visaljs 0.0.2 → 0.0.3

package/runtime/third_party/boringssl/gen/crypto/chacha20_poly1305_armv8-win.S

@@ -0,0 +1,3014 @@
+// This file is generated from a similarly-named Perl script in the BoringSSL
+// source tree. Do not edit by hand.
+
+#include <openssl/asm_base.h>
+
+#if !defined(OPENSSL_NO_ASM) && defined(OPENSSL_AARCH64) && defined(_WIN32)
+.section	.rodata
+
+.align	7
+chacha20_poly1305_constants:
+Lchacha20_consts:
+.byte	'e','x','p','a','n','d',' ','3','2','-','b','y','t','e',' ','k'
+Linc:
+.long	1,2,3,4
+Lrol8:
+.byte	3,0,1,2, 7,4,5,6, 11,8,9,10, 15,12,13,14
+Lclamp:
+.quad	0x0FFFFFFC0FFFFFFF, 0x0FFFFFFC0FFFFFFC
+
+.text
+
+.def Lpoly_hash_ad_internal
+   .type 32
+.endef
+.align	6
+chacha20_poly1305_helpers:
+Lpoly_hash_ad_internal:
+.cfi_startproc
+	cbnz	x4, Lpoly_hash_intro
+	ret
+
+Lpoly_hash_intro:
+	cmp	x4, #16
+	b.lt	Lpoly_hash_ad_tail
+	ldp	x11, x12, [x3], 16
+	adds	x8, x8, x11
+	adcs	x9, x9, x12
+	adc	x10, x10, x15
+	mul	x11, x8, x16     // [t2:t1:t0] = [acc2:acc1:acc0] * r0
+	umulh	x12, x8, x16
+	mul	x13, x9, x16
+	umulh	x14, x9, x16
+	adds	x12, x12, x13
+	mul	x13, x10, x16
+	adc	x13, x13, x14
+	mul	x14, x8, x17       // [t3:t2:t1:t0] = [acc2:acc1:acc0] * [r1:r0]
+	umulh	x8, x8, x17
+	adds	x12, x12, x14
+	mul	x14, x9, x17
+	umulh	x9, x9, x17
+	adcs	x14, x14, x8
+	mul	x10, x10, x17
+	adc	x10, x10, x9
+	adds	x13, x13, x14
+	adc	x14, x10, xzr
+	and	x10, x13, #3         // At this point acc2 is 2 bits at most (value of 3)
+	and	x8, x13, #-4
+	extr	x13, x14, x13, #2
+	adds	x8, x8, x11
+	lsr	x11, x14, #2
+	adc	x9, x14, x11        // No carry out since t0 is 61 bits and t3 is 63 bits
+	adds	x8, x8, x13
+	adcs	x9, x9, x12
+	adc	x10, x10, xzr      // At this point acc2 has the value of 4 at most
+	sub	x4, x4, #16
+	b	Lpoly_hash_ad_internal
+
+Lpoly_hash_ad_tail:
+	cbz	x4, Lpoly_hash_ad_ret
+
+	eor	v20.16b, v20.16b, v20.16b // Use T0 to load the AAD
+	sub	x4, x4, #1
+
+Lpoly_hash_tail_16_compose:
+	ext	v20.16b, v20.16b, v20.16b, #15
+	ldrb	w11, [x3, x4]
+	mov	v20.b[0], w11
+	subs	x4, x4, #1
+	b.ge	Lpoly_hash_tail_16_compose
+	mov	x11, v20.d[0]
+	mov	x12, v20.d[1]
+	adds	x8, x8, x11
+	adcs	x9, x9, x12
+	adc	x10, x10, x15
+	mul	x11, x8, x16     // [t2:t1:t0] = [acc2:acc1:acc0] * r0
+	umulh	x12, x8, x16
+	mul	x13, x9, x16
+	umulh	x14, x9, x16
+	adds	x12, x12, x13
+	mul	x13, x10, x16
+	adc	x13, x13, x14
+	mul	x14, x8, x17       // [t3:t2:t1:t0] = [acc2:acc1:acc0] * [r1:r0]
+	umulh	x8, x8, x17
+	adds	x12, x12, x14
+	mul	x14, x9, x17
+	umulh	x9, x9, x17
+	adcs	x14, x14, x8
+	mul	x10, x10, x17
+	adc	x10, x10, x9
+	adds	x13, x13, x14
+	adc	x14, x10, xzr
+	and	x10, x13, #3         // At this point acc2 is 2 bits at most (value of 3)
+	and	x8, x13, #-4
+	extr	x13, x14, x13, #2
+	adds	x8, x8, x11
+	lsr	x11, x14, #2
+	adc	x9, x14, x11        // No carry out since t0 is 61 bits and t3 is 63 bits
+	adds	x8, x8, x13
+	adcs	x9, x9, x12
+	adc	x10, x10, xzr      // At this point acc2 has the value of 4 at most
+
+Lpoly_hash_ad_ret:
+	ret
+.cfi_endproc
+
+
+/////////////////////////////////
+//
+// void chacha20_poly1305_seal(uint8_t *pt, uint8_t *ct, size_t len_in, uint8_t *ad, size_t len_ad, union open_data *seal_data);
+//
+.globl	chacha20_poly1305_seal
+
+.def chacha20_poly1305_seal
+   .type 32
+.endef
+.align	6
+chacha20_poly1305_seal:
+	AARCH64_SIGN_LINK_REGISTER
+.cfi_startproc
+	stp	x29, x30, [sp, #-80]!
+.cfi_def_cfa_offset	80
+.cfi_offset	w30, -72
+.cfi_offset	w29, -80
+	mov	x29, sp
+    // We probably could do .cfi_def_cfa w29, 80 at this point, but since
+    // we don't actually use the frame pointer like that, it's probably not
+    // worth bothering.
+	stp	d8, d9, [sp, #16]
+	stp	d10, d11, [sp, #32]
+	stp	d12, d13, [sp, #48]
+	stp	d14, d15, [sp, #64]
+.cfi_offset	b15, -8
+.cfi_offset	b14, -16
+.cfi_offset	b13, -24
+.cfi_offset	b12, -32
+.cfi_offset	b11, -40
+.cfi_offset	b10, -48
+.cfi_offset	b9, -56
+.cfi_offset	b8, -64
+
+	adrp	x11, Lchacha20_consts
+	add	x11, x11, :lo12:Lchacha20_consts
+
+	ld1	{v24.16b - v27.16b}, [x11] // Load the CONSTS, INC, ROL8 and CLAMP values
+	ld1	{v28.16b - v30.16b}, [x5]
+
+	mov	x15, #1 // Prepare the Poly1305 state
+	mov	x8, #0
+	mov	x9, #0
+	mov	x10, #0
+
+	ldr	x12, [x5, #56]   // The total cipher text length includes extra_in_len
+	add	x12, x12, x2
+	mov	v31.d[0], x4  // Store the input and aad lengths
+	mov	v31.d[1], x12
+
+	cmp	x2, #128
+	b.le	Lseal_128 // Optimization for smaller buffers
+
+    // Initially we prepare 5 ChaCha20 blocks. Four to encrypt up to 4 blocks (256 bytes) of plaintext,
+    // and one for the Poly1305 R and S keys. The first four blocks (A0-A3..D0-D3) are computed vertically,
+    // the fifth block (A4-D4) horizontally.
+	ld4r	{v0.4s,v1.4s,v2.4s,v3.4s}, [x11]
+	mov	v4.16b, v24.16b
+
+	ld4r	{v5.4s,v6.4s,v7.4s,v8.4s}, [x5], #16
+	mov	v9.16b, v28.16b
+
+	ld4r	{v10.4s,v11.4s,v12.4s,v13.4s}, [x5], #16
+	mov	v14.16b, v29.16b
+
+	ld4r	{v15.4s,v16.4s,v17.4s,v18.4s}, [x5]
+	add	v15.4s, v15.4s, v25.4s
+	mov	v19.16b, v30.16b
+
+	sub	x5, x5, #32
+
+	mov	x6, #10
+
+.align	5
+Lseal_init_rounds:
+	add	v0.4s, v0.4s, v5.4s
+	add	v1.4s, v1.4s, v6.4s
+	add	v2.4s, v2.4s, v7.4s
+	add	v3.4s, v3.4s, v8.4s
+	add	v4.4s, v4.4s, v9.4s
+
+	eor	v15.16b, v15.16b, v0.16b
+	eor	v16.16b, v16.16b, v1.16b
+	eor	v17.16b, v17.16b, v2.16b
+	eor	v18.16b, v18.16b, v3.16b
+	eor	v19.16b, v19.16b, v4.16b
+
+	rev32	v15.8h, v15.8h
+	rev32	v16.8h, v16.8h
+	rev32	v17.8h, v17.8h
+	rev32	v18.8h, v18.8h
+	rev32	v19.8h, v19.8h
+
+	add	v10.4s, v10.4s, v15.4s
+	add	v11.4s, v11.4s, v16.4s
+	add	v12.4s, v12.4s, v17.4s
+	add	v13.4s, v13.4s, v18.4s
+	add	v14.4s, v14.4s, v19.4s
+
+	eor	v5.16b, v5.16b, v10.16b
+	eor	v6.16b, v6.16b, v11.16b
+	eor	v7.16b, v7.16b, v12.16b
+	eor	v8.16b, v8.16b, v13.16b
+	eor	v9.16b, v9.16b, v14.16b
+
+	ushr	v20.4s, v5.4s, #20
+	sli	v20.4s, v5.4s, #12
+	ushr	v5.4s, v6.4s, #20
+	sli	v5.4s, v6.4s, #12
+	ushr	v6.4s, v7.4s, #20
+	sli	v6.4s, v7.4s, #12
+	ushr	v7.4s, v8.4s, #20
+	sli	v7.4s, v8.4s, #12
+	ushr	v8.4s, v9.4s, #20
+	sli	v8.4s, v9.4s, #12
+
+	add	v0.4s, v0.4s, v20.4s
+	add	v1.4s, v1.4s, v5.4s
+	add	v2.4s, v2.4s, v6.4s
+	add	v3.4s, v3.4s, v7.4s
+	add	v4.4s, v4.4s, v8.4s
+
+	eor	v15.16b, v15.16b, v0.16b
+	eor	v16.16b, v16.16b, v1.16b
+	eor	v17.16b, v17.16b, v2.16b
+	eor	v18.16b, v18.16b, v3.16b
+	eor	v19.16b, v19.16b, v4.16b
+
+	tbl	v15.16b, {v15.16b}, v26.16b
+	tbl	v16.16b, {v16.16b}, v26.16b
+	tbl	v17.16b, {v17.16b}, v26.16b
+	tbl	v18.16b, {v18.16b}, v26.16b
+	tbl	v19.16b, {v19.16b}, v26.16b
+
+	add	v10.4s, v10.4s, v15.4s
+	add	v11.4s, v11.4s, v16.4s
+	add	v12.4s, v12.4s, v17.4s
+	add	v13.4s, v13.4s, v18.4s
+	add	v14.4s, v14.4s, v19.4s
+
+	eor	v20.16b, v20.16b, v10.16b
+	eor	v5.16b, v5.16b, v11.16b
+	eor	v6.16b, v6.16b, v12.16b
+	eor	v7.16b, v7.16b, v13.16b
+	eor	v8.16b, v8.16b, v14.16b
+
+	ushr	v9.4s, v8.4s, #25
+	sli	v9.4s, v8.4s, #7
+	ushr	v8.4s, v7.4s, #25
+	sli	v8.4s, v7.4s, #7
+	ushr	v7.4s, v6.4s, #25
+	sli	v7.4s, v6.4s, #7
+	ushr	v6.4s, v5.4s, #25
+	sli	v6.4s, v5.4s, #7
+	ushr	v5.4s, v20.4s, #25
+	sli	v5.4s, v20.4s, #7
+
+	ext	v9.16b, v9.16b, v9.16b, #4
+	ext	v14.16b, v14.16b, v14.16b, #8
+	ext	v19.16b, v19.16b, v19.16b, #12
+	add	v0.4s, v0.4s, v6.4s
+	add	v1.4s, v1.4s, v7.4s
+	add	v2.4s, v2.4s, v8.4s
+	add	v3.4s, v3.4s, v5.4s
+	add	v4.4s, v4.4s, v9.4s
+
+	eor	v18.16b, v18.16b, v0.16b
+	eor	v15.16b, v15.16b, v1.16b
+	eor	v16.16b, v16.16b, v2.16b
+	eor	v17.16b, v17.16b, v3.16b
+	eor	v19.16b, v19.16b, v4.16b
+
+	rev32	v18.8h, v18.8h
+	rev32	v15.8h, v15.8h
+	rev32	v16.8h, v16.8h
+	rev32	v17.8h, v17.8h
+	rev32	v19.8h, v19.8h
+
+	add	v12.4s, v12.4s, v18.4s
+	add	v13.4s, v13.4s, v15.4s
+	add	v10.4s, v10.4s, v16.4s
+	add	v11.4s, v11.4s, v17.4s
+	add	v14.4s, v14.4s, v19.4s
+
+	eor	v6.16b, v6.16b, v12.16b
+	eor	v7.16b, v7.16b, v13.16b
+	eor	v8.16b, v8.16b, v10.16b
+	eor	v5.16b, v5.16b, v11.16b
+	eor	v9.16b, v9.16b, v14.16b
+
+	ushr	v20.4s, v6.4s, #20
+	sli	v20.4s, v6.4s, #12
+	ushr	v6.4s, v7.4s, #20
+	sli	v6.4s, v7.4s, #12
+	ushr	v7.4s, v8.4s, #20
+	sli	v7.4s, v8.4s, #12
+	ushr	v8.4s, v5.4s, #20
+	sli	v8.4s, v5.4s, #12
+	ushr	v5.4s, v9.4s, #20
+	sli	v5.4s, v9.4s, #12
+
+	add	v0.4s, v0.4s, v20.4s
+	add	v1.4s, v1.4s, v6.4s
+	add	v2.4s, v2.4s, v7.4s
+	add	v3.4s, v3.4s, v8.4s
+	add	v4.4s, v4.4s, v5.4s
+
+	eor	v18.16b, v18.16b, v0.16b
+	eor	v15.16b, v15.16b, v1.16b
+	eor	v16.16b, v16.16b, v2.16b
+	eor	v17.16b, v17.16b, v3.16b
+	eor	v19.16b, v19.16b, v4.16b
+
+	tbl	v18.16b, {v18.16b}, v26.16b
+	tbl	v15.16b, {v15.16b}, v26.16b
+	tbl	v16.16b, {v16.16b}, v26.16b
+	tbl	v17.16b, {v17.16b}, v26.16b
+	tbl	v19.16b, {v19.16b}, v26.16b
+
+	add	v12.4s, v12.4s, v18.4s
+	add	v13.4s, v13.4s, v15.4s
+	add	v10.4s, v10.4s, v16.4s
+	add	v11.4s, v11.4s, v17.4s
+	add	v14.4s, v14.4s, v19.4s
+
+	eor	v20.16b, v20.16b, v12.16b
+	eor	v6.16b, v6.16b, v13.16b
+	eor	v7.16b, v7.16b, v10.16b
+	eor	v8.16b, v8.16b, v11.16b
+	eor	v5.16b, v5.16b, v14.16b
+
+	ushr	v9.4s, v5.4s, #25
+	sli	v9.4s, v5.4s, #7
+	ushr	v5.4s, v8.4s, #25
+	sli	v5.4s, v8.4s, #7
+	ushr	v8.4s, v7.4s, #25
+	sli	v8.4s, v7.4s, #7
+	ushr	v7.4s, v6.4s, #25
+	sli	v7.4s, v6.4s, #7
+	ushr	v6.4s, v20.4s, #25
+	sli	v6.4s, v20.4s, #7
+
+	ext	v9.16b, v9.16b, v9.16b, #12
+	ext	v14.16b, v14.16b, v14.16b, #8
+	ext	v19.16b, v19.16b, v19.16b, #4
+	subs	x6, x6, #1
+	b.hi	Lseal_init_rounds
+
+	add	v15.4s, v15.4s, v25.4s
+	mov	x11, #4
+	dup	v20.4s, w11
+	add	v25.4s, v25.4s, v20.4s
+
+	zip1	v20.4s, v0.4s, v1.4s
+	zip2	v21.4s, v0.4s, v1.4s
+	zip1	v22.4s, v2.4s, v3.4s
+	zip2	v23.4s, v2.4s, v3.4s
+
+	zip1	v0.2d, v20.2d, v22.2d
+	zip2	v1.2d, v20.2d, v22.2d
+	zip1	v2.2d, v21.2d, v23.2d
