crypto: sha512 - Optimized SHA512 x86_64 assembly routine using AVX instructions.

Provides SHA512 x86_64 assembly routine optimized with SSE and AVX instructions.
Speedup of 60% or more has been measured over the generic implementation.
Signed-off-by: NTim Chen <tim.c.chen@linux.intel.com>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

arch/x86/crypto/sha512-avx-asm.S

+########################################################################
+# Implement fast SHA-512 with AVX instructions. (x86_64)
+#
+# Copyright (C) 2013 Intel Corporation.
+#
+# Authors:
+#     James Guilford <james.guilford@intel.com>
+#     Kirk Yap <kirk.s.yap@intel.com>
+#     David Cote <david.m.cote@intel.com>
+#     Tim Chen <tim.c.chen@linux.intel.com>
+#
+# This software is available to you under a choice of one of two
+# licenses.  You may choose to be licensed under the terms of the GNU
+# General Public License (GPL) Version 2, available from the file
+# COPYING in the main directory of this source tree, or the
+# OpenIB.org BSD license below:
+#
+#     Redistribution and use in source and binary forms, with or
+#     without modification, are permitted provided that the following
+#     conditions are met:
+#
+#      - Redistributions of source code must retain the above
+#        copyright notice, this list of conditions and the following
+#        disclaimer.
+#
+#      - Redistributions in binary form must reproduce the above
+#        copyright notice, this list of conditions and the following
+#        disclaimer in the documentation and/or other materials
+#        provided with the distribution.
+#
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+# EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+# NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+# BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+# ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+# CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+# SOFTWARE.
+#
+########################################################################
+#
+# This code is described in an Intel White-Paper:
+# "Fast SHA-512 Implementations on Intel Architecture Processors"
+#
+# To find it, surf to http://www.intel.com/p/en_US/embedded
+# and search for that title.
+#
+########################################################################
+
+#ifdef CONFIG_AS_AVX
+#include <linux/linkage.h>
+
+.text
+
+# Virtual Registers
+# ARG1
+msg	= %rdi
+# ARG2
+digest	= %rsi
+# ARG3
+msglen	= %rdx
+T1	= %rcx
+T2	= %r8
+a_64	= %r9
+b_64	= %r10
+c_64	= %r11
+d_64	= %r12
+e_64	= %r13
+f_64	= %r14
+g_64	= %r15
+h_64	= %rbx
+tmp0	= %rax
+
+# Local variables (stack frame)
+
+# Message Schedule
+W_SIZE = 80*8
+# W[t] + K[t] | W[t+1] + K[t+1]
+WK_SIZE = 2*8
+RSPSAVE_SIZE = 1*8
+GPRSAVE_SIZE = 5*8
+
+frame_W = 0
+frame_WK = frame_W + W_SIZE
+frame_RSPSAVE = frame_WK + WK_SIZE
+frame_GPRSAVE = frame_RSPSAVE + RSPSAVE_SIZE
+frame_size = frame_GPRSAVE + GPRSAVE_SIZE
+
+# Useful QWORD "arrays" for simpler memory references
+# MSG, DIGEST, K_t, W_t are arrays
+# WK_2(t) points to 1 of 2 qwords at frame.WK depdending on t being odd/even
+
+# Input message (arg1)
+#define MSG(i)    8*i(msg)
+
+# Output Digest (arg2)
+#define DIGEST(i) 8*i(digest)
+
+# SHA Constants (static mem)
+#define K_t(i)    8*i+K512(%rip)
+
+# Message Schedule (stack frame)
+#define W_t(i)    8*i+frame_W(%rsp)
+
+# W[t]+K[t] (stack frame)
+#define WK_2(i)   8*((i%2))+frame_WK(%rsp)
+
+.macro RotateState
+	# Rotate symbols a..h right
+	TMP   = h_64
+	h_64  = g_64
+	g_64  = f_64
+	f_64  = e_64
+	e_64  = d_64
+	d_64  = c_64
+	c_64  = b_64
+	b_64  = a_64
+	a_64  = TMP
+.endm
+
+.macro RORQ p1 p2
+	# shld is faster than ror on Sandybridge
+	shld	$(64-\p2), \p1, \p1
+.endm
+
+.macro SHA512_Round rnd
+	# Compute Round %%t
+	mov     f_64, T1          # T1 = f
+	mov     e_64, tmp0        # tmp = e
+	xor     g_64, T1          # T1 = f ^ g
+	RORQ    tmp0, 23   # 41    # tmp = e ror 23
+	and     e_64, T1          # T1 = (f ^ g) & e
+	xor     e_64, tmp0        # tmp = (e ror 23) ^ e
+	xor     g_64, T1          # T1 = ((f ^ g) & e) ^ g = CH(e,f,g)
+	idx = \rnd
+	add     WK_2(idx), T1     # W[t] + K[t] from message scheduler
+	RORQ    tmp0, 4   # 18    # tmp = ((e ror 23) ^ e) ror 4
+	xor     e_64, tmp0        # tmp = (((e ror 23) ^ e) ror 4) ^ e
+	mov     a_64, T2          # T2 = a
+	add     h_64, T1          # T1 = CH(e,f,g) + W[t] + K[t] + h
+	RORQ    tmp0, 14  # 14    # tmp = ((((e ror23)^e)ror4)^e)ror14 = S1(e)
+	add     tmp0, T1          # T1 = CH(e,f,g) + W[t] + K[t] + S1(e)
+	mov     a_64, tmp0        # tmp = a
+	xor     c_64, T2          # T2 = a ^ c
+	and     c_64, tmp0        # tmp = a & c
+	and     b_64, T2          # T2 = (a ^ c) & b
+	xor     tmp0, T2          # T2 = ((a ^ c) & b) ^ (a & c) = Maj(a,b,c)
+	mov     a_64, tmp0        # tmp = a
+	RORQ    tmp0, 5  # 39     # tmp = a ror 5
+	xor     a_64, tmp0        # tmp = (a ror 5) ^ a
+	add     T1, d_64          # e(next_state) = d + T1
+	RORQ    tmp0, 6  # 34     # tmp = ((a ror 5) ^ a) ror 6
+	xor     a_64, tmp0        # tmp = (((a ror 5) ^ a) ror 6) ^ a
+	lea     (T1, T2), h_64    # a(next_state) = T1 + Maj(a,b,c)
+	RORQ    tmp0, 28  # 28    # tmp = ((((a ror5)^a)ror6)^a)ror28 = S0(a)
+	add     tmp0, h_64        # a(next_state) = T1 + Maj(a,b,c) S0(a)
+	RotateState
+.endm
+
+.macro SHA512_2Sched_2Round_avx rnd
+	# Compute rounds t-2 and t-1
+	# Compute message schedule QWORDS t and t+1
+
+	#   Two rounds are computed based on the values for K[t-2]+W[t-2] and
+	# K[t-1]+W[t-1] which were previously stored at WK_2 by the message
+	# scheduler.
+	#   The two new schedule QWORDS are stored at [W_t(t)] and [W_t(t+1)].
+	# They are then added to their respective SHA512 constants at
+	# [K_t(t)] and [K_t(t+1)] and stored at dqword [WK_2(t)]
+	#   For brievity, the comments following vectored instructions only refer to
+	# the first of a pair of QWORDS.
+	# Eg. XMM4=W[t-2] really means XMM4={W[t-2]|W[t-1]}
+	#   The computation of the message schedule and the rounds are tightly
+	# stitched to take advantage of instruction-level parallelism.
