Revision control
Copy as Markdown
Other Tools
/* serpent-sse2-amd64.S - SSE2 implementation of Serpent cipher
*
* Copyright (C) 2013-2015 Jussi Kivilinna <jussi.kivilinna@iki.fi>
*
* 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
*/
#ifdef __x86_64
#include <config.h>
#if (defined(HAVE_COMPATIBLE_GCC_AMD64_PLATFORM_AS) || \
defined(HAVE_COMPATIBLE_GCC_WIN64_PLATFORM_AS)) && defined(USE_SERPENT)
#include "asm-common-amd64.h"
/* struct serpent_context: */
#define ctx_keys 0
/* register macros */
#define CTX %rdi
/* vector registers */
#define RA0 %xmm0
#define RA1 %xmm1
#define RA2 %xmm2
#define RA3 %xmm3
#define RA4 %xmm4
#define RB0 %xmm5
#define RB1 %xmm6
#define RB2 %xmm7
#define RB3 %xmm8
#define RB4 %xmm9
#define RNOT %xmm10
#define RTMP0 %xmm11
#define RTMP1 %xmm12
#define RTMP2 %xmm13
/**********************************************************************
helper macros
**********************************************************************/
/* vector 32-bit rotation to left */
#define vec_rol(reg, nleft, tmp) \
movdqa reg, tmp; \
pslld $(nleft), tmp; \
psrld $(32 - (nleft)), reg; \
por tmp, reg;
/* vector 32-bit rotation to right */
#define vec_ror(reg, nright, tmp) \
vec_rol(reg, 32 - nright, tmp)
/* 4x4 32-bit integer matrix transpose */
#define transpose_4x4(x0, x1, x2, x3, t1, t2, t3) \
movdqa x0, t2; \
punpckhdq x1, t2; \
punpckldq x1, x0; \
\
movdqa x2, t1; \
punpckldq x3, t1; \
punpckhdq x3, x2; \
\
movdqa x0, x1; \
punpckhqdq t1, x1; \
punpcklqdq t1, x0; \
\
movdqa t2, x3; \
punpckhqdq x2, x3; \
punpcklqdq x2, t2; \
movdqa t2, x2;
/* fill xmm register with 32-bit value from memory */
#define pbroadcastd(mem32, xreg) \
movd mem32, xreg; \
pshufd $0, xreg, xreg;
/* xor with unaligned memory operand */
#define pxor_u(umem128, xreg, t) \
movdqu umem128, t; \
pxor t, xreg;
/* 128-bit wide byte swap */
#define pbswap(xreg, t0) \
/* reorder 32-bit words, [a,b,c,d] => [d,c,b,a] */ \
pshufd $0x1b, xreg, xreg; \
/* reorder high&low 16-bit words, [d0,d1,c0,c1] => [d1,d0,c1,c0] */ \
pshuflw $0xb1, xreg, xreg; \
pshufhw $0xb1, xreg, xreg; \
/* reorder bytes in 16-bit words */ \
movdqa xreg, t0; \
psrlw $8, t0; \
psllw $8, xreg; \
por t0, xreg;
/**********************************************************************
8-way serpent
**********************************************************************/
/*
* These are the S-Boxes of Serpent from following research paper.
*
* D. A. Osvik, “Speeding up Serpent,” in Third AES Candidate Conference,
* (New York, New York, USA), p. 317–329, National Institute of Standards and
* Technology, 2000.
*
*
*/
#define SBOX0(r0, r1, r2, r3, r4) \
pxor r0, r3; movdqa r1, r4; \
pand r3, r1; pxor r2, r4; \
pxor r0, r1; por r3, r0; \
pxor r4, r0; pxor r3, r4; \
pxor r2, r3; por r1, r2; \
pxor r4, r2; pxor RNOT, r4; \
por r1, r4; pxor r3, r1; \
pxor r4, r1; por r0, r3; \
pxor r3, r1; pxor r3, r4;
#define SBOX0_INVERSE(r0, r1, r2, r3, r4) \
pxor RNOT, r2; movdqa r1, r4; \
por r0, r1; pxor RNOT, r4; \
pxor r2, r1; por r4, r2; \
pxor r3, r1; pxor r4, r0; \
pxor r0, r2; pand r3, r0; \
pxor r0, r4; por r1, r0; \
pxor r2, r0; pxor r4, r3; \
pxor r1, r2; pxor r0, r3; \
pxor r1, r3; \
pand r3, r2; \
pxor r2, r4;
#define SBOX1(r0, r1, r2, r3, r4) \
pxor RNOT, r0; pxor RNOT, r2; \
movdqa r0, r4; pand r1, r0; \
pxor r0, r2; por r3, r0; \
pxor r2, r3; pxor r0, r1; \
pxor r4, r0; por r1, r4; \
pxor r3, r1; por r0, r2; \
pand r4, r2; pxor r1, r0; \
pand r2, r1; \
pxor r0, r1; pand r2, r0; \
pxor r4, r0;
#define SBOX1_INVERSE(r0, r1, r2, r3, r4) \
movdqa r1, r4; pxor r3, r1; \
pand r1, r3; pxor r2, r4; \
pxor r0, r3; por r1, r0; \
pxor r3, r2; pxor r4, r0; \
por r2, r0; pxor r3, r1; \
pxor r1, r0; por r3, r1; \
pxor r0, r1; pxor RNOT, r4; \
pxor r1, r4; por r0, r1; \
pxor r0, r1; \
por r4, r1; \
pxor r1, r3;
#define SBOX2(r0, r1, r2, r3, r4) \
movdqa r0, r4; pand r2, r0; \
pxor r3, r0; pxor r1, r2; \
pxor r0, r2; por r4, r3; \
pxor r1, r3; pxor r2, r4; \
movdqa r3, r1; por r4, r3; \
pxor r0, r3; pand r1, r0; \
pxor r0, r4; pxor r3, r1; \
pxor r4, r1; pxor RNOT, r4;
#define SBOX2_INVERSE(r0, r1, r2, r3, r4) \
pxor r3, r2; pxor r0, r3; \
movdqa r3, r4; pand r2, r3; \
pxor r1, r3; por r2, r1; \
pxor r4, r1; pand r3, r4; \
pxor r3, r2; pand r0, r4; \
pxor r2, r4; pand r1, r2; \
por r0, r2; pxor RNOT, r3; \
pxor r3, r2; pxor r3, r0; \
pand r1, r0; pxor r4, r3; \
pxor r0, r3;
#define SBOX3(r0, r1, r2, r3, r4) \
movdqa r0, r4; por r3, r0; \
pxor r1, r3; pand r4, r1; \
pxor r2, r4; pxor r3, r2; \
pand r0, r3; por r1, r4; \
pxor r4, r3; pxor r1, r0; \
pand r0, r4; pxor r3, r1; \
pxor r2, r4; por r0, r1; \
pxor r2, r1; pxor r3, r0; \
movdqa r1, r2; por r3, r1; \
pxor r0, r1;
#define SBOX3_INVERSE(r0, r1, r2, r3, r4) \
movdqa r2, r4; pxor r1, r2; \
pxor r2, r0; pand r2, r4; \
pxor r0, r4; pand r1, r0; \
pxor r3, r1; por r4, r3; \
pxor r3, r2; pxor r3, r0; \
pxor r4, r1; pand r2, r3; \
pxor r1, r3; pxor r0, r1; \
por r2, r1; pxor r3, r0; \
pxor r4, r1; \
pxor r1, r0;
#define SBOX4(r0, r1, r2, r3, r4) \
pxor r3, r1; pxor RNOT, r3; \
pxor r3, r2; pxor r0, r3; \
movdqa r1, r4; pand r3, r1; \
pxor r2, r1; pxor r3, r4; \
pxor r4, r0; pand r4, r2; \
pxor r0, r2; pand r1, r0; \
pxor r0, r3; por r1, r4; \
pxor r0, r4; por r3, r0; \
pxor r2, r0; pand r3, r2; \
pxor RNOT, r0; pxor r2, r4;
#define SBOX4_INVERSE(r0, r1, r2, r3, r4) \
movdqa r2, r4; pand r3, r2; \
pxor r1, r2; por r3, r1; \
pand r0, r1; pxor r2, r4; \
pxor r1, r4; pand r2, r1; \
pxor RNOT, r0; pxor r4, r3; \
pxor r3, r1; pand r0, r3; \
pxor r2, r3; pxor r1, r0; \
pand r0, r2; pxor r0, r3; \
pxor r4, r2; \
por r3, r2; pxor r0, r3; \
pxor r1, r2;
#define SBOX5(r0, r1, r2, r3, r4) \
pxor r1, r0; pxor r3, r1; \
pxor RNOT, r3; movdqa r1, r4; \
pand r0, r1; pxor r3, r2; \
pxor r2, r1; por r4, r2; \
pxor r3, r4; pand r1, r3; \
pxor r0, r3; pxor r1, r4; \
pxor r2, r4; pxor r0, r2; \
pand r3, r0; pxor RNOT, r2; \
pxor r4, r0; por r3, r4; \
pxor r4, r2;
#define SBOX5_INVERSE(r0, r1, r2, r3, r4) \
pxor RNOT, r1; movdqa r3, r4; \
pxor r1, r2; por r0, r3; \
pxor r2, r3; por r1, r2; \
pand r0, r2; pxor r3, r4; \
pxor r4, r2; por r0, r4; \
pxor r1, r4; pand r2, r1; \
pxor r3, r1; pxor r2, r4; \
pand r4, r3; pxor r1, r4; \
pxor r4, r3; pxor RNOT, r4; \
pxor r0, r3;
#define SBOX6(r0, r1, r2, r3, r4) \
pxor RNOT, r2; movdqa r3, r4; \
pand r0, r3; pxor r4, r0; \
pxor r2, r3; por r4, r2; \
pxor r3, r1; pxor r0, r2; \
por r1, r0; pxor r1, r2; \
pxor r0, r4; por r3, r0; \
pxor r2, r0; pxor r3, r4; \
pxor r0, r4; pxor RNOT, r3; \
pand r4, r2; \
pxor r3, r2;
#define SBOX6_INVERSE(r0, r1, r2, r3, r4) \
pxor r2, r0; movdqa r2, r4; \
pand r0, r2; pxor r3, r4; \
pxor RNOT, r2; pxor r1, r3; \
pxor r3, r2; por r0, r4; \
pxor r2, r0; pxor r4, r3; \
pxor r1, r4; pand r3, r1; \
pxor r0, r1; pxor r3, r0; \
por r2, r0; pxor r1, r3; \
pxor r0, r4;
#define SBOX7(r0, r1, r2, r3, r4) \
movdqa r1, r4; por r2, r1; \
pxor r3, r1; pxor r2, r4; \
pxor r1, r2; por r4, r3; \
pand r0, r3; pxor r2, r4; \
pxor r1, r3; por r4, r1; \
pxor r0, r1; por r4, r0; \
pxor r2, r0; pxor r4, r1; \
pxor r1, r2; pand r0, r1; \
pxor r4, r1; pxor RNOT, r2; \
por r0, r2; \
pxor r2, r4;
#define SBOX7_INVERSE(r0, r1, r2, r3, r4) \
movdqa r2, r4; pxor r0, r2; \
pand r3, r0; por r3, r4; \
pxor RNOT, r2; pxor r1, r3; \
por r0, r1; pxor r2, r0; \
pand r4, r2; pand r4, r3; \
pxor r2, r1; pxor r0, r2; \
por r2, r0; pxor r1, r4; \
pxor r3, r0; pxor r4, r3; \
por r0, r4; pxor r2, r3; \
pxor r2, r4;
/* Apply SBOX number WHICH to to the block. */
#define SBOX(which, r0, r1, r2, r3, r4) \
SBOX##which (r0, r1, r2, r3, r4)
/* Apply inverse SBOX number WHICH to to the block. */
#define SBOX_INVERSE(which, r0, r1, r2, r3, r4) \
SBOX##which##_INVERSE (r0, r1, r2, r3, r4)
/* XOR round key into block state in r0,r1,r2,r3. r4 used as temporary. */
#define BLOCK_XOR_KEY(r0, r1, r2, r3, r4, round) \
pbroadcastd ((ctx_keys + (round) * 16 + 0 * 4)(CTX), r4); \
pxor r4, r0; \
pbroadcastd ((ctx_keys + (round) * 16 + 1 * 4)(CTX), r4); \
pxor r4, r1; \
pbroadcastd ((ctx_keys + (round) * 16 + 2 * 4)(CTX), r4); \
pxor r4, r2; \
pbroadcastd ((ctx_keys + (round) * 16 + 3 * 4)(CTX), r4); \
pxor r4, r3;
/* Apply the linear transformation to BLOCK. */
#define LINEAR_TRANSFORMATION(r0, r1, r2, r3, r4) \
vec_rol(r0, 13, r4); \
vec_rol(r2, 3, r4); \
pxor r0, r1; \
pxor r2, r1; \
movdqa r0, r4; \
pslld $3, r4; \
pxor r2, r3; \
pxor r4, r3; \
vec_rol(r1, 1, r4); \
vec_rol(r3, 7, r4); \
pxor r1, r0; \
pxor r3, r0; \
movdqa r1, r4; \
pslld $7, r4; \
pxor r3, r2; \
pxor r4, r2; \
vec_rol(r0, 5, r4); \
vec_rol(r2, 22, r4);
/* Apply the inverse linear transformation to BLOCK. */
#define LINEAR_TRANSFORMATION_INVERSE(r0, r1, r2, r3, r4) \
vec_ror(r2, 22, r4); \
vec_ror(r0, 5, r4); \
movdqa r1, r4; \
pslld $7, r4; \
pxor r3, r2; \
pxor r4, r2; \
pxor r1, r0; \
pxor r3, r0; \
vec_ror(r3, 7, r4); \
vec_ror(r1, 1, r4); \
movdqa r0, r4; \
pslld $3, r4; \
pxor r2, r3; \
pxor r4, r3; \
pxor r0, r1; \
pxor r2, r1; \
vec_ror(r2, 3, r4); \
vec_ror(r0, 13, r4);
/* Apply a Serpent round to eight parallel blocks. This macro increments
`round'. */
#define ROUND(round, which, a0, a1, a2, a3, a4, na0, na1, na2, na3, na4, \
b0, b1, b2, b3, b4, nb0, nb1, nb2, nb3, nb4) \
BLOCK_XOR_KEY (a0, a1, a2, a3, a4, round); \
SBOX (which, a0, a1, a2, a3, a4); \
BLOCK_XOR_KEY (b0, b1, b2, b3, b4, round); \
SBOX (which, b0, b1, b2, b3, b4); \
LINEAR_TRANSFORMATION (na0, na1, na2, na3, na4); \
LINEAR_TRANSFORMATION (nb0, nb1, nb2, nb3, nb4);
/* Apply the last Serpent round to eight parallel blocks. This macro increments
`round'. */
#define ROUND_LAST(round, which, a0, a1, a2, a3, a4, na0, na1, na2, na3, na4, \
b0, b1, b2, b3, b4, nb0, nb1, nb2, nb3, nb4) \
BLOCK_XOR_KEY (a0, a1, a2, a3, a4, round); \
SBOX (which, a0, a1, a2, a3, a4); \
BLOCK_XOR_KEY (b0, b1, b2, b3, b4, round); \
SBOX (which, b0, b1, b2, b3, b4); \
BLOCK_XOR_KEY (na0, na1, na2, na3, na4, ((round) + 1)); \
BLOCK_XOR_KEY (nb0, nb1, nb2, nb3, nb4, ((round) + 1));
/* Apply an inverse Serpent round to eight parallel blocks. This macro
increments `round'. */
#define ROUND_INVERSE(round, which, a0, a1, a2, a3, a4, \
na0, na1, na2, na3, na4, \
b0, b1, b2, b3, b4, \
nb0, nb1, nb2, nb3, nb4) \
LINEAR_TRANSFORMATION_INVERSE (a0, a1, a2, a3, a4); \
LINEAR_TRANSFORMATION_INVERSE (b0, b1, b2, b3, b4); \
SBOX_INVERSE (which, a0, a1, a2, a3, a4); \
BLOCK_XOR_KEY (na0, na1, na2, na3, na4, round); \
SBOX_INVERSE (which, b0, b1, b2, b3, b4); \
BLOCK_XOR_KEY (nb0, nb1, nb2, nb3, nb4, round);
/* Apply the first inverse Serpent round to eight parallel blocks. This macro
increments `round'. */
#define ROUND_FIRST_INVERSE(round, which, a0, a1, a2, a3, a4, \
na0, na1, na2, na3, na4, \
b0, b1, b2, b3, b4, \
nb0, nb1, nb2, nb3, nb4) \
BLOCK_XOR_KEY (a0, a1, a2, a3, a4, ((round) + 1)); \
BLOCK_XOR_KEY (b0, b1, b2, b3, b4, ((round) + 1)); \
SBOX_INVERSE (which, a0, a1, a2, a3, a4); \
BLOCK_XOR_KEY (na0, na1, na2, na3, na4, round); \
SBOX_INVERSE (which, b0, b1, b2, b3, b4); \
BLOCK_XOR_KEY (nb0, nb1, nb2, nb3, nb4, round);
.text
.align 8
ELF(.type __serpent_enc_blk8,@function;)
__serpent_enc_blk8:
/* input:
* %rdi: ctx, CTX
* RA0, RA1, RA2, RA3, RB0, RB1, RB2, RB3: eight parallel plaintext
* blocks
* output:
* RA4, RA1, RA2, RA0, RB4, RB1, RB2, RB0: eight parallel
* ciphertext blocks
*/
CFI_STARTPROC();
pcmpeqd RNOT, RNOT;
transpose_4x4(RA0, RA1, RA2, RA3, RA4, RTMP0, RTMP1);
transpose_4x4(RB0, RB1, RB2, RB3, RB4, RTMP0, RTMP1);
ROUND (0, 0, RA0, RA1, RA2, RA3, RA4, RA1, RA4, RA2, RA0, RA3,
RB0, RB1, RB2, RB3, RB4, RB1, RB4, RB2, RB0, RB3);
ROUND (1, 1, RA1, RA4, RA2, RA0, RA3, RA2, RA1, RA0, RA4, RA3,
RB1, RB4, RB2, RB0, RB3, RB2, RB1, RB0, RB4, RB3);
ROUND (2, 2, RA2, RA1, RA0, RA4, RA3, RA0, RA4, RA1, RA3, RA2,
RB2, RB1, RB0, RB4, RB3, RB0, RB4, RB1, RB3, RB2);
ROUND (3, 3, RA0, RA4, RA1, RA3, RA2, RA4, RA1, RA3, RA2, RA0,
RB0, RB4, RB1, RB3, RB2, RB4, RB1, RB3, RB2, RB0);
ROUND (4, 4, RA4, RA1, RA3, RA2, RA0, RA1, RA0, RA4, RA2, RA3,
RB4, RB1, RB3, RB2, RB0, RB1, RB0, RB4, RB2, RB3);
ROUND (5, 5, RA1, RA0, RA4, RA2, RA3, RA0, RA2, RA1, RA4, RA3,
RB1, RB0, RB4, RB2, RB3, RB0, RB2, RB1, RB4, RB3);
ROUND (6, 6, RA0, RA2, RA1, RA4, RA3, RA0, RA2, RA3, RA1, RA4,
RB0, RB2, RB1, RB4, RB3, RB0, RB2, RB3, RB1, RB4);
ROUND (7, 7, RA0, RA2, RA3, RA1, RA4, RA4, RA1, RA2, RA0, RA3,
RB0, RB2, RB3, RB1, RB4, RB4, RB1, RB2, RB0, RB3);
ROUND (8, 0, RA4, RA1, RA2, RA0, RA3, RA1, RA3, RA2, RA4, RA0,
RB4, RB1, RB2, RB0, RB3, RB1, RB3, RB2, RB4, RB0);
ROUND (9, 1, RA1, RA3, RA2, RA4, RA0, RA2, RA1, RA4, RA3, RA0,
RB1, RB3, RB2, RB4, RB0, RB2, RB1, RB4, RB3, RB0);
ROUND (10, 2, RA2, RA1, RA4, RA3, RA0, RA4, RA3, RA1, RA0, RA2,
RB2, RB1, RB4, RB3, RB0, RB4, RB3, RB1, RB0, RB2);
ROUND (11, 3, RA4, RA3, RA1, RA0, RA2, RA3, RA1, RA0, RA2, RA4,
RB4, RB3, RB1, RB0, RB2, RB3, RB1, RB0, RB2, RB4);
ROUND (12, 4, RA3, RA1, RA0, RA2, RA4, RA1, RA4, RA3, RA2, RA0,
RB3, RB1, RB0, RB2, RB4, RB1, RB4, RB3, RB2, RB0);
ROUND (13, 5, RA1, RA4, RA3, RA2, RA0, RA4, RA2, RA1, RA3, RA0,
RB1, RB4, RB3, RB2, RB0, RB4, RB2, RB1, RB3, RB0);
ROUND (14, 6, RA4, RA2, RA1, RA3, RA0, RA4, RA2, RA0, RA1, RA3,
RB4, RB2, RB1, RB3, RB0, RB4, RB2, RB0, RB1, RB3);
ROUND (15, 7, RA4, RA2, RA0, RA1, RA3, RA3, RA1, RA2, RA4, RA0,
RB4, RB2, RB0, RB1, RB3, RB3, RB1, RB2, RB4, RB0);
ROUND (16, 0, RA3, RA1, RA2, RA4, RA0, RA1, RA0, RA2, RA3, RA4,
RB3, RB1, RB2, RB4, RB0, RB1, RB0, RB2, RB3, RB4);
ROUND (17, 1, RA1, RA0, RA2, RA3, RA4, RA2, RA1, RA3, RA0, RA4,
RB1, RB0, RB2, RB3, RB4, RB2, RB1, RB3, RB0, RB4);
ROUND (18, 2, RA2, RA1, RA3, RA0, RA4, RA3, RA0, RA1, RA4, RA2,
RB2, RB1, RB3, RB0, RB4, RB3, RB0, RB1, RB4, RB2);
ROUND (19, 3, RA3, RA0, RA1, RA4, RA2, RA0, RA1, RA4, RA2, RA3,
RB3, RB0, RB1, RB4, RB2, RB0, RB1, RB4, RB2, RB3);
ROUND (20, 4, RA0, RA1, RA4, RA2, RA3, RA1, RA3, RA0, RA2, RA4,
RB0, RB1, RB4, RB2, RB3, RB1, RB3, RB0, RB2, RB4);
ROUND (21, 5, RA1, RA3, RA0, RA2, RA4, RA3, RA2, RA1, RA0, RA4,
RB1, RB3, RB0, RB2, RB4, RB3, RB2, RB1, RB0, RB4);
ROUND (22, 6, RA3, RA2, RA1, RA0, RA4, RA3, RA2, RA4, RA1, RA0,
RB3, RB2, RB1, RB0, RB4, RB3, RB2, RB4, RB1, RB0);
ROUND (23, 7, RA3, RA2, RA4, RA1, RA0, RA0, RA1, RA2, RA3, RA4,
RB3, RB2, RB4, RB1, RB0, RB0, RB1, RB2, RB3, RB4);
ROUND (24, 0, RA0, RA1, RA2, RA3, RA4, RA1, RA4, RA2, RA0, RA3,
RB0, RB1, RB2, RB3, RB4, RB1, RB4, RB2, RB0, RB3);
ROUND (25, 1, RA1, RA4, RA2, RA0, RA3, RA2, RA1, RA0, RA4, RA3,
RB1, RB4, RB2, RB0, RB3, RB2, RB1, RB0, RB4, RB3);
ROUND (26, 2, RA2, RA1, RA0, RA4, RA3, RA0, RA4, RA1, RA3, RA2,
RB2, RB1, RB0, RB4, RB3, RB0, RB4, RB1, RB3, RB2);
ROUND (27, 3, RA0, RA4, RA1, RA3, RA2, RA4, RA1, RA3, RA2, RA0,
RB0, RB4, RB1, RB3, RB2, RB4, RB1, RB3, RB2, RB0);
ROUND (28, 4, RA4, RA1, RA3, RA2, RA0, RA1, RA0, RA4, RA2, RA3,
RB4, RB1, RB3, RB2, RB0, RB1, RB0, RB4, RB2, RB3);
ROUND (29, 5, RA1, RA0, RA4, RA2, RA3, RA0, RA2, RA1, RA4, RA3,
RB1, RB0, RB4, RB2, RB3, RB0, RB2, RB1, RB4, RB3);
ROUND (30, 6, RA0, RA2, RA1, RA4, RA3, RA0, RA2, RA3, RA1, RA4,
RB0, RB2, RB1, RB4, RB3, RB0, RB2, RB3, RB1, RB4);
ROUND_LAST (31, 7, RA0, RA2, RA3, RA1, RA4, RA4, RA1, RA2, RA0, RA3,
RB0, RB2, RB3, RB1, RB4, RB4, RB1, RB2, RB0, RB3);
transpose_4x4(RA4, RA1, RA2, RA0, RA3, RTMP0, RTMP1);
transpose_4x4(RB4, RB1, RB2, RB0, RB3, RTMP0, RTMP1);
ret;
CFI_ENDPROC();
ELF(.size __serpent_enc_blk8,.-__serpent_enc_blk8;)
.align 8
ELF(.type __serpent_dec_blk8,@function;)
__serpent_dec_blk8:
/* input:
* %rdi: ctx, CTX
* RA0, RA1, RA2, RA3, RB0, RB1, RB2, RB3: eight parallel
* ciphertext blocks
* output:
* RA0, RA1, RA2, RA3, RB0, RB1, RB2, RB3: eight parallel plaintext
* blocks
*/
CFI_STARTPROC();
pcmpeqd RNOT, RNOT;
transpose_4x4(RA0, RA1, RA2, RA3, RA4, RTMP0, RTMP1);
transpose_4x4(RB0, RB1, RB2, RB3, RB4, RTMP0, RTMP1);
ROUND_FIRST_INVERSE (31, 7, RA0, RA1, RA2, RA3, RA4,
RA3, RA0, RA1, RA4, RA2,
RB0, RB1, RB2, RB3, RB4,
RB3, RB0, RB1, RB4, RB2);
ROUND_INVERSE (30, 6, RA3, RA0, RA1, RA4, RA2, RA0, RA1, RA2, RA4, RA3,
RB3, RB0, RB1, RB4, RB2, RB0, RB1, RB2, RB4, RB3);
ROUND_INVERSE (29, 5, RA0, RA1, RA2, RA4, RA3, RA1, RA3, RA4, RA2, RA0,
RB0, RB1, RB2, RB4, RB3, RB1, RB3, RB4, RB2, RB0);
ROUND_INVERSE (28, 4, RA1, RA3, RA4, RA2, RA0, RA1, RA2, RA4, RA0, RA3,
RB1, RB3, RB4, RB2, RB0, RB1, RB2, RB4, RB0, RB3);
ROUND_INVERSE (27, 3, RA1, RA2, RA4, RA0, RA3, RA4, RA2, RA0, RA1, RA3,
RB1, RB2, RB4, RB0, RB3, RB4, RB2, RB0, RB1, RB3);
ROUND_INVERSE (26, 2, RA4, RA2, RA0, RA1, RA3, RA2, RA3, RA0, RA1, RA4,
RB4, RB2, RB0, RB1, RB3, RB2, RB3, RB0, RB1, RB4);
ROUND_INVERSE (25, 1, RA2, RA3, RA0, RA1, RA4, RA4, RA2, RA1, RA0, RA3,
RB2, RB3, RB0, RB1, RB4, RB4, RB2, RB1, RB0, RB3);
ROUND_INVERSE (24, 0, RA4, RA2, RA1, RA0, RA3, RA4, RA3, RA2, RA0, RA1,
RB4, RB2, RB1, RB0, RB3, RB4, RB3, RB2, RB0, RB1);
ROUND_INVERSE (23, 7, RA4, RA3, RA2, RA0, RA1, RA0, RA4, RA3, RA1, RA2,
RB4, RB3, RB2, RB0, RB1, RB0, RB4, RB3, RB1, RB2);
ROUND_INVERSE (22, 6, RA0, RA4, RA3, RA1, RA2, RA4, RA3, RA2, RA1, RA0,
RB0, RB4, RB3, RB1, RB2, RB4, RB3, RB2, RB1, RB0);
ROUND_INVERSE (21, 5, RA4, RA3, RA2, RA1, RA0, RA3, RA0, RA1, RA2, RA4,
RB4, RB3, RB2, RB1, RB0, RB3, RB0, RB1, RB2, RB4);
ROUND_INVERSE (20, 4, RA3, RA0, RA1, RA2, RA4, RA3, RA2, RA1, RA4, RA0,
RB3, RB0, RB1, RB2, RB4, RB3, RB2, RB1, RB4, RB0);
ROUND_INVERSE (19, 3, RA3, RA2, RA1, RA4, RA0, RA1, RA2, RA4, RA3, RA0,
RB3, RB2, RB1, RB4, RB0, RB1, RB2, RB4, RB3, RB0);
ROUND_INVERSE (18, 2, RA1, RA2, RA4, RA3, RA0, RA2, RA0, RA4, RA3, RA1,
RB1, RB2, RB4, RB3, RB0, RB2, RB0, RB4, RB3, RB1);
ROUND_INVERSE (17, 1, RA2, RA0, RA4, RA3, RA1, RA1, RA2, RA3, RA4, RA0,
RB2, RB0, RB4, RB3, RB1, RB1, RB2, RB3, RB4, RB0);
ROUND_INVERSE (16, 0, RA1, RA2, RA3, RA4, RA0, RA1, RA0, RA2, RA4, RA3,
RB1, RB2, RB3, RB4, RB0, RB1, RB0, RB2, RB4, RB3);
ROUND_INVERSE (15, 7, RA1, RA0, RA2, RA4, RA3, RA4, RA1, RA0, RA3, RA2,
RB1, RB0, RB2, RB4, RB3, RB4, RB1, RB0, RB3, RB2);
ROUND_INVERSE (14, 6, RA4, RA1, RA0, RA3, RA2, RA1, RA0, RA2, RA3, RA4,
RB4, RB1, RB0, RB3, RB2, RB1, RB0, RB2, RB3, RB4);
ROUND_INVERSE (13, 5, RA1, RA0, RA2, RA3, RA4, RA0, RA4, RA3, RA2, RA1,
RB1, RB0, RB2, RB3, RB4, RB0, RB4, RB3, RB2, RB1);
ROUND_INVERSE (12, 4, RA0, RA4, RA3, RA2, RA1, RA0, RA2, RA3, RA1, RA4,
RB0, RB4, RB3, RB2, RB1, RB0, RB2, RB3, RB1, RB4);
ROUND_INVERSE (11, 3, RA0, RA2, RA3, RA1, RA4, RA3, RA2, RA1, RA0, RA4,
RB0, RB2, RB3, RB1, RB4, RB3, RB2, RB1, RB0, RB4);
ROUND_INVERSE (10, 2, RA3, RA2, RA1, RA0, RA4, RA2, RA4, RA1, RA0, RA3,
RB3, RB2, RB1, RB0, RB4, RB2, RB4, RB1, RB0, RB3);
ROUND_INVERSE (9, 1, RA2, RA4, RA1, RA0, RA3, RA3, RA2, RA0, RA1, RA4,
RB2, RB4, RB1, RB0, RB3, RB3, RB2, RB0, RB1, RB4);
ROUND_INVERSE (8, 0, RA3, RA2, RA0, RA1, RA4, RA3, RA4, RA2, RA1, RA0,
RB3, RB2, RB0, RB1, RB4, RB3, RB4, RB2, RB1, RB0);
ROUND_INVERSE (7, 7, RA3, RA4, RA2, RA1, RA0, RA1, RA3, RA4, RA0, RA2,
RB3, RB4, RB2, RB1, RB0, RB1, RB3, RB4, RB0, RB2);
ROUND_INVERSE (6, 6, RA1, RA3, RA4, RA0, RA2, RA3, RA4, RA2, RA0, RA1,
RB1, RB3, RB4, RB0, RB2, RB3, RB4, RB2, RB0, RB1);
ROUND_INVERSE (5, 5, RA3, RA4, RA2, RA0, RA1, RA4, RA1, RA0, RA2, RA3,
RB3, RB4, RB2, RB0, RB1, RB4, RB1, RB0, RB2, RB3);
ROUND_INVERSE (4, 4, RA4, RA1, RA0, RA2, RA3, RA4, RA2, RA0, RA3, RA1,
RB4, RB1, RB0, RB2, RB3, RB4, RB2, RB0, RB3, RB1);
ROUND_INVERSE (3, 3, RA4, RA2, RA0, RA3, RA1, RA0, RA2, RA3, RA4, RA1,
RB4, RB2, RB0, RB3, RB1, RB0, RB2, RB3, RB4, RB1);
ROUND_INVERSE (2, 2, RA0, RA2, RA3, RA4, RA1, RA2, RA1, RA3, RA4, RA0,
RB0, RB2, RB3, RB4, RB1, RB2, RB1, RB3, RB4, RB0);
ROUND_INVERSE (1, 1, RA2, RA1, RA3, RA4, RA0, RA0, RA2, RA4, RA3, RA1,
RB2, RB1, RB3, RB4, RB0, RB0, RB2, RB4, RB3, RB1);
ROUND_INVERSE (0, 0, RA0, RA2, RA4, RA3, RA1, RA0, RA1, RA2, RA3, RA4,
RB0, RB2, RB4, RB3, RB1, RB0, RB1, RB2, RB3, RB4);
transpose_4x4(RA0, RA1, RA2, RA3, RA4, RTMP0, RTMP1);
transpose_4x4(RB0, RB1, RB2, RB3, RB4, RTMP0, RTMP1);
ret;
CFI_ENDPROC();
ELF(.size __serpent_dec_blk8,.-__serpent_dec_blk8;)
.align 8
.globl _gcry_serpent_sse2_ctr_enc
ELF(.type _gcry_serpent_sse2_ctr_enc,@function;)
_gcry_serpent_sse2_ctr_enc:
/* input:
* %rdi: ctx, CTX
* %rsi: dst (8 blocks)
* %rdx: src (8 blocks)
* %rcx: iv (big endian, 128bit)
*/
CFI_STARTPROC();
/* load IV and byteswap */
movdqu (%rcx), RA0;
movdqa RA0, RTMP0;
pbswap(RTMP0, RTMP1); /* be => le */
pcmpeqd RNOT, RNOT;
psrldq $8, RNOT; /* low: -1, high: 0 */
