arm/aes.m4

+C Loads one word, and adds it to the subkey. Uses T0
+C AES_LOAD(SRC, KEY, REG, INCR)
+define(`AES_LOAD_INCR', `
+	ldrb	$3, [$1], #+1
+	ldrb	T0, [$1], #+1
+	orr	$3, T0, lsl #8
+	ldrb	T0, [$1], #+1
+	orr	$3, T0, lsl #16
+	ldrb	T0, [$1], #+1
+	orr	$3, T0, lsl #24
+	ldr	T0, [$2], #$4
+	eor	$3, T0
+')
+C Loads one word, and adds it to the subkey. Uses T0
+C AES_LOAD(SRC, KEY, REG)
+define(`AES_LOAD', `AES_LOAD_INCR($1, $2, $3, +4)')
+
+C Stores one word. Destroys input.
+C AES_STORE(DST, X)
+define(`AES_STORE', `
+	strb	$2, [$1], #+1
+	ror	$2, $2, #8
+	strb	$2, [$1], #+1
+	ror	$2, $2, #8
+	strb	$2, [$1], #+1
+	ror	$2, $2, #8
+	strb	$2, [$1], #+1
+')
+
+C AES_FINAL_ROUND_V6(a,b,c,d,key,res)
+define(`AES_FINAL_ROUND_V6', `
+	uxtb	T0, $1
+	ldrb	$6, [TABLE, T0]
+	uxtb	T0, $2, ror #8
+	ldrb	T0, [TABLE, T0]
+	eor	$6, $6, T0, lsl #8
+	uxtb	T0, $3, ror #16
+	ldrb	T0, [TABLE, T0]
+	eor	$6, $6, T0, lsl #16
+	ldrb	T0, [TABLE, $4, lsr #24]
+	eor	$6, $6, T0, lsl #24
+	ldr	T0, [$5], #+4
+	eor	$6, $6, T0
+')
+
+C AES_FINAL_ROUND_V5(a,b,c,d,key,res,mask)
+C Avoids the uxtb instruction, introduced in ARMv6.
+C The mask argument should hold the constant 0xff
+define(`AES_FINAL_ROUND_V5', `
+	and	T0, $7, $1
+	ldrb	$6, [TABLE, T0]
+	and	T0, $7, $2, ror #8
+	ldrb	T0, [TABLE, T0]
+	eor	$6, $6, T0, lsl #8
+	and	T0, $7, $3, ror #16
+	ldrb	T0, [TABLE, T0]
+	eor	$6, $6, T0, lsl #16
+	ldrb	T0, [TABLE, $4, lsr #24]
+	eor	$6, $6, T0, lsl #24
+	ldr	T0, [$5], #+4
+	eor	$6, T0
+')

arm/ecc-secp192r1-modp.asm

+C arm/ecc-secp192r1-modp.asm
+
+ifelse(`
+   Copyright (C) 2013 Niels Möller
+
+   This file is part of GNU Nettle.
+
+   GNU Nettle is free software: you can redistribute it and/or
+   modify it under the terms of either:
+
+     * the GNU Lesser General Public License as published by the Free
+       Software Foundation; either version 3 of the License, or (at your
+       option) any later version.
+
+   or
+
+     * the GNU General Public License as published by the Free
+       Software Foundation; either version 2 of the License, or (at your
+       option) any later version.
+
+   or both in parallel, as here.
+
+   GNU Nettle is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   General Public License for more details.
+
+   You should have received copies of the GNU General Public License and
+   the GNU Lesser General Public License along with this program.  If
+   not, see http://www.gnu.org/licenses/.
+')
+
+	.file "ecc-secp192r1-modp.asm"
+	.arm
+
+define(`HP', `r0') C Overlaps unused modulo argument
+define(`RP', `r1')
+define(`XP', `r2')
+
+define(`T0', `r3')
+define(`T1', `r4')
+define(`T2', `r5')
+define(`T3', `r6')
+define(`T4', `r7')
+define(`T5', `r8')
+define(`T6', `r10')
+define(`T7', `r11')
+define(`H0', `T0') C Overlaps T0 and T1
+define(`H1', `T1')
+define(`C2', `HP')
+define(`C4', `r12')
+
+	C ecc_secp192r1_modp (const struct ecc_modulo *m, mp_limb_t *rp)
+	.text
+	.align 2
+
+PROLOGUE(_nettle_ecc_secp192r1_modp)
+	push	{r4,r5,r6,r7,r8,r10,r11}
+	C Reduce two words at a time
+	add	HP, XP, #48
+	add	XP, XP, #8
+	ldmdb	HP!, {H0,H1}
+	ldm	XP, {T2,T3,T4,T5,T6,T7}
+	mov	C4, #0
+	adds	T4, T4, H0
+	adcs	T5, T5, H1
+	adcs	T6, T6, H0
+	adcs	T7, T7, H1
+	C Need to add carry to T2 and T4, do T4 later.
+	adc	C4, C4, #0
+
+	ldmdb	HP!, {H0,H1}
+	mov	C2, #0
+	adcs	T2, T2, H0
+	adcs	T3, T3, H1
+	adcs	T4, T4, H0
+	adcs	T5, T5, H1
+	C Need to add carry to T0 and T2, do T2 later
+	adc	C2, C2, #0
+
+	ldmdb	XP!, {T0, T1}
+	adcs	T0, T0, T6
+	adcs	T1, T1, T7
+	adcs	T2, T2, T6
+	adcs	T3, T3, T7
+	adc	C4, C4, #0
+
+	adds	T2, T2, C2
+	adcs	T3, T3, #0
+	adcs	T4, T4, C4
+	adcs	T5, T5, #0
+	mov	C2, #0
+	adc	C2, C2, #0
+
+	C Add in final carry
+	adcs	T0, T0, #0
+	adcs	T1, T1, #0
+	adcs	T2, T2, C2
+	adcs	T3, T3, #0
+	adcs	T4, T4, #0
+	adc	T5, T5, #0
+
+	stm	RP, {T0,T1,T2,T3,T4,T5}
+
+	pop	{r4,r5,r6,r7,r8,r10,r11}
+	bx	lr
+EPILOGUE(_nettle_ecc_secp192r1_modp)