+
+	idx = \rnd - 2
+	vmovdqa	W_t(idx), %xmm4		# XMM4 = W[t-2]
+	idx = \rnd - 15
+	vmovdqu	W_t(idx), %xmm5		# XMM5 = W[t-15]
+	mov	f_64, T1
+	vpsrlq	$61, %xmm4, %xmm0	# XMM0 = W[t-2]>>61
+	mov	e_64, tmp0
+	vpsrlq	$1, %xmm5, %xmm6	# XMM6 = W[t-15]>>1
+	xor	g_64, T1
+	RORQ	tmp0, 23 # 41
+	vpsrlq	$19, %xmm4, %xmm1	# XMM1 = W[t-2]>>19
+	and	e_64, T1
+	xor	e_64, tmp0
+	vpxor	%xmm1, %xmm0, %xmm0	# XMM0 = W[t-2]>>61 ^ W[t-2]>>19
+	xor	g_64, T1
+	idx = \rnd
+	add	WK_2(idx), T1#
+	vpsrlq	$8, %xmm5, %xmm7	# XMM7 = W[t-15]>>8
+	RORQ	tmp0, 4 # 18
+	vpsrlq	$6, %xmm4, %xmm2	# XMM2 = W[t-2]>>6
+	xor	e_64, tmp0
+	mov	a_64, T2
+	add	h_64, T1
+	vpxor	%xmm7, %xmm6, %xmm6	# XMM6 = W[t-15]>>1 ^ W[t-15]>>8
+	RORQ	tmp0, 14 # 14
+	add	tmp0, T1
+	vpsrlq	$7, %xmm5, %xmm8	# XMM8 = W[t-15]>>7
+	mov	a_64, tmp0
+	xor	c_64, T2
+	vpsllq	$(64-61), %xmm4, %xmm3  # XMM3 = W[t-2]<<3
+	and	c_64, tmp0
+	and	b_64, T2
+	vpxor	%xmm3, %xmm2, %xmm2	# XMM2 = W[t-2]>>6 ^ W[t-2]<<3
+	xor	tmp0, T2
+	mov	a_64, tmp0
+	vpsllq	$(64-1), %xmm5, %xmm9	# XMM9 = W[t-15]<<63
+	RORQ	tmp0, 5 # 39
+	vpxor	%xmm9, %xmm8, %xmm8	# XMM8 = W[t-15]>>7 ^ W[t-15]<<63
+	xor	a_64, tmp0
+	add	T1, d_64
+	RORQ	tmp0, 6 # 34
+	xor	a_64, tmp0
+	vpxor	%xmm8, %xmm6, %xmm6	# XMM6 = W[t-15]>>1 ^ W[t-15]>>8 ^
+					#  W[t-15]>>7 ^ W[t-15]<<63
+	lea	(T1, T2), h_64
+	RORQ	tmp0, 28 # 28
+	vpsllq	$(64-19), %xmm4, %xmm4  # XMM4 = W[t-2]<<25
+	add	tmp0, h_64
+	RotateState
+	vpxor	%xmm4, %xmm0, %xmm0     # XMM0 = W[t-2]>>61 ^ W[t-2]>>19 ^
+					#        W[t-2]<<25
+	mov	f_64, T1
+	vpxor	%xmm2, %xmm0, %xmm0     # XMM0 = s1(W[t-2])
+	mov	e_64, tmp0
+	xor	g_64, T1
+	idx = \rnd - 16
+	vpaddq	W_t(idx), %xmm0, %xmm0  # XMM0 = s1(W[t-2]) + W[t-16]
+	idx = \rnd - 7
+	vmovdqu	W_t(idx), %xmm1		# XMM1 = W[t-7]
+	RORQ	tmp0, 23 # 41
+	and	e_64, T1
+	xor	e_64, tmp0
+	xor	g_64, T1
+	vpsllq	$(64-8), %xmm5, %xmm5   # XMM5 = W[t-15]<<56
+	idx = \rnd + 1
+	add	WK_2(idx), T1
+	vpxor	%xmm5, %xmm6, %xmm6     # XMM6 = s0(W[t-15])
+	RORQ	tmp0, 4 # 18
+	vpaddq	%xmm6, %xmm0, %xmm0     # XMM0 = s1(W[t-2]) + W[t-16] + s0(W[t-15])
+	xor	e_64, tmp0
+	vpaddq	%xmm1, %xmm0, %xmm0     # XMM0 = W[t] = s1(W[t-2]) + W[t-7] +
+					#               s0(W[t-15]) + W[t-16]
+	mov	a_64, T2
+	add	h_64, T1
+	RORQ	tmp0, 14 # 14
+	add	tmp0, T1
+	idx = \rnd
+	vmovdqa	%xmm0, W_t(idx)		# Store W[t]
+	vpaddq	K_t(idx), %xmm0, %xmm0  # Compute W[t]+K[t]
+	vmovdqa	%xmm0, WK_2(idx)	# Store W[t]+K[t] for next rounds
+	mov	a_64, tmp0
+	xor	c_64, T2
+	and	c_64, tmp0
+	and	b_64, T2
+	xor	tmp0, T2
+	mov	a_64, tmp0
+	RORQ	tmp0, 5 # 39
+	xor	a_64, tmp0
+	add	T1, d_64
+	RORQ	tmp0, 6 # 34
+	xor	a_64, tmp0
+	lea	(T1, T2), h_64
+	RORQ	tmp0, 28 # 28
+	add	tmp0, h_64
+	RotateState
+.endm
+
+########################################################################
+# void sha512_transform_avx(const void* M, void* D, u64 L)
+# Purpose: Updates the SHA512 digest stored at D with the message stored in M.
