Revision control
Copy as Markdown
Other Tools
/* sha1-avx2-bmi2-amd64.S - Intel AVX2/BMI2 accelerated SHA-1 transform function
* Copyright (C) 2019 Jussi Kivilinna <jussi.kivilinna@iki.fi>
*
* Based on sha1.c:
* Copyright (C) 1998, 2001, 2002, 2003, 2008 Free Software Foundation, Inc.
*
* This file is part of Libgcrypt.
*
* Libgcrypt is free software; you can redistribute it and/or modify
* it under the terms of the GNU Lesser General Public License as
* published by the Free Software Foundation; either version 2.1 of
* the License, or (at your option) any later version.
*
* Libgcrypt is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
*/
/*
* Intel SSSE3 accelerated SHA-1 implementation based on white paper:
* "Improving the Performance of the Secure Hash Algorithm (SHA-1)"
*/
#ifdef __x86_64__
#include <config.h>
#if (defined(HAVE_COMPATIBLE_GCC_AMD64_PLATFORM_AS) || \
defined(HAVE_COMPATIBLE_GCC_WIN64_PLATFORM_AS)) && \
defined(HAVE_GCC_INLINE_ASM_BMI2) && defined(HAVE_GCC_INLINE_ASM_AVX) && \
defined(HAVE_GCC_INLINE_ASM_AVX2) && defined(USE_SHA1)
#include "asm-common-amd64.h"
/* Context structure */
#define state_h0 0
#define state_h1 4
#define state_h2 8
#define state_h3 12
#define state_h4 16
/* Constants */
#define WK_STACK_WORDS (80 * 2)
.text
.align 16
.Lbswap_shufb_ctl:
.long 0x00010203, 0x04050607, 0x08090a0b, 0x0c0d0e0f
.LK1: .long 0x5A827999
.LK2: .long 0x6ED9EBA1
.LK3: .long 0x8F1BBCDC
.LK4: .long 0xCA62C1D6
/* Register macros */
#define RSTATE %r8
#define RDATA %r9
#define ROLDSTACK %r10
#define RNBLKS %r11
#define a %eax
#define b %ebx
#define c %ecx
#define d %edx
#define e %edi
#define ne %r12d
#define RT0 %esi
#define RT1 %ebp
#define Wtmp0 %ymm0
#define Wtmp1 %ymm1
#define Wtmp0x %xmm0
#define Wtmp1x %xmm1
#define W0 %ymm2
#define W1 %ymm3
#define W2 %ymm4
#define W3 %ymm5
#define W4 %ymm6
#define W5 %ymm7
#define W6 %ymm8
#define W7 %ymm9
#define BSWAP_REG %ymm10
#define K1 %ymm11
#define K2 %ymm12
#define K3 %ymm13
#define K4 %ymm14
/* Round function macros. */
#define WK(i,block) ((block) * 16 + ((i) / 4) * 32 + ((i) % 4) * 4)(%rsp)
#define PRE_WK(i) ((i) * 4 * 2)(%rsp)
#define R_F1(a,b,c,d,e,i,block) \
movl c, RT0; \
andn d, b, RT1; \
addl WK(i,block), e; \
andl b, RT0; \
leal (a,ne), a; \
rorxl $2, b, b; \
addl RT1, e; \
rorxl $27, a, ne; \
addl RT0, e;
#define R_F2(a,b,c,d,e,i,block) \
addl WK(i,block), e; \
movl c, RT0; \
xorl b, RT0; \
leal (a,ne), a; \
rorxl $2, b, b; \
xorl d, RT0; \
addl RT0, e; \
rorxl $27, a, ne;
#define R_F3(a,b,c,d,e,i,block) \
movl c, RT0; \
addl WK(i,block), e; \
movl b, RT1; \
xorl b, RT0; \
leal (a,ne), a; \
rorxl $2, b, b; \
andl c, RT1; \
addl RT1, e; \
andl d, RT0; \
rorxl $27, a, ne; \
addl RT0, e;
#define R_F4(a,b,c,d,e,i,block) R_F2(a,b,c,d,e,i,block)
#define R(a,b,c,d,e,f,i,block) \
R_##f(a,b,c,d,e,i,block)
/* Input expansion macros. */
#define W_PRECALC_00_15_0(i, W, tmp0) \
vmovdqu (4*(i))(RDATA), tmp0##x; \
vinserti128 $1, (4*(i) + 64)(RDATA), tmp0, tmp0;
#define W_PRECALC_00_15_1(i, W, tmp0) \
vpshufb BSWAP_REG, tmp0, W;
#define W_PRECALC_00_15_2(i, W, tmp0, K) \
vpaddd K, W, tmp0;
#define W_PRECALC_00_15_3(i, W, tmp0) \
vmovdqa tmp0, PRE_WK((i)&~3);
#define W_PRECALC_16_31_0(i, W, W_m04, W_m08, W_m12, W_m16, tmp0, tmp1) \
vpalignr $8, W_m16, W_m12, W; \
vpsrldq $4, W_m04, tmp0; \
vpxor W_m08, W, W;
#define W_PRECALC_16_31_1(i, W, W_m04, W_m08, W_m12, W_m16, tmp0, tmp1) \
vpxor W_m16, tmp0, tmp0; \
vpxor tmp0, W, W; \
vpslld $1, W, tmp0; \
vpslldq $12, W, tmp1; \
vpsrld $31, W, W;
#define W_PRECALC_16_31_2(i, W, W_m04, W_m08, W_m12, W_m16, tmp0, tmp1) \
vpor W, tmp0, tmp0; \
vpsrld $30, tmp1, W; \
vpslld $2, tmp1, tmp1;
#define W_PRECALC_16_31_3(i, W, W_m04, W_m08, W_m12, W_m16, tmp0, tmp1, K) \
vpxor W, tmp0, tmp0; \
vpxor tmp1, tmp0, W; \
vpaddd K, W, tmp0; \
vmovdqa tmp0, PRE_WK((i)&~3);
#define W_PRECALC_32_79_0(i, W, W_m04, W_m08, W_m12, W_m16, W_m20, W_m24, W_m28, tmp0) \
vpxor W_m28, W, W; \
vpalignr $8, W_m08, W_m04, tmp0;
#define W_PRECALC_32_79_1(i, W, W_m04, W_m08, W_m12, W_m16, W_m20, W_m24, W_m28, tmp0) \
vpxor W_m16, W, W; \
vpxor tmp0, W, W;
#define W_PRECALC_32_79_2(i, W, W_m04, W_m08, W_m12, W_m16, W_m20, W_m24, W_m28, tmp0) \
vpsrld $30, W, tmp0; \
vpslld $2, W, W;
#define W_PRECALC_32_79_3(i, W, W_m04, W_m08, W_m12, W_m16, W_m20, W_m24, W_m28, tmp0, K) \
vpor W, tmp0, W; \
vpaddd K, W, tmp0; \
vmovdqa tmp0, PRE_WK((i)&~3);
/*
* Transform 2*nblks*64 bytes (2*nblks*16 32-bit words) at DATA.
*
* unsigned int
* _gcry_sha1_transform_amd64_avx2_bmi2 (void *ctx, const unsigned char *data,
* size_t nblks)
*/
.globl _gcry_sha1_transform_amd64_avx2_bmi2
ELF(.type _gcry_sha1_transform_amd64_avx2_bmi2,@function)
.align 16
_gcry_sha1_transform_amd64_avx2_bmi2:
/* input:
* %rdi: ctx, CTX
* %rsi: data (64*nblks bytes)
* %rdx: nblks (multiple of 2, larger than 0)
*/
CFI_STARTPROC();
vzeroupper;
movq %rdx, RNBLKS;
movq %rdi, RSTATE;
movq %rsi, RDATA;
pushq %rbx;
CFI_PUSH(%rbx);
pushq %rbp;
CFI_PUSH(%rbp);
pushq %r12;
CFI_PUSH(%r12);
movq %rsp, ROLDSTACK;
CFI_DEF_CFA_REGISTER(ROLDSTACK);
subq $(WK_STACK_WORDS*4), %rsp;
andq $(~63), %rsp;
/* Get the values of the chaining variables. */
movl state_h0(RSTATE), a;
movl state_h1(RSTATE), b;
movl state_h2(RSTATE), c;
movl state_h3(RSTATE), d;
movl state_h4(RSTATE), e;
xorl ne, ne;
vbroadcasti128 .Lbswap_shufb_ctl rRIP, BSWAP_REG;
vpbroadcastd .LK1 rRIP, K1;
vpbroadcastd .LK2 rRIP, K2;
vpbroadcastd .LK3 rRIP, K3;
vpbroadcastd .LK4 rRIP, K4;
/* Precalc 0-31 for block 1 & 2. */
W_PRECALC_00_15_0(0, W0, Wtmp0);
W_PRECALC_00_15_1(1, W0, Wtmp0);
W_PRECALC_00_15_2(2, W0, Wtmp0, K1);
W_PRECALC_00_15_3(3, W0, Wtmp0);
W_PRECALC_00_15_0(4, W7, Wtmp0);
W_PRECALC_00_15_1(5, W7, Wtmp0);
W_PRECALC_00_15_2(6, W7, Wtmp0, K1);
W_PRECALC_00_15_3(7, W7, Wtmp0);
W_PRECALC_00_15_0(8, W6, Wtmp0);
W_PRECALC_00_15_1(9, W6, Wtmp0);
W_PRECALC_00_15_2(10, W6, Wtmp0, K1);
W_PRECALC_00_15_3(11, W6, Wtmp0);
W_PRECALC_00_15_0(12, W5, Wtmp0);
W_PRECALC_00_15_1(13, W5, Wtmp0);
W_PRECALC_00_15_2(14, W5, Wtmp0, K1);
W_PRECALC_00_15_3(15, W5, Wtmp0);
W_PRECALC_16_31_0(16, W4, W5, W6, W7, W0, Wtmp0, Wtmp1);
W_PRECALC_16_31_1(17, W4, W5, W6, W7, W0, Wtmp0, Wtmp1);
W_PRECALC_16_31_2(18, W4, W5, W6, W7, W0, Wtmp0, Wtmp1);
W_PRECALC_16_31_3(19, W4, W5, W6, W7, W0, Wtmp0, Wtmp1, K1);
W_PRECALC_16_31_0(20, W3, W4, W5, W6, W7, Wtmp0, Wtmp1);
W_PRECALC_16_31_1(21, W3, W4, W5, W6, W7, Wtmp0, Wtmp1);
W_PRECALC_16_31_2(22, W3, W4, W5, W6, W7, Wtmp0, Wtmp1);
W_PRECALC_16_31_3(23, W3, W4, W5, W6, W7, Wtmp0, Wtmp1, K2);
W_PRECALC_16_31_0(24, W2, W3, W4, W5, W6, Wtmp0, Wtmp1);
W_PRECALC_16_31_1(25, W2, W3, W4, W5, W6, Wtmp0, Wtmp1);
W_PRECALC_16_31_2(26, W2, W3, W4, W5, W6, Wtmp0, Wtmp1);
W_PRECALC_16_31_3(27, W2, W3, W4, W5, W6, Wtmp0, Wtmp1, K2);
W_PRECALC_16_31_0(28, W1, W2, W3, W4, W5, Wtmp0, Wtmp1);
W_PRECALC_16_31_1(29, W1, W2, W3, W4, W5, Wtmp0, Wtmp1);
W_PRECALC_16_31_2(30, W1, W2, W3, W4, W5, Wtmp0, Wtmp1);
W_PRECALC_16_31_3(31, W1, W2, W3, W4, W5, Wtmp0, Wtmp1, K2);
.align 8
.Loop:
addq $(2 * 64), RDATA;
/* Transform 0-15 for block 1 + Precalc 32-47 for block 1 & 2. */
R( a, b, c, d, e, F1, 0, 0 ); W_PRECALC_32_79_0(32, W0, W1, W2, W3, W4, W5, W6, W7, Wtmp0);
R( e, a, b, c, d, F1, 1, 0 ); W_PRECALC_32_79_1(33, W0, W1, W2, W3, W4, W5, W6, W7, Wtmp0);
R( d, e, a, b, c, F1, 2, 0 ); W_PRECALC_32_79_2(34, W0, W1, W2, W3, W4, W5, W6, W7, Wtmp0);
R( c, d, e, a, b, F1, 3, 0 ); W_PRECALC_32_79_3(35, W0, W1, W2, W3, W4, W5, W6, W7, Wtmp0, K2);
R( b, c, d, e, a, F1, 4, 0 ); W_PRECALC_32_79_0(36, W7, W0, W1, W2, W3, W4, W5, W6, Wtmp0);
R( a, b, c, d, e, F1, 5, 0 ); W_PRECALC_32_79_1(37, W7, W0, W1, W2, W3, W4, W5, W6, Wtmp0);
R( e, a, b, c, d, F1, 6, 0 ); W_PRECALC_32_79_2(38, W7, W0, W1, W2, W3, W4, W5, W6, Wtmp0);
R( d, e, a, b, c, F1, 7, 0 ); W_PRECALC_32_79_3(39, W7, W0, W1, W2, W3, W4, W5, W6, Wtmp0, K2);
R( c, d, e, a, b, F1, 8, 0 ); W_PRECALC_32_79_0(40, W6, W7, W0, W1, W2, W3, W4, W5, Wtmp0);
R( b, c, d, e, a, F1, 9, 0 ); W_PRECALC_32_79_1(41, W6, W7, W0, W1, W2, W3, W4, W5, Wtmp0);
R( a, b, c, d, e, F1, 10, 0 ); W_PRECALC_32_79_2(42, W6, W7, W0, W1, W2, W3, W4, W5, Wtmp0);
R( e, a, b, c, d, F1, 11, 0 ); W_PRECALC_32_79_3(43, W6, W7, W0, W1, W2, W3, W4, W5, Wtmp0, K3);
R( d, e, a, b, c, F1, 12, 0 ); W_PRECALC_32_79_0(44, W5, W6, W7, W0, W1, W2, W3, W4, Wtmp0);
R( c, d, e, a, b, F1, 13, 0 ); W_PRECALC_32_79_1(45, W5, W6, W7, W0, W1, W2, W3, W4, Wtmp0);
R( b, c, d, e, a, F1, 14, 0 ); W_PRECALC_32_79_2(46, W5, W6, W7, W0, W1, W2, W3, W4, Wtmp0);
R( a, b, c, d, e, F1, 15, 0 ); W_PRECALC_32_79_3(47, W5, W6, W7, W0, W1, W2, W3, W4, Wtmp0, K3);
/* Transform 16-47 for block 1 + Precalc 48-79 for block 1 & 2. */
R( e, a, b, c, d, F1, 16, 0 ); W_PRECALC_32_79_0(48, W4, W5, W6, W7, W0, W1, W2, W3, Wtmp0);
R( d, e, a, b, c, F1, 17, 0 ); W_PRECALC_32_79_1(49, W4, W5, W6, W7, W0, W1, W2, W3, Wtmp0);
R( c, d, e, a, b, F1, 18, 0 ); W_PRECALC_32_79_2(50, W4, W5, W6, W7, W0, W1, W2, W3, Wtmp0);
R( b, c, d, e, a, F1, 19, 0 ); W_PRECALC_32_79_3(51, W4, W5, W6, W7, W0, W1, W2, W3, Wtmp0, K3);
R( a, b, c, d, e, F2, 20, 0 ); W_PRECALC_32_79_0(52, W3, W4, W5, W6, W7, W0, W1, W2, Wtmp0);
R( e, a, b, c, d, F2, 21, 0 ); W_PRECALC_32_79_1(53, W3, W4, W5, W6, W7, W0, W1, W2, Wtmp0);
R( d, e, a, b, c, F2, 22, 0 ); W_PRECALC_32_79_2(54, W3, W4, W5, W6, W7, W0, W1, W2, Wtmp0);
R( c, d, e, a, b, F2, 23, 0 ); W_PRECALC_32_79_3(55, W3, W4, W5, W6, W7, W0, W1, W2, Wtmp0, K3);
R( b, c, d, e, a, F2, 24, 0 ); W_PRECALC_32_79_0(56, W2, W3, W4, W5, W6, W7, W0, W1, Wtmp0);
R( a, b, c, d, e, F2, 25, 0 ); W_PRECALC_32_79_1(57, W2, W3, W4, W5, W6, W7, W0, W1, Wtmp0);
R( e, a, b, c, d, F2, 26, 0 ); W_PRECALC_32_79_2(58, W2, W3, W4, W5, W6, W7, W0, W1, Wtmp0);
R( d, e, a, b, c, F2, 27, 0 ); W_PRECALC_32_79_3(59, W2, W3, W4, W5, W6, W7, W0, W1, Wtmp0, K3);
R( c, d, e, a, b, F2, 28, 0 ); W_PRECALC_32_79_0(60, W1, W2, W3, W4, W5, W6, W7, W0, Wtmp0);
R( b, c, d, e, a, F2, 29, 0 ); W_PRECALC_32_79_1(61, W1, W2, W3, W4, W5, W6, W7, W0, Wtmp0);
R( a, b, c, d, e, F2, 30, 0 ); W_PRECALC_32_79_2(62, W1, W2, W3, W4, W5, W6, W7, W0, Wtmp0);
R( e, a, b, c, d, F2, 31, 0 ); W_PRECALC_32_79_3(63, W1, W2, W3, W4, W5, W6, W7, W0, Wtmp0, K4);
R( d, e, a, b, c, F2, 32, 0 ); W_PRECALC_32_79_0(64, W0, W1, W2, W3, W4, W5, W6, W7, Wtmp0);
R( c, d, e, a, b, F2, 33, 0 ); W_PRECALC_32_79_1(65, W0, W1, W2, W3, W4, W5, W6, W7, Wtmp0);
R( b, c, d, e, a, F2, 34, 0 ); W_PRECALC_32_79_2(66, W0, W1, W2, W3, W4, W5, W6, W7, Wtmp0);
R( a, b, c, d, e, F2, 35, 0 ); W_PRECALC_32_79_3(67, W0, W1, W2, W3, W4, W5, W6, W7, Wtmp0, K4);
R( e, a, b, c, d, F2, 36, 0 ); W_PRECALC_32_79_0(68, W7, W0, W1, W2, W3, W4, W5, W6, Wtmp0);
R( d, e, a, b, c, F2, 37, 0 ); W_PRECALC_32_79_1(69, W7, W0, W1, W2, W3, W4, W5, W6, Wtmp0);
R( c, d, e, a, b, F2, 38, 0 ); W_PRECALC_32_79_2(70, W7, W0, W1, W2, W3, W4, W5, W6, Wtmp0);
R( b, c, d, e, a, F2, 39, 0 ); W_PRECALC_32_79_3(71, W7, W0, W1, W2, W3, W4, W5, W6, Wtmp0, K4);
R( a, b, c, d, e, F3, 40, 0 ); W_PRECALC_32_79_0(72, W6, W7, W0, W1, W2, W3, W4, W5, Wtmp0);
R( e, a, b, c, d, F3, 41, 0 ); W_PRECALC_32_79_1(73, W6, W7, W0, W1, W2, W3, W4, W5, Wtmp0);
R( d, e, a, b, c, F3, 42, 0 ); W_PRECALC_32_79_2(74, W6, W7, W0, W1, W2, W3, W4, W5, Wtmp0);
R( c, d, e, a, b, F3, 43, 0 ); W_PRECALC_32_79_3(75, W6, W7, W0, W1, W2, W3, W4, W5, Wtmp0, K4);
R( b, c, d, e, a, F3, 44, 0 ); W_PRECALC_32_79_0(76, W5, W6, W7, W0, W1, W2, W3, W4, Wtmp0);
R( a, b, c, d, e, F3, 45, 0 ); W_PRECALC_32_79_1(77, W5, W6, W7, W0, W1, W2, W3, W4, Wtmp0);
R( e, a, b, c, d, F3, 46, 0 ); W_PRECALC_32_79_2(78, W5, W6, W7, W0, W1, W2, W3, W4, Wtmp0);
R( d, e, a, b, c, F3, 47, 0 ); W_PRECALC_32_79_3(79, W5, W6, W7, W0, W1, W2, W3, W4, Wtmp0, K4);
/* Transform 48-79 for block 1. */
R( c, d, e, a, b, F3, 48, 0 );
R( b, c, d, e, a, F3, 49, 0 );
R( a, b, c, d, e, F3, 50, 0 );
R( e, a, b, c, d, F3, 51, 0 );
R( d, e, a, b, c, F3, 52, 0 );
R( c, d, e, a, b, F3, 53, 0 );
R( b, c, d, e, a, F3, 54, 0 );
R( a, b, c, d, e, F3, 55, 0 );
R( e, a, b, c, d, F3, 56, 0 );
R( d, e, a, b, c, F3, 57, 0 );
R( c, d, e, a, b, F3, 58, 0 );
R( b, c, d, e, a, F3, 59, 0 );
R( a, b, c, d, e, F4, 60, 0 );
R( e, a, b, c, d, F4, 61, 0 );
R( d, e, a, b, c, F4, 62, 0 );
R( c, d, e, a, b, F4, 63, 0 );
R( b, c, d, e, a, F4, 64, 0 );
R( a, b, c, d, e, F4, 65, 0 );
R( e, a, b, c, d, F4, 66, 0 );
R( d, e, a, b, c, F4, 67, 0 );
R( c, d, e, a, b, F4, 68, 0 );
