a05248ed2d
This patch adds AVX optimized XTS-mode helper functions/macros and converts
serpent-avx to use the new facilities. Benefits are slightly improved speed
and reduced stack usage as use of temporary IV-array is avoided.
tcrypt results, with Intel i5-2450M:
enc dec
16B 1.00x 1.00x
64B 1.00x 1.00x
256B 1.04x 1.06x
1024B 1.09x 1.09x
8192B 1.10x 1.09x
Signed-off-by: Jussi Kivilinna <jussi.kivilinna@iki.fi>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
151 lines
4.2 KiB
ArmAsm
151 lines
4.2 KiB
ArmAsm
/*
|
|
* Shared glue code for 128bit block ciphers, AVX assembler macros
|
|
*
|
|
* Copyright © 2012-2013 Jussi Kivilinna <jussi.kivilinna@iki.fi>
|
|
*
|
|
* This program is free software; you can redistribute it and/or modify
|
|
* it under the terms of the GNU General Public License as published by
|
|
* the Free Software Foundation; either version 2 of the License, or
|
|
* (at your option) any later version.
|
|
*
|
|
* This program 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 General Public License for more details.
|
|
*
|
|
*/
|
|
|
|
#define load_8way(src, x0, x1, x2, x3, x4, x5, x6, x7) \
|
|
vmovdqu (0*16)(src), x0; \
|
|
vmovdqu (1*16)(src), x1; \
|
|
vmovdqu (2*16)(src), x2; \
|
|
vmovdqu (3*16)(src), x3; \
|
|
vmovdqu (4*16)(src), x4; \
|
|
vmovdqu (5*16)(src), x5; \
|
|
vmovdqu (6*16)(src), x6; \
|
|
vmovdqu (7*16)(src), x7;
|
|
|
|
#define store_8way(dst, x0, x1, x2, x3, x4, x5, x6, x7) \
|
|
vmovdqu x0, (0*16)(dst); \
|
|
vmovdqu x1, (1*16)(dst); \
|
|
vmovdqu x2, (2*16)(dst); \
|
|
vmovdqu x3, (3*16)(dst); \
|
|
vmovdqu x4, (4*16)(dst); \
|
|
vmovdqu x5, (5*16)(dst); \
|
|
vmovdqu x6, (6*16)(dst); \
|
|
vmovdqu x7, (7*16)(dst);
|
|
|
|
#define store_cbc_8way(src, dst, x0, x1, x2, x3, x4, x5, x6, x7) \
|
|
vpxor (0*16)(src), x1, x1; \
|
|
vpxor (1*16)(src), x2, x2; \
|
|
vpxor (2*16)(src), x3, x3; \
|
|
vpxor (3*16)(src), x4, x4; \
|
|
vpxor (4*16)(src), x5, x5; \
|
|
vpxor (5*16)(src), x6, x6; \
|
|
vpxor (6*16)(src), x7, x7; \
|
|
store_8way(dst, x0, x1, x2, x3, x4, x5, x6, x7);
|
|
|
|
#define inc_le128(x, minus_one, tmp) \
|
|
vpcmpeqq minus_one, x, tmp; \
|
|
vpsubq minus_one, x, x; \
|
|
vpslldq $8, tmp, tmp; \
|
|
vpsubq tmp, x, x;
|
|
|
|
#define load_ctr_8way(iv, bswap, x0, x1, x2, x3, x4, x5, x6, x7, t0, t1, t2) \
|
|
vpcmpeqd t0, t0, t0; \
|
|
vpsrldq $8, t0, t0; /* low: -1, high: 0 */ \
|
|
vmovdqa bswap, t1; \
|
|
\
|
|
/* load IV and byteswap */ \
|
|
vmovdqu (iv), x7; \
|
|
vpshufb t1, x7, x0; \
|
|
\
|
|
/* construct IVs */ \
|
|
inc_le128(x7, t0, t2); \
