| 1 |
/* |
| 2 |
* (C) 2026 Jack Lloyd |
| 3 |
* |
| 4 |
* Botan is released under the Simplified BSD License (see license.txt) |
| 5 |
*/ |
| 6 |
|
| 7 |
#include <botan/internal/camellia.h> |
| 8 |
|
| 9 |
#include <botan/mem_ops.h> |
| 10 |
#include <botan/internal/isa_extn.h> |
| 11 |
#include <botan/internal/simd_8x64.h> |
| 12 |
#include <botan/internal/simd_avx2_gfni.h> |
| 13 |
|
| 14 |
namespace Botan { |
| 15 |
|
| 16 |
namespace Camellia_AVX512 { |
| 17 |
|
| 18 |
namespace { |
| 19 |
|
| 20 |
constexpr uint64_t pre123_a = gfni_matrix(R"( |
| 21 |
1 1 1 0 1 1 0 1 |
| 22 |
0 0 1 1 0 0 1 0 |
| 23 |
1 1 0 1 0 0 0 0 |
| 24 |
1 0 1 1 0 0 1 1 |
| 25 |
0 0 0 0 1 1 0 0 |
| 26 |
1 0 1 0 0 1 0 0 |
| 27 |
0 0 1 0 1 1 0 0 |
| 28 |
1 0 0 0 0 1 1 0)"); |
| 29 |
|
| 30 |
constexpr uint64_t pre4_a = gfni_matrix(R"( |
| 31 |
1 1 0 1 1 0 1 1 |
| 32 |
0 1 1 0 0 1 0 0 |
| 33 |
1 0 1 0 0 0 0 1 |
| 34 |
0 1 1 0 0 1 1 1 |
| 35 |
0 0 0 1 1 0 0 0 |
| 36 |
0 1 0 0 1 0 0 1 |
| 37 |
0 1 0 1 1 0 0 0 |
| 38 |
0 0 0 0 1 1 0 1)"); |
| 39 |
|
| 40 |
constexpr uint8_t pre_c = 0b01000101; |
| 41 |
|
| 42 |
constexpr uint64_t post2_a = gfni_matrix(R"( |
| 43 |
0 0 0 1 1 1 0 0 |
| 44 |
0 0 0 0 0 0 0 1 |
| 45 |
0 1 1 0 0 1 1 0 |
| 46 |
1 0 1 1 1 1 1 0 |
| 47 |
0 0 0 1 1 0 1 1 |
| 48 |
1 0 0 0 1 1 1 0 |
| 49 |
0 1 0 1 1 1 1 0 |
| 50 |
0 1 1 1 1 1 1 1)"); |
| 51 |
|
| 52 |
constexpr uint64_t post3_a = gfni_matrix(R"( |
| 53 |
0 1 1 0 0 1 1 0 |
| 54 |
1 0 1 1 1 1 1 0 |
| 55 |
0 0 0 1 1 0 1 1 |
| 56 |
1 0 0 0 1 1 1 0 |
| 57 |
0 1 0 1 1 1 1 0 |
| 58 |
0 1 1 1 1 1 1 1 |
| 59 |
0 0 0 1 1 1 0 0 |
| 60 |
0 0 0 0 0 0 0 1)"); |
| 61 |
|
| 62 |
constexpr uint64_t post14_a = gfni_matrix(R"( |
| 63 |
0 0 0 0 0 0 0 1 |
| 64 |
0 1 1 0 0 1 1 0 |
| 65 |
1 0 1 1 1 1 1 0 |
| 66 |
0 0 0 1 1 0 1 1 |
| 67 |
1 0 0 0 1 1 1 0 |
| 68 |
0 1 0 1 1 1 1 0 |
| 69 |
0 1 1 1 1 1 1 1 |
| 70 |
0 0 0 1 1 1 0 0)"); |
| 71 |
|
| 72 |
// NOLINTBEGIN(portability-simd-intrinsics) |
| 73 |
|
| 74 |
BOTAN_FORCE_INLINE BOTAN_FN_ISA_AVX512_GFNI SIMD_8x64 camellia_f(SIMD_8x64 x) { |
| 75 |
const __m512i xr = x.raw(); |
| 76 |
|
| 77 |
/* |
| 78 |
* Camellia sends different bytes of each word through different sboxes; we |
| 79 |
* resolve this like cavemen by computing all 4 S-box variants over the full |
| 80 |
* vector in parallel, then blending the results. |
| 81 |
*/ |
| 82 |
|
| 83 |
// Compute S1(x), S2(x), S3(x), S4(x) for all bytes |
| 84 |
const __m512i y123 = _mm512_gf2p8affine_epi64_epi8(xr, _mm512_set1_epi64(pre123_a), pre_c); |
| 85 |
const __m512i y4 = _mm512_gf2p8affine_epi64_epi8(xr, _mm512_set1_epi64(pre4_a), pre_c); |
| 86 |
|
| 87 |
const __m512i s1 = _mm512_gf2p8affineinv_epi64_epi8(y123, _mm512_set1_epi64(post14_a), 0x6E); |
| 88 |
const __m512i s2 = _mm512_gf2p8affineinv_epi64_epi8(y123, _mm512_set1_epi64(post2_a), 0xDC); |
| 89 |
const __m512i s3 = _mm512_gf2p8affineinv_epi64_epi8(y123, _mm512_set1_epi64(post3_a), 0x37); |
| 90 |
const __m512i s4 = _mm512_gf2p8affineinv_epi64_epi8(y4, _mm512_set1_epi64(post14_a), 0x6E); |
| 91 |
|
| 92 |
// Blend to find correct S(x) for each byte position |
| 93 |
|
| 94 |
auto sx = s1; |
| 95 |
sx = _mm512_mask_blend_epi8(__mmask64(0x4848484848484848), sx, s2); // s2 at bytes {3,6} |
| 96 |
sx = _mm512_mask_blend_epi8(__mmask64(0x2424242424242424), sx, s3); // s3 at bytes {2,5} |
| 97 |
sx = _mm512_mask_blend_epi8(__mmask64(0x1212121212121212), sx, s4); // s4 at bytes {1,4} |
| 98 |
|
| 99 |
// Linear mixing layer |
| 100 |
const auto P1 = _mm512_set_epi64(0x0808080908080809, |
| 101 |
0x0000000100000001, |
| 102 |
0x0808080908080809, |
| 103 |
0x0000000100000001, |
| 104 |
0x0808080908080809, |
| 105 |
0x0000000100000001, |
| 106 |
0x0808080908080809, |
| 107 |
0x0000000100000001); |
| 108 |
const auto P2 = _mm512_set_epi64(0x09090A0A09090A0A, |
| 109 |
0x0101020201010202, |
| 110 |
0x09090A0A09090A0A, |
| 111 |
0x0101020201010202, |
| 112 |
0x09090A0A09090A0A, |
| 113 |
0x0101020201010202, |
| 114 |
0x09090A0A09090A0A, |
| 115 |
0x0101020201010202); |
| 116 |
const auto P3 = _mm512_set_epi64(0x0A0B0B0B0A0B0B0B, |
| 117 |
0x0203030302030303, |
| 118 |
0x0A0B0B0B0A0B0B0B, |
| 119 |
0x0203030302030303, |
| 120 |
0x0A0B0B0B0A0B0B0B, |
| 121 |
0x0203030302030303, |
| 122 |
0x0A0B0B0B0A0B0B0B, |
| 123 |
0x0203030302030303); |
| 124 |
const auto P4 = _mm512_set_epi64(0x0C0C0D0C0E0D0C0C, |
| 125 |
0x0404050406050404, |
| 126 |
0x0C0C0D0C0E0D0C0C, |
| 127 |
0x0404050406050404, |
| 128 |
0x0C0C0D0C0E0D0C0C, |
| 129 |
0x0404050406050404, |
| 130 |
0x0C0C0D0C0E0D0C0C, |
| 131 |
0x0404050406050404); |
| 132 |
const auto P5 = _mm512_set_epi64(0x0D0E0E0D0F0E0D0F, |
| 133 |
0x0506060507060507, |
| 134 |
0x0D0E0E0D0F0E0D0F, |
| 135 |
0x0506060507060507, |
| 136 |
0x0D0E0E0D0F0E0D0F, |
| 137 |
0x0506060507060507, |
| 138 |
0x0D0E0E0D0F0E0D0F, |
| 139 |
0x0506060507060507); |
| 140 |
const auto P6 = _mm512_set_epi64(0x0F0F0F0EFFFFFFFF, |
| 141 |
0x07070706FFFFFFFF, |
| 142 |
0x0F0F0F0EFFFFFFFF, |
| 143 |
0x07070706FFFFFFFF, |
| 144 |
0x0F0F0F0EFFFFFFFF, |
| 145 |
0x07070706FFFFFFFF, |
| 146 |
0x0F0F0F0EFFFFFFFF, |
| 147 |
0x07070706FFFFFFFF); |
| 148 |
|
| 149 |
const auto t1 = SIMD_8x64(_mm512_shuffle_epi8(sx, P1)); |
| 150 |
const auto t2 = SIMD_8x64(_mm512_shuffle_epi8(sx, P2)); |
| 151 |
const auto t3 = SIMD_8x64(_mm512_shuffle_epi8(sx, P3)); |
| 152 |
const auto t4 = SIMD_8x64(_mm512_shuffle_epi8(sx, P4)); |
| 153 |
const auto t5 = SIMD_8x64(_mm512_shuffle_epi8(sx, P5)); |
| 154 |
const auto t6 = SIMD_8x64(_mm512_shuffle_epi8(sx, P6)); |
| 155 |
|
| 156 |
return (t1 ^ t2 ^ t3 ^ t4 ^ t5 ^ t6); |