arm/ecc-secp224r1-modp.asm

+C arm/ecc-secp224r1-modp.asm
+
+ifelse(`
+   Copyright (C) 2013 Niels Möller
+
+   This file is part of GNU Nettle.
+
+   GNU Nettle is free software: you can redistribute it and/or
+   modify it under the terms of either:
+
+     * the GNU Lesser General Public License as published by the Free
+       Software Foundation; either version 3 of the License, or (at your
+       option) any later version.
+
+   or
+
+     * the GNU General Public License as published by the Free
+       Software Foundation; either version 2 of the License, or (at your
+       option) any later version.
+
+   or both in parallel, as here.
+
+   GNU Nettle is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   General Public License for more details.
+
+   You should have received copies of the GNU General Public License and
+   the GNU Lesser General Public License along with this program.  If
+   not, see http://www.gnu.org/licenses/.
+')
+
+	.file "ecc-secp224r1-modp.asm"
+	.arm
+
+define(`RP', `r1')	C Overlaps T0
+define(`XP', `r2')
+define(`H', `r0')	C Overlaps unused modulo argument
+
+define(`T0', `r1')
+define(`T1', `r3')
+define(`T2', `r4')
+define(`T3', `r5')
+define(`T4', `r6')
+define(`T5', `r7')
+define(`T6', `r8')
+define(`N3', `r10')
+define(`L0', `r11')
+define(`L1', `r12')
+define(`L2', `lr')
+
+	C ecc_secp224r1_modp (const struct ecc_modulo *m, mp_limb_t *rp)
+	.text
+	.align 2
+
+PROLOGUE(_nettle_ecc_secp224r1_modp)
+	C Pushes RP last
+	push	{r1,r4,r5,r6,r7,r8,r10,r11,lr}
+
+	add	L2, XP, #28
+	ldm	L2, {T0,T1,T2,T3,T4,T5,T6}
+	mov	H, #0
+
+	adds	T0, T0, T4
+	adcs	T1, T1, T5
+	adcs	T2, T2, T6
+	adc	H, H, #0
+
+	C This switch from adcs to sbcs takes carry into account with
+	C correct sign, but it always subtracts 1 too much. We arrange
+	C to also add B^7 + 1 below, so the effect is adding p. This
+	C addition of p also ensures that the result never is
+	C negative.
+
+	sbcs	N3, T3, T0
+	sbcs	T4, T4, T1
+	sbcs	T5, T5, T2
+	sbcs	T6, T6, H
+	mov	H, #1		C This is the B^7
+	sbc	H, #0
+	subs	T6, T6, T3
+	sbc	H, #0
+
+	C Now subtract from low half
+	ldm	XP!, {L0,L1,L2}
+
+	C Clear carry, with the sbcs, this is the 1.
+	adds	XP, #0
+
+	sbcs	T0, L0, T0
+	sbcs	T1, L1, T1
+	sbcs	T2, L2, T2
+	ldm	XP!, {T3,L0,L1,L2}
+	sbcs	T3, T3, N3
+	sbcs	T4, L0, T4
+	sbcs	T5, L1, T5
+	sbcs	T6, L2, T6
+	rsc	H, H, #0
+
+	C Now -2 <= H <= 0 is the borrow, so subtract (B^3 - 1) |H|
+	C Use (B^3 - 1) H = <H, H, H> if -1 <=H <= 0, and
+	C     (B^3 - 1) H = <1,B-1, B-1, B-2> if H = -2
+	subs	T0, T0, H
+	asr	L1, H, #1
+	sbcs	T1, T1, L1
+	eor	H, H, L1
+	sbcs	T2, T2, L1
+	sbcs	T3, T3, H
+	sbcs	T4, T4, #0
+	sbcs	T5, T5, #0
+	sbcs	T6, T6, #0
+	sbcs	H, H, H
+
+	pop	{XP}	C Original RP
+
+	C Final borrow, subtract (B^3 - 1) |H|
+	subs	T0, T0, H
+	sbcs	T1, T1, H
+	sbcs	T2, T2, H
+	sbcs	T3, T3, #0
+	sbcs	T4, T4, #0
+	sbcs	T5, T5, #0
+	sbcs	T6, T6, #0
+
+	stm	XP, {T0,T1,T2,T3,T4,T5,T6}
+
+	pop	{r4,r5,r6,r7,r8,r10,r11,pc}
+EPILOGUE(_nettle_ecc_secp224r1_modp)