|
|
vpshufb t1, x7, x1; \
|
|
inc_le128(x7, t0, t2); \
|
|
vpshufb t1, x7, x2; \
|
|
inc_le128(x7, t0, t2); \
|
|
vpshufb t1, x7, x3; \
|
|
inc_le128(x7, t0, t2); \
|
|
vpshufb t1, x7, x4; \
|
|
inc_le128(x7, t0, t2); \
|
|
vpshufb t1, x7, x5; \
|
|
inc_le128(x7, t0, t2); \
|
|
vpshufb t1, x7, x6; \
|
|
inc_le128(x7, t0, t2); \
|
|
vmovdqa x7, t2; \
|
|
vpshufb t1, x7, x7; \
|
|
inc_le128(t2, t0, t1); \
|
|
vmovdqu t2, (iv);
|
|
|
|
#define store_ctr_8way(src, dst, x0, x1, x2, x3, x4, x5, x6, x7) \
|
|
vpxor (0*16)(src), x0, x0; \
|
|
vpxor (1*16)(src), x1, x1; \
|
|
vpxor (2*16)(src), x2, x2; \
|
|
vpxor (3*16)(src), x3, x3; \
|
|
vpxor (4*16)(src), x4, x4; \
|
|
vpxor (5*16)(src), x5, x5; \
|
|
vpxor (6*16)(src), x6, x6; \
|
|
vpxor (7*16)(src), x7, x7; \
|
|
store_8way(dst, x0, x1, x2, x3, x4, x5, x6, x7);
|
|
|
|
#define gf128mul_x_ble(iv, mask, tmp) \
|
|
vpsrad $31, iv, tmp; \
|
|
vpaddq iv, iv, iv; \
|
|
vpshufd $0x13, tmp, tmp; \
|
|
vpand mask, tmp, tmp; \
|
|
vpxor tmp, iv, iv;
|
|
|
|
#define load_xts_8way(iv, src, dst, x0, x1, x2, x3, x4, x5, x6, x7, tiv, t0, \
|
|
t1, xts_gf128mul_and_shl1_mask) \
|
|
vmovdqa xts_gf128mul_and_shl1_mask, t0; \
|
|
\
|
|
/* load IV */ \
|
|
vmovdqu (iv), tiv; \
|
|
vpxor (0*16)(src), tiv, x0; \
|
|
vmovdqu tiv, (0*16)(dst); \
|
|
\
|
|
/* construct and store IVs, also xor with source */ \
|
|
gf128mul_x_ble(tiv, t0, t1); \
|
|
vpxor (1*16)(src), tiv, x1; \
|
|
vmovdqu tiv, (1*16)(dst); \
|
|
\
|
|
gf128mul_x_ble(tiv, t0, t1); \
|
|
vpxor (2*16)(src), tiv, x2; \
|
|
vmovdqu tiv, (2*16)(dst); \
|
|
\
|
|
gf128mul_x_ble(tiv, t0, t1); \
|
|
vpxor (3*16)(src), tiv, x3; \
|
|
vmovdqu tiv, (3*16)(dst); \
|
|
\
|
|
gf128mul_x_ble(tiv, t0, t1); \
|
|
vpxor (4*16)(src), tiv, x4; \
|
|
vmovdqu tiv, (4*16)(dst); \
|
|
\
|
|
gf128mul_x_ble(tiv, t0, t1); \
|
|
vpxor (5*16)(src), tiv, x5; \
|
|
vmovdqu tiv, (5*16)(dst); \
|
|
\
|
|
gf128mul_x_ble(tiv, t0, t1); \
|
|
vpxor (6*16)(src), tiv, x6; \
|
|
vmovdqu tiv, (6*16)(dst); \
|
|
\
|
|
gf128mul_x_ble(tiv, t0, t1); \
|
|
vpxor (7*16)(src), tiv, x7; \
|
|
vmovdqu tiv, (7*16)(dst); \
|
|
\
|
|
gf128mul_x_ble(tiv, t0, t1); \
|
|
vmovdqu tiv, (iv);
|
|
|
|
#define store_xts_8way(dst, x0, x1, x2, x3, x4, x5, x6, x7) \
|
|
vpxor (0*16)(dst), x0, x0; \
|
|
vpxor (1*16)(dst), x1, x1; \
|
|
vpxor (2*16)(dst), x2, x2; \
|
|
vpxor (3*16)(dst), x3, x3; \
|
|
vpxor (4*16)(dst), x4, x4; \
|
|
vpxor (5*16)(dst), x5, x5; \
|
|
vpxor (6*16)(dst), x6, x6; \
|
|
vpxor (7*16)(dst), x7, x7; \
|
|
store_8way(dst, x0, x1, x2, x3, x4, x5, x6, x7);
|