ASM Community

.386
.model flat,c
.stack 100h

includelib msvcrt 

.data
printf proto arg1:ptr byte

pointBuffer byte "Welcome",0
pointBuffer2 byte 12 dup(?),0
.code
public main

main proc
   
   mov ecx,8
   mov esi,offset pointBuffer
   mov edi,offset pointBuffer2
   
while1: cmp ecx,0
        je endloop
	mov al,
	mov ,al
	
	inc edi
	inc esi
	dec ecx
	cmp ecx,4
	jge takepath1
        jmp while1
endloop: nop
takepath1:  invoke printf,addr pointBuffer2
	jmp while1
	ret
main endp

end main

im trying to do something so simple and i dont see why i cant have a invoke printf in a loop while having an if statment to execute it and jump back into loop

Posted on 2012-06-15 20:57:04 by hellfix

Right. And, in contrast, esi, edi, and ebx are "non-volatile" (esp and ebp too, but we're not concerned with them right now). They need to be "preserved". Hutch didn't like the term "preserved". Let's say: if you alter 'em, you have to put 'em back the way they were, before your function returns. Failure to do so may cause a crash! Might not, but we're "supposed" to do it. (the "proc" and "endproc" macros may do this for you?) Since you're using esi and edi, your choice for a loop counter is pretty obvious. If you have an "abnormal termination", "preserve" 'em!

An alternative would be to push and pop ecx around the call to printf.

Is ".stack 100h" correct in this context?

Best,
Frank

Thanks, Scali! Having any luck, Hellfix?

Best,
Frank

Easiest way, probably...

endloop: nop
takepath1:
        push ecx
        invoke printf,addr pointBuffer2
        pop ecx
	jmp while1
; need a label here to jump to when ecx=0?
	ret  ; never reached?

That should clear up the problem that the call to printf is trashing your ecx. You still don't seem to have an exit condition to your loop.

Does calling your entrypoint "main" imply that this is called by C startup code? If not, "ret" may not provide a clean "ExitProcess". If so, you may still need to "preserve" (push/pop) esi and edi. This appears to be 32-bit code. Why is it in the dos/bios section? I may be confused about what you're trying to do...

Best,
Frank

First things first... "code tags". Put the word "code" between '[]'s at the beginning of your code and "/code" between '[]'s at the end of your code. The management thanks you. :)

.386
.model flat,c
.stack 100h

includelib msvcrt

.data
printf proto arg1:ptr byte

pointBuffer byte "Welcome",0
pointBuffer2 byte 12 dup(?),0

.code
public main

main proc

I think you'll want that "uses ebx, esi, edi" here. I'm not sure what the syntax is - right after the "proc" or on a separate line? Try it until Masm doesn't complain. :) (or RTFM)

   mov ebx,0
   mov esi,offset pointBuffer
   mov edi,offset pointBuffer2
   
while1: cmp ebx,8
        jge endloop

Strictly speaking, "jge" is for signed comparisons, "jae" is for unsigned. Won't hurt.

   mov al,
   mov ,al
   
   inc edi
   inc esi
   inc ebx

   cmp ebx,4
   je printOnce

Okay... you've copied 3 characters - ebx incremented afterwards, remember. Might want 5 here?

        jmp while1

printOnce: push offset pointBuffer2
   call printf

Caller's responsibility to clean up stack!

   add esp, 4
   cmp ebx,4
   jbe while1

We kinda knew it was 4 (or 5) or we wouldn't be here! An unconditional "jmp" should do here.

endloop: nop   
   push offset pointBuffer2
   call printf
   add esp, 4 ; clean up stack!
   ret
main endp
end main

Outside of not "cleaning up the stack" (removing the parameter) after the call to printf, this doesn't look too bad. Another way to do this would be to "pop" a dummy register (ecx would do, since printf trashes it anyway). Also, I think you want to "preserve" ebx, esi, and edi. "Calling convention"! I think Agner Fog's got a document on it:

http://www.agner.org/optimize/

I'm not able to test this. There may be things I don't see, but I think it'll work. I think it'll print "WelWelcome" - might look better with a newline in there...

Best,
Frank