arm/ecc-secp256r1-redc.asm

+C arm/ecc-secp256r1-redc.asm
+
+ifelse(`
+   Copyright (C) 2013 Niels Möller
+
+   This file is part of GNU Nettle.
+
+   GNU Nettle is free software: you can redistribute it and/or
+   modify it under the terms of either:
+
+     * the GNU Lesser General Public License as published by the Free
+       Software Foundation; either version 3 of the License, or (at your
+       option) any later version.
+
+   or
+
+     * the GNU General Public License as published by the Free
+       Software Foundation; either version 2 of the License, or (at your
+       option) any later version.
+
+   or both in parallel, as here.
+
+   GNU Nettle is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   General Public License for more details.
+
+   You should have received copies of the GNU General Public License and
+   the GNU Lesser General Public License along with this program.  If
+   not, see http://www.gnu.org/licenses/.
+')
+
+	.file "ecc-secp256r1-redc.asm"
+	.arm
+
+define(`RP', `r1')	C Overlaps T1 below
+define(`XP', `r2')
+
+define(`T0', `r0')	C Overlaps unused modulo argument
+define(`T1', `r1')
+define(`T2', `r3')
+define(`T3', `r4')
+define(`T4', `r5')
+define(`T5', `r6')
+define(`T6', `r7')
+define(`T7', `r8')
+define(`F0', `r10')
+define(`F1', `r11')
+define(`F2', `r12')
+define(`F3', `lr')
+
+	C ecc_secp256r1_redc (const struct ecc_modulo *m, mp_limb_t *rp)
+	.text
+	.align 2
+
+PROLOGUE(_nettle_ecc_secp256r1_redc)
+	C Pushes RP last
+	push	{r1, r4,r5,r6,r7,r8,r10,r11,lr}
+
+	ldm	XP!, {T0,T1,T2,T3,T4,T5,T6,T7}
+
+	C Set <F3,F2,F1> to the high 4 limbs of (B^2-B+1)<T2,T1,T0>
+	C         T2 T1
+	C   T2 T1 T0
+	C   -  T2 T1 T0
+	C -------------
+	C   F3 F2 F1 F0
+
+
+	adds	F1, T0, T2
+	adcs	F2, T1, #0
+	adc	F3, T2, #0
+
+	subs	F0, T1, T0
+	sbcs	F1, F1, T1	C Could also be rsc ?
+	sbcs	F2, F2, T2
+	sbc	F3, F3, #0
+
+	C Add:
+	C   T10 T9 T8 T7 T6 T5 T4 T3
+	C +  F3 F2 F1 F0 T0 T2 T1 T0
+	C --------------------------
+	C    T7 T6 T5 T4 T3 T2 T1 T0
+
+	adds	T3, T3, T0
+	adcs	T1, T4, T1
+	adcs	T2, T5, T2
+	adcs	T6, T6, T0
+	mov	T0, T3		C FIXME: Be more clever?
+	mov	T3, T6
+	adcs	T4, T7, F0
+
+	ldm	XP!, {T5,T6,T7}
+	adcs	T5, T5, F1
+	adcs	T6, T6, F2
+	adcs	T7, T7, F3
+
+	C New F3, F2, F1, F0, also adding in carry
+	adcs	F1, T0, T2
+	adcs	F2, T1, #0
+	adc	F3, T2, #0
+
+	subs	F0, T1, T0
+	sbcs	F1, F1, T1	C Could also be rsc ?
+	sbcs	F2, F2, T2
+	sbc	F3, F3, #0
+
+	C Start adding
+	adds	T3, T3, T0
+	adcs	T1, T4, T1
+	adcs	T2, T5, T2
+	adcs	T6, T6, T0
+	mov	T0, T3		C FIXME: Be more clever?
+	mov	T3, T6
+	adcs	T4, T7, F0
+
+	ldm	XP!, {T5,T6,T7}
+	adcs	T5, T5, F1
+	adcs	T6, T6, F2
+	adcs	T7, T7, F3
+
+	C Final iteration, eliminate only T0, T1
+	C Set <F2, F1, F0> to the high 3 limbs of (B^2-B+1)<T1,T0>
+
+	C      T1 T0 T1
+	C      -  T1 T0
+	C -------------
+	C      F2 F1 F0
+
+	C First add in carry
+	adcs	F1, T0, #0
+	adcs	F2, T1, #0
+	subs	F0, T1, T0
+	sbcs	F1, F1, T1
+	sbc	F2, F2, #0
+
+	C Add:
+	C    T9 T8 T7 T6 T5 T4 T3 T2
+	C +  F2 F1 F0 T0  0 T1 T0  0
+	C --------------------------
+	C    F2 F1 T7 T6 T5 T4 T3 T2
+
+	adds	T3, T3, T0
+	adcs	T4, T4, T1
+	adcs	T5, T5, #0
+	adcs	T6, T6, T0
+	adcs	T7, T7, F0
+	ldm	XP!, {T0, T1}
+	mov	F3, #0
+	adcs	F1, F1, T0
+	adcs	F2, F2, T1
+
+	C Sum is < B^8 + p, so it's enough to fold carry once,
+	C If carry, add in
+	C   B^7 - B^6 - B^3 + 1  = <0, B-2, B-1, B-1, B-1, 0, 0, 1>
+
+	C Mask from carry flag, leaving carry intact
+	adc	F3, F3, #0
+	rsb	F3, F3, #0
+
+	pop	{XP}	C Original RP
+
+	adcs	T0, T2, #0
+	adcs	T1, T3, #0
+	adcs	T2, T4, #0
+	adcs	T3, T5, F3
+	adcs	T4, T6, F3
+	adcs	T5, T7, F3
+	and	F3, F3, #-2
+	adcs	T6, F1, F3
+	adcs	T7, F2, #0
+
+	stm	XP, {T0,T1,T2,T3,T4,T5,T6,T7}
+
+	pop	{r4,r5,r6,r7,r8,r10,r11,pc}
+EPILOGUE(_nettle_ecc_secp256r1_redc)