+	zip2	v3.2d, v21.2d, v23.2d
+
+	zip1	v20.4s, v5.4s, v6.4s
+	zip2	v21.4s, v5.4s, v6.4s
+	zip1	v22.4s, v7.4s, v8.4s
+	zip2	v23.4s, v7.4s, v8.4s
+
+	zip1	v5.2d, v20.2d, v22.2d
+	zip2	v6.2d, v20.2d, v22.2d
+	zip1	v7.2d, v21.2d, v23.2d
+	zip2	v8.2d, v21.2d, v23.2d
+
+	zip1	v20.4s, v10.4s, v11.4s
+	zip2	v21.4s, v10.4s, v11.4s
+	zip1	v22.4s, v12.4s, v13.4s
+	zip2	v23.4s, v12.4s, v13.4s
+
+	zip1	v10.2d, v20.2d, v22.2d
+	zip2	v11.2d, v20.2d, v22.2d
+	zip1	v12.2d, v21.2d, v23.2d
+	zip2	v13.2d, v21.2d, v23.2d
+
+	zip1	v20.4s, v15.4s, v16.4s
+	zip2	v21.4s, v15.4s, v16.4s
+	zip1	v22.4s, v17.4s, v18.4s
+	zip2	v23.4s, v17.4s, v18.4s
+
+	zip1	v15.2d, v20.2d, v22.2d
+	zip2	v16.2d, v20.2d, v22.2d
+	zip1	v17.2d, v21.2d, v23.2d
+	zip2	v18.2d, v21.2d, v23.2d
+
+	add	v4.4s, v4.4s, v24.4s
+	add	v9.4s, v9.4s, v28.4s
+	and	v4.16b, v4.16b, v27.16b
+
+	add	v0.4s, v0.4s, v24.4s
+	add	v5.4s, v5.4s, v28.4s
+	add	v10.4s, v10.4s, v29.4s
+	add	v15.4s, v15.4s, v30.4s
+
+	add	v1.4s, v1.4s, v24.4s
+	add	v6.4s, v6.4s, v28.4s
+	add	v11.4s, v11.4s, v29.4s
+	add	v16.4s, v16.4s, v30.4s
+
+	add	v2.4s, v2.4s, v24.4s
+	add	v7.4s, v7.4s, v28.4s
+	add	v12.4s, v12.4s, v29.4s
+	add	v17.4s, v17.4s, v30.4s
+
+	add	v3.4s, v3.4s, v24.4s
+	add	v8.4s, v8.4s, v28.4s
+	add	v13.4s, v13.4s, v29.4s
+	add	v18.4s, v18.4s, v30.4s
+
+	mov	x16, v4.d[0] // Move the R key to GPRs
+	mov	x17, v4.d[1]
+	mov	v27.16b, v9.16b // Store the S key
+
+	bl	Lpoly_hash_ad_internal
+
+	mov	x3, x0
+	cmp	x2, #256
+	b.le	Lseal_tail
+
+	ld1	{v20.16b - v23.16b}, [x1], #64
+	eor	v20.16b, v20.16b, v0.16b
+	eor	v21.16b, v21.16b, v5.16b
+	eor	v22.16b, v22.16b, v10.16b
+	eor	v23.16b, v23.16b, v15.16b
+	st1	{v20.16b - v23.16b}, [x0], #64
+
+	ld1	{v20.16b - v23.16b}, [x1], #64
+	eor	v20.16b, v20.16b, v1.16b
+	eor	v21.16b, v21.16b, v6.16b
+	eor	v22.16b, v22.16b, v11.16b
+	eor	v23.16b, v23.16b, v16.16b
+	st1	{v20.16b - v23.16b}, [x0], #64
+
+	ld1	{v20.16b - v23.16b}, [x1], #64
+	eor	v20.16b, v20.16b, v2.16b
+	eor	v21.16b, v21.16b, v7.16b
+	eor	v22.16b, v22.16b, v12.16b
+	eor	v23.16b, v23.16b, v17.16b
+	st1	{v20.16b - v23.16b}, [x0], #64
+
+	ld1	{v20.16b - v23.16b}, [x1], #64
+	eor	v20.16b, v20.16b, v3.16b
+	eor	v21.16b, v21.16b, v8.16b
+	eor	v22.16b, v22.16b, v13.16b
+	eor	v23.16b, v23.16b, v18.16b
+	st1	{v20.16b - v23.16b}, [x0], #64
+
+	sub	x2, x2, #256
+
+	mov	x6, #4 // In the first run of the loop we need to hash 256 bytes, therefore we hash one block for the first 4 rounds
+	mov	x7, #6 // and two blocks for the remaining 6, for a total of (1 * 4 + 2 * 6) * 16 = 256
+
+Lseal_main_loop:
+	adrp	x11, Lchacha20_consts
+	add	x11, x11, :lo12:Lchacha20_consts
+
+	ld4r	{v0.4s,v1.4s,v2.4s,v3.4s}, [x11]
+	mov	v4.16b, v24.16b
+
+	ld4r	{v5.4s,v6.4s,v7.4s,v8.4s}, [x5], #16
+	mov	v9.16b, v28.16b
+
+	ld4r	{v10.4s,v11.4s,v12.4s,v13.4s}, [x5], #16
+	mov	v14.16b, v29.16b
+
+	ld4r	{v15.4s,v16.4s,v17.4s,v18.4s}, [x5]
+	add	v15.4s, v15.4s, v25.4s
+	mov	v19.16b, v30.16b
+
+	eor	v20.16b, v20.16b, v20.16b //zero
+	not	v21.16b, v20.16b // -1
+	sub	v21.4s, v25.4s, v21.4s // Add +1
+	ext	v20.16b, v21.16b, v20.16b, #12 // Get the last element (counter)
+	add	v19.4s, v19.4s, v20.4s
+
+	sub	x5, x5, #32
+.align	5
+Lseal_main_loop_rounds:
+	add	v0.4s, v0.4s, v5.4s
+	add	v1.4s, v1.4s, v6.4s
+	add	v2.4s, v2.4s, v7.4s
+	add	v3.4s, v3.4s, v8.4s
+	add	v4.4s, v4.4s, v9.4s
+
+	eor	v15.16b, v15.16b, v0.16b
+	eor	v16.16b, v16.16b, v1.16b
+	eor	v17.16b, v17.16b, v2.16b
+	eor	v18.16b, v18.16b, v3.16b
+	eor	v19.16b, v19.16b, v4.16b
+
+	rev32	v15.8h, v15.8h
+	rev32	v16.8h, v16.8h
+	rev32	v17.8h, v17.8h
+	rev32	v18.8h, v18.8h
+	rev32	v19.8h, v19.8h
+
+	add	v10.4s, v10.4s, v15.4s
+	add	v11.4s, v11.4s, v16.4s
+	add	v12.4s, v12.4s, v17.4s
+	add	v13.4s, v13.4s, v18.4s
+	add	v14.4s, v14.4s, v19.4s
+
+	eor	v5.16b, v5.16b, v10.16b
+	eor	v6.16b, v6.16b, v11.16b
+	eor	v7.16b, v7.16b, v12.16b
+	eor	v8.16b, v8.16b, v13.16b
+	eor	v9.16b, v9.16b, v14.16b
+
+	ushr	v20.4s, v5.4s, #20
+	sli	v20.4s, v5.4s, #12
+	ushr	v5.4s, v6.4s, #20
+	sli	v5.4s, v6.4s, #12
+	ushr	v6.4s, v7.4s, #20
+	sli	v6.4s, v7.4s, #12
+	ushr	v7.4s, v8.4s, #20
+	sli	v7.4s, v8.4s, #12
+	ushr	v8.4s, v9.4s, #20
+	sli	v8.4s, v9.4s, #12
+
+	add	v0.4s, v0.4s, v20.4s
+	add	v1.4s, v1.4s, v5.4s
+	add	v2.4s, v2.4s, v6.4s
+	add	v3.4s, v3.4s, v7.4s
+	add	v4.4s, v4.4s, v8.4s
+
+	eor	v15.16b, v15.16b, v0.16b
+	eor	v16.16b, v16.16b, v1.16b
+	eor	v17.16b, v17.16b, v2.16b
+	eor	v18.16b, v18.16b, v3.16b
+	eor	v19.16b, v19.16b, v4.16b
+
+	tbl	v15.16b, {v15.16b}, v26.16b
+	tbl	v16.16b, {v16.16b}, v26.16b
+	tbl	v17.16b, {v17.16b}, v26.16b
+	tbl	v18.16b, {v18.16b}, v26.16b
+	tbl	v19.16b, {v19.16b}, v26.16b
+
+	add	v10.4s, v10.4s, v15.4s
+	add	v11.4s, v11.4s, v16.4s
+	add	v12.4s, v12.4s, v17.4s
+	add	v13.4s, v13.4s, v18.4s
+	add	v14.4s, v14.4s, v19.4s
+
+	eor	v20.16b, v20.16b, v10.16b
+	eor	v5.16b, v5.16b, v11.16b
+	eor	v6.16b, v6.16b, v12.16b
+	eor	v7.16b, v7.16b, v13.16b
+	eor	v8.16b, v8.16b, v14.16b
+
+	ushr	v9.4s, v8.4s, #25
+	sli	v9.4s, v8.4s, #7
+	ushr	v8.4s, v7.4s, #25
+	sli	v8.4s, v7.4s, #7
+	ushr	v7.4s, v6.4s, #25
+	sli	v7.4s, v6.4s, #7
+	ushr	v6.4s, v5.4s, #25
+	sli	v6.4s, v5.4s, #7
+	ushr	v5.4s, v20.4s, #25
+	sli	v5.4s, v20.4s, #7
+
+	ext	v9.16b, v9.16b, v9.16b, #4
+	ext	v14.16b, v14.16b, v14.16b, #8
+	ext	v19.16b, v19.16b, v19.16b, #12
+	ldp	x11, x12, [x3], 16
+	adds	x8, x8, x11
+	adcs	x9, x9, x12
+	adc	x10, x10, x15
+	mul	x11, x8, x16     // [t2:t1:t0] = [acc2:acc1:acc0] * r0
+	umulh	x12, x8, x16
+	mul	x13, x9, x16
+	umulh	x14, x9, x16
+	adds	x12, x12, x13
+	mul	x13, x10, x16
+	adc	x13, x13, x14
+	mul	x14, x8, x17       // [t3:t2:t1:t0] = [acc2:acc1:acc0] * [r1:r0]
+	umulh	x8, x8, x17
+	adds	x12, x12, x14
+	mul	x14, x9, x17
+	umulh	x9, x9, x17
+	adcs	x14, x14, x8
+	mul	x10, x10, x17
+	adc	x10, x10, x9
+	adds	x13, x13, x14
+	adc	x14, x10, xzr
+	and	x10, x13, #3         // At this point acc2 is 2 bits at most (value of 3)
+	and	x8, x13, #-4
+	extr	x13, x14, x13, #2
+	adds	x8, x8, x11
+	lsr	x11, x14, #2
+	adc	x9, x14, x11        // No carry out since t0 is 61 bits and t3 is 63 bits
+	adds	x8, x8, x13
+	adcs	x9, x9, x12
+	adc	x10, x10, xzr      // At this point acc2 has the value of 4 at most
+	add	v0.4s, v0.4s, v6.4s
+	add	v1.4s, v1.4s, v7.4s
+	add	v2.4s, v2.4s, v8.4s
+	add	v3.4s, v3.4s, v5.4s
+	add	v4.4s, v4.4s, v9.4s
+
+	eor	v18.16b, v18.16b, v0.16b
+	eor	v15.16b, v15.16b, v1.16b
+	eor	v16.16b, v16.16b, v2.16b
+	eor	v17.16b, v17.16b, v3.16b
+	eor	v19.16b, v19.16b, v4.16b
+
+	rev32	v18.8h, v18.8h
+	rev32	v15.8h, v15.8h
+	rev32	v16.8h, v16.8h
+	rev32	v17.8h, v17.8h
+	rev32	v19.8h, v19.8h
+
+	add	v12.4s, v12.4s, v18.4s
+	add	v13.4s, v13.4s, v15.4s
+	add	v10.4s, v10.4s, v16.4s
+	add	v11.4s, v11.4s, v17.4s
+	add	v14.4s, v14.4s, v19.4s
+
+	eor	v6.16b, v6.16b, v12.16b
+	eor	v7.16b, v7.16b, v13.16b
+	eor	v8.16b, v8.16b, v10.16b
+	eor	v5.16b, v5.16b, v11.16b
+	eor	v9.16b, v9.16b, v14.16b
+
+	ushr	v20.4s, v6.4s, #20
+	sli	v20.4s, v6.4s, #12
+	ushr	v6.4s, v7.4s, #20
+	sli	v6.4s, v7.4s, #12
+	ushr	v7.4s, v8.4s, #20
+	sli	v7.4s, v8.4s, #12
+	ushr	v8.4s, v5.4s, #20
+	sli	v8.4s, v5.4s, #12
+	ushr	v5.4s, v9.4s, #20
+	sli	v5.4s, v9.4s, #12
+
+	add	v0.4s, v0.4s, v20.4s
+	add	v1.4s, v1.4s, v6.4s
+	add	v2.4s, v2.4s, v7.4s
+	add	v3.4s, v3.4s, v8.4s
+	add	v4.4s, v4.4s, v5.4s
+
+	eor	v18.16b, v18.16b, v0.16b
+	eor	v15.16b, v15.16b, v1.16b
+	eor	v16.16b, v16.16b, v2.16b
+	eor	v17.16b, v17.16b, v3.16b
+	eor	v19.16b, v19.16b, v4.16b
+
+	tbl	v18.16b, {v18.16b}, v26.16b
+	tbl	v15.16b, {v15.16b}, v26.16b
+	tbl	v16.16b, {v16.16b}, v26.16b
+	tbl	v17.16b, {v17.16b}, v26.16b
+	tbl	v19.16b, {v19.16b}, v26.16b
+
+	add	v12.4s, v12.4s, v18.4s
+	add	v13.4s, v13.4s, v15.4s
+	add	v10.4s, v10.4s, v16.4s
+	add	v11.4s, v11.4s, v17.4s
+	add	v14.4s, v14.4s, v19.4s
+
+	eor	v20.16b, v20.16b, v12.16b
+	eor	v6.16b, v6.16b, v13.16b
+	eor	v7.16b, v7.16b, v10.16b
+	eor	v8.16b, v8.16b, v11.16b
+	eor	v5.16b, v5.16b, v14.16b
+
+	ushr	v9.4s, v5.4s, #25
+	sli	v9.4s, v5.4s, #7
+	ushr	v5.4s, v8.4s, #25
+	sli	v5.4s, v8.4s, #7
+	ushr	v8.4s, v7.4s, #25
+	sli	v8.4s, v7.4s, #7
+	ushr	v7.4s, v6.4s, #25
+	sli	v7.4s, v6.4s, #7
+	ushr	v6.4s, v20.4s, #25
+	sli	v6.4s, v20.4s, #7
+
+	ext	v9.16b, v9.16b, v9.16b, #12
+	ext	v14.16b, v14.16b, v14.16b, #8
+	ext	v19.16b, v19.16b, v19.16b, #4
+	subs	x6, x6, #1
+	b.ge	Lseal_main_loop_rounds
+	ldp	x11, x12, [x3], 16
+	adds	x8, x8, x11
+	adcs	x9, x9, x12
+	adc	x10, x10, x15
+	mul	x11, x8, x16     // [t2:t1:t0] = [acc2:acc1:acc0] * r0
+	umulh	x12, x8, x16
+	mul	x13, x9, x16
+	umulh	x14, x9, x16
+	adds	x12, x12, x13
+	mul	x13, x10, x16
+	adc	x13, x13, x14
+	mul	x14, x8, x17       // [t3:t2:t1:t0] = [acc2:acc1:acc0] * [r1:r0]
+	umulh	x8, x8, x17
+	adds	x12, x12, x14
+	mul	x14, x9, x17
+	umulh	x9, x9, x17
+	adcs	x14, x14, x8
+	mul	x10, x10, x17
+	adc	x10, x10, x9
+	adds	x13, x13, x14
+	adc	x14, x10, xzr
+	and	x10, x13, #3         // At this point acc2 is 2 bits at most (value of 3)
+	and	x8, x13, #-4
+	extr	x13, x14, x13, #2
+	adds	x8, x8, x11
+	lsr	x11, x14, #2
+	adc	x9, x14, x11        // No carry out since t0 is 61 bits and t3 is 63 bits
+	adds	x8, x8, x13
+	adcs	x9, x9, x12
+	adc	x10, x10, xzr      // At this point acc2 has the value of 4 at most
+	subs	x7, x7, #1
+	b.gt	Lseal_main_loop_rounds
+
+	eor	v20.16b, v20.16b, v20.16b //zero
+	not	v21.16b, v20.16b // -1
+	sub	v21.4s, v25.4s, v21.4s // Add +1
+	ext	v20.16b, v21.16b, v20.16b, #12 // Get the last element (counter)
+	add	v19.4s, v19.4s, v20.4s
+
+	add	v15.4s, v15.4s, v25.4s
+	mov	x11, #5
+	dup	v20.4s, w11
+	add	v25.4s, v25.4s, v20.4s
+
+	zip1	v20.4s, v0.4s, v1.4s
+	zip2	v21.4s, v0.4s, v1.4s
+	zip1	v22.4s, v2.4s, v3.4s
+	zip2	v23.4s, v2.4s, v3.4s
+
+	zip1	v0.2d, v20.2d, v22.2d
+	zip2	v1.2d, v20.2d, v22.2d
+	zip1	v2.2d, v21.2d, v23.2d
+	zip2	v3.2d, v21.2d, v23.2d
+
+	zip1	v20.4s, v5.4s, v6.4s
+	zip2	v21.4s, v5.4s, v6.4s
+	zip1	v22.4s, v7.4s, v8.4s
+	zip2	v23.4s, v7.4s, v8.4s
+
+	zip1	v5.2d, v20.2d, v22.2d
+	zip2	v6.2d, v20.2d, v22.2d
+	zip1	v7.2d, v21.2d, v23.2d
+	zip2	v8.2d, v21.2d, v23.2d
+
+	zip1	v20.4s, v10.4s, v11.4s
+	zip2	v21.4s, v10.4s, v11.4s
+	zip1	v22.4s, v12.4s, v13.4s
+	zip2	v23.4s, v12.4s, v13.4s
+
+	zip1	v10.2d, v20.2d, v22.2d
+	zip2	v11.2d, v20.2d, v22.2d
+	zip1	v12.2d, v21.2d, v23.2d
+	zip2	v13.2d, v21.2d, v23.2d
+
+	zip1	v20.4s, v15.4s, v16.4s
+	zip2	v21.4s, v15.4s, v16.4s
+	zip1	v22.4s, v17.4s, v18.4s
+	zip2	v23.4s, v17.4s, v18.4s
+
+	zip1	v15.2d, v20.2d, v22.2d
+	zip2	v16.2d, v20.2d, v22.2d
+	zip1	v17.2d, v21.2d, v23.2d
+	zip2	v18.2d, v21.2d, v23.2d
+
+	add	v0.4s, v0.4s, v24.4s
+	add	v5.4s, v5.4s, v28.4s
+	add	v10.4s, v10.4s, v29.4s
+	add	v15.4s, v15.4s, v30.4s
+
+	add	v1.4s, v1.4s, v24.4s
+	add	v6.4s, v6.4s, v28.4s
+	add	v11.4s, v11.4s, v29.4s
+	add	v16.4s, v16.4s, v30.4s
+
+	add	v2.4s, v2.4s, v24.4s
+	add	v7.4s, v7.4s, v28.4s
+	add	v12.4s, v12.4s, v29.4s
+	add	v17.4s, v17.4s, v30.4s
+
+	add	v3.4s, v3.4s, v24.4s
+	add	v8.4s, v8.4s, v28.4s
+	add	v13.4s, v13.4s, v29.4s
+	add	v18.4s, v18.4s, v30.4s
+
+	add	v4.4s, v4.4s, v24.4s
+	add	v9.4s, v9.4s, v28.4s
+	add	v14.4s, v14.4s, v29.4s
+	add	v19.4s, v19.4s, v30.4s
+
+	cmp	x2, #320
+	b.le	Lseal_tail
+
+	ld1	{v20.16b - v23.16b}, [x1], #64
+	eor	v20.16b, v20.16b, v0.16b
+	eor	v21.16b, v21.16b, v5.16b
+	eor	v22.16b, v22.16b, v10.16b
+	eor	v23.16b, v23.16b, v15.16b
+	st1	{v20.16b - v23.16b}, [x0], #64
+
+	ld1	{v20.16b - v23.16b}, [x1], #64
+	eor	v20.16b, v20.16b, v1.16b
+	eor	v21.16b, v21.16b, v6.16b
+	eor	v22.16b, v22.16b, v11.16b
+	eor	v23.16b, v23.16b, v16.16b
+	st1	{v20.16b - v23.16b}, [x0], #64
+
+	ld1	{v20.16b - v23.16b}, [x1], #64
+	eor	v20.16b, v20.16b, v2.16b
+	eor	v21.16b, v21.16b, v7.16b
+	eor	v22.16b, v22.16b, v12.16b
+	eor	v23.16b, v23.16b, v17.16b
+	st1	{v20.16b - v23.16b}, [x0], #64
+
+	ld1	{v20.16b - v23.16b}, [x1], #64
+	eor	v20.16b, v20.16b, v3.16b
+	eor	v21.16b, v21.16b, v8.16b
+	eor	v22.16b, v22.16b, v13.16b
+	eor	v23.16b, v23.16b, v18.16b
+	st1	{v20.16b - v23.16b}, [x0], #64
+
+	ld1	{v20.16b - v23.16b}, [x1], #64
+	eor	v20.16b, v20.16b, v4.16b
+	eor	v21.16b, v21.16b, v9.16b
+	eor	v22.16b, v22.16b, v14.16b
+	eor	v23.16b, v23.16b, v19.16b
+	st1	{v20.16b - v23.16b}, [x0], #64
+
+	sub	x2, x2, #320
+
+	mov	x6, #0
+	mov	x7, #10 // For the remainder of the loop we always hash and encrypt 320 bytes per iteration
+
+	b	Lseal_main_loop
+
+Lseal_tail:
+    // This part of the function handles the storage and authentication of the last [0,320) bytes
+    // We assume A0-A4 ... D0-D4 hold at least inl (320 max) bytes of the stream data.
+	cmp	x2, #64
+	b.lt	Lseal_tail_64
+
+    // Store and authenticate 64B blocks per iteration
+	ld1	{v20.16b - v23.16b}, [x1], #64
+
+	eor	v20.16b, v20.16b, v0.16b
+	eor	v21.16b, v21.16b, v5.16b
+	eor	v22.16b, v22.16b, v10.16b
+	eor	v23.16b, v23.16b, v15.16b
+	mov	x11, v20.d[0]
+	mov	x12, v20.d[1]
+	adds	x8, x8, x11
+	adcs	x9, x9, x12
+	adc	x10, x10, x15
+	mul	x11, x8, x16     // [t2:t1:t0] = [acc2:acc1:acc0] * r0
+	umulh	x12, x8, x16
+	mul	x13, x9, x16
+	umulh	x14, x9, x16
+	adds	x12, x12, x13
+	mul	x13, x10, x16
+	adc	x13, x13, x14
+	mul	x14, x8, x17       // [t3:t2:t1:t0] = [acc2:acc1:acc0] * [r1:r0]
+	umulh	x8, x8, x17
+	adds	x12, x12, x14
+	mul	x14, x9, x17
+	umulh	x9, x9, x17
+	adcs	x14, x14, x8
+	mul	x10, x10, x17
+	adc	x10, x10, x9
+	adds	x13, x13, x14
+	adc	x14, x10, xzr
+	and	x10, x13, #3         // At this point acc2 is 2 bits at most (value of 3)
+	and	x8, x13, #-4
+	extr	x13, x14, x13, #2
+	adds	x8, x8, x11
+	lsr	x11, x14, #2
+	adc	x9, x14, x11        // No carry out since t0 is 61 bits and t3 is 63 bits
+	adds	x8, x8, x13
+	adcs	x9, x9, x12
+	adc	x10, x10, xzr      // At this point acc2 has the value of 4 at most
+	mov	x11, v21.d[0]
+	mov	x12, v21.d[1]
+	adds	x8, x8, x11
+	adcs	x9, x9, x12
+	adc	x10, x10, x15
+	mul	x11, x8, x16     // [t2:t1:t0] = [acc2:acc1:acc0] * r0
+	umulh	x12, x8, x16
+	mul	x13, x9, x16
+	umulh	x14, x9, x16
+	adds	x12, x12, x13
+	mul	x13, x10, x16
+	adc	x13, x13, x14
+	mul	x14, x8, x17       // [t3:t2:t1:t0] = [acc2:acc1:acc0] * [r1:r0]
+	umulh	x8, x8, x17
+	adds	x12, x12, x14
+	mul	x14, x9, x17
+	umulh	x9, x9, x17
+	adcs	x14, x14, x8
+	mul	x10, x10, x17
+	adc	x10, x10, x9
+	adds	x13, x13, x14
+	adc	x14, x10, xzr
+	and	x10, x13, #3         // At this point acc2 is 2 bits at most (value of 3)
+	and	x8, x13, #-4
+	extr	x13, x14, x13, #2
+	adds	x8, x8, x11
+	lsr	x11, x14, #2
+	adc	x9, x14, x11        // No carry out since t0 is 61 bits and t3 is 63 bits
+	adds	x8, x8, x13
+	adcs	x9, x9, x12
+	adc	x10, x10, xzr      // At this point acc2 has the value of 4 at most
+	mov	x11, v22.d[0]
+	mov	x12, v22.d[1]
+	adds	x8, x8, x11
+	adcs	x9, x9, x12
+	adc	x10, x10, x15
+	mul	x11, x8, x16     // [t2:t1:t0] = [acc2:acc1:acc0] * r0
+	umulh	x12, x8, x16
+	mul	x13, x9, x16
+	umulh	x14, x9, x16
+	adds	x12, x12, x13
+	mul	x13, x10, x16
+	adc	x13, x13, x14
+	mul	x14, x8, x17       // [t3:t2:t1:t0] = [acc2:acc1:acc0] * [r1:r0]
+	umulh	x8, x8, x17
+	adds	x12, x12, x14
+	mul	x14, x9, x17
+	umulh	x9, x9, x17
+	adcs	x14, x14, x8
+	mul	x10, x10, x17
+	adc	x10, x10, x9
+	adds	x13, x13, x14
+	adc	x14, x10, xzr
+	and	x10, x13, #3         // At this point acc2 is 2 bits at most (value of 3)
+	and	x8, x13, #-4
+	extr	x13, x14, x13, #2
+	adds	x8, x8, x11
+	lsr	x11, x14, #2
+	adc	x9, x14, x11        // No carry out since t0 is 61 bits and t3 is 63 bits
+	adds	x8, x8, x13
+	adcs	x9, x9, x12
+	adc	x10, x10, xzr      // At this point acc2 has the value of 4 at most
+	mov	x11, v23.d[0]
+	mov	x12, v23.d[1]
+	adds	x8, x8, x11
+	adcs	x9, x9, x12
+	adc	x10, x10, x15
+	mul	x11, x8, x16     // [t2:t1:t0] = [acc2:acc1:acc0] * r0
+	umulh	x12, x8, x16
+	mul	x13, x9, x16
+	umulh	x14, x9, x16
+	adds	x12, x12, x13
+	mul	x13, x10, x16
+	adc	x13, x13, x14
+	mul	x14, x8, x17       // [t3:t2:t1:t0] = [acc2:acc1:acc0] * [r1:r0]
+	umulh	x8, x8, x17
+	adds	x12, x12, x14
+	mul	x14, x9, x17
+	umulh	x9, x9, x17
+	adcs	x14, x14, x8
+	mul	x10, x10, x17
+	adc	x10, x10, x9
+	adds	x13, x13, x14
+	adc	x14, x10, xzr
+	and	x10, x13, #3         // At this point acc2 is 2 bits at most (value of 3)
+	and	x8, x13, #-4
+	extr	x13, x14, x13, #2
+	adds	x8, x8, x11
+	lsr	x11, x14, #2
+	adc	x9, x14, x11        // No carry out since t0 is 61 bits and t3 is 63 bits
+	adds	x8, x8, x13
+	adcs	x9, x9, x12
+	adc	x10, x10, xzr      // At this point acc2 has the value of 4 at most
+	st1	{v20.16b - v23.16b}, [x0], #64
+	sub	x2, x2, #64
+
+    // Shift the state left by 64 bytes for the next iteration of the loop
+	mov	v0.16b, v1.16b
+	mov	v5.16b, v6.16b
+	mov	v10.16b, v11.16b
+	mov	v15.16b, v16.16b
+
+	mov	v1.16b, v2.16b
+	mov	v6.16b, v7.16b
+	mov	v11.16b, v12.16b
+	mov	v16.16b, v17.16b
+
+	mov	v2.16b, v3.16b
+	mov	v7.16b, v8.16b
+	mov	v12.16b, v13.16b
+	mov	v17.16b, v18.16b
+
+	mov	v3.16b, v4.16b
+	mov	v8.16b, v9.16b
+	mov	v13.16b, v14.16b
+	mov	v18.16b, v19.16b
+
+	b	Lseal_tail
+
+Lseal_tail_64:
+	ldp	x3, x4, [x5, #48] // extra_in_len and extra_in_ptr
+
+    // Here we handle the last [0,64) bytes of plaintext
+	cmp	x2, #16
+	b.lt	Lseal_tail_16
+    // Each iteration encrypt and authenticate a 16B block
+	ld1	{v20.16b}, [x1], #16
+	eor	v20.16b, v20.16b, v0.16b
+	mov	x11, v20.d[0]
+	mov	x12, v20.d[1]
+	adds	x8, x8, x11
+	adcs	x9, x9, x12
+	adc	x10, x10, x15
+	mul	x11, x8, x16     // [t2:t1:t0] = [acc2:acc1:acc0] * r0
+	umulh	x12, x8, x16
+	mul	x13, x9, x16
+	umulh	x14, x9, x16
+	adds	x12, x12, x13
+	mul	x13, x10, x16
+	adc	x13, x13, x14
+	mul	x14, x8, x17       // [t3:t2:t1:t0] = [acc2:acc1:acc0] * [r1:r0]
+	umulh	x8, x8, x17
+	adds	x12, x12, x14
+	mul	x14, x9, x17
+	umulh	x9, x9, x17
+	adcs	x14, x14, x8
+	mul	x10, x10, x17
+	adc	x10, x10, x9
+	adds	x13, x13, x14
+	adc	x14, x10, xzr
+	and	x10, x13, #3         // At this point acc2 is 2 bits at most (value of 3)
+	and	x8, x13, #-4
+	extr	x13, x14, x13, #2
+	adds	x8, x8, x11
+	lsr	x11, x14, #2
+	adc	x9, x14, x11        // No carry out since t0 is 61 bits and t3 is 63 bits
+	adds	x8, x8, x13
+	adcs	x9, x9, x12
+	adc	x10, x10, xzr      // At this point acc2 has the value of 4 at most
+	st1	{v20.16b}, [x0], #16
+
+	sub	x2, x2, #16
+
+    // Shift the state left by 16 bytes for the next iteration of the loop
+	mov	v0.16b, v5.16b
+	mov	v5.16b, v10.16b
+	mov	v10.16b, v15.16b
+
+	b	Lseal_tail_64
+
+Lseal_tail_16:
+    // Here we handle the last [0,16) bytes of ciphertext that require a padded block
+	cbz	x2, Lseal_hash_extra
+
+	eor	v20.16b, v20.16b, v20.16b // Use T0 to load the plaintext/extra in
+	eor	v21.16b, v21.16b, v21.16b // Use T1 to generate an AND mask that will only mask the ciphertext bytes
+	not	v22.16b, v20.16b
+
+	mov	x6, x2
+	add	x1, x1, x2
+
+	cbz	x4, Lseal_tail_16_compose // No extra data to pad with, zero padding
+
+	mov	x7, #16          // We need to load some extra_in first for padding
+	sub	x7, x7, x2
+	cmp	x4, x7
+	csel	x7, x4, x7, lt // Load the minimum of extra_in_len and the amount needed to fill the register
+	mov	x12, x7
+	add	x3, x3, x7
+	sub	x4, x4, x7
+
+Lseal_tail16_compose_extra_in:
+	ext	v20.16b, v20.16b, v20.16b, #15
+	ldrb	w11, [x3, #-1]!
+	mov	v20.b[0], w11
+	subs	x7, x7, #1
+	b.gt	Lseal_tail16_compose_extra_in
+
+	add	x3, x3, x12
+
+Lseal_tail_16_compose:
+	ext	v20.16b, v20.16b, v20.16b, #15
+	ldrb	w11, [x1, #-1]!
+	mov	v20.b[0], w11
+	ext	v21.16b, v22.16b, v21.16b, #15
+	subs	x2, x2, #1
+	b.gt	Lseal_tail_16_compose
+
+	and	v0.16b, v0.16b, v21.16b
+	eor	v20.16b, v20.16b, v0.16b
+	mov	v21.16b, v20.16b
+
+Lseal_tail_16_store:
+	umov	w11, v20.b[0]
+	strb	w11, [x0], #1
+	ext	v20.16b, v20.16b, v20.16b, #1
+	subs	x6, x6, #1
+	b.gt	Lseal_tail_16_store
+
+    // Hash in the final ct block concatenated with extra_in
+	mov	x11, v21.d[0]
+	mov	x12, v21.d[1]
+	adds	x8, x8, x11
+	adcs	x9, x9, x12
+	adc	x10, x10, x15
+	mul	x11, x8, x16     // [t2:t1:t0] = [acc2:acc1:acc0] * r0
+	umulh	x12, x8, x16
+	mul	x13, x9, x16
+	umulh	x14, x9, x16
+	adds	x12, x12, x13
+	mul	x13, x10, x16
+	adc	x13, x13, x14
+	mul	x14, x8, x17       // [t3:t2:t1:t0] = [acc2:acc1:acc0] * [r1:r0]
+	umulh	x8, x8, x17
+	adds	x12, x12, x14
+	mul	x14, x9, x17
+	umulh	x9, x9, x17
+	adcs	x14, x14, x8
+	mul	x10, x10, x17
+	adc	x10, x10, x9
+	adds	x13, x13, x14
+	adc	x14, x10, xzr
+	and	x10, x13, #3         // At this point acc2 is 2 bits at most (value of 3)
+	and	x8, x13, #-4
+	extr	x13, x14, x13, #2
+	adds	x8, x8, x11
+	lsr	x11, x14, #2
+	adc	x9, x14, x11        // No carry out since t0 is 61 bits and t3 is 63 bits
+	adds	x8, x8, x13
+	adcs	x9, x9, x12
+	adc	x10, x10, xzr      // At this point acc2 has the value of 4 at most
+
+Lseal_hash_extra:
+	cbz	x4, Lseal_finalize
+
+Lseal_hash_extra_loop:
+	cmp	x4, #16
+	b.lt	Lseal_hash_extra_tail
+	ld1	{v20.16b}, [x3], #16
+	mov	x11, v20.d[0]
+	mov	x12, v20.d[1]
+	adds	x8, x8, x11
+	adcs	x9, x9, x12
+	adc	x10, x10, x15
+	mul	x11, x8, x16     // [t2:t1:t0] = [acc2:acc1:acc0] * r0
+	umulh	x12, x8, x16
+	mul	x13, x9, x16
+	umulh	x14, x9, x16
+	adds	x12, x12, x13
+	mul	x13, x10, x16
+	adc	x13, x13, x14
+	mul	x14, x8, x17       // [t3:t2:t1:t0] = [acc2:acc1:acc0] * [r1:r0]
+	umulh	x8, x8, x17
+	adds	x12, x12, x14
+	mul	x14, x9, x17
+	umulh	x9, x9, x17
+	adcs	x14, x14, x8
+	mul	x10, x10, x17
+	adc	x10, x10, x9
+	adds	x13, x13, x14
+	adc	x14, x10, xzr
+	and	x10, x13, #3         // At this point acc2 is 2 bits at most (value of 3)
+	and	x8, x13, #-4
+	extr	x13, x14, x13, #2
+	adds	x8, x8, x11
+	lsr	x11, x14, #2
+	adc	x9, x14, x11        // No carry out since t0 is 61 bits and t3 is 63 bits
+	adds	x8, x8, x13
+	adcs	x9, x9, x12
+	adc	x10, x10, xzr      // At this point acc2 has the value of 4 at most
+	sub	x4, x4, #16
+	b	Lseal_hash_extra_loop
+
+Lseal_hash_extra_tail:
+	cbz	x4, Lseal_finalize
+	eor	v20.16b, v20.16b, v20.16b // Use T0 to load the remaining extra ciphertext
+	add	x3, x3, x4
+
+Lseal_hash_extra_load:
+	ext	v20.16b, v20.16b, v20.16b, #15
+	ldrb	w11, [x3, #-1]!
+	mov	v20.b[0], w11
+	subs	x4, x4, #1
+	b.gt	Lseal_hash_extra_load
+
+    // Hash in the final padded extra_in block
+	mov	x11, v20.d[0]
+	mov	x12, v20.d[1]
+	adds	x8, x8, x11
+	adcs	x9, x9, x12
+	adc	x10, x10, x15
+	mul	x11, x8, x16     // [t2:t1:t0] = [acc2:acc1:acc0] * r0
+	umulh	x12, x8, x16
+	mul	x13, x9, x16
+	umulh	x14, x9, x16
+	adds	x12, x12, x13
+	mul	x13, x10, x16
+	adc	x13, x13, x14
+	mul	x14, x8, x17       // [t3:t2:t1:t0] = [acc2:acc1:acc0] * [r1:r0]
+	umulh	x8, x8, x17
+	adds	x12, x12, x14
+	mul	x14, x9, x17
+	umulh	x9, x9, x17
+	adcs	x14, x14, x8
+	mul	x10, x10, x17
+	adc	x10, x10, x9
+	adds	x13, x13, x14
+	adc	x14, x10, xzr
+	and	x10, x13, #3         // At this point acc2 is 2 bits at most (value of 3)
+	and	x8, x13, #-4
+	extr	x13, x14, x13, #2
+	adds	x8, x8, x11
+	lsr	x11, x14, #2
+	adc	x9, x14, x11        // No carry out since t0 is 61 bits and t3 is 63 bits
+	adds	x8, x8, x13
+	adcs	x9, x9, x12
+	adc	x10, x10, xzr      // At this point acc2 has the value of 4 at most
+
+Lseal_finalize:
+	mov	x11, v31.d[0]
+	mov	x12, v31.d[1]
+	adds	x8, x8, x11
+	adcs	x9, x9, x12
+	adc	x10, x10, x15
+	mul	x11, x8, x16     // [t2:t1:t0] = [acc2:acc1:acc0] * r0
+	umulh	x12, x8, x16
+	mul	x13, x9, x16
+	umulh	x14, x9, x16
+	adds	x12, x12, x13
+	mul	x13, x10, x16
+	adc	x13, x13, x14
+	mul	x14, x8, x17       // [t3:t2:t1:t0] = [acc2:acc1:acc0] * [r1:r0]
+	umulh	x8, x8, x17
+	adds	x12, x12, x14
+	mul	x14, x9, x17
+	umulh	x9, x9, x17
+	adcs	x14, x14, x8
+	mul	x10, x10, x17
+	adc	x10, x10, x9
+	adds	x13, x13, x14
+	adc	x14, x10, xzr
+	and	x10, x13, #3         // At this point acc2 is 2 bits at most (value of 3)
+	and	x8, x13, #-4
+	extr	x13, x14, x13, #2
+	adds	x8, x8, x11
+	lsr	x11, x14, #2
+	adc	x9, x14, x11        // No carry out since t0 is 61 bits and t3 is 63 bits
+	adds	x8, x8, x13
+	adcs	x9, x9, x12
+	adc	x10, x10, xzr      // At this point acc2 has the value of 4 at most
+    // Final reduction step
+	sub	x12, xzr, x15
+	orr	x13, xzr, #3
+	subs	x11, x8, #-5
+	sbcs	x12, x9, x12
+	sbcs	x13, x10, x13
+	csel	x8, x11, x8, cs
+	csel	x9, x12, x9, cs
+	csel	x10, x13, x10, cs
+	mov	x11, v27.d[0]
+	mov	x12, v27.d[1]
+	adds	x8, x8, x11
+	adcs	x9, x9, x12
+	adc	x10, x10, x15
+
+	stp	x8, x9, [x5]
+
+	ldp	d8, d9, [sp, #16]
+	ldp	d10, d11, [sp, #32]
+	ldp	d12, d13, [sp, #48]
+	ldp	d14, d15, [sp, #64]
+.cfi_restore	b15
+.cfi_restore	b14
+.cfi_restore	b13
+.cfi_restore	b12
+.cfi_restore	b11
+.cfi_restore	b10
+.cfi_restore	b9
+.cfi_restore	b8
+	ldp	x29, x30, [sp], 80
+.cfi_restore	w29
+.cfi_restore	w30
+.cfi_def_cfa_offset	0
+	AARCH64_VALIDATE_LINK_REGISTER
+	ret
+
+Lseal_128:
+    // On some architectures preparing 5 blocks for small buffers is wasteful
+	eor	v25.16b, v25.16b, v25.16b
+	mov	x11, #1
+	mov	v25.s[0], w11
+	mov	v0.16b, v24.16b
+	mov	v1.16b, v24.16b
+	mov	v2.16b, v24.16b
+	mov	v5.16b, v28.16b
+	mov	v6.16b, v28.16b
+	mov	v7.16b, v28.16b
+	mov	v10.16b, v29.16b
+	mov	v11.16b, v29.16b
+	mov	v12.16b, v29.16b
+	mov	v17.16b, v30.16b
+	add	v15.4s, v17.4s, v25.4s
+	add	v16.4s, v15.4s, v25.4s
+
+	mov	x6, #10
+
+Lseal_128_rounds:
+	add	v0.4s, v0.4s, v5.4s
+	add	v1.4s, v1.4s, v6.4s
+	add	v2.4s, v2.4s, v7.4s
+	eor	v15.16b, v15.16b, v0.16b
+	eor	v16.16b, v16.16b, v1.16b
+	eor	v17.16b, v17.16b, v2.16b
+	rev32	v15.8h, v15.8h
+	rev32	v16.8h, v16.8h
+	rev32	v17.8h, v17.8h
+
+	add	v10.4s, v10.4s, v15.4s
+	add	v11.4s, v11.4s, v16.4s
+	add	v12.4s, v12.4s, v17.4s
+	eor	v5.16b, v5.16b, v10.16b
+	eor	v6.16b, v6.16b, v11.16b
+	eor	v7.16b, v7.16b, v12.16b
+	ushr	v20.4s, v5.4s, #20
+	sli	v20.4s, v5.4s, #12
+	ushr	v5.4s, v6.4s, #20
+	sli	v5.4s, v6.4s, #12
+	ushr	v6.4s, v7.4s, #20
+	sli	v6.4s, v7.4s, #12
+
+	add	v0.4s, v0.4s, v20.4s
+	add	v1.4s, v1.4s, v5.4s
+	add	v2.4s, v2.4s, v6.4s
+	eor	v15.16b, v15.16b, v0.16b
+	eor	v16.16b, v16.16b, v1.16b
+	eor	v17.16b, v17.16b, v2.16b
+	tbl	v15.16b, {v15.16b}, v26.16b
+	tbl	v16.16b, {v16.16b}, v26.16b
+	tbl	v17.16b, {v17.16b}, v26.16b
+
+	add	v10.4s, v10.4s, v15.4s
+	add	v11.4s, v11.4s, v16.4s
+	add	v12.4s, v12.4s, v17.4s
+	eor	v20.16b, v20.16b, v10.16b
+	eor	v5.16b, v5.16b, v11.16b
+	eor	v6.16b, v6.16b, v12.16b
+	ushr	v7.4s, v6.4s, #25
+	sli	v7.4s, v6.4s, #7
+	ushr	v6.4s, v5.4s, #25
+	sli	v6.4s, v5.4s, #7
+	ushr	v5.4s, v20.4s, #25
+	sli	v5.4s, v20.4s, #7
+
+	ext	v5.16b, v5.16b, v5.16b, #4
+	ext	v6.16b, v6.16b, v6.16b, #4
+	ext	v7.16b, v7.16b, v7.16b, #4
+
+	ext	v10.16b, v10.16b, v10.16b, #8
+	ext	v11.16b, v11.16b, v11.16b, #8
+	ext	v12.16b, v12.16b, v12.16b, #8
+
+	ext	v15.16b, v15.16b, v15.16b, #12
+	ext	v16.16b, v16.16b, v16.16b, #12
+	ext	v17.16b, v17.16b, v17.16b, #12
+	add	v0.4s, v0.4s, v5.4s
+	add	v1.4s, v1.4s, v6.4s
+	add	v2.4s, v2.4s, v7.4s
+	eor	v15.16b, v15.16b, v0.16b
+	eor	v16.16b, v16.16b, v1.16b
+	eor	v17.16b, v17.16b, v2.16b
+	rev32	v15.8h, v15.8h
+	rev32	v16.8h, v16.8h
+	rev32	v17.8h, v17.8h
+
+	add	v10.4s, v10.4s, v15.4s
+	add	v11.4s, v11.4s, v16.4s
+	add	v12.4s, v12.4s, v17.4s
+	eor	v5.16b, v5.16b, v10.16b
+	eor	v6.16b, v6.16b, v11.16b
+	eor	v7.16b, v7.16b, v12.16b
+	ushr	v20.4s, v5.4s, #20
+	sli	v20.4s, v5.4s, #12
+	ushr	v5.4s, v6.4s, #20
+	sli	v5.4s, v6.4s, #12
+	ushr	v6.4s, v7.4s, #20
+	sli	v6.4s, v7.4s, #12
+
+	add	v0.4s, v0.4s, v20.4s
+	add	v1.4s, v1.4s, v5.4s
+	add	v2.4s, v2.4s, v6.4s
+	eor	v15.16b, v15.16b, v0.16b
+	eor	v16.16b, v16.16b, v1.16b
+	eor	v17.16b, v17.16b, v2.16b
+	tbl	v15.16b, {v15.16b}, v26.16b
+	tbl	v16.16b, {v16.16b}, v26.16b
+	tbl	v17.16b, {v17.16b}, v26.16b
+
+	add	v10.4s, v10.4s, v15.4s
+	add	v11.4s, v11.4s, v16.4s
+	add	v12.4s, v12.4s, v17.4s
+	eor	v20.16b, v20.16b, v10.16b
+	eor	v5.16b, v5.16b, v11.16b
+	eor	v6.16b, v6.16b, v12.16b
+	ushr	v7.4s, v6.4s, #25
+	sli	v7.4s, v6.4s, #7
+	ushr	v6.4s, v5.4s, #25
+	sli	v6.4s, v5.4s, #7
+	ushr	v5.4s, v20.4s, #25
+	sli	v5.4s, v20.4s, #7
+
+	ext	v5.16b, v5.16b, v5.16b, #12
+	ext	v6.16b, v6.16b, v6.16b, #12
+	ext	v7.16b, v7.16b, v7.16b, #12
+
+	ext	v10.16b, v10.16b, v10.16b, #8
+	ext	v11.16b, v11.16b, v11.16b, #8
+	ext	v12.16b, v12.16b, v12.16b, #8
+
+	ext	v15.16b, v15.16b, v15.16b, #4
+	ext	v16.16b, v16.16b, v16.16b, #4
+	ext	v17.16b, v17.16b, v17.16b, #4
+	subs	x6, x6, #1
+	b.hi	Lseal_128_rounds
+
+	add	v0.4s, v0.4s, v24.4s
+	add	v1.4s, v1.4s, v24.4s
+	add	v2.4s, v2.4s, v24.4s
+
+	add	v5.4s, v5.4s, v28.4s
+	add	v6.4s, v6.4s, v28.4s
+	add	v7.4s, v7.4s, v28.4s
+
+    // Only the first 32 bytes of the third block (counter = 0) are needed,
+    // so skip updating v12 and v17.
+	add	v10.4s, v10.4s, v29.4s
+	add	v11.4s, v11.4s, v29.4s
+
+	add	v30.4s, v30.4s, v25.4s
+	add	v15.4s, v15.4s, v30.4s
+	add	v30.4s, v30.4s, v25.4s
+	add	v16.4s, v16.4s, v30.4s
+
+	and	v2.16b, v2.16b, v27.16b
+	mov	x16, v2.d[0] // Move the R key to GPRs
+	mov	x17, v2.d[1]
+	mov	v27.16b, v7.16b // Store the S key
+
+	bl	Lpoly_hash_ad_internal
+	b	Lseal_tail
+.cfi_endproc
+
+
+/////////////////////////////////
+//
+// void chacha20_poly1305_open(uint8_t *pt, uint8_t *ct, size_t len_in, uint8_t *ad, size_t len_ad, union open_data *aead_data);
+//
+.globl	chacha20_poly1305_open
+
+.def chacha20_poly1305_open
+   .type 32
+.endef
+.align	6
+chacha20_poly1305_open:
+	AARCH64_SIGN_LINK_REGISTER
+.cfi_startproc
+	stp	x29, x30, [sp, #-80]!
+.cfi_def_cfa_offset	80
+.cfi_offset	w30, -72
+.cfi_offset	w29, -80
+	mov	x29, sp
+    // We probably could do .cfi_def_cfa w29, 80 at this point, but since
+    // we don't actually use the frame pointer like that, it's probably not
+    // worth bothering.
+	stp	d8, d9, [sp, #16]
+	stp	d10, d11, [sp, #32]
+	stp	d12, d13, [sp, #48]
+	stp	d14, d15, [sp, #64]
+.cfi_offset	b15, -8
+.cfi_offset	b14, -16
+.cfi_offset	b13, -24
+.cfi_offset	b12, -32
+.cfi_offset	b11, -40
+.cfi_offset	b10, -48
+.cfi_offset	b9, -56
+.cfi_offset	b8, -64
+
+	adrp	x11, Lchacha20_consts
+	add	x11, x11, :lo12:Lchacha20_consts
+
+	ld1	{v24.16b - v27.16b}, [x11] // Load the CONSTS, INC, ROL8 and CLAMP values
+	ld1	{v28.16b - v30.16b}, [x5]
+
+	mov	x15, #1 // Prepare the Poly1305 state
+	mov	x8, #0
+	mov	x9, #0
+	mov	x10, #0
+
+	mov	v31.d[0], x4  // Store the input and aad lengths
+	mov	v31.d[1], x2
+
+	cmp	x2, #128
+	b.le	Lopen_128 // Optimization for smaller buffers
+
+    // Initially we prepare a single ChaCha20 block for the Poly1305 R and S keys
+	mov	v0.16b, v24.16b
+	mov	v5.16b, v28.16b
+	mov	v10.16b, v29.16b
+	mov	v15.16b, v30.16b
+
+	mov	x6, #10
+
+.align	5
+Lopen_init_rounds:
+	add	v0.4s, v0.4s, v5.4s
+	eor	v15.16b, v15.16b, v0.16b
+	rev32	v15.8h, v15.8h
+
+	add	v10.4s, v10.4s, v15.4s
+	eor	v5.16b, v5.16b, v10.16b
+	ushr	v20.4s, v5.4s, #20
+	sli	v20.4s, v5.4s, #12
+	add	v0.4s, v0.4s, v20.4s
+	eor	v15.16b, v15.16b, v0.16b
+	tbl	v15.16b, {v15.16b}, v26.16b
+
+	add	v10.4s, v10.4s, v15.4s
+	eor	v20.16b, v20.16b, v10.16b
+	ushr	v5.4s, v20.4s, #25
+	sli	v5.4s, v20.4s, #7
+	ext	v5.16b, v5.16b, v5.16b, #4
+	ext	v10.16b, v10.16b, v10.16b, #8
+	ext	v15.16b, v15.16b, v15.16b, #12
+	add	v0.4s, v0.4s, v5.4s
+	eor	v15.16b, v15.16b, v0.16b
+	rev32	v15.8h, v15.8h
+
+	add	v10.4s, v10.4s, v15.4s
+	eor	v5.16b, v5.16b, v10.16b
+	ushr	v20.4s, v5.4s, #20
+	sli	v20.4s, v5.4s, #12
+	add	v0.4s, v0.4s, v20.4s
+	eor	v15.16b, v15.16b, v0.16b
+	tbl	v15.16b, {v15.16b}, v26.16b
+
+	add	v10.4s, v10.4s, v15.4s
+	eor	v20.16b, v20.16b, v10.16b
+	ushr	v5.4s, v20.4s, #25
+	sli	v5.4s, v20.4s, #7
+	ext	v5.16b, v5.16b, v5.16b, #12
+	ext	v10.16b, v10.16b, v10.16b, #8
+	ext	v15.16b, v15.16b, v15.16b, #4
+	subs	x6, x6, #1
+	b.hi	Lopen_init_rounds
+
+	add	v0.4s, v0.4s, v24.4s
+	add	v5.4s, v5.4s, v28.4s
+
+	and	v0.16b, v0.16b, v27.16b
+	mov	x16, v0.d[0] // Move the R key to GPRs
+	mov	x17, v0.d[1]
+	mov	v27.16b, v5.16b // Store the S key
+
+	bl	Lpoly_hash_ad_internal
+
+	mov	x3, x1
+
+// Each iteration of the loop hash 320 bytes, and prepare stream for 320 bytes
+Lopen_main_loop:
+
+	cmp	x2, #192
+	b.lt	Lopen_tail
+
+	adrp	x11, Lchacha20_consts
+	add	x11, x11, :lo12:Lchacha20_consts
+
+	ld4r	{v0.4s,v1.4s,v2.4s,v3.4s}, [x11]
+	mov	v4.16b, v24.16b
+
+	ld4r	{v5.4s,v6.4s,v7.4s,v8.4s}, [x5], #16
+	mov	v9.16b, v28.16b
+
+	ld4r	{v10.4s,v11.4s,v12.4s,v13.4s}, [x5], #16
+	mov	v14.16b, v29.16b
+
+	ld4r	{v15.4s,v16.4s,v17.4s,v18.4s}, [x5]
+	sub	x5, x5, #32
+	add	v15.4s, v15.4s, v25.4s
+	mov	v19.16b, v30.16b
+
+	eor	v20.16b, v20.16b, v20.16b //zero
+	not	v21.16b, v20.16b // -1
+	sub	v21.4s, v25.4s, v21.4s // Add +1
+	ext	v20.16b, v21.16b, v20.16b, #12 // Get the last element (counter)
+	add	v19.4s, v19.4s, v20.4s
+
+	lsr	x4, x2, #4 // How many whole blocks we have to hash, will always be at least 12
+	sub	x4, x4, #10
+
+	mov	x7, #10
+	subs	x6, x7, x4 // itr1 can be negative if we have more than 320 bytes to hash
+	csel	x7, x7, x4, le // if itr1 is zero or less, itr2 should be 10 to indicate all 10 rounds are full
+
+	cbz	x7, Lopen_main_loop_rounds_short
+
+.align	5
+Lopen_main_loop_rounds:
+	ldp	x11, x12, [x3], 16
+	adds	x8, x8, x11
+	adcs	x9, x9, x12
+	adc	x10, x10, x15
+	mul	x11, x8, x16     // [t2:t1:t0] = [acc2:acc1:acc0] * r0
+	umulh	x12, x8, x16
+	mul	x13, x9, x16
+	umulh	x14, x9, x16
+	adds	x12, x12, x13
+	mul	x13, x10, x16
+	adc	x13, x13, x14
+	mul	x14, x8, x17       // [t3:t2:t1:t0] = [acc2:acc1:acc0] * [r1:r0]
+	umulh	x8, x8, x17
+	adds	x12, x12, x14
+	mul	x14, x9, x17
+	umulh	x9, x9, x17
+	adcs	x14, x14, x8
+	mul	x10, x10, x17
+	adc	x10, x10, x9
+	adds	x13, x13, x14
+	adc	x14, x10, xzr
+	and	x10, x13, #3         // At this point acc2 is 2 bits at most (value of 3)
+	and	x8, x13, #-4
+	extr	x13, x14, x13, #2
+	adds	x8, x8, x11
+	lsr	x11, x14, #2
+	adc	x9, x14, x11        // No carry out since t0 is 61 bits and t3 is 63 bits
+	adds	x8, x8, x13
+	adcs	x9, x9, x12
+	adc	x10, x10, xzr      // At this point acc2 has the value of 4 at most
+Lopen_main_loop_rounds_short:
+	add	v0.4s, v0.4s, v5.4s
+	add	v1.4s, v1.4s, v6.4s
+	add	v2.4s, v2.4s, v7.4s
+	add	v3.4s, v3.4s, v8.4s
+	add	v4.4s, v4.4s, v9.4s
+
+	eor	v15.16b, v15.16b, v0.16b
+	eor	v16.16b, v16.16b, v1.16b
+	eor	v17.16b, v17.16b, v2.16b
+	eor	v18.16b, v18.16b, v3.16b
+	eor	v19.16b, v19.16b, v4.16b
+
+	rev32	v15.8h, v15.8h
+	rev32	v16.8h, v16.8h
+	rev32	v17.8h, v17.8h
+	rev32	v18.8h, v18.8h
+	rev32	v19.8h, v19.8h
+
+	add	v10.4s, v10.4s, v15.4s
+	add	v11.4s, v11.4s, v16.4s
+	add	v12.4s, v12.4s, v17.4s
+	add	v13.4s, v13.4s, v18.4s
+	add	v14.4s, v14.4s, v19.4s
+
+	eor	v5.16b, v5.16b, v10.16b
+	eor	v6.16b, v6.16b, v11.16b
+	eor	v7.16b, v7.16b, v12.16b
+	eor	v8.16b, v8.16b, v13.16b
+	eor	v9.16b, v9.16b, v14.16b
+
+	ushr	v20.4s, v5.4s, #20
+	sli	v20.4s, v5.4s, #12
+	ushr	v5.4s, v6.4s, #20
+	sli	v5.4s, v6.4s, #12
+	ushr	v6.4s, v7.4s, #20
+	sli	v6.4s, v7.4s, #12
+	ushr	v7.4s, v8.4s, #20
+	sli	v7.4s, v8.4s, #12
+	ushr	v8.4s, v9.4s, #20
+	sli	v8.4s, v9.4s, #12
+
+	add	v0.4s, v0.4s, v20.4s
+	add	v1.4s, v1.4s, v5.4s
+	add	v2.4s, v2.4s, v6.4s
+	add	v3.4s, v3.4s, v7.4s
+	add	v4.4s, v4.4s, v8.4s
+
+	eor	v15.16b, v15.16b, v0.16b
+	eor	v16.16b, v16.16b, v1.16b
+	eor	v17.16b, v17.16b, v2.16b
+	eor	v18.16b, v18.16b, v3.16b
+	eor	v19.16b, v19.16b, v4.16b
+
+	tbl	v15.16b, {v15.16b}, v26.16b
+	tbl	v16.16b, {v16.16b}, v26.16b
+	tbl	v17.16b, {v17.16b}, v26.16b
+	tbl	v18.16b, {v18.16b}, v26.16b
+	tbl	v19.16b, {v19.16b}, v26.16b
+
+	add	v10.4s, v10.4s, v15.4s
+	add	v11.4s, v11.4s, v16.4s
+	add	v12.4s, v12.4s, v17.4s
+	add	v13.4s, v13.4s, v18.4s
+	add	v14.4s, v14.4s, v19.4s
+
+	eor	v20.16b, v20.16b, v10.16b
+	eor	v5.16b, v5.16b, v11.16b
+	eor	v6.16b, v6.16b, v12.16b
+	eor	v7.16b, v7.16b, v13.16b
+	eor	v8.16b, v8.16b, v14.16b
+
+	ushr	v9.4s, v8.4s, #25
+	sli	v9.4s, v8.4s, #7
+	ushr	v8.4s, v7.4s, #25
+	sli	v8.4s, v7.4s, #7
+	ushr	v7.4s, v6.4s, #25
+	sli	v7.4s, v6.4s, #7
+	ushr	v6.4s, v5.4s, #25
+	sli	v6.4s, v5.4s, #7
+	ushr	v5.4s, v20.4s, #25
+	sli	v5.4s, v20.4s, #7
+
+	ext	v9.16b, v9.16b, v9.16b, #4
+	ext	v14.16b, v14.16b, v14.16b, #8
+	ext	v19.16b, v19.16b, v19.16b, #12
+	ldp	x11, x12, [x3], 16
+	adds	x8, x8, x11
+	adcs	x9, x9, x12
+	adc	x10, x10, x15
+	mul	x11, x8, x16     // [t2:t1:t0] = [acc2:acc1:acc0] * r0
+	umulh	x12, x8, x16
+	mul	x13, x9, x16
+	umulh	x14, x9, x16
+	adds	x12, x12, x13
+	mul	x13, x10, x16
+	adc	x13, x13, x14
+	mul	x14, x8, x17       // [t3:t2:t1:t0] = [acc2:acc1:acc0] * [r1:r0]
+	umulh	x8, x8, x17
+	adds	x12, x12, x14
+	mul	x14, x9, x17
+	umulh	x9, x9, x17
+	adcs	x14, x14, x8
+	mul	x10, x10, x17
+	adc	x10, x10, x9
+	adds	x13, x13, x14
+	adc	x14, x10, xzr
+	and	x10, x13, #3         // At this point acc2 is 2 bits at most (value of 3)
+	and	x8, x13, #-4
+	extr	x13, x14, x13, #2
+	adds	x8, x8, x11
+	lsr	x11, x14, #2
+	adc	x9, x14, x11        // No carry out since t0 is 61 bits and t3 is 63 bits
+	adds	x8, x8, x13
+	adcs	x9, x9, x12
+	adc	x10, x10, xzr      // At this point acc2 has the value of 4 at most
+	add	v0.4s, v0.4s, v6.4s
+	add	v1.4s, v1.4s, v7.4s
+	add	v2.4s, v2.4s, v8.4s
+	add	v3.4s, v3.4s, v5.4s
+	add	v4.4s, v4.4s, v9.4s
+
+	eor	v18.16b, v18.16b, v0.16b
+	eor	v15.16b, v15.16b, v1.16b
+	eor	v16.16b, v16.16b, v2.16b
+	eor	v17.16b, v17.16b, v3.16b
+	eor	v19.16b, v19.16b, v4.16b
+
+	rev32	v18.8h, v18.8h
+	rev32	v15.8h, v15.8h
+	rev32	v16.8h, v16.8h
+	rev32	v17.8h, v17.8h
+	rev32	v19.8h, v19.8h
+
+	add	v12.4s, v12.4s, v18.4s
+	add	v13.4s, v13.4s, v15.4s
+	add	v10.4s, v10.4s, v16.4s
+	add	v11.4s, v11.4s, v17.4s
+	add	v14.4s, v14.4s, v19.4s
+
+	eor	v6.16b, v6.16b, v12.16b
+	eor	v7.16b, v7.16b, v13.16b
+	eor	v8.16b, v8.16b, v10.16b
+	eor	v5.16b, v5.16b, v11.16b
+	eor	v9.16b, v9.16b, v14.16b
+
+	ushr	v20.4s, v6.4s, #20
+	sli	v20.4s, v6.4s, #12
+	ushr	v6.4s, v7.4s, #20
+	sli	v6.4s, v7.4s, #12
+	ushr	v7.4s, v8.4s, #20
+	sli	v7.4s, v8.4s, #12
+	ushr	v8.4s, v5.4s, #20
+	sli	v8.4s, v5.4s, #12
+	ushr	v5.4s, v9.4s, #20
+	sli	v5.4s, v9.4s, #12
+
+	add	v0.4s, v0.4s, v20.4s
+	add	v1.4s, v1.4s, v6.4s
+	add	v2.4s, v2.4s, v7.4s
+	add	v3.4s, v3.4s, v8.4s
+	add	v4.4s, v4.4s, v5.4s
+
+	eor	v18.16b, v18.16b, v0.16b
+	eor	v15.16b, v15.16b, v1.16b
+	eor	v16.16b, v16.16b, v2.16b
+	eor	v17.16b, v17.16b, v3.16b
+	eor	v19.16b, v19.16b, v4.16b
+
+	tbl	v18.16b, {v18.16b}, v26.16b
+	tbl	v15.16b, {v15.16b}, v26.16b
+	tbl	v16.16b, {v16.16b}, v26.16b
+	tbl	v17.16b, {v17.16b}, v26.16b
+	tbl	v19.16b, {v19.16b}, v26.16b
+
+	add	v12.4s, v12.4s, v18.4s
+	add	v13.4s, v13.4s, v15.4s
+	add	v10.4s, v10.4s, v16.4s
+	add	v11.4s, v11.4s, v17.4s
+	add	v14.4s, v14.4s, v19.4s
+
+	eor	v20.16b, v20.16b, v12.16b
+	eor	v6.16b, v6.16b, v13.16b
+	eor	v7.16b, v7.16b, v10.16b
+	eor	v8.16b, v8.16b, v11.16b
+	eor	v5.16b, v5.16b, v14.16b
+
+	ushr	v9.4s, v5.4s, #25
