diff mbox series

[1/5] An optimized Chacha20 implementation with 8-way unrolling for ppc64le.

Message ID 20230424184726.2091-2-dtsen@linux.ibm.com (mailing list archive)
State Superseded
Delegated to: Herbert Xu
Headers show
Series crypto: Accelerated Chacha20/Poly1305 implementation | expand

Commit Message

Danny Tsen April 24, 2023, 6:47 p.m. UTC 
Improve overall performance of chacha20 encrypt and decrypt operations
for Power10 or later CPU.

Signed-off-by: Danny Tsen <dtsen@linux.ibm.com>
---
 arch/powerpc/crypto/chacha-p10le-8x.S | 842 ++++++++++++++++++++++++++
 1 file changed, 842 insertions(+)
 create mode 100644 arch/powerpc/crypto/chacha-p10le-8x.S

Comments

Elliott, Robert (Servers) April 24, 2023, 8:40 p.m. UTC | #1 
> +# Copyright 2023- IBM Inc. All rights reserved

I don't think any such entity exists - you probably mean IBM Corporation.
Danny Tsen April 24, 2023, 11:07 p.m. UTC | #2 
This is recommended template to use for IBM copyright.

Thanks.

-Danny

On 4/24/23 3:40 PM, Elliott, Robert (Servers) wrote:
>> +# Copyright 2023- IBM Inc. All rights reserved
> I don't think any such entity exists - you probably mean IBM Corporation.
Michael Ellerman April 25, 2023, 12:02 p.m. UTC | #3 
Danny Tsen <dtsen@linux.ibm.com> writes:
> This is recommended template to use for IBM copyright.

According to who?

The documentation I've seen specifies "IBM Corp." or "IBM Corporation".

cheers
Danny Tsen April 25, 2023, 12:08 p.m. UTC | #4 
Hi Michael,

It's in IBM repo.

Thanks.

-Danny

On 4/25/23 7:02 AM, Michael Ellerman wrote:
> Danny Tsen <dtsen@linux.ibm.com> writes:
>> This is recommended template to use for IBM copyright.
> According to who?
>
> The documentation I've seen specifies "IBM Corp." or "IBM Corporation".
>
> cheers
David Laight April 26, 2023, 8:19 a.m. UTC | #5 
From: Danny Tsen
> Sent: 24 April 2023 19:47
> 
> Improve overall performance of chacha20 encrypt and decrypt operations
> for Power10 or later CPU.
> 
> Signed-off-by: Danny Tsen <dtsen@linux.ibm.com>
> ---
>  arch/powerpc/crypto/chacha-p10le-8x.S | 842 ++++++++++++++++++++++++++
>  1 file changed, 842 insertions(+)
>  create mode 100644 arch/powerpc/crypto/chacha-p10le-8x.S
...
> +.macro QT_loop_8x
> +	# QR(v0, v4,  v8, v12, v1, v5,  v9, v13, v2, v6, v10, v14, v3, v7, v11, v15)
> +	xxlor	0, 32+25, 32+25
> +	xxlor	32+25, 20, 20
> +	vadduwm 0, 0, 4
> +	vadduwm 1, 1, 5
> +	vadduwm 2, 2, 6
> +	vadduwm 3, 3, 7
> +	  vadduwm 16, 16, 20
> +	  vadduwm 17, 17, 21
> +	  vadduwm 18, 18, 22
> +	  vadduwm 19, 19, 23
> +
> +	  vpermxor 12, 12, 0, 25
> +	  vpermxor 13, 13, 1, 25
> +	  vpermxor 14, 14, 2, 25
> +	  vpermxor 15, 15, 3, 25
> +	  vpermxor 28, 28, 16, 25
> +	  vpermxor 29, 29, 17, 25
> +	  vpermxor 30, 30, 18, 25
> +	  vpermxor 31, 31, 19, 25
> +	xxlor	32+25, 0, 0
> +	vadduwm 8, 8, 12
> +	vadduwm 9, 9, 13
> +	vadduwm 10, 10, 14
> +	vadduwm 11, 11, 15
...

Is it just me or is all this code just complete jibberish?

There really ought to be enough comments so that it is possible
to check that the code is doing something that looks like chacha20
without spending all day tracking register numbers through
hundreds of lines of assembler.

I also wonder how much faster the 8-way unroll is?
On modern cpu with 'out of order' execute (etc) it is
not impossible to get the loop operations 'for free'
because they use execution units that are otherwise idle.

Massive loop unrolling is so 1980's.

	David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
Registration No: 1397386 (Wales)
diff mbox series

Patch

diff --git a/arch/powerpc/crypto/chacha-p10le-8x.S b/arch/powerpc/crypto/chacha-p10le-8x.S
new file mode 100644
index 000000000000..7c15d17101d7
--- /dev/null
+++ b/arch/powerpc/crypto/chacha-p10le-8x.S
@@ -0,0 +1,842 @@ 
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+#
+# Accelerated chacha20 implementation for ppc64le.
+#
+# Copyright 2023- IBM Inc. All rights reserved
+#
+#===================================================================================
+# Written by Danny Tsen <dtsen@us.ibm.com>
+#
+# chacha_p10le_8x(u32 *state, byte *dst, const byte *src,
+#				 size_t len, int nrounds);
+#
+# do rounds,  8 quarter rounds
+# 1.  a += b; d ^= a; d <<<= 16;
+# 2.  c += d; b ^= c; b <<<= 12;
+# 3.  a += b; d ^= a; d <<<= 8;
+# 4.  c += d; b ^= c; b <<<= 7
+#
+# row1 = (row1 + row2),  row4 = row1 xor row4,  row4 rotate each word by 16
+# row3 = (row3 + row4),  row2 = row3 xor row2,  row2 rotate each word by 12
+# row1 = (row1 + row2), row4 = row1 xor row4,  row4 rotate each word by 8
+# row3 = (row3 + row4), row2 = row3 xor row2,  row2 rotate each word by 7
+#
+# 4 blocks (a b c d)
+#
+# a0 b0 c0 d0
+# a1 b1 c1 d1
+# ...
+# a4 b4 c4 d4
+# ...
+# a8 b8 c8 d8
+# ...
+# a12 b12 c12 d12
+# a13 ...
+# a14 ...
+# a15 b15 c15 d15
+#
+# Column round (v0, v4,  v8, v12, v1, v5,  v9, v13, v2, v6, v10, v14, v3, v7, v11, v15)
+# Diagnal round (v0, v5, v10, v15, v1, v6, v11, v12, v2, v7,  v8, v13, v3, v4,  v9, v14)
+#
+
+#include <asm/ppc_asm.h>
+#include <asm/asm-offsets.h>
+#include <asm/asm-compat.h>
+#include <linux/linkage.h>
+
+.machine	"any"
+.text
+
+.macro	SAVE_GPR GPR OFFSET FRAME
+	std	\GPR,\OFFSET(\FRAME)
+.endm
+
+.macro	SAVE_VRS VRS OFFSET FRAME
+	li	16, \OFFSET
+	stvx	\VRS, 16, \FRAME
+.endm
+
+.macro	SAVE_VSX VSX OFFSET FRAME
+	li	16, \OFFSET
+	stxvx	\VSX, 16, \FRAME
+.endm
+
+.macro	RESTORE_GPR GPR OFFSET FRAME
+	ld	\GPR,\OFFSET(\FRAME)
+.endm
+
+.macro	RESTORE_VRS VRS OFFSET FRAME
+	li	16, \OFFSET
+	lvx	\VRS, 16, \FRAME
+.endm
+
+.macro	RESTORE_VSX VSX OFFSET FRAME
+	li	16, \OFFSET
+	lxvx	\VSX, 16, \FRAME
+.endm
+
+.macro SAVE_REGS
+	mflr 0
+	std 0, 16(1)
+	stdu 1,-752(1)
+
+	SAVE_GPR 14, 112, 1
+	SAVE_GPR 15, 120, 1
+	SAVE_GPR 16, 128, 1
+	SAVE_GPR 17, 136, 1
+	SAVE_GPR 18, 144, 1
+	SAVE_GPR 19, 152, 1
+	SAVE_GPR 20, 160, 1
+	SAVE_GPR 21, 168, 1
+	SAVE_GPR 22, 176, 1
+	SAVE_GPR 23, 184, 1
+	SAVE_GPR 24, 192, 1
+	SAVE_GPR 25, 200, 1
+	SAVE_GPR 26, 208, 1
+	SAVE_GPR 27, 216, 1
+	SAVE_GPR 28, 224, 1
+	SAVE_GPR 29, 232, 1
+	SAVE_GPR 30, 240, 1
+	SAVE_GPR 31, 248, 1
+
+	addi	9, 1, 256
+	SAVE_VRS 20, 0, 9
+	SAVE_VRS 21, 16, 9
+	SAVE_VRS 22, 32, 9
+	SAVE_VRS 23, 48, 9
+	SAVE_VRS 24, 64, 9
+	SAVE_VRS 25, 80, 9
+	SAVE_VRS 26, 96, 9
+	SAVE_VRS 27, 112, 9
+	SAVE_VRS 28, 128, 9
+	SAVE_VRS 29, 144, 9
+	SAVE_VRS 30, 160, 9
+	SAVE_VRS 31, 176, 9
+
+	SAVE_VSX 14, 192, 9
+	SAVE_VSX 15, 208, 9
+	SAVE_VSX 16, 224, 9
+	SAVE_VSX 17, 240, 9
+	SAVE_VSX 18, 256, 9
+	SAVE_VSX 19, 272, 9
+	SAVE_VSX 20, 288, 9
+	SAVE_VSX 21, 304, 9
+	SAVE_VSX 22, 320, 9
+	SAVE_VSX 23, 336, 9
+	SAVE_VSX 24, 352, 9
+	SAVE_VSX 25, 368, 9
+	SAVE_VSX 26, 384, 9
+	SAVE_VSX 27, 400, 9
+	SAVE_VSX 28, 416, 9
+	SAVE_VSX 29, 432, 9
+	SAVE_VSX 30, 448, 9
+	SAVE_VSX 31, 464, 9
+.endm # SAVE_REGS
+
+.macro RESTORE_REGS
+	addi	9, 1, 256
+	RESTORE_VRS 20, 0, 9
+	RESTORE_VRS 21, 16, 9
+	RESTORE_VRS 22, 32, 9
+	RESTORE_VRS 23, 48, 9
+	RESTORE_VRS 24, 64, 9
+	RESTORE_VRS 25, 80, 9
+	RESTORE_VRS 26, 96, 9
+	RESTORE_VRS 27, 112, 9
+	RESTORE_VRS 28, 128, 9
+	RESTORE_VRS 29, 144, 9
+	RESTORE_VRS 30, 160, 9
+	RESTORE_VRS 31, 176, 9
+
+	RESTORE_VSX 14, 192, 9
+	RESTORE_VSX 15, 208, 9
+	RESTORE_VSX 16, 224, 9
+	RESTORE_VSX 17, 240, 9
+	RESTORE_VSX 18, 256, 9
+	RESTORE_VSX 19, 272, 9
+	RESTORE_VSX 20, 288, 9
+	RESTORE_VSX 21, 304, 9
+	RESTORE_VSX 22, 320, 9
+	RESTORE_VSX 23, 336, 9
+	RESTORE_VSX 24, 352, 9
+	RESTORE_VSX 25, 368, 9
+	RESTORE_VSX 26, 384, 9
+	RESTORE_VSX 27, 400, 9
+	RESTORE_VSX 28, 416, 9
+	RESTORE_VSX 29, 432, 9
+	RESTORE_VSX 30, 448, 9
+	RESTORE_VSX 31, 464, 9
+
+	RESTORE_GPR 14, 112, 1
+	RESTORE_GPR 15, 120, 1
+	RESTORE_GPR 16, 128, 1
+	RESTORE_GPR 17, 136, 1
+	RESTORE_GPR 18, 144, 1
+	RESTORE_GPR 19, 152, 1
+	RESTORE_GPR 20, 160, 1
+	RESTORE_GPR 21, 168, 1
+	RESTORE_GPR 22, 176, 1
+	RESTORE_GPR 23, 184, 1
+	RESTORE_GPR 24, 192, 1
+	RESTORE_GPR 25, 200, 1
+	RESTORE_GPR 26, 208, 1
+	RESTORE_GPR 27, 216, 1
+	RESTORE_GPR 28, 224, 1
+	RESTORE_GPR 29, 232, 1
+	RESTORE_GPR 30, 240, 1
+	RESTORE_GPR 31, 248, 1
+
+	addi    1, 1, 752
+	ld 0, 16(1)
+	mtlr 0
+.endm # RESTORE_REGS
+
+.macro QT_loop_8x
+	# QR(v0, v4,  v8, v12, v1, v5,  v9, v13, v2, v6, v10, v14, v3, v7, v11, v15)
+	xxlor	0, 32+25, 32+25
+	xxlor	32+25, 20, 20
+	vadduwm 0, 0, 4
+	vadduwm 1, 1, 5
+	vadduwm 2, 2, 6
+	vadduwm 3, 3, 7
+	  vadduwm 16, 16, 20
+	  vadduwm 17, 17, 21
+	  vadduwm 18, 18, 22
+	  vadduwm 19, 19, 23
+
+	  vpermxor 12, 12, 0, 25
+	  vpermxor 13, 13, 1, 25
+	  vpermxor 14, 14, 2, 25
+	  vpermxor 15, 15, 3, 25
+	  vpermxor 28, 28, 16, 25
+	  vpermxor 29, 29, 17, 25
+	  vpermxor 30, 30, 18, 25
+	  vpermxor 31, 31, 19, 25
+	xxlor	32+25, 0, 0
+	vadduwm 8, 8, 12
+	vadduwm 9, 9, 13
+	vadduwm 10, 10, 14
+	vadduwm 11, 11, 15
+	  vadduwm 24, 24, 28
+	  vadduwm 25, 25, 29
+	  vadduwm 26, 26, 30
+	  vadduwm 27, 27, 31
+	vxor 4, 4, 8
+	vxor 5, 5, 9
+	vxor 6, 6, 10
+	vxor 7, 7, 11
+	  vxor 20, 20, 24
+	  vxor 21, 21, 25
+	  vxor 22, 22, 26
+	  vxor 23, 23, 27
+
+	xxlor	0, 32+25, 32+25
+	xxlor	32+25, 21, 21
+	vrlw 4, 4, 25  #
+	vrlw 5, 5, 25
+	vrlw 6, 6, 25
+	vrlw 7, 7, 25
+	  vrlw 20, 20, 25  #
+	  vrlw 21, 21, 25
+	  vrlw 22, 22, 25
+	  vrlw 23, 23, 25
+	xxlor	32+25, 0, 0
+	vadduwm 0, 0, 4
+	vadduwm 1, 1, 5
+	vadduwm 2, 2, 6
+	vadduwm 3, 3, 7
+	  vadduwm 16, 16, 20
+	  vadduwm 17, 17, 21
+	  vadduwm 18, 18, 22
+	  vadduwm 19, 19, 23
+
+	xxlor	0, 32+25, 32+25
+	xxlor	32+25, 22, 22
+	  vpermxor 12, 12, 0, 25
+	  vpermxor 13, 13, 1, 25
+	  vpermxor 14, 14, 2, 25
+	  vpermxor 15, 15, 3, 25
+	  vpermxor 28, 28, 16, 25
+	  vpermxor 29, 29, 17, 25
+	  vpermxor 30, 30, 18, 25
+	  vpermxor 31, 31, 19, 25
+	xxlor	32+25, 0, 0
+	vadduwm 8, 8, 12
+	vadduwm 9, 9, 13
+	vadduwm 10, 10, 14
+	vadduwm 11, 11, 15
+	  vadduwm 24, 24, 28
+	  vadduwm 25, 25, 29
+	  vadduwm 26, 26, 30
+	  vadduwm 27, 27, 31
+	xxlor	0, 32+28, 32+28
+	xxlor	32+28, 23, 23
+	vxor 4, 4, 8
+	vxor 5, 5, 9
+	vxor 6, 6, 10
+	vxor 7, 7, 11
+	  vxor 20, 20, 24
+	  vxor 21, 21, 25
+	  vxor 22, 22, 26
+	  vxor 23, 23, 27
+	vrlw 4, 4, 28  #
+	vrlw 5, 5, 28
+	vrlw 6, 6, 28
+	vrlw 7, 7, 28
+	  vrlw 20, 20, 28  #
+	  vrlw 21, 21, 28
+	  vrlw 22, 22, 28
+	  vrlw 23, 23, 28
+	xxlor	32+28, 0, 0
+
+	# QR(v0, v5, v10, v15, v1, v6, v11, v12, v2, v7,  v8, v13, v3, v4,  v9, v14)
+	xxlor	0, 32+25, 32+25
+	xxlor	32+25, 20, 20
+	vadduwm 0, 0, 5
+	vadduwm 1, 1, 6
+	vadduwm 2, 2, 7
+	vadduwm 3, 3, 4
+	  vadduwm 16, 16, 21
+	  vadduwm 17, 17, 22
+	  vadduwm 18, 18, 23
+	  vadduwm 19, 19, 20
+
+	  vpermxor 15, 15, 0, 25
+	  vpermxor 12, 12, 1, 25
+	  vpermxor 13, 13, 2, 25
+	  vpermxor 14, 14, 3, 25
+	  vpermxor 31, 31, 16, 25
+	  vpermxor 28, 28, 17, 25
+	  vpermxor 29, 29, 18, 25
+	  vpermxor 30, 30, 19, 25
+
+	xxlor	32+25, 0, 0
+	vadduwm 10, 10, 15
+	vadduwm 11, 11, 12
+	vadduwm 8, 8, 13
+	vadduwm 9, 9, 14
+	  vadduwm 26, 26, 31
+	  vadduwm 27, 27, 28
+	  vadduwm 24, 24, 29
+	  vadduwm 25, 25, 30
+	vxor 5, 5, 10
+	vxor 6, 6, 11
+	vxor 7, 7, 8
+	vxor 4, 4, 9
+	  vxor 21, 21, 26
+	  vxor 22, 22, 27
+	  vxor 23, 23, 24
+	  vxor 20, 20, 25
+
+	xxlor	0, 32+25, 32+25
+	xxlor	32+25, 21, 21
+	vrlw 5, 5, 25
+	vrlw 6, 6, 25
+	vrlw 7, 7, 25
+	vrlw 4, 4, 25
+	  vrlw 21, 21, 25
+	  vrlw 22, 22, 25
+	  vrlw 23, 23, 25
+	  vrlw 20, 20, 25
+	xxlor	32+25, 0, 0
+
+	vadduwm 0, 0, 5
+	vadduwm 1, 1, 6
+	vadduwm 2, 2, 7
+	vadduwm 3, 3, 4
+	  vadduwm 16, 16, 21
+	  vadduwm 17, 17, 22
+	  vadduwm 18, 18, 23
+	  vadduwm 19, 19, 20
+
+	xxlor	0, 32+25, 32+25
+	xxlor	32+25, 22, 22
+	  vpermxor 15, 15, 0, 25
+	  vpermxor 12, 12, 1, 25
+	  vpermxor 13, 13, 2, 25
+	  vpermxor 14, 14, 3, 25
+	  vpermxor 31, 31, 16, 25
+	  vpermxor 28, 28, 17, 25
+	  vpermxor 29, 29, 18, 25
+	  vpermxor 30, 30, 19, 25
+	xxlor	32+25, 0, 0
+
+	vadduwm 10, 10, 15
+	vadduwm 11, 11, 12
+	vadduwm 8, 8, 13
+	vadduwm 9, 9, 14
+	  vadduwm 26, 26, 31
+	  vadduwm 27, 27, 28
+	  vadduwm 24, 24, 29
+	  vadduwm 25, 25, 30
+
+	xxlor	0, 32+28, 32+28
+	xxlor	32+28, 23, 23
+	vxor 5, 5, 10
+	vxor 6, 6, 11
+	vxor 7, 7, 8
+	vxor 4, 4, 9
+	  vxor 21, 21, 26
+	  vxor 22, 22, 27
+	  vxor 23, 23, 24
+	  vxor 20, 20, 25
+	vrlw 5, 5, 28
+	vrlw 6, 6, 28
+	vrlw 7, 7, 28
+	vrlw 4, 4, 28
+	  vrlw 21, 21, 28
+	  vrlw 22, 22, 28
+	  vrlw 23, 23, 28
+	  vrlw 20, 20, 28
+	xxlor	32+28, 0, 0
+.endm
+
+.macro QT_loop_4x
+	# QR(v0, v4,  v8, v12, v1, v5,  v9, v13, v2, v6, v10, v14, v3, v7, v11, v15)
+	vadduwm 0, 0, 4
+	vadduwm 1, 1, 5
+	vadduwm 2, 2, 6
+	vadduwm 3, 3, 7
+	  vpermxor 12, 12, 0, 20
+	  vpermxor 13, 13, 1, 20
+	  vpermxor 14, 14, 2, 20
+	  vpermxor 15, 15, 3, 20
+	vadduwm 8, 8, 12
+	vadduwm 9, 9, 13
+	vadduwm 10, 10, 14
+	vadduwm 11, 11, 15
+	vxor 4, 4, 8
+	vxor 5, 5, 9
+	vxor 6, 6, 10
+	vxor 7, 7, 11
+	vrlw 4, 4, 21
+	vrlw 5, 5, 21
+	vrlw 6, 6, 21
+	vrlw 7, 7, 21
+	vadduwm 0, 0, 4
+	vadduwm 1, 1, 5
+	vadduwm 2, 2, 6
+	vadduwm 3, 3, 7
+	  vpermxor 12, 12, 0, 22
+	  vpermxor 13, 13, 1, 22
+	  vpermxor 14, 14, 2, 22
+	  vpermxor 15, 15, 3, 22
+	vadduwm 8, 8, 12
+	vadduwm 9, 9, 13
+	vadduwm 10, 10, 14
+	vadduwm 11, 11, 15
+	vxor 4, 4, 8
+	vxor 5, 5, 9
+	vxor 6, 6, 10
+	vxor 7, 7, 11
+	vrlw 4, 4, 23
+	vrlw 5, 5, 23
+	vrlw 6, 6, 23
+	vrlw 7, 7, 23
+
+	# QR(v0, v5, v10, v15, v1, v6, v11, v12, v2, v7,  v8, v13, v3, v4,  v9, v14)
+	vadduwm 0, 0, 5
+	vadduwm 1, 1, 6
+	vadduwm 2, 2, 7
+	vadduwm 3, 3, 4
+	  vpermxor 15, 15, 0, 20
+	  vpermxor 12, 12, 1, 20
+	  vpermxor 13, 13, 2, 20
+	  vpermxor 14, 14, 3, 20
+	vadduwm 10, 10, 15
+	vadduwm 11, 11, 12
+	vadduwm 8, 8, 13
+	vadduwm 9, 9, 14
+	vxor 5, 5, 10
+	vxor 6, 6, 11
+	vxor 7, 7, 8
+	vxor 4, 4, 9
+	vrlw 5, 5, 21
+	vrlw 6, 6, 21
+	vrlw 7, 7, 21
+	vrlw 4, 4, 21
+	vadduwm 0, 0, 5
+	vadduwm 1, 1, 6
+	vadduwm 2, 2, 7
+	vadduwm 3, 3, 4
+	  vpermxor 15, 15, 0, 22
+	  vpermxor 12, 12, 1, 22
+	  vpermxor 13, 13, 2, 22
+	  vpermxor 14, 14, 3, 22
+	vadduwm 10, 10, 15
+	vadduwm 11, 11, 12
+	vadduwm 8, 8, 13
+	vadduwm 9, 9, 14
+	vxor 5, 5, 10
+	vxor 6, 6, 11
+	vxor 7, 7, 8
+	vxor 4, 4, 9
+	vrlw 5, 5, 23
+	vrlw 6, 6, 23
+	vrlw 7, 7, 23
+	vrlw 4, 4, 23
+.endm
+
+# Transpose
+.macro TP_4x a0 a1 a2 a3
+	xxmrghw  10, 32+\a0, 32+\a1	# a0, a1, b0, b1
+	xxmrghw  11, 32+\a2, 32+\a3	# a2, a3, b2, b3
+	xxmrglw  12, 32+\a0, 32+\a1	# c0, c1, d0, d1
+	xxmrglw  13, 32+\a2, 32+\a3	# c2, c3, d2, d3
+	xxpermdi	32+\a0, 10, 11, 0	# a0, a1, a2, a3
+	xxpermdi	32+\a1, 10, 11, 3	# b0, b1, b2, b3
+	xxpermdi	32+\a2, 12, 13, 0	# c0, c1, c2, c3
+	xxpermdi	32+\a3, 12, 13, 3	# d0, d1, d2, d3
+.endm
+
+# key stream = working state + state
+.macro Add_state S
+	vadduwm \S+0, \S+0, 16-\S
+	vadduwm \S+4, \S+4, 17-\S
+	vadduwm \S+8, \S+8, 18-\S
+	vadduwm \S+12, \S+12, 19-\S
+
+	vadduwm \S+1, \S+1, 16-\S
+	vadduwm \S+5, \S+5, 17-\S
+	vadduwm \S+9, \S+9, 18-\S
+	vadduwm \S+13, \S+13, 19-\S
+
+	vadduwm \S+2, \S+2, 16-\S
+	vadduwm \S+6, \S+6, 17-\S
+	vadduwm \S+10, \S+10, 18-\S
+	vadduwm \S+14, \S+14, 19-\S
+
+	vadduwm	\S+3, \S+3, 16-\S
+	vadduwm	\S+7, \S+7, 17-\S
+	vadduwm	\S+11, \S+11, 18-\S
+	vadduwm	\S+15, \S+15, 19-\S
+.endm
+
+#
+# write 256 bytes
+#
+.macro Write_256 S
+	add 9, 14, 5
+	add 16, 14, 4
+	lxvw4x 0, 0, 9
+	lxvw4x 1, 17, 9
+	lxvw4x 2, 18, 9
+	lxvw4x 3, 19, 9
+	lxvw4x 4, 20, 9
+	lxvw4x 5, 21, 9
+	lxvw4x 6, 22, 9
+	lxvw4x 7, 23, 9
+	lxvw4x 8, 24, 9
+	lxvw4x 9, 25, 9
+	lxvw4x 10, 26, 9
+	lxvw4x 11, 27, 9
+	lxvw4x 12, 28, 9
+	lxvw4x 13, 29, 9
+	lxvw4x 14, 30, 9
+	lxvw4x 15, 31, 9
+
+	xxlxor \S+32, \S+32, 0
+	xxlxor \S+36, \S+36, 1
+	xxlxor \S+40, \S+40, 2
+	xxlxor \S+44, \S+44, 3
+	xxlxor \S+33, \S+33, 4
+	xxlxor \S+37, \S+37, 5
+	xxlxor \S+41, \S+41, 6
+	xxlxor \S+45, \S+45, 7
+	xxlxor \S+34, \S+34, 8
+	xxlxor \S+38, \S+38, 9
+	xxlxor \S+42, \S+42, 10
+	xxlxor \S+46, \S+46, 11
+	xxlxor \S+35, \S+35, 12
+	xxlxor \S+39, \S+39, 13
+	xxlxor \S+43, \S+43, 14
+	xxlxor \S+47, \S+47, 15
+
+	stxvw4x \S+32, 0, 16
+	stxvw4x \S+36, 17, 16
+	stxvw4x \S+40, 18, 16
+	stxvw4x \S+44, 19, 16
+
+	stxvw4x \S+33, 20, 16
+	stxvw4x \S+37, 21, 16
+	stxvw4x \S+41, 22, 16
+	stxvw4x \S+45, 23, 16
+
+	stxvw4x \S+34, 24, 16
+	stxvw4x \S+38, 25, 16
+	stxvw4x \S+42, 26, 16
+	stxvw4x \S+46, 27, 16
+
+	stxvw4x \S+35, 28, 16
+	stxvw4x \S+39, 29, 16
+	stxvw4x \S+43, 30, 16
+	stxvw4x \S+47, 31, 16
+
+.endm
+
+#
+# chacha20_p10le_8x(u32 *state, byte *dst, const byte *src, size_t len, int nrounds);
+#
+SYM_FUNC_START(chacha_p10le_8x)
+.align 5
+	cmpdi	6, 0
+	ble	Out_no_chacha
+
+	SAVE_REGS
+
+	# r17 - r31 mainly for Write_256 macro.
+	li	17, 16
+	li	18, 32
+	li	19, 48
+	li	20, 64
+	li	21, 80
+	li	22, 96
+	li	23, 112
+	li	24, 128
+	li	25, 144
+	li	26, 160
+	li	27, 176
+	li	28, 192
+	li	29, 208
+	li	30, 224
+	li	31, 240
+
+	mr 15, 6			# len
+	li 14, 0			# offset to inp and outp
+
+        lxvw4x	48, 0, 3		#  vr16, constants
+	lxvw4x	49, 17, 3		#  vr17, key 1
+	lxvw4x	50, 18, 3		#  vr18, key 2
+	lxvw4x	51, 19, 3		#  vr19, counter, nonce
+
+	# create (0, 1, 2, 3) counters
+	vspltisw 0, 0
+	vspltisw 1, 1
+	vspltisw 2, 2
+	vspltisw 3, 3
+	vmrghw	4, 0, 1
+	vmrglw	5, 2, 3
+	vsldoi	30, 4, 5, 8		# vr30 counter, 4 (0, 1, 2, 3)
+
+	vspltisw 21, 12
+	vspltisw 23, 7
+
+	addis	11, 2, permx@toc@ha
+	addi	11, 11, permx@toc@l
+	lxvw4x	32+20, 0, 11
+	lxvw4x	32+22, 17, 11
+
+	sradi	8, 7, 1
+
+	mtctr 8
+
+	# save constants to vsx
+	xxlor	16, 48, 48
+	xxlor	17, 49, 49
+	xxlor	18, 50, 50
+	xxlor	19, 51, 51
+
+	vspltisw 25, 4
+	vspltisw 26, 8
+
+	xxlor	25, 32+26, 32+26
+	xxlor	24, 32+25, 32+25
+
+	vadduwm	31, 30, 25		# counter = (0, 1, 2, 3) + (4, 4, 4, 4)
+	xxlor	30, 32+30, 32+30
+	xxlor	31, 32+31, 32+31
+
+	xxlor	20, 32+20, 32+20
+	xxlor	21, 32+21, 32+21
+	xxlor	22, 32+22, 32+22
+	xxlor	23, 32+23, 32+23
+
+	cmpdi	6, 512
+	blt	Loop_last
+
+Loop_8x:
+	xxspltw  32+0, 16, 0
+	xxspltw  32+1, 16, 1
+	xxspltw  32+2, 16, 2
+	xxspltw  32+3, 16, 3
+
+	xxspltw  32+4, 17, 0
+	xxspltw  32+5, 17, 1
+	xxspltw  32+6, 17, 2
+	xxspltw  32+7, 17, 3
+	xxspltw  32+8, 18, 0
+	xxspltw  32+9, 18, 1
+	xxspltw  32+10, 18, 2
+	xxspltw  32+11, 18, 3
+	xxspltw  32+12, 19, 0
+	xxspltw  32+13, 19, 1
+	xxspltw  32+14, 19, 2
+	xxspltw  32+15, 19, 3
+	vadduwm	12, 12, 30	# increase counter
+
+	xxspltw  32+16, 16, 0
+	xxspltw  32+17, 16, 1
+	xxspltw  32+18, 16, 2
+	xxspltw  32+19, 16, 3
+
+	xxspltw  32+20, 17, 0
+	xxspltw  32+21, 17, 1
+	xxspltw  32+22, 17, 2
+	xxspltw  32+23, 17, 3
+	xxspltw  32+24, 18, 0
+	xxspltw  32+25, 18, 1
+	xxspltw  32+26, 18, 2
+	xxspltw  32+27, 18, 3
+	xxspltw  32+28, 19, 0
+	xxspltw  32+29, 19, 1
+	vadduwm	28, 28, 31	# increase counter
+	xxspltw  32+30, 19, 2
+	xxspltw  32+31, 19, 3
+
+.align 5
+quarter_loop_8x:
+	QT_loop_8x
+
+	bdnz	quarter_loop_8x
+
+	xxlor	0, 32+30, 32+30
+	xxlor	32+30, 30, 30
+	vadduwm	12, 12, 30
+	xxlor	32+30, 0, 0
+	TP_4x 0, 1, 2, 3
+	TP_4x 4, 5, 6, 7
+	TP_4x 8, 9, 10, 11
+	TP_4x 12, 13, 14, 15
+
+	xxlor	0, 48, 48
+	xxlor	1, 49, 49
+	xxlor	2, 50, 50
+	xxlor	3, 51, 51
+	xxlor	48, 16, 16
+	xxlor	49, 17, 17
+	xxlor	50, 18, 18
+	xxlor	51, 19, 19
+	Add_state 0
+	xxlor	48, 0, 0
+	xxlor	49, 1, 1
+	xxlor	50, 2, 2
+	xxlor	51, 3, 3
+	Write_256 0
+	addi	14, 14, 256	# offset +=256
+	addi	15, 15, -256	# len -=256
+
+	xxlor	5, 32+31, 32+31
+	xxlor	32+31, 31, 31
+	vadduwm	28, 28, 31
+	xxlor	32+31, 5, 5
+	TP_4x 16+0, 16+1, 16+2, 16+3
+	TP_4x 16+4, 16+5, 16+6, 16+7
+	TP_4x 16+8, 16+9, 16+10, 16+11
+	TP_4x 16+12, 16+13, 16+14, 16+15
+
+	xxlor	32, 16, 16
+	xxlor	33, 17, 17
+	xxlor	34, 18, 18
+	xxlor	35, 19, 19
+	Add_state 16
+	Write_256 16
+	addi	14, 14, 256	# offset +=256
+	addi	15, 15, -256	# len +=256
+
+	xxlor	32+24, 24, 24
+	xxlor	32+25, 25, 25
+	xxlor	32+30, 30, 30
+	vadduwm	30, 30, 25
+	vadduwm	31, 30, 24
+	xxlor	30, 32+30, 32+30
+	xxlor	31, 32+31, 32+31
+
+	cmpdi	15, 0
+	beq	Out_loop
+
+	cmpdi	15, 512
+	blt	Loop_last
+
+	mtctr 8
+	b Loop_8x
+
+Loop_last:
+        lxvw4x	48, 0, 3		#  vr16, constants
+	lxvw4x	49, 17, 3		#  vr17, key 1
+	lxvw4x	50, 18, 3		#  vr18, key 2
+	lxvw4x	51, 19, 3		#  vr19, counter, nonce
+
+	vspltisw 21, 12
+	vspltisw 23, 7
+	addis	11, 2, permx@toc@ha
+	addi	11, 11, permx@toc@l
+	lxvw4x	32+20, 0, 11
+	lxvw4x	32+22, 17, 11
+
+	sradi	8, 7, 1
+	mtctr 8
+
+Loop_4x:
+	vspltw  0, 16, 0
+	vspltw  1, 16, 1
+	vspltw  2, 16, 2
+	vspltw  3, 16, 3
+
+	vspltw  4, 17, 0
+	vspltw  5, 17, 1
+	vspltw  6, 17, 2
+	vspltw  7, 17, 3
+	vspltw  8, 18, 0
+	vspltw  9, 18, 1
+	vspltw  10, 18, 2
+	vspltw  11, 18, 3
+	vspltw  12, 19, 0
+	vadduwm	12, 12, 30	# increase counter
+	vspltw  13, 19, 1
+	vspltw  14, 19, 2
+	vspltw  15, 19, 3
+
+.align 5
+quarter_loop:
+	QT_loop_4x
+
+	bdnz	quarter_loop
+
+	vadduwm	12, 12, 30
+	TP_4x 0, 1, 2, 3
+	TP_4x 4, 5, 6, 7
+	TP_4x 8, 9, 10, 11
+	TP_4x 12, 13, 14, 15
+
+	Add_state 0
+	Write_256 0
+	addi	14, 14, 256	# offset += 256
+	addi	15, 15, -256	# len += 256
+
+	# Update state counter
+	vspltisw 25, 4
+	vadduwm	30, 30, 25
+
+	cmpdi	15, 0
+	beq	Out_loop
+	cmpdi	15, 256
+	blt	Out_loop
+
+	mtctr 8
+	b Loop_4x
+
+Out_loop:
+	RESTORE_REGS
+	blr
+
+Out_no_chacha:
+	li	3, 0
+	blr
+SYM_FUNC_END(chacha_p10le_8x)
+
+SYM_DATA_START_LOCAL(PERMX)
+.align 5
+permx:
+.long 0x22330011, 0x66774455, 0xaabb8899, 0xeeffccdd
+.long 0x11223300, 0x55667744, 0x99aabb88, 0xddeeffcc
+SYM_DATA_END(PERMX)