R( b, c, d, e, a, F4, 69, 0 );
R( a, b, c, d, e, F4, 70, 0 );
R( e, a, b, c, d, F4, 71, 0 );
R( d, e, a, b, c, F4, 72, 0 );
R( c, d, e, a, b, F4, 73, 0 );
R( b, c, d, e, a, F4, 74, 0 );
R( a, b, c, d, e, F4, 75, 0 );
R( e, a, b, c, d, F4, 76, 0 );
R( d, e, a, b, c, F4, 77, 0 );
R( c, d, e, a, b, F4, 78, 0 );
addl state_h0(RSTATE), a;
R( b, c, d, e, a, F4, 79, 0 );
addl ne, a;
xorl ne, ne;
/* Update the chaining variables. */
addl state_h3(RSTATE), d;
addl state_h2(RSTATE), c;
addl state_h1(RSTATE), b;
addl state_h4(RSTATE), e;
movl d, state_h3(RSTATE);
movl c, state_h2(RSTATE);
movl b, state_h1(RSTATE);
movl a, state_h0(RSTATE);
movl e, state_h4(RSTATE);
/* Transform 0-47 for block 2. */
R( a, b, c, d, e, F1, 0, 1 );
R( e, a, b, c, d, F1, 1, 1 );
R( d, e, a, b, c, F1, 2, 1 );
R( c, d, e, a, b, F1, 3, 1 );
R( b, c, d, e, a, F1, 4, 1 );
R( a, b, c, d, e, F1, 5, 1 );
R( e, a, b, c, d, F1, 6, 1 );
R( d, e, a, b, c, F1, 7, 1 );
R( c, d, e, a, b, F1, 8, 1 );
R( b, c, d, e, a, F1, 9, 1 );
R( a, b, c, d, e, F1, 10, 1 );
R( e, a, b, c, d, F1, 11, 1 );
R( d, e, a, b, c, F1, 12, 1 );
R( c, d, e, a, b, F1, 13, 1 );
R( b, c, d, e, a, F1, 14, 1 );
R( a, b, c, d, e, F1, 15, 1 );
R( e, a, b, c, d, F1, 16, 1 );
R( d, e, a, b, c, F1, 17, 1 );
R( c, d, e, a, b, F1, 18, 1 );
R( b, c, d, e, a, F1, 19, 1 );
R( a, b, c, d, e, F2, 20, 1 );
R( e, a, b, c, d, F2, 21, 1 );
R( d, e, a, b, c, F2, 22, 1 );
R( c, d, e, a, b, F2, 23, 1 );
R( b, c, d, e, a, F2, 24, 1 );
R( a, b, c, d, e, F2, 25, 1 );
R( e, a, b, c, d, F2, 26, 1 );
R( d, e, a, b, c, F2, 27, 1 );
R( c, d, e, a, b, F2, 28, 1 );
R( b, c, d, e, a, F2, 29, 1 );
R( a, b, c, d, e, F2, 30, 1 );
R( e, a, b, c, d, F2, 31, 1 );
R( d, e, a, b, c, F2, 32, 1 );
R( c, d, e, a, b, F2, 33, 1 );
R( b, c, d, e, a, F2, 34, 1 );
R( a, b, c, d, e, F2, 35, 1 );
R( e, a, b, c, d, F2, 36, 1 );
R( d, e, a, b, c, F2, 37, 1 );
R( c, d, e, a, b, F2, 38, 1 );
R( b, c, d, e, a, F2, 39, 1 );
R( a, b, c, d, e, F3, 40, 1 );
R( e, a, b, c, d, F3, 41, 1 );
R( d, e, a, b, c, F3, 42, 1 );
R( c, d, e, a, b, F3, 43, 1 );
R( b, c, d, e, a, F3, 44, 1 );
R( a, b, c, d, e, F3, 45, 1 );
R( e, a, b, c, d, F3, 46, 1 );
R( d, e, a, b, c, F3, 47, 1 );
addq $-2, RNBLKS;
jz .Lend;
/* Transform 48-79 for block 2 + Precalc 0-31 for next two blocks. */
R( c, d, e, a, b, F3, 48, 1 ); W_PRECALC_00_15_0(0, W0, Wtmp0);
R( b, c, d, e, a, F3, 49, 1 ); W_PRECALC_00_15_1(1, W0, Wtmp0);
R( a, b, c, d, e, F3, 50, 1 ); W_PRECALC_00_15_2(2, W0, Wtmp0, K1);
