Справочник Пользователя для AMD 250

25112

Rev. 3.06

September 2005

The following section contains a collection of code snippets and subroutines showing the efficient 
implementation of 64-bit arithmetic in 32-bit mode. Note that these are 32-bit recommendations, in 
64-bit mode it is important to use 64-bit integer instructions for best performance.

Addition, subtraction, negation, and shifting are best handled by inline code. Multiplication, division, 
and the computation of remainders are less common operations and are usually implemented as 
subroutines. If these subroutines are used often, the programmer should consider inlining them. 
Except for division and remainder calculations, the following code works for both signed and 
unsigned integers. The division and remainder code shown works for unsigned integers, but can easily 
be extended to handle signed integers.

; Add ECX:EBX to EDX:EAX, and place sum in EDX:EAX.

add eax, ebx

adc edx, ecx

; Subtract ECX:EBX from EDX:EAX and place difference in EDX:EAX.

sub eax, ebx

sbb edx, ecx

; Negate EDX:EAX.

not edx

neg eax

sbb edx, -1   ; Fix: Increment high word if low word was 0.

; Shift EDX:EAX left, shift count in ECX (count

;  applied modulo 64).

   shld edx, eax, cl   ; First apply shift count.

   shl  eax, cl        ; mod 32 to EDX:EAX

   test ecx, 32        ; Need to shift by another 32?

   jz   lshift_done    ; No, done.

   mov  edx, eax       ; Left shift EDX:EAX

   xor  eax, eax       ;  by 32 bits

lshift_done: