Selection Sort Assembly x86 [Nasm]: Strange incrementation of value j

Question

I'm having problems with this selection sort, the problem is when vecnums[j] is minor to (<) vecnums[min] that I have to put j in min and then increment j by 1, after that when I load j in SI, j is no longer the value it must be (using debugger I figured out that goes to 0102, when it should be 0002). I don't know why happens that, if someone could help me I would really appreciate it :)

I apologize if I didn't express myself well, don't doubt in asking me whatever you don't understand.

Thanks for your time!

Vecnums is an array of numbers (2 bytes size) and I loaded it with: 5,-11,3,-4,10,1005,0,5,-1,23,-34,85,-30,-82,1

i       resb        1
j       resb        1
min     resb        1
        db      0
vecnums     times 60 db 0
nlog        resb        1  ;for the example is 15

ssort:      
        mov     byte[i],0
        mov     ch,0
        mov     cl,[nlog]
        sub     cl,1
cicloi:     
        mov     ah,0
        mov     al,[i]
        mov     [min],al ; min=i

        push        cx
        mov     cl,[i]
        add     cl,1
        mov     byte[j],cl
        mov     cl,[nlog]   
        sub     cl,[j]
cicloj:
        mov     si,[j]
        imul        si,2
        mov     ax,word[vecnums+si]
        mov     si,[min]
        imul        si,2
        mov     dx,word[vecnums+si]
        cmp     ax,dx   ;ax=vecnums[j] dx=vecnums[min]
        jnl     noMenor
        mov     ah,0    ;vecnums[j] < vecnums[min]
        mov     al,[j]
        mov     [min],al ; min=j
noMenor:                                     ; vecnums[j] > vecnums[min]
        inc     byte[j]
        loop        cicloj
        mov     si,[j]
        imul        si,2
        mov     ax,word[vecnums+si]
        mov     si,[min]
        imul        si,2
        mov     dx,word[vecnums+si]
        mov     si,[j]
        imul        si,2
        mov     word[vecnums+si],dx
        mov     si,[min]
        imul        si,2
        mov     word[vecnums+si],ax
        inc     byte[i]
        pop     cx
        loop        salto
        jmp     finrut
salto:      jmp     cicloi   ;the reason for this is that the jump is too long to do it with loop (couldn't assemble if I do it directly with loop)
finrut:
    ret

Answer 1

Your j variable is byte sized (same goes for i and min). However when you say mov si, [j] you are loading 2 bytes, hence the high byte will come from the following variable (min in case of j). You should fix your loads, for example by using movzx si, byte [j]

Side note: it is general practice to use shifts for multiplying by powers of two.