Page Clearing Prefetch

[JohnnyTheDon]

I have a assembly function that clears pages (using SSE, but that isn't the point):

Code: Select all

clear_pages:
	shl rsi, 8 ; Multiply number of pages to clear by size of a page (4096) and then divide number of pages to clear by the number of bytes cleared in one iteration (16)
	xorps xmm0, xmm0 ; Clear xmm0
.loop:
	movntdq [rdi], xmm0
	add rdi, 16
	dec rsi
	cmp rsi, 0
	jne .loop
	ret

Would it improve performance to use prefetching (using the prefetchnta instruction) even though I'm writing, not reading?

EDIT: Fixed error in code