+	sli	v9.4s, v5.4s, #7
+	ushr	v5.4s, v8.4s, #25
+	sli	v5.4s, v8.4s, #7
+	ushr	v8.4s, v7.4s, #25
+	sli	v8.4s, v7.4s, #7
+	ushr	v7.4s, v6.4s, #25
+	sli	v7.4s, v6.4s, #7
+	ushr	v6.4s, v20.4s, #25
+	sli	v6.4s, v20.4s, #7
+
+	ext	v9.16b, v9.16b, v9.16b, #12
+	ext	v14.16b, v14.16b, v14.16b, #8
+	ext	v19.16b, v19.16b, v19.16b, #4
+	subs	x7, x7, #1
+	b.gt	Lopen_main_loop_rounds
+	subs	x6, x6, #1
+	b.ge	Lopen_main_loop_rounds_short
+
+	eor	v20.16b, v20.16b, v20.16b //zero
+	not	v21.16b, v20.16b // -1
+	sub	v21.4s, v25.4s, v21.4s // Add +1
+	ext	v20.16b, v21.16b, v20.16b, #12 // Get the last element (counter)
+	add	v19.4s, v19.4s, v20.4s
+
+	add	v15.4s, v15.4s, v25.4s
+	mov	x11, #5
+	dup	v20.4s, w11
+	add	v25.4s, v25.4s, v20.4s
+
+	zip1	v20.4s, v0.4s, v1.4s
+	zip2	v21.4s, v0.4s, v1.4s
+	zip1	v22.4s, v2.4s, v3.4s
+	zip2	v23.4s, v2.4s, v3.4s
+
+	zip1	v0.2d, v20.2d, v22.2d
+	zip2	v1.2d, v20.2d, v22.2d
+	zip1	v2.2d, v21.2d, v23.2d
+	zip2	v3.2d, v21.2d, v23.2d
+
+	zip1	v20.4s, v5.4s, v6.4s
+	zip2	v21.4s, v5.4s, v6.4s
+	zip1	v22.4s, v7.4s, v8.4s
+	zip2	v23.4s, v7.4s, v8.4s
+
+	zip1	v5.2d, v20.2d, v22.2d
+	zip2	v6.2d, v20.2d, v22.2d
+	zip1	v7.2d, v21.2d, v23.2d
+	zip2	v8.2d, v21.2d, v23.2d
+
+	zip1	v20.4s, v10.4s, v11.4s
+	zip2	v21.4s, v10.4s, v11.4s
+	zip1	v22.4s, v12.4s, v13.4s
+	zip2	v23.4s, v12.4s, v13.4s
+
+	zip1	v10.2d, v20.2d, v22.2d
+	zip2	v11.2d, v20.2d, v22.2d
+	zip1	v12.2d, v21.2d, v23.2d
+	zip2	v13.2d, v21.2d, v23.2d
+
+	zip1	v20.4s, v15.4s, v16.4s
+	zip2	v21.4s, v15.4s, v16.4s
+	zip1	v22.4s, v17.4s, v18.4s
+	zip2	v23.4s, v17.4s, v18.4s
+
+	zip1	v15.2d, v20.2d, v22.2d
+	zip2	v16.2d, v20.2d, v22.2d
+	zip1	v17.2d, v21.2d, v23.2d
+	zip2	v18.2d, v21.2d, v23.2d
+
+	add	v0.4s, v0.4s, v24.4s
+	add	v5.4s, v5.4s, v28.4s
+	add	v10.4s, v10.4s, v29.4s
+	add	v15.4s, v15.4s, v30.4s
+
+	add	v1.4s, v1.4s, v24.4s
+	add	v6.4s, v6.4s, v28.4s
+	add	v11.4s, v11.4s, v29.4s
+	add	v16.4s, v16.4s, v30.4s
+
+	add	v2.4s, v2.4s, v24.4s
+	add	v7.4s, v7.4s, v28.4s
+	add	v12.4s, v12.4s, v29.4s
+	add	v17.4s, v17.4s, v30.4s
+
+	add	v3.4s, v3.4s, v24.4s
+	add	v8.4s, v8.4s, v28.4s
+	add	v13.4s, v13.4s, v29.4s
+	add	v18.4s, v18.4s, v30.4s
+
+	add	v4.4s, v4.4s, v24.4s
+	add	v9.4s, v9.4s, v28.4s
+	add	v14.4s, v14.4s, v29.4s
+	add	v19.4s, v19.4s, v30.4s
+
+    // We can always safely store 192 bytes
+	ld1	{v20.16b - v23.16b}, [x1], #64
+	eor	v20.16b, v20.16b, v0.16b
+	eor	v21.16b, v21.16b, v5.16b
+	eor	v22.16b, v22.16b, v10.16b
+	eor	v23.16b, v23.16b, v15.16b
+	st1	{v20.16b - v23.16b}, [x0], #64
+
+	ld1	{v20.16b - v23.16b}, [x1], #64
+	eor	v20.16b, v20.16b, v1.16b
+	eor	v21.16b, v21.16b, v6.16b
+	eor	v22.16b, v22.16b, v11.16b
+	eor	v23.16b, v23.16b, v16.16b
+	st1	{v20.16b - v23.16b}, [x0], #64
+
+	ld1	{v20.16b - v23.16b}, [x1], #64
+	eor	v20.16b, v20.16b, v2.16b
+	eor	v21.16b, v21.16b, v7.16b
+	eor	v22.16b, v22.16b, v12.16b
+	eor	v23.16b, v23.16b, v17.16b
+	st1	{v20.16b - v23.16b}, [x0], #64
+
+	sub	x2, x2, #192
+
+	mov	v0.16b, v3.16b
+	mov	v5.16b, v8.16b
+	mov	v10.16b, v13.16b
+	mov	v15.16b, v18.16b
+
+	cmp	x2, #64
+	b.lt	Lopen_tail_64_store
+
+	ld1	{v20.16b - v23.16b}, [x1], #64
+	eor	v20.16b, v20.16b, v3.16b
+	eor	v21.16b, v21.16b, v8.16b
+	eor	v22.16b, v22.16b, v13.16b
+	eor	v23.16b, v23.16b, v18.16b
+	st1	{v20.16b - v23.16b}, [x0], #64
+
+	sub	x2, x2, #64
+
+	mov	v0.16b, v4.16b
+	mov	v5.16b, v9.16b
+	mov	v10.16b, v14.16b
+	mov	v15.16b, v19.16b
+
+	cmp	x2, #64
+	b.lt	Lopen_tail_64_store
+
+	ld1	{v20.16b - v23.16b}, [x1], #64
+	eor	v20.16b, v20.16b, v4.16b
+	eor	v21.16b, v21.16b, v9.16b
+	eor	v22.16b, v22.16b, v14.16b
+	eor	v23.16b, v23.16b, v19.16b
+	st1	{v20.16b - v23.16b}, [x0], #64
+
+	sub	x2, x2, #64
+	b	Lopen_main_loop
+
+Lopen_tail:
+
+	cbz	x2, Lopen_finalize
+
+	lsr	x4, x2, #4 // How many whole blocks we have to hash
+
+	cmp	x2, #64
+	b.le	Lopen_tail_64
+	cmp	x2, #128
+	b.le	Lopen_tail_128
+
+Lopen_tail_192:
+     // We need three more blocks
+	mov	v0.16b, v24.16b
+	mov	v1.16b, v24.16b
+	mov	v2.16b, v24.16b
+	mov	v5.16b, v28.16b
+	mov	v6.16b, v28.16b
+	mov	v7.16b, v28.16b
+	mov	v10.16b, v29.16b
+	mov	v11.16b, v29.16b
+	mov	v12.16b, v29.16b
+	mov	v15.16b, v30.16b
+	mov	v16.16b, v30.16b
+	mov	v17.16b, v30.16b
+	eor	v23.16b, v23.16b, v23.16b
+	eor	v21.16b, v21.16b, v21.16b
+	ins	v23.s[0], v25.s[0]
+	ins	v21.d[0], x15
+
+	add	v22.4s, v23.4s, v21.4s
+	add	v21.4s, v22.4s, v21.4s
+
+	add	v15.4s, v15.4s, v21.4s
+	add	v16.4s, v16.4s, v23.4s
+	add	v17.4s, v17.4s, v22.4s
+
+	mov	x7, #10
+	subs	x6, x7, x4 // itr1 can be negative if we have more than 160 bytes to hash
+	csel	x7, x7, x4, le // if itr1 is zero or less, itr2 should be 10 to indicate all 10 rounds are hashing
+	sub	x4, x4, x7
+
+	cbz	x7, Lopen_tail_192_rounds_no_hash
+
+Lopen_tail_192_rounds:
+	ldp	x11, x12, [x3], 16
+	adds	x8, x8, x11
+	adcs	x9, x9, x12
+	adc	x10, x10, x15
+	mul	x11, x8, x16     // [t2:t1:t0] = [acc2:acc1:acc0] * r0
+	umulh	x12, x8, x16
+	mul	x13, x9, x16
+	umulh	x14, x9, x16
+	adds	x12, x12, x13
+	mul	x13, x10, x16
+	adc	x13, x13, x14
+	mul	x14, x8, x17       // [t3:t2:t1:t0] = [acc2:acc1:acc0] * [r1:r0]
+	umulh	x8, x8, x17
+	adds	x12, x12, x14
+	mul	x14, x9, x17
+	umulh	x9, x9, x17
+	adcs	x14, x14, x8
+	mul	x10, x10, x17
+	adc	x10, x10, x9
+	adds	x13, x13, x14
+	adc	x14, x10, xzr
+	and	x10, x13, #3         // At this point acc2 is 2 bits at most (value of 3)
+	and	x8, x13, #-4
+	extr	x13, x14, x13, #2
+	adds	x8, x8, x11
+	lsr	x11, x14, #2
+	adc	x9, x14, x11        // No carry out since t0 is 61 bits and t3 is 63 bits
+	adds	x8, x8, x13
+	adcs	x9, x9, x12
+	adc	x10, x10, xzr      // At this point acc2 has the value of 4 at most
+Lopen_tail_192_rounds_no_hash:
+	add	v0.4s, v0.4s, v5.4s
+	add	v1.4s, v1.4s, v6.4s
+	add	v2.4s, v2.4s, v7.4s
+	eor	v15.16b, v15.16b, v0.16b
+	eor	v16.16b, v16.16b, v1.16b
+	eor	v17.16b, v17.16b, v2.16b
+	rev32	v15.8h, v15.8h
+	rev32	v16.8h, v16.8h
+	rev32	v17.8h, v17.8h
+
+	add	v10.4s, v10.4s, v15.4s
+	add	v11.4s, v11.4s, v16.4s
+	add	v12.4s, v12.4s, v17.4s
+	eor	v5.16b, v5.16b, v10.16b
+	eor	v6.16b, v6.16b, v11.16b
+	eor	v7.16b, v7.16b, v12.16b
+	ushr	v20.4s, v5.4s, #20
+	sli	v20.4s, v5.4s, #12
+	ushr	v5.4s, v6.4s, #20
+	sli	v5.4s, v6.4s, #12
+	ushr	v6.4s, v7.4s, #20
+	sli	v6.4s, v7.4s, #12
+
+	add	v0.4s, v0.4s, v20.4s
+	add	v1.4s, v1.4s, v5.4s
+	add	v2.4s, v2.4s, v6.4s
+	eor	v15.16b, v15.16b, v0.16b
+	eor	v16.16b, v16.16b, v1.16b
+	eor	v17.16b, v17.16b, v2.16b
+	tbl	v15.16b, {v15.16b}, v26.16b
+	tbl	v16.16b, {v16.16b}, v26.16b
+	tbl	v17.16b, {v17.16b}, v26.16b
+
+	add	v10.4s, v10.4s, v15.4s
+	add	v11.4s, v11.4s, v16.4s
+	add	v12.4s, v12.4s, v17.4s
+	eor	v20.16b, v20.16b, v10.16b
+	eor	v5.16b, v5.16b, v11.16b
+	eor	v6.16b, v6.16b, v12.16b
+	ushr	v7.4s, v6.4s, #25
+	sli	v7.4s, v6.4s, #7
+	ushr	v6.4s, v5.4s, #25
+	sli	v6.4s, v5.4s, #7
+	ushr	v5.4s, v20.4s, #25
+	sli	v5.4s, v20.4s, #7
+
+	ext	v5.16b, v5.16b, v5.16b, #4
+	ext	v6.16b, v6.16b, v6.16b, #4
+	ext	v7.16b, v7.16b, v7.16b, #4
+
+	ext	v10.16b, v10.16b, v10.16b, #8
+	ext	v11.16b, v11.16b, v11.16b, #8
+	ext	v12.16b, v12.16b, v12.16b, #8
+
+	ext	v15.16b, v15.16b, v15.16b, #12
+	ext	v16.16b, v16.16b, v16.16b, #12
+	ext	v17.16b, v17.16b, v17.16b, #12
+	add	v0.4s, v0.4s, v5.4s
+	add	v1.4s, v1.4s, v6.4s
+	add	v2.4s, v2.4s, v7.4s
+	eor	v15.16b, v15.16b, v0.16b
+	eor	v16.16b, v16.16b, v1.16b
+	eor	v17.16b, v17.16b, v2.16b
+	rev32	v15.8h, v15.8h
+	rev32	v16.8h, v16.8h
+	rev32	v17.8h, v17.8h
+
+	add	v10.4s, v10.4s, v15.4s
+	add	v11.4s, v11.4s, v16.4s
+	add	v12.4s, v12.4s, v17.4s
+	eor	v5.16b, v5.16b, v10.16b
+	eor	v6.16b, v6.16b, v11.16b
+	eor	v7.16b, v7.16b, v12.16b
+	ushr	v20.4s, v5.4s, #20
+	sli	v20.4s, v5.4s, #12
+	ushr	v5.4s, v6.4s, #20
+	sli	v5.4s, v6.4s, #12
+	ushr	v6.4s, v7.4s, #20
+	sli	v6.4s, v7.4s, #12
+
+	add	v0.4s, v0.4s, v20.4s
+	add	v1.4s, v1.4s, v5.4s
+	add	v2.4s, v2.4s, v6.4s
+	eor	v15.16b, v15.16b, v0.16b
+	eor	v16.16b, v16.16b, v1.16b
+	eor	v17.16b, v17.16b, v2.16b
+	tbl	v15.16b, {v15.16b}, v26.16b
+	tbl	v16.16b, {v16.16b}, v26.16b
+	tbl	v17.16b, {v17.16b}, v26.16b
+
+	add	v10.4s, v10.4s, v15.4s
+	add	v11.4s, v11.4s, v16.4s
+	add	v12.4s, v12.4s, v17.4s
+	eor	v20.16b, v20.16b, v10.16b
+	eor	v5.16b, v5.16b, v11.16b
+	eor	v6.16b, v6.16b, v12.16b
+	ushr	v7.4s, v6.4s, #25
+	sli	v7.4s, v6.4s, #7
+	ushr	v6.4s, v5.4s, #25
+	sli	v6.4s, v5.4s, #7
+	ushr	v5.4s, v20.4s, #25
+	sli	v5.4s, v20.4s, #7
+
+	ext	v5.16b, v5.16b, v5.16b, #12
+	ext	v6.16b, v6.16b, v6.16b, #12
+	ext	v7.16b, v7.16b, v7.16b, #12
+
+	ext	v10.16b, v10.16b, v10.16b, #8
+	ext	v11.16b, v11.16b, v11.16b, #8
+	ext	v12.16b, v12.16b, v12.16b, #8
+
+	ext	v15.16b, v15.16b, v15.16b, #4
+	ext	v16.16b, v16.16b, v16.16b, #4
+	ext	v17.16b, v17.16b, v17.16b, #4
+	subs	x7, x7, #1
+	b.gt	Lopen_tail_192_rounds
+	subs	x6, x6, #1
+	b.ge	Lopen_tail_192_rounds_no_hash
+
+    // We hashed 160 bytes at most, may still have 32 bytes left
+Lopen_tail_192_hash:
+	cbz	x4, Lopen_tail_192_hash_done
+	ldp	x11, x12, [x3], 16
+	adds	x8, x8, x11
+	adcs	x9, x9, x12
+	adc	x10, x10, x15