R( e, a, b, c, d, F3, 51, 1 ); W_PRECALC_00_15_3(3, W0, Wtmp0);
R( d, e, a, b, c, F3, 52, 1 ); W_PRECALC_00_15_0(4, W7, Wtmp0);
R( c, d, e, a, b, F3, 53, 1 ); W_PRECALC_00_15_1(5, W7, Wtmp0);
R( b, c, d, e, a, F3, 54, 1 ); W_PRECALC_00_15_2(6, W7, Wtmp0, K1);
R( a, b, c, d, e, F3, 55, 1 ); W_PRECALC_00_15_3(7, W7, Wtmp0);
R( e, a, b, c, d, F3, 56, 1 ); W_PRECALC_00_15_0(8, W6, Wtmp0);
R( d, e, a, b, c, F3, 57, 1 ); W_PRECALC_00_15_1(9, W6, Wtmp0);
R( c, d, e, a, b, F3, 58, 1 ); W_PRECALC_00_15_2(10, W6, Wtmp0, K1);
R( b, c, d, e, a, F3, 59, 1 ); W_PRECALC_00_15_3(11, W6, Wtmp0);
R( a, b, c, d, e, F4, 60, 1 ); W_PRECALC_00_15_0(12, W5, Wtmp0);
R( e, a, b, c, d, F4, 61, 1 ); W_PRECALC_00_15_1(13, W5, Wtmp0);
R( d, e, a, b, c, F4, 62, 1 ); W_PRECALC_00_15_2(14, W5, Wtmp0, K1);
R( c, d, e, a, b, F4, 63, 1 ); W_PRECALC_00_15_3(15, W5, Wtmp0);
R( b, c, d, e, a, F4, 64, 1 ); W_PRECALC_16_31_0(16, W4, W5, W6, W7, W0, Wtmp0, Wtmp1);
R( a, b, c, d, e, F4, 65, 1 ); W_PRECALC_16_31_1(17, W4, W5, W6, W7, W0, Wtmp0, Wtmp1);
R( e, a, b, c, d, F4, 66, 1 ); W_PRECALC_16_31_2(18, W4, W5, W6, W7, W0, Wtmp0, Wtmp1);
R( d, e, a, b, c, F4, 67, 1 ); W_PRECALC_16_31_3(19, W4, W5, W6, W7, W0, Wtmp0, Wtmp1, K1);
R( c, d, e, a, b, F4, 68, 1 ); W_PRECALC_16_31_0(20, W3, W4, W5, W6, W7, Wtmp0, Wtmp1);
R( b, c, d, e, a, F4, 69, 1 ); W_PRECALC_16_31_1(21, W3, W4, W5, W6, W7, Wtmp0, Wtmp1);
R( a, b, c, d, e, F4, 70, 1 ); W_PRECALC_16_31_2(22, W3, W4, W5, W6, W7, Wtmp0, Wtmp1);
R( e, a, b, c, d, F4, 71, 1 ); W_PRECALC_16_31_3(23, W3, W4, W5, W6, W7, Wtmp0, Wtmp1, K2);
R( d, e, a, b, c, F4, 72, 1 ); W_PRECALC_16_31_0(24, W2, W3, W4, W5, W6, Wtmp0, Wtmp1);
R( c, d, e, a, b, F4, 73, 1 ); W_PRECALC_16_31_1(25, W2, W3, W4, W5, W6, Wtmp0, Wtmp1);
R( b, c, d, e, a, F4, 74, 1 ); W_PRECALC_16_31_2(26, W2, W3, W4, W5, W6, Wtmp0, Wtmp1);
R( a, b, c, d, e, F4, 75, 1 ); W_PRECALC_16_31_3(27, W2, W3, W4, W5, W6, Wtmp0, Wtmp1, K2);
R( e, a, b, c, d, F4, 76, 1 ); W_PRECALC_16_31_0(28, W1, W2, W3, W4, W5, Wtmp0, Wtmp1);
R( d, e, a, b, c, F4, 77, 1 ); W_PRECALC_16_31_1(29, W1, W2, W3, W4, W5, Wtmp0, Wtmp1);
R( c, d, e, a, b, F4, 78, 1 ); W_PRECALC_16_31_2(30, W1, W2, W3, W4, W5, Wtmp0, Wtmp1);
addl state_h0(RSTATE), a; W_PRECALC_16_31_3(31, W1, W2, W3, W4, W5, Wtmp0, Wtmp1, K2);
R( b, c, d, e, a, F4, 79, 1 );
addl ne, a;
xorl ne, ne;
/* Update the chaining variables. */
addl state_h3(RSTATE), d;
addl state_h2(RSTATE), c;
addl state_h1(RSTATE), b;
addl state_h4(RSTATE), e;
