for searching and folding such very frequently seen pieces of inline code:

GitHub.

3.12 C99 restrict

Here is a reason why Fortran programs, in some cases, work faster than C/C++ ones.

void f1 (int x, int y, int sum, int product, int sum_product, int update_me, size_t s)
{
for (int i=0; i<s; i++)
{
sum[i]=x[i]+y[i];
product[i]=x[i]y[i];
update_me[i]=i123; // some dummy value
sum_product[i]=sum[i]+product[i];
};
};

That’s very simple example with one specific thing in it: the pointer to theupdate_mearray could be a
pointer to thesumarray,productarray, or even thesum_productarray—nothing forbids that, right?

The compiler is fully aware of this, so it generates code with four stages in the loop body:

• calculate nextsum[i]


• calculate nextproduct[i]


• calculate nextupdate_me[i]