1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
|
/* This file is part of the gf2x library.
Copyright 2007, 2008, 2009
Richard Brent, Pierrick Gaudry, Emmanuel Thome', Paul Zimmermann
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.
You should have received a copy of the GNU General Public License along
with this program; see the file COPYING. If not, write to the Free
Software Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA
02111-1307, USA.
*/
#ifndef GF2X_MUL9_H_
#define GF2X_MUL9_H_
#include "gf2x.h"
/* All gf2x source files for lowlevel functions must include gf2x-small.h
* This is mandatory for the tuning mechanism. */
#include "gf2x/gf2x-small.h"
#include <emmintrin.h>
#include <wmmintrin.h>
#if GF2X_WORDSIZE != 64
#error "This code is for 64-bit only"
#endif
#include "gf2x/gf2x-config.h"
#ifndef HAVE_PCLMUL_SUPPORT
#error "This code needs pclmul support"
#endif
/* TODO: if somebody comes up with a neat way to improve the interface so
* as to remove the false dependency on pclmul, that would be nice.
*/
static inline __v2di
GF2X_FUNC(mul9k3_mul1) (unsigned long a, unsigned long b)
{
__v2di aa = (__v2di) { a, 0 };
__v2di bb = (__v2di) { b, 0 };
return _mm_clmulepi64_si128(aa, bb, 0);
}
/* same as mul5, but stores a[4]*b[4] into {d,2} */
GF2X_STORAGE_CLASS_mul5
void GF2X_FUNC(mul9k3_mul5c) (unsigned long *c, const unsigned long *a,
const unsigned long *b, unsigned long *d)
{
/* Montgomery formulae with 13 multiplications */
unsigned long ta[3], tb[3], pa[8], pb[8];
__v2di p0, p2, p4, p6, p8, p10, p12, p14, p16, p18, p20, p22, p24;
__v2di t0, t2, t4, t6, t8, t10, t12;
ta[0] = a[0] ^ a[4] ; tb[0] = b[0] ^ b[4];
ta[1] = a[1] ^ a[2] ; tb[1] = b[1] ^ b[2];
ta[2] = a[3] ^ ta[0] ; tb[2] = b[3] ^ tb[0];
pa[0] = ta[1] ^ ta[2] ; pb[0] = tb[1] ^ tb[2];
pa[1] = a[2] ^ ta[2] ; pb[1] = b[2] ^ tb[2];
pa[2] = ta[0] ^ ta[1] ; pb[2] = tb[0] ^ tb[1];
pa[3] = a[1] ^ ta[2] ; pb[3] = b[1] ^ tb[2];
pa[4] = a[0] ^ a[2] ^ a[3] ; pb[4] = b[0] ^ b[2] ^ b[3];
pa[5] = a[4] ^ ta[1] ; pb[5] = b[4] ^ tb[1];
pa[6] = a[3] ^ a[4] ; pb[6] = b[3] ^ b[4];
pa[7] = a[0] ^ a[1] ; pb[7] = b[0] ^ b[1];
p0 = GF2X_FUNC(mul9k3_mul1)(pa[0], pb[0]);
p2 = GF2X_FUNC(mul9k3_mul1)(pa[1], pb[1]);
p4 = GF2X_FUNC(mul9k3_mul1)(pa[2], pb[2]);