movl d, state_h3(RSTATE);
movl c, state_h2(RSTATE);
movl b, state_h1(RSTATE);
movl a, state_h0(RSTATE);
movl e, state_h4(RSTATE);
jmp .Loop;
.align 16
.Lend:
vzeroall;
/* Transform 48-79 for block 2 + burn stack */
R( c, d, e, a, b, F3, 48, 1 );
R( b, c, d, e, a, F3, 49, 1 );
R( a, b, c, d, e, F3, 50, 1 );
R( e, a, b, c, d, F3, 51, 1 );
R( d, e, a, b, c, F3, 52, 1 );
R( c, d, e, a, b, F3, 53, 1 );
R( b, c, d, e, a, F3, 54, 1 );
R( a, b, c, d, e, F3, 55, 1 );
R( e, a, b, c, d, F3, 56, 1 );
R( d, e, a, b, c, F3, 57, 1 );
R( c, d, e, a, b, F3, 58, 1 );
R( b, c, d, e, a, F3, 59, 1 );
R( a, b, c, d, e, F4, 60, 1 ); vmovdqa %ymm0, (0*32)(%rsp);
R( e, a, b, c, d, F4, 61, 1 ); vmovdqa %ymm0, (1*32)(%rsp);
R( d, e, a, b, c, F4, 62, 1 ); vmovdqa %ymm0, (2*32)(%rsp);
R( c, d, e, a, b, F4, 63, 1 ); vmovdqa %ymm0, (3*32)(%rsp);
R( b, c, d, e, a, F4, 64, 1 ); vmovdqa %ymm0, (4*32)(%rsp);
R( a, b, c, d, e, F4, 65, 1 ); vmovdqa %ymm0, (5*32)(%rsp);
R( e, a, b, c, d, F4, 66, 1 ); vmovdqa %ymm0, (6*32)(%rsp);
R( d, e, a, b, c, F4, 67, 1 ); vmovdqa %ymm0, (7*32)(%rsp);
R( c, d, e, a, b, F4, 68, 1 ); vmovdqa %ymm0, (8*32)(%rsp);
R( b, c, d, e, a, F4, 69, 1 ); vmovdqa %ymm0, (9*32)(%rsp);
R( a, b, c, d, e, F4, 70, 1 ); vmovdqa %ymm0, (10*32)(%rsp);
R( e, a, b, c, d, F4, 71, 1 ); vmovdqa %ymm0, (11*32)(%rsp);
R( d, e, a, b, c, F4, 72, 1 ); vmovdqa %ymm0, (12*32)(%rsp);
R( c, d, e, a, b, F4, 73, 1 ); vmovdqa %ymm0, (13*32)(%rsp);
R( b, c, d, e, a, F4, 74, 1 ); vmovdqa %ymm0, (14*32)(%rsp);
R( a, b, c, d, e, F4, 75, 1 ); vmovdqa %ymm0, (15*32)(%rsp);
R( e, a, b, c, d, F4, 76, 1 ); vmovdqa %ymm0, (16*32)(%rsp);
R( d, e, a, b, c, F4, 77, 1 ); vmovdqa %ymm0, (17*32)(%rsp);
R( c, d, e, a, b, F4, 78, 1 ); vmovdqa %ymm0, (18*32)(%rsp);
addl state_h0(RSTATE), a;
R( b, c, d, e, a, F4, 79, 1 );
addl ne, a;
xorl ne, ne;
/* WK_STACK_WORDS*4/32-1 = 19 */
vmovdqa %ymm0, (19*32)(%rsp);
/* Update the chaining variables. */
addl state_h3(RSTATE), d;
addl state_h2(RSTATE), c;
addl state_h1(RSTATE), b;
addl state_h4(RSTATE), e;
movl d, state_h3(RSTATE);
movl c, state_h2(RSTATE);
movl b, state_h1(RSTATE);
movl a, state_h0(RSTATE);
movl e, state_h4(RSTATE);
movq ROLDSTACK, %rsp;
CFI_REGISTER(ROLDSTACK, %rsp);
CFI_DEF_CFA_REGISTER(%rsp);
popq %r12;
CFI_POP(%r12);
popq %rbp;
CFI_POP(%rbp);
popq %rbx;
CFI_POP(%rbx);
/* stack already burned */
xorl %eax, %eax;
ret;
CFI_ENDPROC();
ELF(.size _gcry_sha1_transform_amd64_avx2_bmi2,
.-_gcry_sha1_transform_amd64_avx2_bmi2;)
#endif
#endif