newbie :adding larger value

newbie :adding larger value

[Vadiraj C S]

  Hello Everyone..

   I'm a newbie. Has any one worked on adding a large value in assembly. I mean
 a value which is larger than 32 bit. How is this mathematics taken care...





-- 
Regards
Vadiraj C S

Re: newbie :adding larger value

[Frederic Marmond]

hum, as a float ar as an integer?

i don't have much time to take care of all cases, here is the basic:

64bits integer not signed:
you must add the lower 32 bits, take care of the carry, add this to the 
upper 32 bits and add then the upper 32 bits. (but you may loose the 
upper carry...)
something like this:

;add 64 bits eax(upper 32bits)ebx(lower 32bits)
;to 64 bits ecx(upper) edx(lower)
add edx,ebx      ;add the lower 32 bits
jnc no_lowcarry      : if no carry, just add the upper 32bits
    inc ecx            ;if carry, increments the upper 32bits
no_lowcarry:
add ecx,eax   ;add the upper 32bits

jnc no_upcarry
    ;error: this would result to a 65 bits number (too big to stands in 
a 64 bits number)
    ;handle it here
no_upcarry:

Fred


Vadiraj C S wrote:

>  Hello Everyone..
>
>   I'm a newbie. Has any one worked on adding a large value in assembly. I mean
> a value which is larger than 32 bit. How is this mathematics taken care...
>
>
>
>
>
>  
>

Re: newbie :adding larger value

[Philip Jacob Smith]

> jnc no_lowcarry      : if no carry, just add the upper 32bits
>     inc ecx            ;if carry, increments the upper 32bits
> no_lowcarry:

Oh my god...

Ok, this is real easy now.  Say we have two 64-bit numbers, as such:

big_one dd 1, 2
big_two dd 3, 4

In normal intel backwardness (the right way, btw) 1 and 3 are the least signifigant dwords.

It's as easy as this:

add [big_one+0], [big_two+0]
adc [big_one+4], [big_two+4]

or, if you prefer code that compiles:

mov eax, [big_two]
add [big_one], eax
mov eax, [big_two+4]
adc [big_one+4], eax

jc overflow_handler

You just go from least signifigant to most signifigant adding them together, and use 'add' for the first and 'adc' for each one after that, and when you're done, the carry flag is the same as if you had added just one number, the 'jc overflow_handler' will jump only if the answer in this case would not fit into 64 bits.  To subtract you do exactly the same thing, just using sub and sbc.

Re: newbie :adding larger value

[Frederic Marmond]

! yes !
Hum, I feel like I have to play more often with x86 asm, I'm forgetting 
all but the basic... (my work makes me let the asm in the side and focus 
on C/C++ progs, ...)
Of course you are right!
Thanks to refresh my (so volatile?) memory! ;)

Fred

Philip Jacob Smith wrote:

>> jnc no_lowcarry      : if no carry, just add the upper 32bits
>>     inc ecx            ;if carry, increments the upper 32bits
>> no_lowcarry:
>
>
> Oh my god...
>
> Ok, this is real easy now.  Say we have two 64-bit numbers, as such:
>
> big_one dd 1, 2
> big_two dd 3, 4
>
> In normal intel backwardness (the right way, btw) 1 and 3 are the 
> least signifigant dwords.
>
> It's as easy as this:
>
> add [big_one+0], [big_two+0]
> adc [big_one+4], [big_two+4]
>
> or, if you prefer code that compiles:
>
> mov eax, [big_two]
> add [big_one], eax
> mov eax, [big_two+4]
> adc [big_one+4], eax
>
> jc overflow_handler
>
> You just go from least signifigant to most signifigant adding them 
> together, and use 'add' for the first and 'adc' for each one after 
> that, and when you're done, the carry flag is the same as if you had 
> added just one number, the 'jc overflow_handler' will jump only if the 
> answer in this case would not fit into 64 bits.  To subtract you do 
> exactly the same thing, just using sub and sbc.
> -
> To unsubscribe from this list: send the line "unsubscribe 
> linux-assembly" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>

Re: newbie :adding larger value

[peter willy krause]

On Thursday 27 November 2003 07:17, Vadiraj C S wrote:
>   Hello Everyone..
>
>    I'm a newbie. Has any one worked on adding a large value in
> assembly. I mean a value which is larger than 32 bit. How is this
> mathematics taken care...