p6 = GF2X_FUNC(mul9k3_mul1)(pa[3], pb[3]);
p8 = GF2X_FUNC(mul9k3_mul1)(pa[4], pb[4]);
p10 = GF2X_FUNC(mul9k3_mul1)(pa[5], pb[5]);
p12 = GF2X_FUNC(mul9k3_mul1)(pa[6], pb[6]);
p14 = GF2X_FUNC(mul9k3_mul1)(pa[7], pb[7]);
p16 = GF2X_FUNC(mul9k3_mul1)(ta[0], tb[0]);
p18 = GF2X_FUNC(mul9k3_mul1)(a[4], b[4]);
p20 = GF2X_FUNC(mul9k3_mul1)(a[3], b[3]);
p22 = GF2X_FUNC(mul9k3_mul1)(a[1], b[1]);
p24 = GF2X_FUNC(mul9k3_mul1)(a[0], b[0]);
t0 = p14 ^ p24;
t2 = p12 ^ p18;
t4 = p2 ^ p16;
t6 = p0 ^ p6;
t8 = p4 ^ p16;
t10 = p10 ^ t0;
t12 = p8 ^ t2;
__v2di ce0 = p24;
__v2di ce2 = p18 ^ t8 ^ t10;
__v2di ce4 = p0 ^ p20 ^ p22 ^ t10 ^ t12;
__v2di ce6 = p24 ^ t4 ^ t12;
__v2di ce8 = p18;
__v2di co1 = p22 ^ t0;
__v2di co3 = t2 ^ t4 ^ t6;
__v2di co5 = t0 ^ t6 ^ t8;
__v2di co7 = p20 ^ t2;
_mm_storeu_si128((__v2di*)(d), p18); /* a[4] * b[4] */
_mm_storeu_si128((__v2di*)(c), ce0 ^ _mm_slli_si128(co1, 8));
_mm_storeu_si128((__v2di*)(c+2), ce2 ^ _mm_srli_si128(co1, 8) ^ _mm_slli_si128(co3, 8));
_mm_storeu_si128((__v2di*)(c+4), ce4 ^ _mm_srli_si128(co3, 8) ^ _mm_slli_si128(co5, 8));
_mm_storeu_si128((__v2di*)(c+6), ce6 ^ _mm_srli_si128(co5, 8) ^ _mm_slli_si128(co7, 8));
_mm_storeu_si128((__v2di*)(c+8), ce8 ^ _mm_srli_si128(co7, 8));
}
/* same as mul5, but assumes {d,2} contains a[4]*b[4] */
GF2X_STORAGE_CLASS_mul5
void GF2X_FUNC(mul9k3_mul5b) (unsigned long *c, const unsigned long *a,
const unsigned long *b, const unsigned long *d)
{
/* Montgomery formulae with 13 multiplications */
unsigned long ta[3], tb[3], pa[8], pb[8];
__v2di p0, p2, p4, p6, p8, p10, p12, p14, p16, p18, p20, p22, p24;
__v2di t0, t2, t4, t6, t8, t10, t12;
ta[0] = a[0] ^ a[4] ; tb[0] = b[0] ^ b[4];
ta[1] = a[1] ^ a[2] ; tb[1] = b[1] ^ b[2];
ta[2] = a[3] ^ ta[0] ; tb[2] = b[3] ^ tb[0];
pa[0] = ta[1] ^ ta[2] ; pb[0] = tb[1] ^ tb[2];
pa[1] = a[2] ^ ta[2] ; pb[1] = b[2] ^ tb[2];
pa[2] = ta[0] ^ ta[1] ; pb[2] = tb[0] ^ tb[1];
pa[3] = a[1] ^ ta[2] ; pb[3] = b[1] ^ tb[2];
pa[4] = a[0] ^ a[2] ^ a[3] ; pb[4] = b[0] ^ b[2] ^ b[3];
pa[5] = a[4] ^ ta[1] ; pb[5] = b[4] ^ tb[1];
pa[6] = a[3] ^ a[4] ; pb[6] = b[3] ^ b[4];
pa[7] = a[0] ^ a[1] ; pb[7] = b[0] ^ b[1];
p0 = GF2X_FUNC(mul9k3_mul1)(pa[0], pb[0]);
p2 = GF2X_FUNC(mul9k3_mul1)(pa[1], pb[1]);
p4 = GF2X_FUNC(mul9k3_mul1)(pa[2], pb[2]);