+# The size of the message pointed to by M must be an integer multiple of SHA512
+# message blocks.
+# L is the message length in SHA512 blocks
+########################################################################
+ENTRY(sha512_transform_avx)
+	cmp $0, msglen
+	je nowork
+
+	# Allocate Stack Space
+	mov	%rsp, %rax
+	sub     $frame_size, %rsp
+	and	$~(0x20 - 1), %rsp
+	mov	%rax, frame_RSPSAVE(%rsp)
+
+	# Save GPRs
+	mov     %rbx, frame_GPRSAVE(%rsp)
+	mov     %r12, frame_GPRSAVE +8*1(%rsp)
+	mov     %r13, frame_GPRSAVE +8*2(%rsp)
+	mov     %r14, frame_GPRSAVE +8*3(%rsp)
+	mov     %r15, frame_GPRSAVE +8*4(%rsp)
+
+updateblock:
+
+	# Load state variables
+	mov     DIGEST(0), a_64
+	mov     DIGEST(1), b_64
+	mov     DIGEST(2), c_64
+	mov     DIGEST(3), d_64
+	mov     DIGEST(4), e_64
+	mov     DIGEST(5), f_64
+	mov     DIGEST(6), g_64
+	mov     DIGEST(7), h_64
+
+	t = 0
+	.rept 80/2 + 1
+	# (80 rounds) / (2 rounds/iteration) + (1 iteration)
+	# +1 iteration because the scheduler leads hashing by 1 iteration
+		.if t < 2
+			# BSWAP 2 QWORDS
+			vmovdqa  XMM_QWORD_BSWAP(%rip), %xmm1
+			vmovdqu  MSG(t), %xmm0
+			vpshufb  %xmm1, %xmm0, %xmm0    # BSWAP
+			vmovdqa  %xmm0, W_t(t) # Store Scheduled Pair
+			vpaddq   K_t(t), %xmm0, %xmm0 # Compute W[t]+K[t]
+			vmovdqa  %xmm0, WK_2(t) # Store into WK for rounds
+		.elseif t < 16
+			# BSWAP 2 QWORDS# Compute 2 Rounds
+			vmovdqu  MSG(t), %xmm0
+			vpshufb  %xmm1, %xmm0, %xmm0    # BSWAP
+			SHA512_Round t-2    # Round t-2
+			vmovdqa  %xmm0, W_t(t) # Store Scheduled Pair
+			vpaddq   K_t(t), %xmm0, %xmm0 # Compute W[t]+K[t]
+			SHA512_Round t-1    # Round t-1
+			vmovdqa  %xmm0, WK_2(t)# Store W[t]+K[t] into WK
+		.elseif t < 79
+			# Schedule 2 QWORDS# Compute 2 Rounds
+			SHA512_2Sched_2Round_avx t
+		.else
+			# Compute 2 Rounds
+			SHA512_Round t-2
+			SHA512_Round t-1
+		.endif
+		t = t+2
+	.endr
+
+	# Update digest
+	add     a_64, DIGEST(0)
+	add     b_64, DIGEST(1)
+	add     c_64, DIGEST(2)
+	add     d_64, DIGEST(3)
+	add     e_64, DIGEST(4)
+	add     f_64, DIGEST(5)
+	add     g_64, DIGEST(6)
+	add     h_64, DIGEST(7)
+
+	# Advance to next message block
+	add     $16*8, msg
+	dec     msglen
+	jnz     updateblock
+
+	# Restore GPRs
+	mov     frame_GPRSAVE(%rsp),      %rbx
+	mov     frame_GPRSAVE +8*1(%rsp), %r12
+	mov     frame_GPRSAVE +8*2(%rsp), %r13
+	mov     frame_GPRSAVE +8*3(%rsp), %r14
+	mov     frame_GPRSAVE +8*4(%rsp), %r15
+
+	# Restore Stack Pointer
+	mov	frame_RSPSAVE(%rsp), %rsp
+
+nowork:
+	ret
+ENDPROC(sha512_transform_avx)
+
+########################################################################
+### Binary Data
+
+.data
+
+.align 16
+
+# Mask for byte-swapping a couple of qwords in an XMM register using (v)pshufb.
+XMM_QWORD_BSWAP:
+	.octa 0x08090a0b0c0d0e0f0001020304050607
+
+# K[t] used in SHA512 hashing
+K512:
+	.quad 0x428a2f98d728ae22,0x7137449123ef65cd
+	.quad 0xb5c0fbcfec4d3b2f,0xe9b5dba58189dbbc
+	.quad 0x3956c25bf348b538,0x59f111f1b605d019
+	.quad 0x923f82a4af194f9b,0xab1c5ed5da6d8118
+	.quad 0xd807aa98a3030242,0x12835b0145706fbe
+	.quad 0x243185be4ee4b28c,0x550c7dc3d5ffb4e2
+	.quad 0x72be5d74f27b896f,0x80deb1fe3b1696b1
+	.quad 0x9bdc06a725c71235,0xc19bf174cf692694
+	.quad 0xe49b69c19ef14ad2,0xefbe4786384f25e3
+	.quad 0x0fc19dc68b8cd5b5,0x240ca1cc77ac9c65
+	.quad 0x2de92c6f592b0275,0x4a7484aa6ea6e483
+	.quad 0x5cb0a9dcbd41fbd4,0x76f988da831153b5
+	.quad 0x983e5152ee66dfab,0xa831c66d2db43210
+	.quad 0xb00327c898fb213f,0xbf597fc7beef0ee4
+	.quad 0xc6e00bf33da88fc2,0xd5a79147930aa725
+	.quad 0x06ca6351e003826f,0x142929670a0e6e70
+	.quad 0x27b70a8546d22ffc,0x2e1b21385c26c926
+	.quad 0x4d2c6dfc5ac42aed,0x53380d139d95b3df
+	.quad 0x650a73548baf63de,0x766a0abb3c77b2a8
+	.quad 0x81c2c92e47edaee6,0x92722c851482353b
+	.quad 0xa2bfe8a14cf10364,0xa81a664bbc423001
+	.quad 0xc24b8b70d0f89791,0xc76c51a30654be30
+	.quad 0xd192e819d6ef5218,0xd69906245565a910
+	.quad 0xf40e35855771202a,0x106aa07032bbd1b8
+	.quad 0x19a4c116b8d2d0c8,0x1e376c085141ab53
+	.quad 0x2748774cdf8eeb99,0x34b0bcb5e19b48a8
+	.quad 0x391c0cb3c5c95a63,0x4ed8aa4ae3418acb
+	.quad 0x5b9cca4f7763e373,0x682e6ff3d6b2b8a3
+	.quad 0x748f82ee5defb2fc,0x78a5636f43172f60
+	.quad 0x84c87814a1f0ab72,0x8cc702081a6439ec
+	.quad 0x90befffa23631e28,0xa4506cebde82bde9
+	.quad 0xbef9a3f7b2c67915,0xc67178f2e372532b
+	.quad 0xca273eceea26619c,0xd186b8c721c0c207
+	.quad 0xeada7dd6cde0eb1e,0xf57d4f7fee6ed178
+	.quad 0x06f067aa72176fba,0x0a637dc5a2c898a6
+	.quad 0x113f9804bef90dae,0x1b710b35131c471b
+	.quad 0x28db77f523047d84,0x32caab7b40c72493
+	.quad 0x3c9ebe0a15c9bebc,0x431d67c49c100d4c
+	.quad 0x4cc5d4becb3e42b6,0x597f299cfc657e2a
+	.quad 0x5fcb6fab3ad6faec,0x6c44198c4a475817
+#endif