Bubble sort in 16 bytes of x86 assembly language

bsort32.asm

;;
;; The "worlds smallest" bubble sort (16 bytes)
;;
;; Copyright (c) 1998 by Joergen Ibsen / Jibz
;; All Rights Reserved
;;

bits 32

section .text

;=====================================================================
;     enter:  edi -> array
;             ecx =  #elements - 1
;     exit:   edi -> array
;             ecx =  0
;     modify: ecx
;=====================================================================
bsort32:
.outerloop:
    pusha                         ; save edi and ecx
    mov     esi, edi
.innerloop:
    lodsd
    cmp     eax, [esi]            ; do we need to swap?
    jge     short .order_ok
    xchg    eax, [esi]            ; if so, this is first step
.order_ok:
    stosd                         ; second step, or just write back eax
    loop    .innerloop
    popa                          ; pop edi and ecx
    loop    .outerloop            ; ecx is used for both loops