arm/ecc-secp384r1-modp.asm

+C arm/ecc-secp384r1-modp.asm
+
+ifelse(`
+   Copyright (C) 2013 Niels Möller
+
+   This file is part of GNU Nettle.
+
+   GNU Nettle is free software: you can redistribute it and/or
+   modify it under the terms of either:
+
+     * the GNU Lesser General Public License as published by the Free
+       Software Foundation; either version 3 of the License, or (at your
+       option) any later version.
+
+   or
+
+     * the GNU General Public License as published by the Free
+       Software Foundation; either version 2 of the License, or (at your
+       option) any later version.
+
+   or both in parallel, as here.
+
+   GNU Nettle is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   General Public License for more details.
+
+   You should have received copies of the GNU General Public License and
+   the GNU Lesser General Public License along with this program.  If
+   not, see http://www.gnu.org/licenses/.
+')
+
+	.file "ecc-secp384r1-modp.asm"
+	.arm
+
+define(`RP', `r1')
+define(`XP', `r2')
+
+define(`T0', `r0')
+define(`T1', `r3')
+define(`T2', `r4')
+define(`T3', `r5')
+define(`F0', `r6')
+define(`F1', `r7')
+define(`F2', `r8')
+define(`F3', `r10')
+define(`F4', `r11')
+define(`N', `r12')
+define(`H', `lr')
+	
+	C ecc_secp384r1_modp (const struct ecc_modulo *m, mp_limb_t *rp)
+	.text
+	.align 2
+
+PROLOGUE(_nettle_ecc_secp384r1_modp)
+	push	{r4,r5,r6,r7,r8,r10,r11,lr}
+
+	add	XP, XP, #80
+	ldm	XP, {T0, T1, T2, T3}	C 20-23
+
+	C First get top 4 limbs, which need folding twice, as
+	C
+	C     T3 T2 T1 T0
+	C        T3 T2 T1
+	C             -T3
+	C ----------------
+	C  F4 F3 F2 F1 F0
+	C
+	C Start with
+	C
+	C   T3 T1 T0
+	C         T1
+	C        -T3
+	C -----------
+	C   F2 F1 F0   Always fits
+	
+	adds	F0, T0, T1
+	adcs	F1, T1, #0
+	adcs	F2, T3, #0
+	subs	F0, F0, T3
+	sbcs	F1, F1, #0
+	sbcs	F2, F2, #0
+
+	C      T3 T2 T2  0
+	C         F2 F1 F0
+	C  ----------------
+	C   F4 F3 F2 F1 F0
+
+	mov	F4, #0
+	adds	F1, F1, T2
+	adcs	F2, F2, T2
+	adcs	F3, T3, #0
+	adcs	F4, F4, #0
+
+	C Add in to high part
+	sub	XP, XP, #32
+	ldm	XP, {T0, T1, T2, T3}	C 12-15
+	mov	H, #0
+	adds	F0, T0, F0
+	adcs	F1, T1, F1
+	adcs	F2, T2, F2
+	adcs	F3, T3, F3
+	adcs	F4, F4, #0			C Do F4 later
+
+	C Add to low part, keeping carry (positive or negative) in H
+	sub	XP, XP, #48
+	ldm	XP, {T0, T1, T2, T3}	C 0-3
+	mov	H, #0
+	adds	T0, T0, F0
+	adcs	T1, T1, F1
+	adcs	T2, T2, F2
+	adcs	T3, T3, F3
+	adc	H, H, #0
+	subs	T1, T1, F0
+	sbcs	T2, T2, F1
+	sbcs	T3, T3, F2
+	sbc	H, H, #0
+	adds	T3, T3, F0
+	adc	H, H, #0
+	
+	stm	XP!, {T0,T1,T2,T3}	C 0-3
+	mov	N, #2
+.Loop:
+	ldm	XP, {T0,T1,T2,T3}	C 4-7
+
+	C First, propagate carry
+	adds	T0, T0, H
+	asr	H, #31		C Sign extend
+	adcs	T1, T1, H
+	adcs	T2, T2, H
+	adcs	T3, T3, H
+	adc	H, H, #0
+
+	C +B^4 term
+	adds	T0, T0, F0
+	adcs	T1, T1, F1
+	adcs	T2, T2, F2
+	adcs	T3, T3, F3
+	adc	H, H, #0
+
+	C +B^3 terms
+	ldr	F0, [XP, #+48]		C 16
+	adds	T0, T0, F1
+	adcs	T1, T1, F2
+	adcs	T2, T2, F3
+	adcs	T3, T3, F0
+	adc	H, H, #0
+
+	C -B
+	ldr	F1, [XP, #+52]		C 17-18
+	ldr	F2, [XP, #+56]
+	subs	T0, T0, F3
+	sbcs	T1, T1, F0
+	sbcs	T2, T2, F1
+	sbcs	T3, T3, F2
+	sbcs	H, H, #0
+
+	C +1
+	ldr	F3, [XP, #+60]		C 19
+	adds	T0, T0, F0
+	adcs	T1, T1, F1
+	adcs	T2, T2, F2
+	adcs	T3, T3, F3
+	adc	H, H, #0
+	subs	N, N, #1
+	stm	XP!, {T0,T1,T2,T3}
+	bne	.Loop
+
+	C Fold high limbs, we need to add in
+	C
+	C F4 F4 0 -F4 F4 H H 0 -H H
+	C
+	C We always have F4 >= 0, but we can have H < 0.
+	C Sign extension gets tricky when F4 = 0 and H < 0.
+	sub	XP, XP, #48
+
+	ldm	XP, {T0,T1,T2,T3}	C 0-3
+
+	C     H  H  0 -H  H
+	C  ----------------
+	C  S  H F3 F2 F1 F0
+	C
+	C Define S = H >> 31 (asr), we then have
+	C
+	C  F0 = H
+	C  F1 = S - H
+	C  F2 = - [H > 0]
+	C  F3 = H - [H > 0]
+	C   H = H + S
+	C 
+	C And we get underflow in S - H iff H > 0
+
+	C				H = 0	H > 0	H = -1
+	mov	F0, H		C	0	H	-1
+	asr	H, #31
+	subs	F1, H, F0	C	0,C=1	-H,C=0	0,C=1
+	sbc	F2, F2, F2	C	0	-1	0
+	sbc	F3, F0, #0	C	0	H-1	-1
+
+	adds	T0, T0, F0
+	adcs	T1, T1, F1
+	adcs	T2, T2, F2
+	adcs	T3, T3, F3
+	adc	H, H, F0	C	0+cy	H+cy	-2+cy
+
+	stm	XP!, {T0,T1,T2,T3}	C 0-3
+	ldm	XP, {T0,T1,T2,T3}	C 4-7
+	
+	C   F4  0 -F4
+	C ---------
+	C   F3 F2  F1
+	
+	rsbs	F1, F4, #0
+	sbc	F2, F2, F2
+	sbc	F3, F4, #0
+
+	C Sign extend H
+	adds	F0, F4, H
+	asr	H, H, #31	
+	adcs	F1, F1, H
+	adcs	F2, F2, H
+	adcs	F3, F3, H
+	adcs	F4, F4, H
+	adc	H, H, #0
+	
+	adds	T0, T0, F0
+	adcs	T1, T1, F1
+	adcs	T2, T2, F2
+	adcs	T3, T3, F3
+
+	stm	XP!, {T0,T1,T2,T3}	C 4-7
+	ldm	XP, {T0,T1,T2,T3}	C 8-11
+
+	adcs	T0, T0, F4
+	adcs	T1, T1, H
+	adcs	T2, T2, H
+	adcs	T3, T3, H
+	adc	H, H, #0
+	
+	stm	XP, {T0,T1,T2,T3}	C 8-11
+
+	C Final (unlikely) carry
+	sub	XP, XP, #32
+	ldm	XP!, {T0,T1,T2,T3}	C 0-3
+	C Fold H into F0-F4
+	mov	F0, H
+	asr	H, #31
+	subs	F1, H, F0
+	sbc	F2, F2, F2
+	sbc	F3, F0, #0
+	add	F4, F0, H
+
+	adds	T0, T0, F0
+	adcs	T1, T1, F1
+	adcs	T2, T2, F2
+	adcs	T3, T3, F3
+	
+	stm	RP!, {T0,T1,T2,T3}	C 0-3
+	ldm	XP!, {T0,T1,T2,T3}	C 4-7
+	adcs	T0, T0, F4
+	adcs	T1, T1, H
+	adcs	T2, T2, H
+	adcs	T3, T3, H
+	stm	RP!, {T0,T1,T2,T3}	C 4-7
+	ldm	XP, {T0,T1,T2,T3}	C 8-11
+	adcs	T0, T0, H
+	adcs	T1, T1, H
+	adcs	T2, T2, H
+	adcs	T3, T3, H
+	stm	RP, {T0,T1,T2,T3}	C 8-11
+	pop	{r4,r5,r6,r7,r8,r10,r11,pc}
+EPILOGUE(_nettle_ecc_secp384r1_modp)

