Assembly output of the three functions from https://lemire.me/blog/2021/01/06/memory-access-on-the-apple-m1-processor/#comments

two_or_three.s

	.globl	_compute_two            ; -- Begin function compute_two
	.p2align	2
_compute_two:                           ; @compute_two
	.cfi_startproc
; %bb.0:
Lloh0:
	adrp	x8, _M@PAGE
Lloh1:
	ldr	x8, [x8, _M@PAGEOFF]
	tst	x8, #0x7fffffffffffffff
	b.eq	LBB0_5
; %bb.1:
	lsl	x9, x8, #1
	cmp	x9, #2                  ; =2
	mov	w8, #2
	csel	x8, x9, x8, hi
	sub	x8, x8, #1              ; =1
	lsr	x10, x8, #1
	cbz	x10, LBB0_6
; %bb.2:
	mov	w8, #0
	mov	w13, #0
	add	x11, x10, #1            ; =1
	and	x12, x11, #0xfffffffffffffffe
	lsl	x10, x12, #1
	add	x14, x2, #8             ; =8
	mov	x15, x12
LBB0_3:                                 ; =>This Inner Loop Header: Depth=1
	ldpsw	x16, x17, [x14, #-8]
	ldrb	w16, [x0, x16]
	ldpsw	x1, x3, [x14], #16
	ldrb	w1, [x0, x1]
	ldrb	w17, [x0, x17]
	ldrb	w3, [x0, x3]
	eor	w16, w17, w16
	eor	w17, w3, w1
	add	w8, w8, w16, sxtb
	add	w13, w13, w17, sxtb
	subs	x15, x15, #2            ; =2
	b.ne	LBB0_3
; %bb.4:
	add	w8, w13, w8
	cmp	x11, x12
	b.ne	LBB0_7
	b	LBB0_8
LBB0_5:
	mov	w0, #0
	ret
LBB0_6:
	mov	w8, #0
LBB0_7:                                 ; =>This Inner Loop Header: Depth=1
	add	x11, x2, x10, lsl #2
	ldpsw	x12, x11, [x11]
	ldrb	w12, [x0, x12]
	ldrb	w11, [x0, x11]
	eor	w11, w11, w12
	add	w8, w8, w11, sxtb
	add	x10, x10, #2            ; =2
	cmp	x10, x9
	b.lo	LBB0_7
LBB0_8:
	mov	x0, x8
	ret
	.loh AdrpLdr	Lloh0, Lloh1
	.cfi_endproc
                                        ; -- End function
                                        
                                        
                                        
                                        
					
					
					
					
					
					
					
                                        
                                        
	.globl	_compute_two_plus       ; -- Begin function compute_two_plus
	.p2align	2
_compute_two_plus:                      ; @compute_two_plus
	.cfi_startproc
; %bb.0:
Lloh2:
	adrp	x8, _M@PAGE
Lloh3:
	ldr	x10, [x8, _M@PAGEOFF]
	tst	x10, #0x7fffffffffffffff
	b.eq	LBB1_4
; %bb.1:
	mov	w8, #0
	mov	x9, #0
	lsl	x10, x10, #1
LBB1_2:                                 ; =>This Inner Loop Header: Depth=1
	add	x11, x2, x9, lsl #2
	ldpsw	x12, x11, [x11]
	add	x12, x0, x12
	ldrb	w13, [x12]
	ldrb	w12, [x12, #1]
	eor	w12, w12, w13
	add	x11, x0, x11
	ldrb	w13, [x11]
	eor	w12, w12, w13
	ldrb	w11, [x11, #1]
	eor	w11, w12, w11
	add	w8, w8, w11, sxtb
	add	x9, x9, #2              ; =2
	cmp	x9, x10
	b.lo	LBB1_2
; %bb.3:
	mov	x0, x8
	ret
LBB1_4:
	mov	w8, #0
	mov	x0, x8
	ret
	.loh AdrpLdr	Lloh2, Lloh3
	.cfi_endproc
                                        ; -- End function
                                   
                                   
                                   
                                   
                                   
                                   
                                   
                                   
                                   
                                   
	.globl	_compute_three          ; -- Begin function compute_three
	.p2align	2
_compute_three:                         ; @compute_three
	.cfi_startproc
; %bb.0:
Lloh4:
	adrp	x8, _M@PAGE
Lloh5:
	ldr	x8, [x8, _M@PAGEOFF]
	adds	x9, x8, x8, lsl #1
	b.eq	LBB2_4
; %bb.1:
	mov	x10, #0
	mov	w8, #0
LBB2_2:                                 ; =>This Inner Loop Header: Depth=1
	add	x11, x2, x10, lsl #2
	ldpsw	x12, x13, [x11]
	ldrb	w12, [x0, x12]
	ldrb	w13, [x0, x13]
	eor	w12, w13, w12
	ldrsw	x11, [x11, #8]
	ldrb	w11, [x0, x11]
	eor	w11, w12, w11
	add	w8, w8, w11, sxtb
	add	x10, x10, #3            ; =3
	cmp	x10, x9
	b.lo	LBB2_2
; %bb.3:
	mov	x0, x8
	ret
LBB2_4:
	mov	w8, #0
	mov	x0, x8
	ret
	.loh AdrpLdr	Lloh4, Lloh5
	.cfi_endproc
                                        ; -- End function