Aligned vs unaligned

Andy wrote:

Why? Isn't this the way other instructions work?

The instructions work like this, but the addresses continue to number bytes. To do the same with the pointer, you will have to scale the addresses to 32-byte values at each step of the loop.

Andy wrote:

And would be better like saving the offset in a different register and use it something like that?

Then it will be more difficult to determine the alignment of addresses. The value of the sum of the two registers must be aligned.

Code:

esi rdx
 0   0 = 0
 1  31 = 32, head loop 31 times
 2  30 = 32, ...
 3  29 = 32
  ...

Code:

    test esi, 00011111b ; 31
    jz ok
hl:
    ...
    inc esi
    test esi, 00011111b; 31, address
    loopnz hl ; counter
ok:

Code:

    xor edx, edx
    lea eax, [esi + edx]
    test al, 00011111b
    jz ok
hl:
    ...
    inc edx
    lea eax, [esi + edx]
    test al, 00011111b; address
    loopnz hl; counter
ok:

But it may become easier if you go to the end address instead of the byte counter. Then you will get rid of the need for double inc/dec.

Code:

add ecx, esi

Maybe.