arm/ecc-secp521r1-modp.asm

+C arm/ecc-secp521r1-modp.asm
+
+ifelse(`
+   Copyright (C) 2013 Niels Möller
+
+   This file is part of GNU Nettle.
+
+   GNU Nettle is free software: you can redistribute it and/or
+   modify it under the terms of either:
+
+     * the GNU Lesser General Public License as published by the Free
+       Software Foundation; either version 3 of the License, or (at your
+       option) any later version.
+
+   or
+
+     * the GNU General Public License as published by the Free
+       Software Foundation; either version 2 of the License, or (at your
+       option) any later version.
+
+   or both in parallel, as here.
+
+   GNU Nettle is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   General Public License for more details.
+
+   You should have received copies of the GNU General Public License and
+   the GNU Lesser General Public License along with this program.  If
+   not, see http://www.gnu.org/licenses/.
+')
+
+	.file "ecc-secp521r1-modp.asm"
+	.arm
+
+define(`HP', `r0')
+define(`RP', `r1')
+define(`XP', `r2')
+define(`T0', `r3')
+define(`T1', `r4')
+define(`T2', `r5')
+define(`F0', `r6')
+define(`F1', `r7')
+define(`F2', `r8')
+define(`F3', `r10')
+define(`H', `r12')
+define(`N', `lr')
+
+	C ecc_secp521r1_modp (const struct ecc_modulo *m, mp_limb_t *rp)
+	.text
+.Lc511:
+	.int 511
+
+	.align 2
+
+PROLOGUE(_nettle_ecc_secp521r1_modp)
+	push	{r4,r5,r6,r7,r8,r10,lr}
+
+	C Use that B^17 = 2^23 (mod p)
+	ldr	F3, [XP, #+68]		C 17
+	add	HP, XP, #72		C 18
+	ldr	T0, [XP]		C 0
+	adds	T0, T0, F3, lsl	#23
+	str	T0, [XP], #+4
+	mov	N, #5
+
+	C 5 iterations, reading limbs 18-20, 21-23, 24-26, 27-29, 30-32
+	C and adding to limbs          1-3,    4-6,   7-9, 19-12, 13-15
+.Loop:
+	ldm	XP, {T0,T1,T2}		C  1+3*k --  3+3*k
+	lsr	F0, F3, #9
+	ldm	HP!, {F1,F2,F3}		C 18+3*k -- 20+3*k
+	orr	F0, F0, F1, lsl #23
+	lsr	F1, F1, #9
+	orr	F1, F1, F2, lsl #23
+	lsr	F2, F2, #9
+	orr	F2, F2, F3, lsl #23
+	adcs	T0, T0, F0
+	adcs	T1, T1, F1
+	adcs	T2, T2, F2
+	sub	N, N, #1
+	stm	XP!,{T0,T1,T2}
+	teq	N, #0
+	bne	.Loop
+
+	ldr	F0, [XP], #-64		C 16
+	ldr	F1, [HP]		C 33
+	ldr	T0, .Lc511
+
+	C Handling of high limbs
+	C F0 = rp[16] + carry in + F3 >> 9
+	adcs	F0, F0, F3, lsr #9
+	C Copy low 9 bits to H, then shift right including carry
+	and	H, F0, T0
+	mov	F0, F0, rrx
+	lsr	F0, F0, #8
+	C Add in F1 = rp[33], with weight 2^1056 = 2^14
+	adds	F0, F0, F1, lsl #14
+	lsr	F1, F1, #18
+	adc	F1, F1, #0
+
+	ldm	XP!, {T0, T1}		C 0-1
+	adds	T0, T0, F0
+	adcs	T1, T1, F1
+	stm	RP!, {T0, T1}
+
+	ldm	XP!, {T0,T1,T2,F0,F1,F2,F3}	C 2-8
+	adcs	T0, T0, #0
+	adcs	T1, T1, #0
+	adcs	T2, T2, #0
+	adcs	F0, F0, #0
+	adcs	F1, F1, #0
+	adcs	F2, F2, #0
+	adcs	F3, F3, #0
+	stm	RP!, {T0,T1,T2,F0,F1,F2,F3}	C 2-8
+	ldm	XP, {T0,T1,T2,F0,F1,F2,F3}	C 9-15
+	adcs	T0, T0, #0
+	adcs	T1, T1, #0
+	adcs	T2, T2, #0
+	adcs	F0, F0, #0
+	adcs	F1, F1, #0
+	adcs	F2, F2, #0
+	adcs	F3, F3, #0
+	adcs	H, H, #0
+	stm	RP, {T0,T1,T2,F0,F1,F2,F3,H}	C 9-16
+
+	pop	{r4,r5,r6,r7,r8,r10,pc}
+EPILOGUE(_nettle_ecc_secp521r1_modp)

arm/fat/aes-decrypt-internal-2.asm

+C arm/fat/aes-decrypt-internal-2.asm
+
+
+ifelse(`
+   Copyright (C) 2015 Niels Möller
+
+   This file is part of GNU Nettle.
+
+   GNU Nettle is free software: you can redistribute it and/or
+   modify it under the terms of either:
+
+     * the GNU Lesser General Public License as published by the Free
+       Software Foundation; either version 3 of the License, or (at your
+       option) any later version.
+
+   or
+
+     * the GNU General Public License as published by the Free
+       Software Foundation; either version 2 of the License, or (at your
+       option) any later version.
+
+   or both in parallel, as here.
+
+   GNU Nettle is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   General Public License for more details.
+
+   You should have received copies of the GNU General Public License and
+   the GNU Lesser General Public License along with this program.  If
+   not, see http://www.gnu.org/licenses/.
+')
+
+define(`fat_transform', `$1_armv6')
+include_src(`arm/v6/aes-decrypt-internal.asm')

arm/fat/aes-decrypt-internal.asm

+C arm/fat/aes-decrypt-internal.asm
+
+
+ifelse(`
+   Copyright (C) 2015 Niels Möller
+
+   This file is part of GNU Nettle.
+
+   GNU Nettle is free software: you can redistribute it and/or
+   modify it under the terms of either:
+
+     * the GNU Lesser General Public License as published by the Free
+       Software Foundation; either version 3 of the License, or (at your
+       option) any later version.
+
+   or
+
+     * the GNU General Public License as published by the Free
+       Software Foundation; either version 2 of the License, or (at your
+       option) any later version.
+
+   or both in parallel, as here.
+
+   GNU Nettle is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   General Public License for more details.
+
+   You should have received copies of the GNU General Public License and
+   the GNU Lesser General Public License along with this program.  If
+   not, see http://www.gnu.org/licenses/.
+')
+
+define(`fat_transform', `$1_arm')
+include_src(`arm/aes-decrypt-internal.asm')

