.text
	.globl dotproduct
	.type dotproduct,@function

dotproduct:
	# rowloop
	mov	$8, %cl
1:
	rol	$8, %rdi	# get next row of A into low 8 bits

	# columnloop
	mov	$8, %ch
2:
	# extract first column
	movq	%rsi, %mm0
	pmovmskb %mm0, %edx

	# dot with current row, clears CF, product in PF
	and	%dil, %dl

	jp	3f		# set CF to inverse of PF
	stc
3:	adc	%rax, %rax	# shift CF into LSB of %rax

	# shift over columns of B
	rol	%rsi

	dec	%ch
	jnz	2b

	# restore B
	ror	$8, %rsi

	loop	1b		# %ch will be zero here
	ret

	.size dotproduct,.-dotproduct