Is there a simple way to add 4 unsigned bytes at a time with 255 saturation? This is one way:

;
; add 4 bytes with 255 saturation
;
; input:
;
; output:
;
; destroys:
;

        add     al,bl
        sbb     cl,cl
        add     ah,bh

        sbb     ch,ch
        or      al,cl

        ror     ebx,16
        or      ah,ch
        ror     eax,16          ; 5 cycles

Repeat the code above to handle the two upper bytes. To be any faster, we must handle more than one byte simultaneously. If we have data in the first and third byte (00ff00ff) of EAX and EBX, then we can do something like this: add eax,ebx mov ebx,eax and ebx,0100100h ; Get the carry bits from both additions mov ecx,ebx shr ecx,8 ; Carry bit (0 or 1) in lowest position sub ebx,ecx ; EBX is 00ff00ff, 00000000, 000000ff or 00ff0000 or eax,ebx
This method uses 7 instructions to handle two bytes, so it will be faster to use the naive code above. This gem can not run on any 16 bit machine, as of the name 4 bytes at a time means a 32 bits and a 16 bit machine does not have that, you can do it but not the way shown here.

Gem writer: Terje Mathisen
last updated: 1998-03-16