[Chaper Nine][Previous] [Next] [Art of Assembly][Randall Hyde]

Art of Assembly: Chapter Nine


9.3.4 - Extended Precision Multiplication
9.3.5 - Extended Precision Division

9.3.4 Extended Precision Multiplication


Although a 16x16 or 32x32 multiply is usually sufficient, there are times when you may want to multiply larger values together. You will use the 80x86 single operand mul and imul instructions for extended precision multiplication.

Not surprisingly (in view of how adc and sbb work), you use the same techniques to perform extended precision multiplication on the 80x86 that you employ when manually multiplying two values.

Consider a simplified form of the way you perform multi-digit multiplication by hand:




1) Multiply the first two               2) Multiply 5*2: 
   digits together (5*3):

        123                                     123
         45                                      45
        ---                                      ---
         15                                      15
                                                10

 3) Multiply 5*1:                       4) 4*3:

        123                                     123
         45                                      45
        ---                                      ---
         15                                      15
        10                                      10
        5                                       5
                                                12

 5) Multiply 4*2:                       6) 4*1:

         123                                     123
          45                                      45
         ---                                     ---
          15                                      15
         10                                      10
         5                                       5
         12                                      12
         8                                       8
                                                4

 7) Add all the partial products together:

         123
          45
         ---
          15
         10
         5
         12
         8
        4
        ------
        5535

The 80x86 does extended precision multiplication in the same manner except that it works with bytes, words, and double words rather than digits. The figure below shows how this works.

Probably the most important thing to remember when performing an extended precision multiplication is that you must also perform a multiple precision addition at the same time. Adding up all the partial products requires several additions that will produce the result. The following listing demonstrates the proper way to multiply two 32 bit values on a 16 bit processor:

Note: Multiplier and Multiplicand are 32 bit variables declared in the data segment via the dword directive. Product is a 64 bit variable declared in the data segment via the qword directive.





Multiply        proc    near 
                push    ax
                push    dx
                push    cx
                push    bx

; Multiply the L.O. word of Multiplier times Multiplicand:

                mov     ax, word ptr Multiplier
                mov     bx, ax                          ;Save Multiplier val
                mul     word ptr Multiplicand           ;Multiply L.O. words
                mov     word ptr Product, ax            ;Save partial product
                mov     cx, dx                          ;Save H.O. word

                mov     ax, bx                          ;Get Multiplier in BX
                mul     word ptr Multiplicand+2         ;Multiply L.O. * H.O.
                add     ax, cx                          ;Add partial product
                adc     dx, 0                           ;Don't forget carry!
                mov     bx, ax                          ;Save partial product
                mov     cx, dx                          ; for now.

; Multiply the H.O. word of Multiplier times Multiplicand:

                mov     ax, word ptr Multiplier+2       ;Get H.O. Multiplier
                mul     word ptr Multiplicand           ;Times L.O. word
                add     ax, bx                          ;Add partial product
                mov     word ptr product+2, ax          ;Save partial product
                adc     cx, dx                          ;Add in carry/H.O.!

                mov     ax, word ptr Multiplier+2       ;Multiply the H.O.
                mul     word ptr Multiplicand+2         ; words together.
                add     ax, cx                          ;Add partial product
                adc     dx, 0                           ;Don't forget carry!
                mov     word ptr Product+4, ax          ;Save partial product
                mov     word ptr Product+6, dx

                pop     bx
                pop     cx
                pop     dx
                pop     ax
                ret
Multiply        endp

One thing you must keep in mind concerning this code, it only works for unsigned operands. Multiplication of signed operands appears in the exercises.


9.3.5 Extended Precision Division


You cannot synthesize a general n-bit/m-bit division operation using the div and idiv instructions. Such an operation must be performed using a sequence of shift and subtract instructions. Such an operation is extremely messy. A less general operation, dividing an n bit quantity by a 32 bit (on the 80386 or later) or 16 bit quantity is easily synthesized using the div instruction. The following code demonstrates how to divide a 64 bit quantity by a 16 bit divisor, producing a 64 bit quotient and a 16 bit remainder:




dseg            segment para public 'DATA'
dividend        dword   0FFFFFFFFh, 12345678h
divisor         word    16
Quotient        dword   0,0
Modulo          word    0
dseg            ends

cseg            segment para public 'CODE'
                assume  cs:cseg, ds:dseg

; Divide a 64 bit quantity by a 16 bit quantity:

Divide64        proc    near

                mov     ax, word ptr dividend+6
                sub     dx, dx
                div     divisor
                mov     word ptr Quotient+6, ax
                mov     ax, word ptr dividend+4
                div     divisor
                mov     word ptr Quotient+4, ax
                mov     ax, word ptr dividend+2
                div     divisor
                mov     word ptr Quotient+2, ax
                mov     ax, word ptr dividend
                div     divisor
                mov     word ptr Quotient, ax
                mov     Modulo, dx
                ret
Divide64        endp
cseg            ends

This code can be extended to any number of bits by simply adding additional mov / div / mov instructions at the beginning of the sequence. Of course, on the 80386 and later processors you can divide by a 32 bit value by using edx and eax in the above sequence (with a few other appropriate adjustments).

If you need to use a divisor larger than 16 bits (32 bits on an 80386 or later), you're going to have to implement the division using a shift and subtract strategy. Unfortunately, such algorithms are very slow. In this section we'll develop two division algorithms that operate on an arbitrary number of bits. The first is slow but easier to understand, the second is quite a bit faster (in general).

As for multiplication, the best way to understand how the computer performs division is to study how you were taught to perform long division by hand. Consider the operation 3456/12 and the steps you would take to manually perform this operation:

This algorithm is actually easier in binary since at each step you do not have to guess how many times 12 goes into the remainder nor do you have to multiply 12 by your guess to obtain the amount to subtract. At each step in the binary algorithm the divisor goes into the remainder exactly zero or one times. As an example, consider the division of 27 (11011) by three (11):























There is a novel way to implement this binary division algorithm that computes the quotient and the remainder at the same time. The algorithm is the following:





Quotient := Dividend;
Remainder := 0;
for i:= 1 to NumberBits do

	Remainder:Quotient := Remainder:Quotient SHL 1;
	if Remainder >= Divisor then

		Remainder := Remainder - Divisor;
		Quotient := Quotient + 1;

	endif
endfor

NumberBits is the number of bits in the Remainder, Quotient, Divisor, and Dividend variables. Note that the Quotient := Quotient + 1 statement sets the L.O. bit of Quotient to one since this algorithm previously shifts Quotient one bit to the left. The 80x86 code to implement this algorithm is





; Assume Dividend (and Quotient) is DX:AX, Divisor is in CX:BX,
; and Remainder is in SI:DI.

                mov     bp, 32          ;Count off 32 bits in BP
                sub     si, si          ;Set remainder to zero
                sub     di, di
BitLoop:        shl     ax, 1           ;See the section on shifts
                rcl     dx, 1           ; that describes how this
                rcl     di, 1           ; 64 bit SHL operation works
                rcl     si, 1
                cmp     si, cx          ;Compare H.O. words of Rem,
                ja      GoesInto        ; Divisor.
                jb      TryNext
                cmp     di, bx          ;Compare L.O. words.
                jb      TryNext

GoesInto:       sub     di, bx          ;Remainder := Remainder -
                sbb     si, cx          ;                Divisor
                inc     ax              ;Set L.O. bit of AX
TryNext:        dec     bp              ;Repeat 32 times.
                jne     BitLoop

This code looks short and simple, but there are a few problems with it. First, it does not check for division by zero (it will produce the value 0FFFFFFFFh if you attempt to divide by zero), it only handles unsigned values, and it is very slow. Handling division by zero is very simple, just check the divisor against zero prior to running this code and return an appropriate error code if the divisor is zero. Dealing with signed values is equally simple, you'll see how to do that in a little bit. The performance of this algorithm, however, leaves a lot to be desired. Assuming one pass through the loop takes about 30 clock cycles[2], this algorithm would require almost 1,000 clock cycles to complete! That's an order of magnitude worse than the DIV/IDIV instructions on the 80x86 that are among the slowest instructions on the 80x86.

There is a technique you can use to boost the performance of this division by a fair amount: check to see if the divisor variable really uses 32 bits. Often, even though the divisor is a 32 bit variable, the value itself fits just fine into 16 bits (i.e., the H.O. word of Divisor is zero). In this special case, that occurs frequently, you can use the DIV instruction which is much faster.


[2] This will vary depending upon your choice of processor.
9.3.4 - Extended Precision Multiplication
9.3.5 - Extended Precision Division


Art of Assembly: Chapter Nine - 27 SEP 1996

[Chapter Nine][Previous] [Next] [Art of Assembly][Randall Hyde]



Number of Web Site Hits since Jan 1, 2000: