6
\$\begingroup\$

I'm trying to work up to being able to print the hailstone sequence in assembly. To do that though, I first needed to learn how to actually print a number out using the write system call.

After a few design iterations, I ended up using division to get each digit one at a time by using the remainder. Each remainder gets pushed onto the stack, then the address of the stack is given to the system call to print out.

What I'd like advice on:

  • In a couple places, I have math split out over multiple lines, like:

    sub edx, esp  ; Calculate how many were pushed
    dec edx
    

    It doesn't seem like it's possible to combine that into something like sub edx, esp - 0, but if there is a neater way, I'd like to know.

  • This is my first time using div. Is there anything wrong with how I'm using it? Also, is there a sane way of not using div altogether? Apparently it's stupid slow and should be avoided if possible.

  • Anything else notable. I'm a super-beginner, and this code is quite verbose.

I'm using n to represent the number that I want to print out. In theory though, that number could come from anywhere. It's just a placeholder for the exercise.

global _start

section .data
    n: dd 123456

section .text
    _start:
        mov ebp, esp  ; So we can tell how many were pushed

        mov ecx, [n]

        .loop:
            mov edx, 0  ; Zeroing out edx for div
            mov eax, ecx  ; Num to be divided
            mov ebx, 10  ; Divide by 10
            div ebx
            mov ecx, eax  ; Quotient

            add edx, '0'
            push edx  ; Remainder

            cmp ecx, 0
            jne .loop

        mov eax, 4  ; Write system call
        mov ebx, 1  ; STDOUT
        mov ecx, esp  ; The string on the stack
        mov edx, ebp
        sub edx, esp  ; Calculate how many were pushed
        dec edx
        int 0x80

        mov eax, 1
        mov ebx, 0
        int 0x80

Assembled and linked using:

nasm numprint2.asm -g -f elf32 -Wall -o numprint2.o  
ld numprint2.o -m elf_i386 -o numprint2
\$\endgroup\$
3
  • \$\begingroup\$ Just curious, how come you went with 32 bit over 64 bit? \$\endgroup\$ Oct 5, 2020 at 0:43
  • 1
    \$\begingroup\$ @JoseFernandoLopezFernandez Our school decided to go with 32-bit. I believe it had to do with tools we used later (like immunity) being easier to use. We did all the debugging later on a 32-bit VM. \$\endgroup\$ Oct 5, 2020 at 0:48
  • \$\begingroup\$ Consider implementing the division by reciprocal multiplication if you want to increase speed and/or reduce energy requirement. \$\endgroup\$ May 18 at 12:25

1 Answer 1

2
\$\begingroup\$

print a number out using the write system call.
... using division to get each digit one at a time by using the remainder. Each remainder gets pushed onto the stack, then the address of the stack is given to the system call to print out.

Next is what the stack looks like in your program where you convert the number 123456. Every push will have written a dword on the stack:

'1',0,0,0,'2',0,0,0,'3',0,0,0,'4',0,0,0,'5',0,0,0,'6',0,0,0
^                                                           ^  
ESP                                                         EBP

Therefore, the code:

mov edx, ebp
sub edx, esp  ; Calculate how many were pushed
dec edx

will make EDX equal to 23. However, the write system call expects from you the number of characters (bytes) in the string which should be 6.

... using div. Is there anything wrong with how I'm using it? Also, is there a sane way of not using div altogether? Apparently it's stupid slow and should be avoided if possible.

You are using it correctly. On modern hardware division is no longer the extremely slow operation that it used to be (same goes for multiplication). I would not seek to avoid it for this humble number to text conversion.

.loop:
  mov  edx, 0    ; Zeroing out edx for div
  mov  eax, ecx  ; Num to be divided
  mov  ebx, 10   ; Divide by 10
  div  ebx
  mov  ecx, eax  ; Quotient
  add  edx, '0'
  push edx       ; Remainder
  cmp  ecx, 0
  jne  .loop

In the division loop, you can improve a lot:

  • zero a register using the xor reg, reg instruction
  • hoist initializing the loop-invariant const 10 out of the loop
  • don't copy back and forth between EAX and ECX
  • check a register for being zero through test reg, reg

Suggestion for a re-write

  mov  ecx, esp
  sub  esp, 16        ; Room for at most 10 digits (keep ESP dword-aligned)
  mov  eax, [n]
  mov  ebx, 10        ; CONST

.loop:
  xor  edx, edx
  div  ebx            ; Divide EDX:EAX by 10
  dec  ecx
  add  edx, '0'       ; From remainder [0,9] to character ['0','9']
  mov  [ecx], dl
  test eax, eax       ; Is quotient zero?
  jnz  .loop

  mov  eax, 4         ; Write system call
  mov  ebx, 1         ; STDOUT
  lea  edx, [esp+16]
  sub  edx, ecx
  int  0x80
  add  esp, 16        ; Clean-up

This is how the above version of the code uses the stack:

0,0,0,0,0,0,0,0,0,0,'1','2','3','4','5','6'
^                   ^                       ^  
ESP                 ECX (after)             ECX (before)
                    <------- EDX = 6 ------>

Anything else notable.

Everyone is allowed their own programming style of course, but having the mnemonics, operands, and especially the tail comments aligned (like in my re-write) will certainly add to the readability of the code...

\$\endgroup\$
1
  • \$\begingroup\$ Even though modern hardware is better at division than older processors, decimal printing is a very common operation, so it might be worth reducing power draw by replacing it with a multiplication. That's arguably no longer a beginner implementation, though! \$\endgroup\$ May 18 at 12:23

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.