arm/fat/aes-encrypt-internal-2.asm

+C arm/fat/aes-encrypt-internal-2.asm
+
+
+ifelse(`
+   Copyright (C) 2015 Niels Möller
+
+   This file is part of GNU Nettle.
+
+   GNU Nettle is free software: you can redistribute it and/or
+   modify it under the terms of either:
+
+     * the GNU Lesser General Public License as published by the Free
+       Software Foundation; either version 3 of the License, or (at your
+       option) any later version.
+
+   or
+
+     * the GNU General Public License as published by the Free
+       Software Foundation; either version 2 of the License, or (at your
+       option) any later version.
+
+   or both in parallel, as here.
+
+   GNU Nettle is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   General Public License for more details.
+
+   You should have received copies of the GNU General Public License and
+   the GNU Lesser General Public License along with this program.  If
+   not, see http://www.gnu.org/licenses/.
+')
+
+define(`fat_transform', `$1_armv6')
+include_src(`arm/v6/aes-encrypt-internal.asm')

arm/fat/aes-encrypt-internal.asm

+C arm/fat/aes-encrypt-internal.asm
+
+
+ifelse(`
+   Copyright (C) 2015 Niels Möller
+
+   This file is part of GNU Nettle.
+
+   GNU Nettle is free software: you can redistribute it and/or
+   modify it under the terms of either:
+
+     * the GNU Lesser General Public License as published by the Free
+       Software Foundation; either version 3 of the License, or (at your
+       option) any later version.
+
+   or
+
+     * the GNU General Public License as published by the Free
+       Software Foundation; either version 2 of the License, or (at your
+       option) any later version.
+
+   or both in parallel, as here.
+
+   GNU Nettle is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   General Public License for more details.
+
+   You should have received copies of the GNU General Public License and
+   the GNU Lesser General Public License along with this program.  If
+   not, see http://www.gnu.org/licenses/.
+')
+
+define(`fat_transform', `$1_arm')
+include_src(`arm/aes-encrypt-internal.asm')

arm/fat/chacha-3core.asm

+C arm/fat/chacha-3core.asm
+
+
+ifelse(`
+   Copyright (C) 2020 Niels Möller
+
+   This file is part of GNU Nettle.
+
+   GNU Nettle is free software: you can redistribute it and/or
+   modify it under the terms of either:
+
+     * the GNU Lesser General Public License as published by the Free
+       Software Foundation; either version 3 of the License, or (at your
+       option) any later version.
+
+   or
+
+     * the GNU General Public License as published by the Free
+       Software Foundation; either version 2 of the License, or (at your
+       option) any later version.
+
+   or both in parallel, as here.
+
+   GNU Nettle is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   General Public License for more details.
+
+   You should have received copies of the GNU General Public License and
+   the GNU Lesser General Public License along with this program.  If
+   not, see http://www.gnu.org/licenses/.
+')
+
+dnl PROLOGUE(_nettle_fat_chacha_3core) picked up by configure
+
+include_src(`arm/neon/chacha-3core.asm')

arm/fat/salsa20-2core.asm

+C arm/fat/salsa20-2core.asm
+
+
+ifelse(`
+   Copyright (C) 2020 Niels Möller
+
+   This file is part of GNU Nettle.
+
+   GNU Nettle is free software: you can redistribute it and/or
+   modify it under the terms of either:
+
+     * the GNU Lesser General Public License as published by the Free
+       Software Foundation; either version 3 of the License, or (at your
+       option) any later version.
+
+   or
+
+     * the GNU General Public License as published by the Free
+       Software Foundation; either version 2 of the License, or (at your
+       option) any later version.
+
+   or both in parallel, as here.
+
+   GNU Nettle is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   General Public License for more details.
+
+   You should have received copies of the GNU General Public License and
+   the GNU Lesser General Public License along with this program.  If
+   not, see http://www.gnu.org/licenses/.
+')
+
+dnl PROLOGUE(_nettle_fat_salsa20_2core) picked up by configure
+
+include_src(`arm/neon/salsa20-2core.asm')

arm/fat/sha1-compress-2.asm

+C arm/fat/sha1-compress-2.asm
+
+
+ifelse(`
+   Copyright (C) 2015 Niels Möller
+
+   This file is part of GNU Nettle.
+
+   GNU Nettle is free software: you can redistribute it and/or
+   modify it under the terms of either:
+
+     * the GNU Lesser General Public License as published by the Free
+       Software Foundation; either version 3 of the License, or (at your
+       option) any later version.
+
+   or
+
+     * the GNU General Public License as published by the Free
+       Software Foundation; either version 2 of the License, or (at your
+       option) any later version.
+
+   or both in parallel, as here.
+
+   GNU Nettle is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   General Public License for more details.
+
+   You should have received copies of the GNU General Public License and
+   the GNU Lesser General Public License along with this program.  If
+   not, see http://www.gnu.org/licenses/.
+')
+
+dnl PROLOGUE(nettle_sha1_compress) picked up by configure
+
+define(`fat_transform', `_$1_armv6')
+include_src(`arm/v6/sha1-compress.asm')

arm/fat/sha256-compress-n-2.asm

+C arm/fat/sha256-compress-n-2.asm
+
+
+ifelse(`
+   Copyright (C) 2015 Niels Möller
+
+   This file is part of GNU Nettle.
+
+   GNU Nettle is free software: you can redistribute it and/or
+   modify it under the terms of either:
+
+     * the GNU Lesser General Public License as published by the Free
+       Software Foundation; either version 3 of the License, or (at your
+       option) any later version.
+
+   or
+
+     * the GNU General Public License as published by the Free
+       Software Foundation; either version 2 of the License, or (at your
+       option) any later version.
+
+   or both in parallel, as here.
+
+   GNU Nettle is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   General Public License for more details.
+
+   You should have received copies of the GNU General Public License and
+   the GNU Lesser General Public License along with this program.  If
+   not, see http://www.gnu.org/licenses/.
+')
+
+dnl PROLOGUE(_nettle_sha256_compress_n) picked up by configure
+
+define(`fat_transform', `$1_armv6')
+include_src(`arm/v6/sha256-compress-n.asm')

arm/fat/sha3-permute-2.asm

+C arm/fat/sha3-permute-2.asm
+
+
+ifelse(`
+   Copyright (C) 2015 Niels Möller
+
+   This file is part of GNU Nettle.
+
+   GNU Nettle is free software: you can redistribute it and/or
+   modify it under the terms of either:
+
+     * the GNU Lesser General Public License as published by the Free
+       Software Foundation; either version 3 of the License, or (at your
+       option) any later version.
+
+   or
+
+     * the GNU General Public License as published by the Free
+       Software Foundation; either version 2 of the License, or (at your
+       option) any later version.
+
+   or both in parallel, as here.
+
+   GNU Nettle is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   General Public License for more details.
+
+   You should have received copies of the GNU General Public License and
+   the GNU Lesser General Public License along with this program.  If
+   not, see http://www.gnu.org/licenses/.
+')
+
+dnl PROLOGUE(_nettle_sha3_permute) picked up by configure
+
+define(`fat_transform', `_$1_neon')
+include_src(`arm/neon/sha3-permute.asm')