example from 'lib4th', adding two counted (by l.s. cell value) unlimited 
size integers (re http://www.lxhp.in-berlin.de/lhpf8.html ).

hp

--------------
counted integer representation:
	l.s.cell [ count ][ cell-1 ] ... [ celll-n ]
	  EDI--^  EAX [ .. memory             ]
	  tos --^    n     (n) cells           ^-- m.s. cell

'cell' is a 32-bit item in memory.
2-s complement representation for -ve values.

memory arrangement as w. the (lib4th) Forth data stack, pointed to by reg 
EDI, 'push' towards _higher_ addresses; 'top' item also in EAX.
'tos' for 'top of stack', strikt 'little endian' byte order storage.

addition: (nasm syntax):

nadd:
	call nnorm        ; assure equal cells' count
                               ; by signed extending the 'shorter' item
        mov ecx,eax     ; counter, n cells (count of 32bit items)
        neg eax            ; disp to 2nd oprd
        push ecx           ; save count
        push esi            ;      & reg
        lea esi,[edi+eax*4] ; m.s. cell 1st oprnd
        lea edi,[esi+eax*4-4] :    ... 2nd
        push esi            ; save ptr to result 'tos'
        or eax,eax         ; clr initial carry
.l:
        mov eax,[edi]	; 2nd
        adc eax,[esi]	; 1st
        mov [edi],eax	; sum
        lea edi,[edi+4]	; 'lea' and 'dec' 
        lea esi,[esi+4]     ;     preserve the carry flag
        dec ecx		; count cells size of int's
        jg .l
        pop edi            ; new 'tos' ptr
        pop esi            ; saved reg
        pop eax           ; count item of sum integer
        lea edi,[edi+4]   ; drop count item of 1st operand
        ret

above code requires items of equal size. the original 'nnorm' routine  is 
too complicated for an introductory example - it runs some other tests 
and modifies the int's for the shortest pair without loss of praecision.

-----------------
mail to 'hp': - lx  at lxhp : in-berlin : de -
-----------------

Re: newbie :adding larger value

[Vadiraj C S]

> --------------
> counted integer representation:
> 	l.s.cell [ count ][ cell-1 ] ... [ celll-n ]
> 	  EDI--^  EAX [ .. memory             ]
> 	  tos --^    n     (n) cells           ^-- m.s. cell
> 
> 'cell' is a 32-bit item in memory.
> 2-s complement representation for -ve values.
> 


  Thanks a lot, got some idea, i guess you have algos for these type of mathematical calculations..called GMP..

-- 
Regards
Vadiraj C S

Re: newbie :adding larger value

[peter willy krause]

sorry for the confusion...

> example from 'lib4th', adding two counted (by l.s. cell value)
> unlimited size integers (re http://www.lxhp.in-berlin.de/lhpf8.html ).

data in memory (which is of no concern w. the mentioned example):
> counted integer representation:
> 	l.s.cell [ count ][ cell-1 ] ... [ celll-n ]
> 	  EDI--^  EAX [ .. memory             ]
> 	  tos --^    n     (n) cells           ^-- m.s. cell

data on 'stack', on which the example operates, are stored in the opposite 
order, 'count' in EAX & @tos, m.s.cell at 'tos'-4, l.s.cell @tos-(4*eax):

	[ count ][ cell-n ] ... [ cell-1 ]
  EDI--^  EAX    [ .. memory             ]
  tos --^    n   ^-- m.s.cell (n) cells  ^-- l.s. cell

-- 
mail to 'hp': lx at lxhp : in-berlin : de

Re: newbie :adding larger value

[Randall Hyde]

Take a look at the chapter on Advanced Arithmetic in
"The Art of Assembly Language" at http://webster.cs.ucr.edu
Cheers,
Randy Hyde

----- Original Message ----- 
From: "Vadiraj C S" <vadiraj@mail.cyberneme.com>
To: <linux-assembly@vger.kernel.org>
Sent: Wednesday, November 26, 2003 11:17 PM
Subject: newbie :adding larger value


> 
> 
>   Hello Everyone..
> 
>    I'm a newbie. Has any one worked on adding a large value in assembly. I mean
>  a value which is larger than 32 bit. How is this mathematics taken care...
> 
> 
> 
> 
> 
> -- 
> Regards
> Vadiraj C S
> 
> 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-assembly" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html