movdqa RNOT, RTMP2;
paddq RTMP2, RTMP2; /* low: -2, high: 0 */
/* construct IVs */
movdqa RTMP0, RTMP1;
psubq RNOT, RTMP0; /* +1 */
movdqa RTMP0, RA1;
psubq RTMP2, RTMP1; /* +2 */
movdqa RTMP1, RA2;
psubq RTMP2, RTMP0; /* +3 */
movdqa RTMP0, RA3;
psubq RTMP2, RTMP1; /* +4 */
movdqa RTMP1, RB0;
psubq RTMP2, RTMP0; /* +5 */
movdqa RTMP0, RB1;
psubq RTMP2, RTMP1; /* +6 */
movdqa RTMP1, RB2;
psubq RTMP2, RTMP0; /* +7 */
movdqa RTMP0, RB3;
psubq RTMP2, RTMP1; /* +8 */
/* check need for handling 64-bit overflow and carry */
cmpl $0xffffffff, 8(%rcx);
jne .Lno_ctr_carry;
movl 12(%rcx), %eax;
bswapl %eax;
cmpl $-8, %eax;
jb .Lno_ctr_carry;
pslldq $8, RNOT; /* low: 0, high: -1 */
je .Lcarry_RTMP0;
cmpl $-6, %eax;
jb .Lcarry_RB3;
je .Lcarry_RB2;
cmpl $-4, %eax;
jb .Lcarry_RB1;
je .Lcarry_RB0;
cmpl $-2, %eax;
jb .Lcarry_RA3;
je .Lcarry_RA2;
psubq RNOT, RA1;
.Lcarry_RA2:
psubq RNOT, RA2;
.Lcarry_RA3:
psubq RNOT, RA3;
.Lcarry_RB0:
psubq RNOT, RB0;
.Lcarry_RB1:
psubq RNOT, RB1;
.Lcarry_RB2:
psubq RNOT, RB2;
.Lcarry_RB3:
psubq RNOT, RB3;
.Lcarry_RTMP0:
psubq RNOT, RTMP1;
.Lno_ctr_carry:
/* le => be */
pbswap(RA1, RTMP0);
pbswap(RA2, RTMP0);
pbswap(RA3, RTMP0);
pbswap(RB0, RTMP0);
pbswap(RB1, RTMP0);
pbswap(RB2, RTMP0);
pbswap(RB3, RTMP0);
pbswap(RTMP1, RTMP0);
/* store new IV */
movdqu RTMP1, (%rcx);
call __serpent_enc_blk8;
pxor_u((0 * 16)(%rdx), RA4, RTMP0);
pxor_u((1 * 16)(%rdx), RA1, RTMP0);
pxor_u((2 * 16)(%rdx), RA2, RTMP0);
pxor_u((3 * 16)(%rdx), RA0, RTMP0);
pxor_u((4 * 16)(%rdx), RB4, RTMP0);
pxor_u((5 * 16)(%rdx), RB1, RTMP0);
pxor_u((6 * 16)(%rdx), RB2, RTMP0);
pxor_u((7 * 16)(%rdx), RB0, RTMP0);
movdqu RA4, (0 * 16)(%rsi);
movdqu RA1, (1 * 16)(%rsi);
movdqu RA2, (2 * 16)(%rsi);
movdqu RA0, (3 * 16)(%rsi);
movdqu RB4, (4 * 16)(%rsi);
movdqu RB1, (5 * 16)(%rsi);
movdqu RB2, (6 * 16)(%rsi);
movdqu RB0, (7 * 16)(%rsi);
/* clear the used registers */
pxor RA0, RA0;
pxor RA1, RA1;
pxor RA2, RA2;
pxor RA3, RA3;
pxor RA4, RA4;
pxor RB0, RB0;
pxor RB1, RB1;
pxor RB2, RB2;
pxor RB3, RB3;
pxor RB4, RB4;
pxor RTMP0, RTMP0;
pxor RTMP1, RTMP1;
pxor RTMP2, RTMP2;
pxor RNOT, RNOT;
ret;
CFI_ENDPROC();
ELF(.size _gcry_serpent_sse2_ctr_enc,.-_gcry_serpent_sse2_ctr_enc;)
.align 8
.globl _gcry_serpent_sse2_cbc_dec
ELF(.type _gcry_serpent_sse2_cbc_dec,@function;)
_gcry_serpent_sse2_cbc_dec:
/* input:
* %rdi: ctx, CTX
* %rsi: dst (8 blocks)
* %rdx: src (8 blocks)
* %rcx: iv
*/
CFI_STARTPROC();
movdqu (0 * 16)(%rdx), RA0;
movdqu (1 * 16)(%rdx), RA1;
movdqu (2 * 16)(%rdx), RA2;
movdqu (3 * 16)(%rdx), RA3;
movdqu (4 * 16)(%rdx), RB0;
movdqu (5 * 16)(%rdx), RB1;
movdqu (6 * 16)(%rdx), RB2;
movdqu (7 * 16)(%rdx), RB3;
call __serpent_dec_blk8;
movdqu (7 * 16)(%rdx), RNOT;
pxor_u((%rcx), RA0, RTMP0);
pxor_u((0 * 16)(%rdx), RA1, RTMP0);
pxor_u((1 * 16)(%rdx), RA2, RTMP0);
pxor_u((2 * 16)(%rdx), RA3, RTMP0);
pxor_u((3 * 16)(%rdx), RB0, RTMP0);
pxor_u((4 * 16)(%rdx), RB1, RTMP0);
pxor_u((5 * 16)(%rdx), RB2, RTMP0);
pxor_u((6 * 16)(%rdx), RB3, RTMP0);
movdqu RNOT, (%rcx); /* store new IV */
movdqu RA0, (0 * 16)(%rsi);
movdqu RA1, (1 * 16)(%rsi);
movdqu RA2, (2 * 16)(%rsi);
movdqu RA3, (3 * 16)(%rsi);
movdqu RB0, (4 * 16)(%rsi);
movdqu RB1, (5 * 16)(%rsi);
movdqu RB2, (6 * 16)(%rsi);
movdqu RB3, (7 * 16)(%rsi);
/* clear the used registers */
pxor RA0, RA0;
pxor RA1, RA1;
pxor RA2, RA2;
pxor RA3, RA3;
pxor RA4, RA4;
pxor RB0, RB0;
pxor RB1, RB1;
pxor RB2, RB2;
pxor RB3, RB3;
pxor RB4, RB4;
pxor RTMP0, RTMP0;
pxor RTMP1, RTMP1;
pxor RTMP2, RTMP2;
pxor RNOT, RNOT;
ret;
CFI_ENDPROC();
ELF(.size _gcry_serpent_sse2_cbc_dec,.-_gcry_serpent_sse2_cbc_dec;)
.align 8
.globl _gcry_serpent_sse2_cfb_dec
ELF(.type _gcry_serpent_sse2_cfb_dec,@function;)
_gcry_serpent_sse2_cfb_dec:
/* input:
* %rdi: ctx, CTX
* %rsi: dst (8 blocks)
* %rdx: src (8 blocks)
* %rcx: iv
*/
CFI_STARTPROC();
/* Load input */
movdqu (%rcx), RA0;
movdqu 0 * 16(%rdx), RA1;
movdqu 1 * 16(%rdx), RA2;
movdqu 2 * 16(%rdx), RA3;
movdqu 3 * 16(%rdx), RB0;
movdqu 4 * 16(%rdx), RB1;
movdqu 5 * 16(%rdx), RB2;
movdqu 6 * 16(%rdx), RB3;
