Sierpinski fractal in 28 bytes

Some ages ago there was a competition on #nlcoders doing the Sierpiński triangle in as less bytes as possible.

Of course, I was beaten very bad. Shorter versions exists (22 bytes, 19 bytes, even 10 bytes!). Some using rule 90.

; 28 byte Sierpinski-fractal  Joris Zwart (C) 1998

            .model tiny
            .code
            org 100h

start:
            mov al,13h
x:
            cwd
y:
            test cl,dl
            jne skip
            int 10h

            mov ah,0ch
skip:
            inc dl
            jns y
            inc cl
            jns x

            cbw
            int 16h

            mov ax,3
            int 10h
            ret

            end start

This assembles down to these 28 bytes:

b0 13 99 84 ca 75 04
cd 10 b4 0c fe c2 79
f4 fe c1 79 ef 98 cd
16 b8 03 00 cd 10 c3

Which maps to this list file:

0000: b0 13       mov al,0x13
0002: 99          cwd
0003: 84 ca       test dl,cl
0005: 75 04       jne 0x000b
0007: cd 10       int 0x10
0009: b4 0c       mov ah,0xc
000b: fe c2       inc dl
000d: 79 f4       jns 0x0003
000f: fe c1       inc cl
0011: 79 ef       jns 0x0002
0013: 98          cbw
0014: cd 16       int 0x16
0016: b8 03 00    mov ax,3
0019: cd 10       int 0x10
001b: c3          ret

Pseudocode

For the assembly-impaired here is the pseudocode:

FOR X = 0 TO 127
    FOR Y = 0 TO 127
        IF X AND Y THEN PLOT X, Y
    NEXT Y
NEXT X