arm/fat/sha512-compress-2.asm

+C arm/fat/sha3-compress-2.asm
+
+
+ifelse(`
+   Copyright (C) 2015 Niels Möller
+
+   This file is part of GNU Nettle.
+
+   GNU Nettle is free software: you can redistribute it and/or
+   modify it under the terms of either:
+
+     * the GNU Lesser General Public License as published by the Free
+       Software Foundation; either version 3 of the License, or (at your
+       option) any later version.
+
+   or
+
+     * the GNU General Public License as published by the Free
+       Software Foundation; either version 2 of the License, or (at your
+       option) any later version.
+
+   or both in parallel, as here.
+
+   GNU Nettle is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   General Public License for more details.
+
+   You should have received copies of the GNU General Public License and
+   the GNU Lesser General Public License along with this program.  If
+   not, see http://www.gnu.org/licenses/.
+')
+
+dnl PROLOGUE(_nettle_sha512_compress) picked up by configure
+
+define(`fat_transform', `$1_neon')
+include_src(`arm/neon/sha512-compress.asm')

arm/fat/umac-nh-2.asm

+C arm/fat/umac-nh-2.asm
+
+
+ifelse(`
+   Copyright (C) 2015 Niels Möller
+
+   This file is part of GNU Nettle.
+
+   GNU Nettle is free software: you can redistribute it and/or
+   modify it under the terms of either:
+
+     * the GNU Lesser General Public License as published by the Free
+       Software Foundation; either version 3 of the License, or (at your
+       option) any later version.
+
+   or
+
+     * the GNU General Public License as published by the Free
+       Software Foundation; either version 2 of the License, or (at your
+       option) any later version.
+
+   or both in parallel, as here.
+
+   GNU Nettle is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   General Public License for more details.
+
+   You should have received copies of the GNU General Public License and
+   the GNU Lesser General Public License along with this program.  If
+   not, see http://www.gnu.org/licenses/.
+')
+
+dnl PROLOGUE(_nettle_umac_nh) picked up by configure
+
+define(`fat_transform', `$1_neon')
+include_src(`arm/neon/umac-nh.asm')

arm/fat/umac-nh-n-2.asm

+C arm/fat/umac-nh-n-2.asm
+
+
+ifelse(`
+   Copyright (C) 2015 Niels Möller
+
+   This file is part of GNU Nettle.
+
+   GNU Nettle is free software: you can redistribute it and/or
+   modify it under the terms of either:
+
+     * the GNU Lesser General Public License as published by the Free
+       Software Foundation; either version 3 of the License, or (at your
+       option) any later version.
+
+   or
+
+     * the GNU General Public License as published by the Free
+       Software Foundation; either version 2 of the License, or (at your
+       option) any later version.
+
+   or both in parallel, as here.
+
+   GNU Nettle is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   General Public License for more details.
+
+   You should have received copies of the GNU General Public License and
+   the GNU Lesser General Public License along with this program.  If
+   not, see http://www.gnu.org/licenses/.
+')
+
+dnl PROLOGUE(_nettle_umac_nh_n) picked up by configure
+
+define(`fat_transform', `$1_neon')
+include_src(`arm/neon/umac-nh-n.asm')

arm/machine.m4

+define(`QREG', `ifelse(
+	$1, d0, q0,
+	$1, d2, q1,
+	$1, d4, q2,
+	$1, d6, q3,
+	$1, d8, q4,
+	$1, d10, q5,
+	$1, d12, q6,
+	$1, d14, q7,
+	$1, d16, q8,
+	$1, d18, q9,
+	$1, d20, q10,
+	$1, d22, q11,
+	$1, d24, q12,
+	$1, d26, q13,
+	$1, d28, q14,
+	$1, d30, q15,
+	`NO REGISTER')')dnl
+
+define(`D0REG', `ifelse(
+	$1, q0, d0,
+	$1, q1, d2,
+	$1, q2, d4,
+	$1, q3, d6,
+	$1, q4, d8,
+	$1, q5, d10,
+	$1, q6, d12,
+	$1, q7, d14,
+	$1, q8, d16,
+	$1, q9, d18,
+	$1, q10, d20,
+	$1, q11, d22,
+	$1, q12, d24,
+	$1, q13, d26,
+	$1, q14, d28,
+	$1, q15, d30,
+	`NO REGISTER')')dnl
+
+define(`D1REG', `ifelse(
+	$1, q0, d1,
+	$1, q1, d3,
+	$1, q2, d5,
+	$1, q3, d7,
+	$1, q4, d9,
+	$1, q5, d11,
+	$1, q6, d13,
+	$1, q7, d15,
+	$1, q8, d17,
+	$1, q9, d19,
+	$1, q10, d21,
+	$1, q11, d23,
+	$1, q12, d25,
+	$1, q13, d27,
+	$1, q14, d29,
+	$1, q15, d31,
+	`NO REGISTER')')dnl