+	mul	x11, x8, x16     // [t2:t1:t0] = [acc2:acc1:acc0] * r0
+	umulh	x12, x8, x16
+	mul	x13, x9, x16
+	umulh	x14, x9, x16
+	adds	x12, x12, x13
+	mul	x13, x10, x16
+	adc	x13, x13, x14
+	mul	x14, x8, x17       // [t3:t2:t1:t0] = [acc2:acc1:acc0] * [r1:r0]
+	umulh	x8, x8, x17
+	adds	x12, x12, x14
+	mul	x14, x9, x17
+	umulh	x9, x9, x17
+	adcs	x14, x14, x8
+	mul	x10, x10, x17
+	adc	x10, x10, x9
+	adds	x13, x13, x14
+	adc	x14, x10, xzr
+	and	x10, x13, #3         // At this point acc2 is 2 bits at most (value of 3)
+	and	x8, x13, #-4
+	extr	x13, x14, x13, #2
+	adds	x8, x8, x11
+	lsr	x11, x14, #2
+	adc	x9, x14, x11        // No carry out since t0 is 61 bits and t3 is 63 bits
+	adds	x8, x8, x13
+	adcs	x9, x9, x12
+	adc	x10, x10, xzr      // At this point acc2 has the value of 4 at most
+	sub	x4, x4, #1
+	b	Lopen_tail_192_hash
+
+Lopen_tail_192_hash_done:
+
+	add	v0.4s, v0.4s, v24.4s
+	add	v1.4s, v1.4s, v24.4s
+	add	v2.4s, v2.4s, v24.4s
+	add	v5.4s, v5.4s, v28.4s
+	add	v6.4s, v6.4s, v28.4s
+	add	v7.4s, v7.4s, v28.4s
+	add	v10.4s, v10.4s, v29.4s
+	add	v11.4s, v11.4s, v29.4s
+	add	v12.4s, v12.4s, v29.4s
+	add	v15.4s, v15.4s, v30.4s
+	add	v16.4s, v16.4s, v30.4s
+	add	v17.4s, v17.4s, v30.4s
+
+	add	v15.4s, v15.4s, v21.4s
+	add	v16.4s, v16.4s, v23.4s
+	add	v17.4s, v17.4s, v22.4s
+
+	ld1	{v20.16b - v23.16b}, [x1], #64
+
+	eor	v20.16b, v20.16b, v1.16b
+	eor	v21.16b, v21.16b, v6.16b
+	eor	v22.16b, v22.16b, v11.16b
+	eor	v23.16b, v23.16b, v16.16b
+
+	st1	{v20.16b - v23.16b}, [x0], #64
+
+	ld1	{v20.16b - v23.16b}, [x1], #64
+
+	eor	v20.16b, v20.16b, v2.16b
+	eor	v21.16b, v21.16b, v7.16b
+	eor	v22.16b, v22.16b, v12.16b
+	eor	v23.16b, v23.16b, v17.16b
+
+	st1	{v20.16b - v23.16b}, [x0], #64
+
+	sub	x2, x2, #128
+	b	Lopen_tail_64_store
+
+Lopen_tail_128:
+     // We need two more blocks
+	mov	v0.16b, v24.16b
+	mov	v1.16b, v24.16b
+	mov	v5.16b, v28.16b
+	mov	v6.16b, v28.16b
+	mov	v10.16b, v29.16b
+	mov	v11.16b, v29.16b
+	mov	v15.16b, v30.16b
+	mov	v16.16b, v30.16b
+	eor	v23.16b, v23.16b, v23.16b
+	eor	v22.16b, v22.16b, v22.16b
+	ins	v23.s[0], v25.s[0]
+	ins	v22.d[0], x15
+	add	v22.4s, v22.4s, v23.4s
+
+	add	v15.4s, v15.4s, v22.4s
+	add	v16.4s, v16.4s, v23.4s
+
+	mov	x6, #10
+	sub	x6, x6, x4
+
+Lopen_tail_128_rounds:
+	add	v0.4s, v0.4s, v5.4s
+	eor	v15.16b, v15.16b, v0.16b
+	rev32	v15.8h, v15.8h
+
+	add	v10.4s, v10.4s, v15.4s
+	eor	v5.16b, v5.16b, v10.16b
+	ushr	v20.4s, v5.4s, #20
+	sli	v20.4s, v5.4s, #12
+	add	v0.4s, v0.4s, v20.4s
+	eor	v15.16b, v15.16b, v0.16b
+	tbl	v15.16b, {v15.16b}, v26.16b
+
+	add	v10.4s, v10.4s, v15.4s
+	eor	v20.16b, v20.16b, v10.16b
+	ushr	v5.4s, v20.4s, #25
+	sli	v5.4s, v20.4s, #7
+	ext	v5.16b, v5.16b, v5.16b, #4
+	ext	v10.16b, v10.16b, v10.16b, #8
+	ext	v15.16b, v15.16b, v15.16b, #12
+	add	v1.4s, v1.4s, v6.4s
+	eor	v16.16b, v16.16b, v1.16b
+	rev32	v16.8h, v16.8h
+
+	add	v11.4s, v11.4s, v16.4s
+	eor	v6.16b, v6.16b, v11.16b
+	ushr	v20.4s, v6.4s, #20
+	sli	v20.4s, v6.4s, #12
+	add	v1.4s, v1.4s, v20.4s
+	eor	v16.16b, v16.16b, v1.16b
+	tbl	v16.16b, {v16.16b}, v26.16b
+
+	add	v11.4s, v11.4s, v16.4s
+	eor	v20.16b, v20.16b, v11.16b
+	ushr	v6.4s, v20.4s, #25
+	sli	v6.4s, v20.4s, #7
+	ext	v6.16b, v6.16b, v6.16b, #4
+	ext	v11.16b, v11.16b, v11.16b, #8
+	ext	v16.16b, v16.16b, v16.16b, #12
+	add	v0.4s, v0.4s, v5.4s
+	eor	v15.16b, v15.16b, v0.16b
+	rev32	v15.8h, v15.8h
+
+	add	v10.4s, v10.4s, v15.4s
+	eor	v5.16b, v5.16b, v10.16b
+	ushr	v20.4s, v5.4s, #20
+	sli	v20.4s, v5.4s, #12
+	add	v0.4s, v0.4s, v20.4s
+	eor	v15.16b, v15.16b, v0.16b
+	tbl	v15.16b, {v15.16b}, v26.16b
+
+	add	v10.4s, v10.4s, v15.4s
+	eor	v20.16b, v20.16b, v10.16b
+	ushr	v5.4s, v20.4s, #25
+	sli	v5.4s, v20.4s, #7
+	ext	v5.16b, v5.16b, v5.16b, #12
+	ext	v10.16b, v10.16b, v10.16b, #8
+	ext	v15.16b, v15.16b, v15.16b, #4
+	add	v1.4s, v1.4s, v6.4s
+	eor	v16.16b, v16.16b, v1.16b
+	rev32	v16.8h, v16.8h
+
+	add	v11.4s, v11.4s, v16.4s
+	eor	v6.16b, v6.16b, v11.16b
+	ushr	v20.4s, v6.4s, #20
+	sli	v20.4s, v6.4s, #12
+	add	v1.4s, v1.4s, v20.4s
+	eor	v16.16b, v16.16b, v1.16b
+	tbl	v16.16b, {v16.16b}, v26.16b
+
+	add	v11.4s, v11.4s, v16.4s
+	eor	v20.16b, v20.16b, v11.16b
+	ushr	v6.4s, v20.4s, #25
+	sli	v6.4s, v20.4s, #7
+	ext	v6.16b, v6.16b, v6.16b, #12
+	ext	v11.16b, v11.16b, v11.16b, #8
+	ext	v16.16b, v16.16b, v16.16b, #4
+	subs	x6, x6, #1
+	b.gt	Lopen_tail_128_rounds
+	cbz	x4, Lopen_tail_128_rounds_done
+	subs	x4, x4, #1
+	ldp	x11, x12, [x3], 16
+	adds	x8, x8, x11
+	adcs	x9, x9, x12
+	adc	x10, x10, x15
+	mul	x11, x8, x16     // [t2:t1:t0] = [acc2:acc1:acc0] * r0
+	umulh	x12, x8, x16
+	mul	x13, x9, x16
+	umulh	x14, x9, x16
+	adds	x12, x12, x13
+	mul	x13, x10, x16
+	adc	x13, x13, x14
+	mul	x14, x8, x17       // [t3:t2:t1:t0] = [acc2:acc1:acc0] * [r1:r0]
+	umulh	x8, x8, x17
+	adds	x12, x12, x14
+	mul	x14, x9, x17
+	umulh	x9, x9, x17
+	adcs	x14, x14, x8
+	mul	x10, x10, x17
+	adc	x10, x10, x9
+	adds	x13, x13, x14
+	adc	x14, x10, xzr
+	and	x10, x13, #3         // At this point acc2 is 2 bits at most (value of 3)
+	and	x8, x13, #-4
+	extr	x13, x14, x13, #2
+	adds	x8, x8, x11
+	lsr	x11, x14, #2
+	adc	x9, x14, x11        // No carry out since t0 is 61 bits and t3 is 63 bits
+	adds	x8, x8, x13
+	adcs	x9, x9, x12
+	adc	x10, x10, xzr      // At this point acc2 has the value of 4 at most
+	b	Lopen_tail_128_rounds
+
+Lopen_tail_128_rounds_done:
+	add	v0.4s, v0.4s, v24.4s
+	add	v1.4s, v1.4s, v24.4s
+	add	v5.4s, v5.4s, v28.4s
+	add	v6.4s, v6.4s, v28.4s
+	add	v10.4s, v10.4s, v29.4s
+	add	v11.4s, v11.4s, v29.4s
+	add	v15.4s, v15.4s, v30.4s
+	add	v16.4s, v16.4s, v30.4s
+	add	v15.4s, v15.4s, v22.4s
+	add	v16.4s, v16.4s, v23.4s
+
+	ld1	{v20.16b - v23.16b}, [x1], #64
+
+	eor	v20.16b, v20.16b, v1.16b
+	eor	v21.16b, v21.16b, v6.16b
+	eor	v22.16b, v22.16b, v11.16b
+	eor	v23.16b, v23.16b, v16.16b
+
+	st1	{v20.16b - v23.16b}, [x0], #64
+	sub	x2, x2, #64
+
+	b	Lopen_tail_64_store
+
+Lopen_tail_64:
+    // We just need a single block
+	mov	v0.16b, v24.16b
+	mov	v5.16b, v28.16b
+	mov	v10.16b, v29.16b
+	mov	v15.16b, v30.16b
+	eor	v23.16b, v23.16b, v23.16b
+	ins	v23.s[0], v25.s[0]
+	add	v15.4s, v15.4s, v23.4s
+
+	mov	x6, #10
+	sub	x6, x6, x4
+
+Lopen_tail_64_rounds:
+	add	v0.4s, v0.4s, v5.4s
+	eor	v15.16b, v15.16b, v0.16b
+	rev32	v15.8h, v15.8h
+
+	add	v10.4s, v10.4s, v15.4s
+	eor	v5.16b, v5.16b, v10.16b
+	ushr	v20.4s, v5.4s, #20
+	sli	v20.4s, v5.4s, #12
+	add	v0.4s, v0.4s, v20.4s
+	eor	v15.16b, v15.16b, v0.16b
+	tbl	v15.16b, {v15.16b}, v26.16b
+
+	add	v10.4s, v10.4s, v15.4s
+	eor	v20.16b, v20.16b, v10.16b
+	ushr	v5.4s, v20.4s, #25
+	sli	v5.4s, v20.4s, #7
+	ext	v5.16b, v5.16b, v5.16b, #4
+	ext	v10.16b, v10.16b, v10.16b, #8
+	ext	v15.16b, v15.16b, v15.16b, #12
+	add	v0.4s, v0.4s, v5.4s
+	eor	v15.16b, v15.16b, v0.16b
+	rev32	v15.8h, v15.8h
+
+	add	v10.4s, v10.4s, v15.4s
+	eor	v5.16b, v5.16b, v10.16b
+	ushr	v20.4s, v5.4s, #20
+	sli	v20.4s, v5.4s, #12
+	add	v0.4s, v0.4s, v20.4s
+	eor	v15.16b, v15.16b, v0.16b
+	tbl	v15.16b, {v15.16b}, v26.16b
+
+	add	v10.4s, v10.4s, v15.4s
+	eor	v20.16b, v20.16b, v10.16b
+	ushr	v5.4s, v20.4s, #25
+	sli	v5.4s, v20.4s, #7
+	ext	v5.16b, v5.16b, v5.16b, #12
+	ext	v10.16b, v10.16b, v10.16b, #8
+	ext	v15.16b, v15.16b, v15.16b, #4
+	subs	x6, x6, #1
+	b.gt	Lopen_tail_64_rounds
+	cbz	x4, Lopen_tail_64_rounds_done
+	subs	x4, x4, #1
+	ldp	x11, x12, [x3], 16
+	adds	x8, x8, x11
+	adcs	x9, x9, x12
+	adc	x10, x10, x15
+	mul	x11, x8, x16     // [t2:t1:t0] = [acc2:acc1:acc0] * r0
+	umulh	x12, x8, x16
+	mul	x13, x9, x16
+	umulh	x14, x9, x16
+	adds	x12, x12, x13
+	mul	x13, x10, x16
+	adc	x13, x13, x14
+	mul	x14, x8, x17       // [t3:t2:t1:t0] = [acc2:acc1:acc0] * [r1:r0]
+	umulh	x8, x8, x17
+	adds	x12, x12, x14
+	mul	x14, x9, x17
+	umulh	x9, x9, x17
+	adcs	x14, x14, x8
+	mul	x10, x10, x17
+	adc	x10, x10, x9
+	adds	x13, x13, x14
+	adc	x14, x10, xzr
+	and	x10, x13, #3         // At this point acc2 is 2 bits at most (value of 3)
+	and	x8, x13, #-4
+	extr	x13, x14, x13, #2
+	adds	x8, x8, x11
+	lsr	x11, x14, #2
+	adc	x9, x14, x11        // No carry out since t0 is 61 bits and t3 is 63 bits
+	adds	x8, x8, x13
+	adcs	x9, x9, x12
+	adc	x10, x10, xzr      // At this point acc2 has the value of 4 at most
+	b	Lopen_tail_64_rounds
+
+Lopen_tail_64_rounds_done:
+	add	v0.4s, v0.4s, v24.4s
+	add	v5.4s, v5.4s, v28.4s
+	add	v10.4s, v10.4s, v29.4s
+	add	v15.4s, v15.4s, v30.4s
+	add	v15.4s, v15.4s, v23.4s
+
+Lopen_tail_64_store:
+	cmp	x2, #16
+	b.lt	Lopen_tail_16
+
+	ld1	{v20.16b}, [x1], #16
+	eor	v20.16b, v20.16b, v0.16b
+	st1	{v20.16b}, [x0], #16
+	mov	v0.16b, v5.16b
+	mov	v5.16b, v10.16b
+	mov	v10.16b, v15.16b
+	sub	x2, x2, #16
+	b	Lopen_tail_64_store
+
+Lopen_tail_16:
+    // Here we handle the last [0,16) bytes that require a padded block
+	cbz	x2, Lopen_finalize
+
+	eor	v20.16b, v20.16b, v20.16b // Use T0 to load the ciphertext
+	eor	v21.16b, v21.16b, v21.16b // Use T1 to generate an AND mask
+	not	v22.16b, v20.16b
+
+	add	x7, x1, x2
+	mov	x6, x2
+
+Lopen_tail_16_compose:
+	ext	v20.16b, v20.16b, v20.16b, #15
+	ldrb	w11, [x7, #-1]!