p6 = GF2X_FUNC(mul9k3_mul1)(pa[3], pb[3]);
p8 = GF2X_FUNC(mul9k3_mul1)(pa[4], pb[4]);
p10 = GF2X_FUNC(mul9k3_mul1)(pa[5], pb[5]);
p12 = GF2X_FUNC(mul9k3_mul1)(pa[6], pb[6]);
p14 = GF2X_FUNC(mul9k3_mul1)(pa[7], pb[7]);
p16 = GF2X_FUNC(mul9k3_mul1)(ta[0], tb[0]);
/* p18 = GF2X_FUNC(mul9k3_mul1)(a[4], b[4]); */
p18 = _mm_loadu_si128((__v2di *) d);
p20 = GF2X_FUNC(mul9k3_mul1)(a[3], b[3]);
p22 = GF2X_FUNC(mul9k3_mul1)(a[1], b[1]);
p24 = GF2X_FUNC(mul9k3_mul1)(a[0], b[0]);
t0 = p14 ^ p24;
t2 = p12 ^ p18;
t4 = p2 ^ p16;
t6 = p0 ^ p6;
t8 = p4 ^ p16;
t10 = p10 ^ t0;
t12 = p8 ^ t2;
__v2di ce0 = p24;
__v2di ce2 = p18 ^ t8 ^ t10;
__v2di ce4 = p0 ^ p20 ^ p22 ^ t10 ^ t12;
__v2di ce6 = p24 ^ t4 ^ t12;
__v2di ce8 = p18;
__v2di co1 = p22 ^ t0;
__v2di co3 = t2 ^ t4 ^ t6;
__v2di co5 = t0 ^ t6 ^ t8;
__v2di co7 = p20 ^ t2;
_mm_storeu_si128((__v2di*)(c), ce0 ^ _mm_slli_si128(co1, 8));
_mm_storeu_si128((__v2di*)(c+2), ce2 ^ _mm_srli_si128(co1, 8) ^ _mm_slli_si128(co3, 8));
_mm_storeu_si128((__v2di*)(c+4), ce4 ^ _mm_srli_si128(co3, 8) ^ _mm_slli_si128(co5, 8));
_mm_storeu_si128((__v2di*)(c+6), ce6 ^ _mm_srli_si128(co5, 8) ^ _mm_slli_si128(co7, 8));
_mm_storeu_si128((__v2di*)(c+8), ce8 ^ _mm_srli_si128(co7, 8));
}
/* (based on mul9k.c)
1 call to mul4 and 2 calls to mul5 minus one multiplication, i.e.,
34 multiplications with mul5clk_c */
GF2X_STORAGE_CLASS_mul9
void gf2x_mul9 (unsigned long *c, const unsigned long *a, const unsigned long *b)
{
unsigned long aa[5], bb[5], ab[10], ab5, ab6, ab7, ab8, ab9, d[2];
gf2x_mul4 (c+10, a+5, b+5);
GF2X_FUNC(mul9k3_mul5c) (c, a, b, d); /* a[4]*b[4] is cached in d */
aa[0] = a[0] ^ a[5];
aa[1] = a[1] ^ a[6];
aa[2] = a[2] ^ a[7];
aa[3] = a[3] ^ a[8];
aa[4] = a[4];
bb[0] = b[0] ^ b[5];
bb[1] = b[1] ^ b[6];
bb[2] = b[2] ^ b[7];
bb[3] = b[3] ^ b[8];
bb[4] = b[4];
GF2X_FUNC(mul9k3_mul5b) (ab, aa, bb, d);
ab5 = ab[5] ^ c[5];
ab6 = ab[6] ^ c[6];
ab7 = ab[7] ^ c[7];
ab8 = ab[8] ^ c[8];
ab9 = ab[9] ^ c[9];
c[5] ^= ab[0] ^ c[0] ^ c[10];
c[6] ^= ab[1] ^ c[1] ^ c[11];
c[7] ^= ab[2] ^ c[2] ^ c[12];
c[8] ^= ab[3] ^ c[3] ^ c[13];
c[9] ^= ab[4] ^ c[4] ^ c[14];
c[10] ^= ab5 ^ c[15];
c[11] ^= ab6 ^ c[16];
c[12] ^= ab7 ^ c[17];
c[13] ^= ab8;
c[14] ^= ab9;
}
#endif /* GF2X_MUL9_H_ */
|