arm/memxor.asm

+C arm/memxor.asm
+
+ifelse(`
+   Copyright (C) 2013 Niels Möller
+
+   This file is part of GNU Nettle.
+
+   GNU Nettle is free software: you can redistribute it and/or
+   modify it under the terms of either:
+
+     * the GNU Lesser General Public License as published by the Free
+       Software Foundation; either version 3 of the License, or (at your
+       option) any later version.
+
+   or
+
+     * the GNU General Public License as published by the Free
+       Software Foundation; either version 2 of the License, or (at your
+       option) any later version.
+
+   or both in parallel, as here.
+
+   GNU Nettle is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   General Public License for more details.
+
+   You should have received copies of the GNU General Public License and
+   the GNU Lesser General Public License along with this program.  If
+   not, see http://www.gnu.org/licenses/.
+')
+
+C Possible speedups:
+C
+C The ldm instruction can do load two registers per cycle,
+C if the address is two-word aligned. Or three registers in two
+C cycles, regardless of alignment.
+
+C Register usage:
+
+define(`DST', `r0')
+define(`SRC', `r1')
+define(`N', `r2')
+define(`CNT', `r6')
+define(`TNC', `r12')
+
+C little-endian and big-endian need to shift in different directions for
+C alignment correction
+define(`S0ADJ', IF_LE(`lsr', `lsl'))
+define(`S1ADJ', IF_LE(`lsl', `lsr'))
+
+	.syntax unified
+
+	.file "memxor.asm"
+
+	.text
+	.arm
+
+	C memxor(void *dst, const void *src, size_t n)
+	.align 4
+PROLOGUE(nettle_memxor)
+	cmp	N, #0
+	beq	.Lmemxor_done
+
+	cmp	N, #7
+	bcs	.Lmemxor_large
+
+	C Simple byte loop
+.Lmemxor_bytes:
+	ldrb	r3, [SRC], #+1
+	ldrb	r12, [DST]
+	eor	r3, r12
+	strb	r3, [DST], #+1
+	subs	N, #1
+	bne	.Lmemxor_bytes
+
+.Lmemxor_done:
+	bx	lr
+
+.Lmemxor_align_loop:
+	ldrb	r3, [SRC], #+1
+	ldrb	r12, [DST]
+	eor	r3, r12
+	strb	r3, [DST], #+1
+	sub	N, #1
+
+.Lmemxor_large:
+	tst	DST, #3
+	bne	.Lmemxor_align_loop
+
+	C We have at least 4 bytes left to do here.
+	sub	N, #4
+
+	ands	r3, SRC, #3
+	beq	.Lmemxor_same
+
+	C Different alignment case.
+	C     v original SRC
+	C +-------+------+
+	C |SRC    |SRC+4 |
+	C +---+---+------+
+	C     |DST    |
+	C     +-------+
+	C
+	C With little-endian, we need to do
+	C DST[i] ^= (SRC[i] >> CNT) ^ (SRC[i+1] << TNC)
+	C With big-endian, we need to do
+	C DST[i] ^= (SRC[i] << CNT) ^ (SRC[i+1] >> TNC)
+
+	push	{r4,r5,r6}
+	
+	lsl	CNT, r3, #3
+	bic	SRC, #3
+	rsb	TNC, CNT, #32
+
+	ldr	r4, [SRC], #+4
+
+	tst	N, #4
+	itet	eq
+	moveq	r5, r4
+	subne	N, #4
+	beq	.Lmemxor_odd
+
+.Lmemxor_word_loop:
+	ldr	r5, [SRC], #+4
+	ldr	r3, [DST]
+	eor	r3, r3, r4, S0ADJ CNT
+	eor	r3, r3, r5, S1ADJ TNC
+	str	r3, [DST], #+4
+.Lmemxor_odd:
+	ldr	r4, [SRC], #+4
+	ldr	r3, [DST]
+	eor	r3, r3, r5, S0ADJ CNT
+	eor	r3, r3, r4, S1ADJ TNC
+	str	r3, [DST], #+4
+	subs	N, #8
+	bcs	.Lmemxor_word_loop
+	adds	N, #8
+	beq	.Lmemxor_odd_done
+
+	C We have TNC/8 left-over bytes in r4, high end on LE and low end on
+	C BE, excess bits to be discarded by alignment adjustment at the other
+	S0ADJ	r4, CNT
+	C now byte-aligned at low end on LE and high end on BE
+	ldr	r3, [DST]
+	eor	r3, r4
+
+	pop	{r4,r5,r6}
+
+	C Store bytes, one by one.
+.Lmemxor_leftover:
+	C bring uppermost byte down for saving while preserving lower ones
+IF_BE(`	ror	r3, #24')
+	strb	r3, [DST], #+1
+	subs	N, #1
+	beq	.Lmemxor_done
+	subs	TNC, #8
+	C bring down next byte, no need to preserve
+IF_LE(`	lsr	r3, #8')
+	bne	.Lmemxor_leftover
+	b	.Lmemxor_bytes
+.Lmemxor_odd_done:
+	pop	{r4,r5,r6}
+	bx	lr
+
+.Lmemxor_same:
+	push	{r4,r5,r6,r7,r8,r10,r11,r14}	C lr is the link register
+
+	subs	N, #8
+	bcc	.Lmemxor_same_end
+
+	ldmia	SRC!, {r3, r4, r5}
+	C Keep address for loads in r14
+	mov	r14, DST
+	ldmia	r14!, {r6, r7, r8}
+	subs	N, #12
+	eor	r10, r3, r6
+	eor	r11, r4, r7
+	eor	r12, r5, r8
+	bcc	.Lmemxor_same_final_store
+	subs	N, #12
+	ldmia	r14!, {r6, r7, r8}
+	bcc	.Lmemxor_same_wind_down
+
+	C 6 cycles per iteration, 0.50 cycles/byte. For this speed,
+	C loop starts at offset 0x11c in the object file.
+
+.Lmemxor_same_loop:
+	C r10-r12 contains values to be stored at DST
+	C r6-r8 contains values read from r14, in advance
+	ldmia	SRC!, {r3, r4, r5}
+	subs	N, #12
+	stmia	DST!, {r10, r11, r12}
+	eor	r10, r3, r6
+	eor	r11, r4, r7
+	eor	r12, r5, r8
+	ldmia	r14!, {r6, r7, r8}
+	bcs	.Lmemxor_same_loop
+
+.Lmemxor_same_wind_down:
+	C Wind down code
+	ldmia	SRC!, {r3, r4, r5}
+	stmia	DST!, {r10, r11, r12}
+	eor	r10, r3, r6
+	eor	r11, r4, r7
+	eor	r12, r5, r8
+.Lmemxor_same_final_store:
+	stmia	DST!, {r10, r11, r12}
+	
+.Lmemxor_same_end:
+	C We have 0-11 bytes left to do, and N holds number of bytes -12.
+	adds	N, #4
+	bcc	.Lmemxor_same_lt_8
+	C Do 8 bytes more, leftover is in N
+	ldmia	SRC!, {r3, r4}
+	ldmia	DST, {r6, r7}
+	eor	r3, r6
+	eor	r4, r7
+	stmia	DST!, {r3, r4}
+	pop	{r4,r5,r6,r7,r8,r10,r11,r14}
+	beq	.Lmemxor_done
+	b	.Lmemxor_bytes
+
+.Lmemxor_same_lt_8:
+	pop	{r4,r5,r6,r7,r8,r10,r11,r14}
+	adds	N, #4
+	bcc	.Lmemxor_same_lt_4
+
+	ldr	r3, [SRC], #+4
+	ldr	r12, [DST]
+	eor	r3, r12
+	str	r3, [DST], #+4
+	beq	.Lmemxor_done
+	b	.Lmemxor_bytes
+
+.Lmemxor_same_lt_4:
+	adds	N, #4
+	beq	.Lmemxor_done
+	b	.Lmemxor_bytes
+	
+EPILOGUE(nettle_memxor)