Memset and memcpy suck

[Dwedit]

I don't see any ldmia's or stmia's in the implementations of memset or memcpy. Care to provide a better one?

For reference, here is memcpy's inner loop when doing a word-aligned copy:

Code: Select all

 8000b58:	6826      	ldr	r6, [r4, #0]
 8000b5a:	601e      	str	r6, [r3, #0]
 8000b5c:	6866      	ldr	r6, [r4, #4]
 8000b5e:	605e      	str	r6, [r3, #4]
 8000b60:	68a6      	ldr	r6, [r4, #8]
 8000b62:	609e      	str	r6, [r3, #8]
 8000b64:	68e6      	ldr	r6, [r4, #12]
 8000b66:	3d10      	subs	r5, #16
 8000b68:	60de      	str	r6, [r3, #12]
 8000b6a:	3410      	adds	r4, #16
 8000b6c:	3310      	adds	r3, #16
 8000b6e:	2d0f      	cmp	r5, #15
 8000b70:	d8f2      	bhi.n	8000b58 <memcpy+0x30>