AMD 250 manual 175

25112 Rev. 3.06 September 2005 Software Optimization Guide for AMD64 Processors

;

; Destroys: EAX, ECX, EDX, EFlags

_ullrem	PROC
push	ebx		; Save EBX as per calling convention.
mov	ecx, [esp+20]		; divisor_hi
mov	ebx, [esp+16]		; divisor_lo
mov	edx, [esp+12]		; dividend_hi
mov	eax, [esp+8]		; dividend_lo
test	ecx, ecx		; divisor > 2^32 - 1?
jnz	r_big_divisor		; Yes, divisor > 32^32 - 1.
cmp	edx, ebx		; Only one division needed (ECX = 0)?
jae	r_two_divs		; Need two divisions.
div	ebx		; EAX	= quotient_lo
mov	eax, edx		; EAX	= remainder_lo
mov	edx, ecx		; EDX	= remainder_hi = 0
pop	ebx		; Restore EBX per calling convention.
ret			; Done, return to caller.
r_two_divs:
mov ecx, eax		; Save dividend_lo in ECX.
mov eax, edx		; Get dividend_hi.
xor edx, edx		; Zero-extend it into EDX:EAX.
div ebx		; EAX = quotient_hi, EDX = intermediate remainder
mov eax, ecx		; EAX = dividend_lo
div ebx		; EAX = quotient_lo
mov eax, edx		; EAX = remainder_lo
xor edx, edx		; EDX = remainder_hi = 0
pop ebx		; Restore EBX as per calling convention.
ret		; Done, return to caller.
r_big_divisor:
push	edi		;	Save EDI as per calling convention.
mov	edi, ecx		;	Save divisor_hi.
shr	edx, 1		;	Shift both divisor and dividend right
rcr	eax, 1		;	by 1 bit.
ror	edi, 1
rcr	ebx, 1
bsr	ecx, ecx		;	ECX = number of remaining shifts
shrd	ebx, edi, cl		;	Scale down divisor and dividend such
shrd	eax, edx, cl		;	that divisor is less than 2^32
shr	edx, cl		;	(that is, it fits in EBX).
rol	edi, 1		;	Restore original divisor (EDI:ESI).
div	ebx		;	Compute quotient.
mov	ebx, [esp+12]		;	dividend low word
mov	ecx, eax		;	Save quotient.
imul	edi, eax		;	quotient * divisor high word (low only)
mul	DWORD PTR [esp+20] ;			quotient * divisor low word
add	edx, edi		;	EDX:EAX = quotient * divisor
sub	ebx, eax		;	dividend_lo – (quot.*divisor)_lo
mov	ecx, [esp+16]		;	dividend_hi

Chapter 8 Integer Optimizations 175