code golf, bubble sort, 19B, x86-16 asm

bubblesort16:
;; inputs: pointer in ds:si,  size in in cx
;; clobbers: al, cx=0
    dec     cx          ; cx = max valid index.  (Inner loop stops 1 before cx, because it loads i and i+1).
.outer:                 ; do{
    push    cx
.inner:                 ;   do{
    lodsb               ;     al = arr[i++]
    cmp     [si],al     ;     compare with arr[i (new one)]
    jge     .noswap
    rol     word [si-1], 8    ; swap
.noswap:
    loop    .inner      ; } while(i < size);
    pop     cx
    sub     si,cx
    loop    .outer      ; while(--size);
    ret