+	mov	v20.b[0], w11
+	ext	v21.16b, v22.16b, v21.16b, #15
+	subs	x2, x2, #1
+	b.gt	Lopen_tail_16_compose
+
+	and	v20.16b, v20.16b, v21.16b
+    // Hash in the final padded block
+	mov	x11, v20.d[0]
+	mov	x12, v20.d[1]
+	adds	x8, x8, x11
+	adcs	x9, x9, x12
+	adc	x10, x10, x15
+	mul	x11, x8, x16     // [t2:t1:t0] = [acc2:acc1:acc0] * r0
+	umulh	x12, x8, x16
+	mul	x13, x9, x16
+	umulh	x14, x9, x16
+	adds	x12, x12, x13
+	mul	x13, x10, x16
+	adc	x13, x13, x14
+	mul	x14, x8, x17       // [t3:t2:t1:t0] = [acc2:acc1:acc0] * [r1:r0]
+	umulh	x8, x8, x17
+	adds	x12, x12, x14
+	mul	x14, x9, x17
+	umulh	x9, x9, x17
+	adcs	x14, x14, x8
+	mul	x10, x10, x17
+	adc	x10, x10, x9
+	adds	x13, x13, x14
+	adc	x14, x10, xzr
+	and	x10, x13, #3         // At this point acc2 is 2 bits at most (value of 3)
+	and	x8, x13, #-4
+	extr	x13, x14, x13, #2
+	adds	x8, x8, x11
+	lsr	x11, x14, #2
+	adc	x9, x14, x11        // No carry out since t0 is 61 bits and t3 is 63 bits
+	adds	x8, x8, x13
+	adcs	x9, x9, x12
+	adc	x10, x10, xzr      // At this point acc2 has the value of 4 at most
+	eor	v20.16b, v20.16b, v0.16b
+
+Lopen_tail_16_store:
+	umov	w11, v20.b[0]
+	strb	w11, [x0], #1
+	ext	v20.16b, v20.16b, v20.16b, #1
+	subs	x6, x6, #1
+	b.gt	Lopen_tail_16_store
+
+Lopen_finalize:
+	mov	x11, v31.d[0]
+	mov	x12, v31.d[1]
+	adds	x8, x8, x11
+	adcs	x9, x9, x12
+	adc	x10, x10, x15
+	mul	x11, x8, x16     // [t2:t1:t0] = [acc2:acc1:acc0] * r0
+	umulh	x12, x8, x16
+	mul	x13, x9, x16
+	umulh	x14, x9, x16
+	adds	x12, x12, x13
+	mul	x13, x10, x16
+	adc	x13, x13, x14
+	mul	x14, x8, x17       // [t3:t2:t1:t0] = [acc2:acc1:acc0] * [r1:r0]
+	umulh	x8, x8, x17
+	adds	x12, x12, x14
+	mul	x14, x9, x17
+	umulh	x9, x9, x17
+	adcs	x14, x14, x8
+	mul	x10, x10, x17
+	adc	x10, x10, x9
+	adds	x13, x13, x14
+	adc	x14, x10, xzr
+	and	x10, x13, #3         // At this point acc2 is 2 bits at most (value of 3)
+	and	x8, x13, #-4
+	extr	x13, x14, x13, #2
+	adds	x8, x8, x11
+	lsr	x11, x14, #2
+	adc	x9, x14, x11        // No carry out since t0 is 61 bits and t3 is 63 bits
+	adds	x8, x8, x13
+	adcs	x9, x9, x12
+	adc	x10, x10, xzr      // At this point acc2 has the value of 4 at most
+    // Final reduction step
+	sub	x12, xzr, x15
+	orr	x13, xzr, #3
+	subs	x11, x8, #-5
+	sbcs	x12, x9, x12
+	sbcs	x13, x10, x13
+	csel	x8, x11, x8, cs
+	csel	x9, x12, x9, cs
+	csel	x10, x13, x10, cs
+	mov	x11, v27.d[0]
+	mov	x12, v27.d[1]
+	adds	x8, x8, x11
+	adcs	x9, x9, x12
+	adc	x10, x10, x15
+
+	stp	x8, x9, [x5]
+
+	ldp	d8, d9, [sp, #16]
+	ldp	d10, d11, [sp, #32]
+	ldp	d12, d13, [sp, #48]
+	ldp	d14, d15, [sp, #64]
+.cfi_restore	b15
+.cfi_restore	b14
+.cfi_restore	b13
+.cfi_restore	b12
+.cfi_restore	b11
+.cfi_restore	b10
+.cfi_restore	b9
+.cfi_restore	b8
+	ldp	x29, x30, [sp], 80
+.cfi_restore	w29
+.cfi_restore	w30
+.cfi_def_cfa_offset	0
+	AARCH64_VALIDATE_LINK_REGISTER
+	ret
+
+Lopen_128:
+    // On some architectures preparing 5 blocks for small buffers is wasteful
+	eor	v25.16b, v25.16b, v25.16b
+	mov	x11, #1
+	mov	v25.s[0], w11
+	mov	v0.16b, v24.16b
+	mov	v1.16b, v24.16b
+	mov	v2.16b, v24.16b
+	mov	v5.16b, v28.16b
+	mov	v6.16b, v28.16b
+	mov	v7.16b, v28.16b
+	mov	v10.16b, v29.16b
+	mov	v11.16b, v29.16b
+	mov	v12.16b, v29.16b
+	mov	v17.16b, v30.16b
+	add	v15.4s, v17.4s, v25.4s
+	add	v16.4s, v15.4s, v25.4s
+
+	mov	x6, #10
+
+Lopen_128_rounds:
+	add	v0.4s, v0.4s, v5.4s
+	add	v1.4s, v1.4s, v6.4s
+	add	v2.4s, v2.4s, v7.4s
+	eor	v15.16b, v15.16b, v0.16b
+	eor	v16.16b, v16.16b, v1.16b
+	eor	v17.16b, v17.16b, v2.16b
+	rev32	v15.8h, v15.8h
+	rev32	v16.8h, v16.8h
+	rev32	v17.8h, v17.8h
+
+	add	v10.4s, v10.4s, v15.4s
+	add	v11.4s, v11.4s, v16.4s
+	add	v12.4s, v12.4s, v17.4s
+	eor	v5.16b, v5.16b, v10.16b
+	eor	v6.16b, v6.16b, v11.16b
+	eor	v7.16b, v7.16b, v12.16b
+	ushr	v20.4s, v5.4s, #20
+	sli	v20.4s, v5.4s, #12
+	ushr	v5.4s, v6.4s, #20
+	sli	v5.4s, v6.4s, #12
+	ushr	v6.4s, v7.4s, #20
+	sli	v6.4s, v7.4s, #12
+
+	add	v0.4s, v0.4s, v20.4s
+	add	v1.4s, v1.4s, v5.4s
+	add	v2.4s, v2.4s, v6.4s
+	eor	v15.16b, v15.16b, v0.16b
+	eor	v16.16b, v16.16b, v1.16b
+	eor	v17.16b, v17.16b, v2.16b
+	tbl	v15.16b, {v15.16b}, v26.16b
+	tbl	v16.16b, {v16.16b}, v26.16b
+	tbl	v17.16b, {v17.16b}, v26.16b
+
+	add	v10.4s, v10.4s, v15.4s
+	add	v11.4s, v11.4s, v16.4s
+	add	v12.4s, v12.4s, v17.4s
+	eor	v20.16b, v20.16b, v10.16b
+	eor	v5.16b, v5.16b, v11.16b
+	eor	v6.16b, v6.16b, v12.16b
+	ushr	v7.4s, v6.4s, #25
+	sli	v7.4s, v6.4s, #7
+	ushr	v6.4s, v5.4s, #25
+	sli	v6.4s, v5.4s, #7
+	ushr	v5.4s, v20.4s, #25
+	sli	v5.4s, v20.4s, #7
+
+	ext	v5.16b, v5.16b, v5.16b, #4
+	ext	v6.16b, v6.16b, v6.16b, #4
+	ext	v7.16b, v7.16b, v7.16b, #4
+
+	ext	v10.16b, v10.16b, v10.16b, #8
+	ext	v11.16b, v11.16b, v11.16b, #8
+	ext	v12.16b, v12.16b, v12.16b, #8
+
+	ext	v15.16b, v15.16b, v15.16b, #12
+	ext	v16.16b, v16.16b, v16.16b, #12
+	ext	v17.16b, v17.16b, v17.16b, #12
+	add	v0.4s, v0.4s, v5.4s
+	add	v1.4s, v1.4s, v6.4s
+	add	v2.4s, v2.4s, v7.4s
+	eor	v15.16b, v15.16b, v0.16b
+	eor	v16.16b, v16.16b, v1.16b
+	eor	v17.16b, v17.16b, v2.16b
+	rev32	v15.8h, v15.8h
+	rev32	v16.8h, v16.8h
+	rev32	v17.8h, v17.8h
+
+	add	v10.4s, v10.4s, v15.4s
+	add	v11.4s, v11.4s, v16.4s
+	add	v12.4s, v12.4s, v17.4s
+	eor	v5.16b, v5.16b, v10.16b
+	eor	v6.16b, v6.16b, v11.16b
+	eor	v7.16b, v7.16b, v12.16b
+	ushr	v20.4s, v5.4s, #20
+	sli	v20.4s, v5.4s, #12
+	ushr	v5.4s, v6.4s, #20
+	sli	v5.4s, v6.4s, #12
+	ushr	v6.4s, v7.4s, #20
+	sli	v6.4s, v7.4s, #12
+
+	add	v0.4s, v0.4s, v20.4s
+	add	v1.4s, v1.4s, v5.4s
+	add	v2.4s, v2.4s, v6.4s
+	eor	v15.16b, v15.16b, v0.16b
+	eor	v16.16b, v16.16b, v1.16b
+	eor	v17.16b, v17.16b, v2.16b
+	tbl	v15.16b, {v15.16b}, v26.16b
+	tbl	v16.16b, {v16.16b}, v26.16b
+	tbl	v17.16b, {v17.16b}, v26.16b
+
+	add	v10.4s, v10.4s, v15.4s
+	add	v11.4s, v11.4s, v16.4s
+	add	v12.4s, v12.4s, v17.4s
+	eor	v20.16b, v20.16b, v10.16b
+	eor	v5.16b, v5.16b, v11.16b
+	eor	v6.16b, v6.16b, v12.16b
+	ushr	v7.4s, v6.4s, #25
+	sli	v7.4s, v6.4s, #7
+	ushr	v6.4s, v5.4s, #25
+	sli	v6.4s, v5.4s, #7
+	ushr	v5.4s, v20.4s, #25
+	sli	v5.4s, v20.4s, #7
+
+	ext	v5.16b, v5.16b, v5.16b, #12
+	ext	v6.16b, v6.16b, v6.16b, #12
+	ext	v7.16b, v7.16b, v7.16b, #12
+
+	ext	v10.16b, v10.16b, v10.16b, #8
+	ext	v11.16b, v11.16b, v11.16b, #8
+	ext	v12.16b, v12.16b, v12.16b, #8
+
+	ext	v15.16b, v15.16b, v15.16b, #4
+	ext	v16.16b, v16.16b, v16.16b, #4
+	ext	v17.16b, v17.16b, v17.16b, #4
+	subs	x6, x6, #1
+	b.hi	Lopen_128_rounds
+
+	add	v0.4s, v0.4s, v24.4s
+	add	v1.4s, v1.4s, v24.4s
+	add	v2.4s, v2.4s, v24.4s
+
+	add	v5.4s, v5.4s, v28.4s
+	add	v6.4s, v6.4s, v28.4s
+	add	v7.4s, v7.4s, v28.4s
+
+	add	v10.4s, v10.4s, v29.4s
+	add	v11.4s, v11.4s, v29.4s
+
+	add	v30.4s, v30.4s, v25.4s
+	add	v15.4s, v15.4s, v30.4s
+	add	v30.4s, v30.4s, v25.4s
+	add	v16.4s, v16.4s, v30.4s
+
+	and	v2.16b, v2.16b, v27.16b
+	mov	x16, v2.d[0] // Move the R key to GPRs
+	mov	x17, v2.d[1]
+	mov	v27.16b, v7.16b // Store the S key
+
+	bl	Lpoly_hash_ad_internal
+
+Lopen_128_store:
+	cmp	x2, #64
+	b.lt	Lopen_128_store_64
+
+	ld1	{v20.16b - v23.16b}, [x1], #64
+
+	mov	x11, v20.d[0]
+	mov	x12, v20.d[1]
+	adds	x8, x8, x11
+	adcs	x9, x9, x12
+	adc	x10, x10, x15
+	mul	x11, x8, x16     // [t2:t1:t0] = [acc2:acc1:acc0] * r0
+	umulh	x12, x8, x16
+	mul	x13, x9, x16
+	umulh	x14, x9, x16
+	adds	x12, x12, x13
+	mul	x13, x10, x16
+	adc	x13, x13, x14
+	mul	x14, x8, x17       // [t3:t2:t1:t0] = [acc2:acc1:acc0] * [r1:r0]
+	umulh	x8, x8, x17
+	adds	x12, x12, x14
+	mul	x14, x9, x17
+	umulh	x9, x9, x17
+	adcs	x14, x14, x8
+	mul	x10, x10, x17
+	adc	x10, x10, x9
+	adds	x13, x13, x14
+	adc	x14, x10, xzr
+	and	x10, x13, #3         // At this point acc2 is 2 bits at most (value of 3)
+	and	x8, x13, #-4
+	extr	x13, x14, x13, #2
+	adds	x8, x8, x11
+	lsr	x11, x14, #2
+	adc	x9, x14, x11        // No carry out since t0 is 61 bits and t3 is 63 bits
+	adds	x8, x8, x13
+	adcs	x9, x9, x12
+	adc	x10, x10, xzr      // At this point acc2 has the value of 4 at most
+	mov	x11, v21.d[0]
+	mov	x12, v21.d[1]
+	adds	x8, x8, x11
+	adcs	x9, x9, x12
+	adc	x10, x10, x15
+	mul	x11, x8, x16     // [t2:t1:t0] = [acc2:acc1:acc0] * r0
+	umulh	x12, x8, x16
+	mul	x13, x9, x16
+	umulh	x14, x9, x16
+	adds	x12, x12, x13
+	mul	x13, x10, x16
+	adc	x13, x13, x14
+	mul	x14, x8, x17       // [t3:t2:t1:t0] = [acc2:acc1:acc0] * [r1:r0]
+	umulh	x8, x8, x17
+	adds	x12, x12, x14
+	mul	x14, x9, x17
+	umulh	x9, x9, x17
+	adcs	x14, x14, x8
+	mul	x10, x10, x17
+	adc	x10, x10, x9
+	adds	x13, x13, x14
+	adc	x14, x10, xzr
+	and	x10, x13, #3         // At this point acc2 is 2 bits at most (value of 3)
+	and	x8, x13, #-4
+	extr	x13, x14, x13, #2
+	adds	x8, x8, x11
+	lsr	x11, x14, #2
+	adc	x9, x14, x11        // No carry out since t0 is 61 bits and t3 is 63 bits
+	adds	x8, x8, x13
+	adcs	x9, x9, x12
+	adc	x10, x10, xzr      // At this point acc2 has the value of 4 at most
+	mov	x11, v22.d[0]
+	mov	x12, v22.d[1]
+	adds	x8, x8, x11
+	adcs	x9, x9, x12
+	adc	x10, x10, x15
+	mul	x11, x8, x16     // [t2:t1:t0] = [acc2:acc1:acc0] * r0
+	umulh	x12, x8, x16
+	mul	x13, x9, x16
+	umulh	x14, x9, x16
+	adds	x12, x12, x13
+	mul	x13, x10, x16
+	adc	x13, x13, x14
+	mul	x14, x8, x17       // [t3:t2:t1:t0] = [acc2:acc1:acc0] * [r1:r0]
+	umulh	x8, x8, x17
+	adds	x12, x12, x14
+	mul	x14, x9, x17
+	umulh	x9, x9, x17
+	adcs	x14, x14, x8
+	mul	x10, x10, x17
+	adc	x10, x10, x9
+	adds	x13, x13, x14
+	adc	x14, x10, xzr
+	and	x10, x13, #3         // At this point acc2 is 2 bits at most (value of 3)
+	and	x8, x13, #-4
+	extr	x13, x14, x13, #2
+	adds	x8, x8, x11
+	lsr	x11, x14, #2
+	adc	x9, x14, x11        // No carry out since t0 is 61 bits and t3 is 63 bits
+	adds	x8, x8, x13
+	adcs	x9, x9, x12
+	adc	x10, x10, xzr      // At this point acc2 has the value of 4 at most
+	mov	x11, v23.d[0]
+	mov	x12, v23.d[1]
+	adds	x8, x8, x11
+	adcs	x9, x9, x12
+	adc	x10, x10, x15
+	mul	x11, x8, x16     // [t2:t1:t0] = [acc2:acc1:acc0] * r0
+	umulh	x12, x8, x16
+	mul	x13, x9, x16
+	umulh	x14, x9, x16
+	adds	x12, x12, x13
+	mul	x13, x10, x16
+	adc	x13, x13, x14
+	mul	x14, x8, x17       // [t3:t2:t1:t0] = [acc2:acc1:acc0] * [r1:r0]
+	umulh	x8, x8, x17
+	adds	x12, x12, x14
+	mul	x14, x9, x17
+	umulh	x9, x9, x17
+	adcs	x14, x14, x8
+	mul	x10, x10, x17
+	adc	x10, x10, x9
+	adds	x13, x13, x14
+	adc	x14, x10, xzr
+	and	x10, x13, #3         // At this point acc2 is 2 bits at most (value of 3)
+	and	x8, x13, #-4
+	extr	x13, x14, x13, #2
+	adds	x8, x8, x11
+	lsr	x11, x14, #2
+	adc	x9, x14, x11        // No carry out since t0 is 61 bits and t3 is 63 bits
+	adds	x8, x8, x13
+	adcs	x9, x9, x12
+	adc	x10, x10, xzr      // At this point acc2 has the value of 4 at most
+
+	eor	v20.16b, v20.16b, v0.16b
+	eor	v21.16b, v21.16b, v5.16b
+	eor	v22.16b, v22.16b, v10.16b
+	eor	v23.16b, v23.16b, v15.16b
+
+	st1	{v20.16b - v23.16b}, [x0], #64
+
+	sub	x2, x2, #64
+
+	mov	v0.16b, v1.16b
+	mov	v5.16b, v6.16b
+	mov	v10.16b, v11.16b
+	mov	v15.16b, v16.16b
+
+Lopen_128_store_64:
+
+	lsr	x4, x2, #4
+	mov	x3, x1
+
+Lopen_128_hash_64:
+	cbz	x4, Lopen_tail_64_store
+	ldp	x11, x12, [x3], 16
+	adds	x8, x8, x11
+	adcs	x9, x9, x12
+	adc	x10, x10, x15
+	mul	x11, x8, x16     // [t2:t1:t0] = [acc2:acc1:acc0] * r0
+	umulh	x12, x8, x16
+	mul	x13, x9, x16
+	umulh	x14, x9, x16
+	adds	x12, x12, x13
+	mul	x13, x10, x16
+	adc	x13, x13, x14
+	mul	x14, x8, x17       // [t3:t2:t1:t0] = [acc2:acc1:acc0] * [r1:r0]
+	umulh	x8, x8, x17
+	adds	x12, x12, x14
+	mul	x14, x9, x17
+	umulh	x9, x9, x17
+	adcs	x14, x14, x8
+	mul	x10, x10, x17
+	adc	x10, x10, x9
+	adds	x13, x13, x14
+	adc	x14, x10, xzr
+	and	x10, x13, #3         // At this point acc2 is 2 bits at most (value of 3)
+	and	x8, x13, #-4
+	extr	x13, x14, x13, #2
+	adds	x8, x8, x11
+	lsr	x11, x14, #2
+	adc	x9, x14, x11        // No carry out since t0 is 61 bits and t3 is 63 bits
+	adds	x8, x8, x13
+	adcs	x9, x9, x12
+	adc	x10, x10, xzr      // At this point acc2 has the value of 4 at most
+	sub	x4, x4, #1
+	b	Lopen_128_hash_64
+.cfi_endproc
+
+#endif  // !OPENSSL_NO_ASM && defined(OPENSSL_AARCH64) && defined(_WIN32)