/* Update IV */
movdqu 7 * 16(%rdx), RNOT;
movdqu RNOT, (%rcx);
call __serpent_enc_blk8;
pxor_u((0 * 16)(%rdx), RA4, RTMP0);
pxor_u((1 * 16)(%rdx), RA1, RTMP0);
pxor_u((2 * 16)(%rdx), RA2, RTMP0);
pxor_u((3 * 16)(%rdx), RA0, RTMP0);
pxor_u((4 * 16)(%rdx), RB4, RTMP0);
pxor_u((5 * 16)(%rdx), RB1, RTMP0);
pxor_u((6 * 16)(%rdx), RB2, RTMP0);
pxor_u((7 * 16)(%rdx), RB0, RTMP0);
movdqu RA4, (0 * 16)(%rsi);
movdqu RA1, (1 * 16)(%rsi);
movdqu RA2, (2 * 16)(%rsi);
movdqu RA0, (3 * 16)(%rsi);
movdqu RB4, (4 * 16)(%rsi);
movdqu RB1, (5 * 16)(%rsi);
movdqu RB2, (6 * 16)(%rsi);
movdqu RB0, (7 * 16)(%rsi);
/* clear the used registers */
pxor RA0, RA0;
pxor RA1, RA1;
pxor RA2, RA2;
pxor RA3, RA3;
pxor RA4, RA4;
pxor RB0, RB0;
pxor RB1, RB1;
pxor RB2, RB2;
pxor RB3, RB3;
pxor RB4, RB4;
pxor RTMP0, RTMP0;
pxor RTMP1, RTMP1;
pxor RTMP2, RTMP2;
pxor RNOT, RNOT;
ret;
CFI_ENDPROC();
ELF(.size _gcry_serpent_sse2_cfb_dec,.-_gcry_serpent_sse2_cfb_dec;)
.align 8
.globl _gcry_serpent_sse2_ocb_enc
ELF(.type _gcry_serpent_sse2_ocb_enc,@function;)
_gcry_serpent_sse2_ocb_enc:
/* input:
* %rdi: ctx, CTX
* %rsi: dst (8 blocks)
* %rdx: src (8 blocks)
* %rcx: offset
* %r8 : checksum
* %r9 : L pointers (void *L[8])
*/
CFI_STARTPROC();
subq $(4 * 8), %rsp;
CFI_ADJUST_CFA_OFFSET(4 * 8);
movq %r10, (0 * 8)(%rsp);
movq %r11, (1 * 8)(%rsp);
movq %r12, (2 * 8)(%rsp);
movq %r13, (3 * 8)(%rsp);
CFI_REL_OFFSET(%r10, 0 * 8);
CFI_REL_OFFSET(%r11, 1 * 8);
CFI_REL_OFFSET(%r12, 2 * 8);
CFI_REL_OFFSET(%r13, 3 * 8);
movdqu (%rcx), RTMP0;
movdqu (%r8), RTMP1;
/* Offset_i = Offset_{i-1} xor L_{ntz(i)} */
/* Checksum_i = Checksum_{i-1} xor P_i */
/* C_i = Offset_i xor ENCIPHER(K, P_i xor Offset_i) */
#define OCB_INPUT(n, lreg, xreg) \
movdqu (n * 16)(%rdx), xreg; \
movdqu (lreg), RNOT; \
pxor RNOT, RTMP0; \
pxor xreg, RTMP1; \
pxor RTMP0, xreg; \
movdqu RTMP0, (n * 16)(%rsi);
movq (0 * 8)(%r9), %r10;
movq (1 * 8)(%r9), %r11;
movq (2 * 8)(%r9), %r12;
movq (3 * 8)(%r9), %r13;
OCB_INPUT(0, %r10, RA0);
OCB_INPUT(1, %r11, RA1);
OCB_INPUT(2, %r12, RA2);
OCB_INPUT(3, %r13, RA3);
movq (4 * 8)(%r9), %r10;
movq (5 * 8)(%r9), %r11;
movq (6 * 8)(%r9), %r12;
movq (7 * 8)(%r9), %r13;
OCB_INPUT(4, %r10, RB0);
OCB_INPUT(5, %r11, RB1);
OCB_INPUT(6, %r12, RB2);
OCB_INPUT(7, %r13, RB3);
#undef OCB_INPUT
movdqu RTMP0, (%rcx);
movdqu RTMP1, (%r8);
movq (0 * 8)(%rsp), %r10;
movq (1 * 8)(%rsp), %r11;
movq (2 * 8)(%rsp), %r12;
movq (3 * 8)(%rsp), %r13;
CFI_RESTORE(%r10);
CFI_RESTORE(%r11);
CFI_RESTORE(%r12);
CFI_RESTORE(%r13);
call __serpent_enc_blk8;
addq $(4 * 8), %rsp;
CFI_ADJUST_CFA_OFFSET(-4 * 8);
pxor_u((0 * 16)(%rsi), RA4, RTMP0);
pxor_u((1 * 16)(%rsi), RA1, RTMP0);
pxor_u((2 * 16)(%rsi), RA2, RTMP0);
pxor_u((3 * 16)(%rsi), RA0, RTMP0);
pxor_u((4 * 16)(%rsi), RB4, RTMP0);
pxor_u((5 * 16)(%rsi), RB1, RTMP0);
pxor_u((6 * 16)(%rsi), RB2, RTMP0);
pxor_u((7 * 16)(%rsi), RB0, RTMP0);
movdqu RA4, (0 * 16)(%rsi);
movdqu RA1, (1 * 16)(%rsi);
movdqu RA2, (2 * 16)(%rsi);
movdqu RA0, (3 * 16)(%rsi);
movdqu RB4, (4 * 16)(%rsi);
movdqu RB1, (5 * 16)(%rsi);
movdqu RB2, (6 * 16)(%rsi);
movdqu RB0, (7 * 16)(%rsi);
/* clear the used registers */
pxor RA0, RA0;
pxor RA1, RA1;
pxor RA2, RA2;
pxor RA3, RA3;
pxor RA4, RA4;
pxor RB0, RB0;
pxor RB1, RB1;
pxor RB2, RB2;
pxor RB3, RB3;
pxor RB4, RB4;
pxor RTMP0, RTMP0;
pxor RTMP1, RTMP1;
pxor RTMP2, RTMP2;
pxor RNOT, RNOT;
ret;
CFI_ENDPROC();
ELF(.size _gcry_serpent_sse2_ocb_enc,.-_gcry_serpent_sse2_ocb_enc;)
.align 8
.globl _gcry_serpent_sse2_ocb_dec
ELF(.type _gcry_serpent_sse2_ocb_dec,@function;)
_gcry_serpent_sse2_ocb_dec:
/* input:
* %rdi: ctx, CTX
* %rsi: dst (8 blocks)
* %rdx: src (8 blocks)
* %rcx: offset
* %r8 : checksum
* %r9 : L pointers (void *L[8])
*/
CFI_STARTPROC();
subq $(4 * 8), %rsp;
CFI_ADJUST_CFA_OFFSET(4 * 8);
movq %r10, (0 * 8)(%rsp);
movq %r11, (1 * 8)(%rsp);
movq %r12, (2 * 8)(%rsp);
movq %r13, (3 * 8)(%rsp);
CFI_REL_OFFSET(%r10, 0 * 8);
CFI_REL_OFFSET(%r11, 1 * 8);
CFI_REL_OFFSET(%r12, 2 * 8);
CFI_REL_OFFSET(%r13, 3 * 8);
movdqu (%rcx), RTMP0;
/* Offset_i = Offset_{i-1} xor L_{ntz(i)} */
