| 1 |
/* |
| 2 |
* SHA-1 using SIMD instructions |
| 3 |
* Based on public domain code by Dean Gaudet |
| 4 |
* (http://arctic.org/~dean/crypto/sha1.html) |
| 5 |
* (C) 2009-2011,2023,2025 Jack Lloyd |
| 6 |
* |
| 7 |
* Botan is released under the Simplified BSD License (see license.txt) |
| 8 |
*/ |
| 9 |
|
| 10 |
#include <botan/internal/sha1.h> |
| 11 |
|
| 12 |
#include <botan/internal/isa_extn.h> |
| 13 |
#include <botan/internal/sha1_f.h> |
| 14 |
#include <botan/internal/simd_4x32.h> |
| 15 |
|
| 16 |
namespace Botan { |
| 17 |
|
| 18 |
namespace { |
| 19 |
|
| 20 |
/* |
| 21 |
For each multiple of 4, t, we want to calculate this: |
| 22 |
|
| 23 |
W[t+0] = rol(W[t-3] ^ W[t-8] ^ W[t-14] ^ W[t-16], 1); |
| 24 |
W[t+1] = rol(W[t-2] ^ W[t-7] ^ W[t-13] ^ W[t-15], 1); |
| 25 |
W[t+2] = rol(W[t-1] ^ W[t-6] ^ W[t-12] ^ W[t-14], 1); |
| 26 |
W[t+3] = rol(W[t] ^ W[t-5] ^ W[t-11] ^ W[t-13], 1); |
| 27 |
|
| 28 |
we'll actually calculate this: |
| 29 |
|
| 30 |
W[t+0] = rol(W[t-3] ^ W[t-8] ^ W[t-14] ^ W[t-16], 1); |
| 31 |
W[t+1] = rol(W[t-2] ^ W[t-7] ^ W[t-13] ^ W[t-15], 1); |
| 32 |
W[t+2] = rol(W[t-1] ^ W[t-6] ^ W[t-12] ^ W[t-14], 1); |
| 33 |
W[t+3] = rol( 0 ^ W[t-5] ^ W[t-11] ^ W[t-13], 1); |
| 34 |
W[t+3] ^= rol(W[t+0], 1); |
| 35 |
|
| 36 |
the parameters are: |
| 37 |
|
| 38 |
W0 = &W[t-16]; |
| 39 |
W1 = &W[t-12]; |
| 40 |
W2 = &W[t- 8]; |
| 41 |
W3 = &W[t- 4]; |
| 42 |
|
| 43 |
and on output: |
| 44 |
W0 = W[t]..W[t+3] |
| 45 |
*/ |
| 46 |
BOTAN_FORCE_INLINE SIMD_4x32 BOTAN_FN_ISA_SIMD_4X32 sha1_simd_next_w(SIMD_4x32& XW0, |
| 47 |
SIMD_4x32 XW1, |
| 48 |
SIMD_4x32 XW2, |
| 49 |
SIMD_4x32 XW3) { |
| 50 |
SIMD_4x32 T0 = XW0; // W[t-16..t-13] |
| 51 |
T0 ^= SIMD_4x32::alignr8(XW1, XW0); // W[t-14..t-11] |
| 52 |
T0 ^= XW2; // W[t-8..t-5] |
| 53 |
T0 ^= XW3.shift_elems_right<1>(); // W[t-3..t-1] || 0 |
| 54 |
|
| 55 |
/* unrotated W[t]..W[t+2] in T0 ... still need W[t+3] */ |
| 56 |
|
| 57 |
// Extract w[t+0] into T2 |
| 58 |
auto T2 = T0.shift_elems_left<3>(); |
| 59 |
|
| 60 |
// Main rotation |
| 61 |
T0 = T0.rotl<1>(); |
| 62 |
|
| 63 |
// Rotation of W[t+3] has rot by 2 to account for us working on non-rotated words |
| 64 |
T2 = T2.rotl<2>(); |
| 65 |
|
| 66 |
// Merge rol(W[t+0], 1) into W[t+3] |
| 67 |
T0 ^= T2; |
| 68 |
|
| 69 |
XW0 = T0; |
| 70 |
return T0; |
| 71 |
} |
| 72 |
|
| 73 |
} // namespace |
| 74 |
|
| 75 |
/* |
| 76 |
* SHA-1 Compression Function using SIMD for message expansion |
| 77 |
*/ |
| 78 |
//static |
| 79 |
void BOTAN_FN_ISA_SIMD_4X32 SHA_1::simd_compress_n(digest_type& digest, std::span<const uint8_t> input, size_t blocks) { |
| 80 |
using namespace SHA1_F; |
| 81 |
|
| 82 |
const SIMD_4x32 K00_19 = SIMD_4x32::splat(K1); |
| 83 |
const SIMD_4x32 K20_39 = SIMD_4x32::splat(K2); |
| 84 |
const SIMD_4x32 K40_59 = SIMD_4x32::splat(K3); |
| 85 |
const SIMD_4x32 K60_79 = SIMD_4x32::splat(K4); |
| 86 |
|
| 87 |
uint32_t A = digest[0]; |
| 88 |
uint32_t B = digest[1]; |
| 89 |
uint32_t C = digest[2]; |
| 90 |
uint32_t D = digest[3]; |
| 91 |
uint32_t E = digest[4]; |
| 92 |
|
| 93 |
BufferSlicer in(input); |
| 94 |
|
| 95 |
for(size_t i = 0; i != blocks; ++i) { |
| 96 |
uint32_t PT[4]; |
| 97 |
|
| 98 |
const auto block = in.take(block_bytes); |
| 99 |
|
| 100 |
SIMD_4x32 W0 = SIMD_4x32::load_be(&block[0]); // NOLINT(*-container-data-pointer) |
| 101 |
SIMD_4x32 W1 = SIMD_4x32::load_be(&block[16]); |
| 102 |
SIMD_4x32 W2 = SIMD_4x32::load_be(&block[32]); |
| 103 |
SIMD_4x32 W3 = SIMD_4x32::load_be(&block[48]); |
| 104 |
|
| 105 |
SIMD_4x32 P0 = W0 + K00_19; |
| 106 |
SIMD_4x32 P1 = W1 + K00_19; |
| 107 |
SIMD_4x32 P2 = W2 + K00_19; |
| 108 |
SIMD_4x32 P3 = W3 + K00_19; |
| 109 |
|
| 110 |
P0.store_le(PT); |
| 111 |
F1(A, B, C, D, E, PT[0]); |
| 112 |
F1(E, A, B, C, D, PT[1]); |
| 113 |
F1(D, E, A, B, C, PT[2]); |
| 114 |
F1(C, D, E, A, B, PT[3]); |
| 115 |
P0 = sha1_simd_next_w(W0, W1, W2, W3) + K00_19; |
| 116 |
|
| 117 |
P1.store_le(PT); |
| 118 |
F1(B, C, D, E, A, PT[0]); |
| 119 |
F1(A, B, C, D, E, PT[1]); |
| 120 |
F1(E, A, B, C, D, PT[2]); |
| 121 |
F1(D, E, A, B, C, PT[3]); |
| 122 |
P1 = sha1_simd_next_w(W1, W2, W3, W0) + K20_39; |
| 123 |
|
| 124 |
P2.store_le(PT); |
| 125 |
F1(C, D, E, A, B, PT[0]); |
| 126 |
F1(B, C, D, E, A, PT[1]); |
| 127 |
F1(A, B, C, D, E, PT[2]); |
| 128 |
F1(E, A, B, C, D, PT[3]); |
| 129 |
P2 = sha1_simd_next_w(W2, W3, W0, W1) + K20_39; |
| 130 |
|
| 131 |
P3.store_le(PT); |
| 132 |
F1(D, E, A, B, C, PT[0]); |
| 133 |
F1(C, D, E, A, B, PT[1]); |
| 134 |
F1(B, C, D, E, A, PT[2]); |
| 135 |
F1(A, B, C, D, E, PT[3]); |
| 136 |
P3 = sha1_simd_next_w(W3, W0, W1, W2) + K20_39; |
| 137 |
|
| 138 |
P0.store_le(PT); |
| 139 |
F1(E, A, B, C, D, PT[0]); |
| 140 |
F1(D, E, A, B, C, PT[1]); |
| 141 |
F1(C, D, E, A, B, PT[2]); |
| 142 |
F1(B, C, D, E, A, PT[3]); |
| 143 |
P0 = sha1_simd_next_w(W0, W1, W2, W3) + K20_39; |
| 144 |
|
| 145 |
P1.store_le(PT); |
| 146 |
F2(A, B, C, D, E, PT[0]); |
| 147 |
F2(E, A, B, C, D, PT[1]); |
| 148 |
F2(D, E, A, B, C, PT[2]); |
| 149 |
F2(C, D, E, A, B, PT[3]); |
| 150 |
P1 = sha1_simd_next_w(W1, W2, W3, W0) + K20_39; |
| 151 |
|
| 152 |
P2.store_le(PT); |
| 153 |
F2(B, C, D, E, A, PT[0]); |
| 154 |
F2(A, B, C, D, E, PT[1]); |
| 155 |
F2(E, A, B, C, D, PT[2]); |
| 156 |
F2(D, E, A, B, C, PT[3]); |
| 157 |
P2 = sha1_simd_next_w(W2, W3, W0, W1) + K40_59; |
| 158 |
|
| 159 |
P3.store_le(PT); |
| 160 |
F2(C, D, E, A, B, PT[0]); |
| 161 |
F2(B, C, D, E, A, PT[1]); |
| 162 |
F2(A, B, C, D, E, PT[2]); |
| 163 |
F2(E, A, B, C, D, PT[3]); |
| 164 |
P3 = sha1_simd_next_w(W3, W0, W1, W2) + K40_59; |
| 165 |
|
| 166 |
P0.store_le(PT); |
| 167 |
F2(D, E, A, B, C, PT[0]); |
| 168 |
F2(C, D, E, A, B, PT[1]); |
| 169 |
F2(B, C, D, E, A, PT[2]); |
| 170 |
F2(A, B, C, D, E, PT[3]); |
| 171 |
P0 = sha1_simd_next_w(W0, W1, W2, W3) + K40_59; |
| 172 |
|
| 173 |
P1.store_le(PT); |
| 174 |
F2(E, A, B, C, D, PT[0]); |
| 175 |
F2(D, E, A, B, C, PT[1]); |
| 176 |
F2(C, D, E, A, B, PT[2]); |
| 177 |
F2(B, C, D, E, A, PT[3]); |
| 178 |
P1 = sha1_simd_next_w(W1, W2, W3, W0) + K40_59; |
| 179 |
|
| 180 |
P2.store_le(PT); |
| 181 |
F3(A, B, C, D, E, PT[0]); |
| 182 |
F3(E, A, B, C, D, PT[1]); |
| 183 |
F3(D, E, A, B, C, PT[2]); |
| 184 |
F3(C, D, E, A, B, PT[3]); |
| 185 |
P2 = sha1_simd_next_w(W2, W3, W0, W1) + K40_59; |
| 186 |
|
| 187 |
P3.store_le(PT); |
| 188 |
F3(B, C, D, E, A, PT[0]); |
| 189 |
F3(A, B, C, D, E, PT[1]); |
| 190 |
F3(E, A, B, C, D, PT[2]); |
| 191 |
F3(D, E, A, B, C, PT[3]); |
| 192 |
P3 = sha1_simd_next_w(W3, W0, W1, W2) + K60_79; |
| 193 |
|
| 194 |
P0.store_le(PT); |
| 195 |
F3(C, D, E, A, B, PT[0]); |
| 196 |
F3(B, C, D, E, A, PT[1]); |
| 197 |
F3(A, B, C, D, E, PT[2]); |
| 198 |
F3(E, A, B, C, D, PT[3]); |
| 199 |
P0 = sha1_simd_next_w(W0, W1, W2, W3) + K60_79; |
| 200 |
|
| 201 |
P1.store_le(PT); |
| 202 |
F3(D, E, A, B, C, PT[0]); |
| 203 |
F3(C, D, E, A, B, PT[1]); |
| 204 |
F3(B, C, D, E, A, PT[2]); |
| 205 |
F3(A, B, C, D, E, PT[3]); |
| 206 |
P1 = sha1_simd_next_w(W1, W2, W3, W0) + K60_79; |
| 207 |
|
| 208 |
P2.store_le(PT); |
| 209 |
F3(E, A, B, C, D, PT[0]); |
| 210 |
F3(D, E, A, B, C, PT[1]); |
| 211 |
F3(C, D, E, A, B, PT[2]); |
| 212 |
F3(B, C, D, E, A, PT[3]); |
| 213 |
P2 = sha1_simd_next_w(W2, W3, W0, W1) + K60_79; |
| 214 |
|
| 215 |
P3.store_le(PT); |
| 216 |
F4(A, B, C, D, E, PT[0]); |
| 217 |
F4(E, A, B, C, D, PT[1]); |
| 218 |
F4(D, E, A, B, C, PT[2]); |
| 219 |
F4(C, D, E, A, B, PT[3]); |
| 220 |
P3 = sha1_simd_next_w(W3, W0, W1, W2) + K60_79; |
| 221 |
|
| 222 |
P0.store_le(PT); |
| 223 |
F4(B, C, D, E, A, PT[0]); |
| 224 |
F4(A, B, C, D, E, PT[1]); |
| 225 |
F4(E, A, B, C, D, PT[2]); |
| 226 |
F4(D, E, A, B, C, PT[3]); |
| 227 |
|
| 228 |
P1.store_le(PT); |
| 229 |
F4(C, D, E, A, B, PT[0]); |
| 230 |
F4(B, C, D, E, A, PT[1]); |
| 231 |
F4(A, B, C, D, E, PT[2]); |
| 232 |
F4(E, A, B, C, D, PT[3]); |
| 233 |
|
| 234 |
P2.store_le(PT); |
| 235 |
F4(D, E, A, B, C, PT[0]); |
| 236 |
F4(C, D, E, A, B, PT[1]); |
| 237 |
F4(B, C, D, E, A, PT[2]); |
| 238 |
F4(A, B, C, D, E, PT[3]); |
| 239 |
|
| 240 |
P3.store_le(PT); |
| 241 |
F4(E, A, B, C, D, PT[0]); |
| 242 |
F4(D, E, A, B, C, PT[1]); |
| 243 |
F4(C, D, E, A, B, PT[2]); |
| 244 |
F4(B, C, D, E, A, PT[3]); |
| 245 |
|
| 246 |
A = (digest[0] += A); |
| 247 |
B = (digest[1] += B); |
| 248 |
C = (digest[2] += C); |
| 249 |
D = (digest[3] += D); |
| 250 |
E = (digest[4] += E); |
| 251 |
} |
| 252 |
} |
| 253 |
|
| 254 |
} // namespace Botan |