/* P_i = Offset_i xor DECIPHER(K, C_i xor Offset_i) */
#define OCB_INPUT(n, lreg, xreg) \
movdqu (n * 16)(%rdx), xreg; \
movdqu (lreg), RNOT; \
pxor RNOT, RTMP0; \
pxor RTMP0, xreg; \
movdqu RTMP0, (n * 16)(%rsi);
movq (0 * 8)(%r9), %r10;
movq (1 * 8)(%r9), %r11;
movq (2 * 8)(%r9), %r12;
movq (3 * 8)(%r9), %r13;
OCB_INPUT(0, %r10, RA0);
OCB_INPUT(1, %r11, RA1);
OCB_INPUT(2, %r12, RA2);
OCB_INPUT(3, %r13, RA3);
movq (4 * 8)(%r9), %r10;
movq (5 * 8)(%r9), %r11;
movq (6 * 8)(%r9), %r12;
movq (7 * 8)(%r9), %r13;
OCB_INPUT(4, %r10, RB0);
OCB_INPUT(5, %r11, RB1);
OCB_INPUT(6, %r12, RB2);
OCB_INPUT(7, %r13, RB3);
#undef OCB_INPUT
movdqu RTMP0, (%rcx);
movq (0 * 8)(%rsp), %r10;
movq (1 * 8)(%rsp), %r11;
movq (2 * 8)(%rsp), %r12;
movq (3 * 8)(%rsp), %r13;
CFI_RESTORE(%r10);
CFI_RESTORE(%r11);
CFI_RESTORE(%r12);
CFI_RESTORE(%r13);
call __serpent_dec_blk8;
addq $(4 * 8), %rsp;
CFI_ADJUST_CFA_OFFSET(-4 * 8);
movdqu (%r8), RTMP0;
pxor_u((0 * 16)(%rsi), RA0, RTMP1);
pxor_u((1 * 16)(%rsi), RA1, RTMP1);
pxor_u((2 * 16)(%rsi), RA2, RTMP1);
pxor_u((3 * 16)(%rsi), RA3, RTMP1);
pxor_u((4 * 16)(%rsi), RB0, RTMP1);
pxor_u((5 * 16)(%rsi), RB1, RTMP1);
pxor_u((6 * 16)(%rsi), RB2, RTMP1);
pxor_u((7 * 16)(%rsi), RB3, RTMP1);
/* Checksum_i = Checksum_{i-1} xor P_i */
movdqu RA0, (0 * 16)(%rsi);
pxor RA0, RTMP0;
movdqu RA1, (1 * 16)(%rsi);
pxor RA1, RTMP0;
movdqu RA2, (2 * 16)(%rsi);
pxor RA2, RTMP0;
movdqu RA3, (3 * 16)(%rsi);
pxor RA3, RTMP0;
movdqu RB0, (4 * 16)(%rsi);
pxor RB0, RTMP0;
movdqu RB1, (5 * 16)(%rsi);
pxor RB1, RTMP0;
movdqu RB2, (6 * 16)(%rsi);
pxor RB2, RTMP0;
movdqu RB3, (7 * 16)(%rsi);
pxor RB3, RTMP0;
movdqu RTMP0, (%r8);
/* clear the used registers */
pxor RA0, RA0;
pxor RA1, RA1;
pxor RA2, RA2;
pxor RA3, RA3;
pxor RA4, RA4;
pxor RB0, RB0;
pxor RB1, RB1;
pxor RB2, RB2;
pxor RB3, RB3;
pxor RB4, RB4;
pxor RTMP0, RTMP0;
pxor RTMP1, RTMP1;
pxor RTMP2, RTMP2;
pxor RNOT, RNOT;
ret;
CFI_ENDPROC();
ELF(.size _gcry_serpent_sse2_ocb_dec,.-_gcry_serpent_sse2_ocb_dec;)
.align 8
.globl _gcry_serpent_sse2_ocb_auth
ELF(.type _gcry_serpent_sse2_ocb_auth,@function;)
_gcry_serpent_sse2_ocb_auth:
/* input:
* %rdi: ctx, CTX
* %rsi: abuf (8 blocks)
* %rdx: offset
* %rcx: checksum
* %r8 : L pointers (void *L[8])
*/
CFI_STARTPROC();
subq $(4 * 8), %rsp;
CFI_ADJUST_CFA_OFFSET(4 * 8);
movq %r10, (0 * 8)(%rsp);
movq %r11, (1 * 8)(%rsp);
movq %r12, (2 * 8)(%rsp);
movq %r13, (3 * 8)(%rsp);
CFI_REL_OFFSET(%r10, 0 * 8);
CFI_REL_OFFSET(%r11, 1 * 8);
CFI_REL_OFFSET(%r12, 2 * 8);
CFI_REL_OFFSET(%r13, 3 * 8);
movdqu (%rdx), RTMP0;
/* Offset_i = Offset_{i-1} xor L_{ntz(i)} */
/* Sum_i = Sum_{i-1} xor ENCIPHER(K, A_i xor Offset_i) */
#define OCB_INPUT(n, lreg, xreg) \
movdqu (n * 16)(%rsi), xreg; \
movdqu (lreg), RNOT; \
pxor RNOT, RTMP0; \
pxor RTMP0, xreg;
movq (0 * 8)(%r8), %r10;
movq (1 * 8)(%r8), %r11;
movq (2 * 8)(%r8), %r12;
movq (3 * 8)(%r8), %r13;
OCB_INPUT(0, %r10, RA0);
OCB_INPUT(1, %r11, RA1);
OCB_INPUT(2, %r12, RA2);
OCB_INPUT(3, %r13, RA3);
movq (4 * 8)(%r8), %r10;
movq (5 * 8)(%r8), %r11;
movq (6 * 8)(%r8), %r12;
movq (7 * 8)(%r8), %r13;
OCB_INPUT(4, %r10, RB0);
OCB_INPUT(5, %r11, RB1);
OCB_INPUT(6, %r12, RB2);
OCB_INPUT(7, %r13, RB3);
#undef OCB_INPUT
movdqu RTMP0, (%rdx);
movq (0 * 8)(%rsp), %r10;
movq (1 * 8)(%rsp), %r11;
movq (2 * 8)(%rsp), %r12;
movq (3 * 8)(%rsp), %r13;
CFI_RESTORE(%r10);
CFI_RESTORE(%r11);
CFI_RESTORE(%r12);
CFI_RESTORE(%r13);
call __serpent_enc_blk8;
addq $(4 * 8), %rsp;
CFI_ADJUST_CFA_OFFSET(-4 * 8);
movdqu (%rcx), RTMP0;
pxor RB4, RA4;
pxor RB1, RA1;
pxor RB2, RA2;
pxor RB0, RA0;
pxor RTMP0, RA2;
pxor RA4, RA1;
pxor RA2, RA0;
pxor RA1, RA0;
movdqu RA0, (%rcx);
/* clear the used registers */
pxor RA0, RA0;
pxor RA1, RA1;
pxor RA2, RA2;
pxor RA3, RA3;
pxor RA4, RA4;
pxor RB0, RB0;
pxor RB1, RB1;
pxor RB2, RB2;
pxor RB3, RB3;
pxor RB4, RB4;
pxor RTMP0, RTMP0;
pxor RTMP1, RTMP1;
pxor RTMP2, RTMP2;
pxor RNOT, RNOT;
ret;
CFI_ENDPROC();
ELF(.size _gcry_serpent_sse2_ocb_auth,.-_gcry_serpent_sse2_ocb_auth;)
#endif /*defined(USE_SERPENT)*/
#endif /*__x86_64*/