| 157 |
} |
| 158 |
|
| 159 |
BOTAN_FORCE_INLINE BOTAN_FN_ISA_AVX512 SIMD_8x64 FL_8(SIMD_8x64 v, uint64_t K) { |
| 160 |
const auto Kv = _mm512_set1_epi64(K); |
| 161 |
auto vr = v.raw(); |
| 162 |
|
| 163 |
// x2 ^= rotl<1>(x1 & k1): AND, rotate 32-bit elements, shift high->low, XOR |
| 164 |
vr = _mm512_xor_si512(vr, _mm512_srli_epi64(_mm512_rol_epi32(_mm512_and_si512(vr, Kv), 1), 32)); |
| 165 |
|
| 166 |
// x1 ^= (x2 | k2): OR, shift low->high, XOR |
| 167 |
vr = _mm512_xor_si512(vr, _mm512_slli_epi64(_mm512_or_si512(vr, Kv), 32)); |
| 168 |
|
| 169 |
return SIMD_8x64(vr); |
| 170 |
} |
| 171 |
|
| 172 |
BOTAN_FORCE_INLINE BOTAN_FN_ISA_AVX512 SIMD_8x64 FLINV_8(SIMD_8x64 v, uint64_t K) { |
| 173 |
const auto Kv = _mm512_set1_epi64(K); |
| 174 |
auto vr = v.raw(); |
| 175 |
|
| 176 |
// x1 ^= (x2 | k2): OR, shift low->high, XOR |
| 177 |
vr = _mm512_xor_si512(vr, _mm512_slli_epi64(_mm512_or_si512(vr, Kv), 32)); |
| 178 |
|
| 179 |
// x2 ^= rotl<1>(x1 & k1): AND, rotate 32-bit elements, shift high->low, XOR |
| 180 |
vr = _mm512_xor_si512(vr, _mm512_srli_epi64(_mm512_rol_epi32(_mm512_and_si512(vr, Kv), 1), 32)); |
| 181 |
|
| 182 |
return SIMD_8x64(vr); |
| 183 |
} |
| 184 |
|
| 185 |
/* |
| 186 |
* Load 8 blocks, byte-swap, and deinterleave into L (even) and R (odd) halves |
| 187 |
*/ |
| 188 |
BOTAN_FORCE_INLINE BOTAN_FN_ISA_AVX512 void load_and_deinterleave(const uint8_t in[], SIMD_8x64& L, SIMD_8x64& R) { |
| 189 |
const auto idx_l = _mm512_set_epi64(0x0E, 0x0C, 0x0A, 0x08, 0x06, 0x04, 0x02, 0x00); |
| 190 |
const auto idx_r = _mm512_set_epi64(0x0F, 0x0D, 0x0B, 0x09, 0x07, 0x05, 0x03, 0x01); |
| 191 |
|
| 192 |
auto A = SIMD_8x64::load_be(in); |
| 193 |
auto B = SIMD_8x64::load_be(in + 64); |
| 194 |
|
| 195 |
L = SIMD_8x64(_mm512_permutex2var_epi64(A.raw(), idx_l, B.raw())); |
| 196 |
R = SIMD_8x64(_mm512_permutex2var_epi64(A.raw(), idx_r, B.raw())); |
| 197 |
} |
| 198 |
|
| 199 |
/* |
| 200 |
* Interleave R/L halves (note swap), byte-swap, and store 8 blocks |
| 201 |
*/ |
| 202 |
BOTAN_FORCE_INLINE BOTAN_FN_ISA_AVX512 void interleave_and_store(uint8_t out[], SIMD_8x64 L, SIMD_8x64 R) { |
| 203 |
const auto idx_lo = _mm512_set_epi64(0x0B, 0x03, 0x0A, 0x02, 0x09, 0x01, 0x08, 0x00); |
| 204 |
const auto idx_hi = _mm512_set_epi64(0x0F, 0x07, 0x0E, 0x06, 0x0D, 0x05, 0x0C, 0x04); |
| 205 |
|
| 206 |
auto A = SIMD_8x64(_mm512_permutex2var_epi64(R.raw(), idx_lo, L.raw())); |
| 207 |
auto B = SIMD_8x64(_mm512_permutex2var_epi64(R.raw(), idx_hi, L.raw())); |
| 208 |
|
| 209 |
A.store_be(out); |
| 210 |
B.store_be(out + 64); |
| 211 |
} |
| 212 |
|
| 213 |
// NOLINTEND(portability-simd-intrinsics) |
| 214 |
|
| 215 |
// Helpers for 6 round iterations |
| 216 |
|
| 217 |
BOTAN_FORCE_INLINE BOTAN_FN_ISA_AVX512_GFNI void six_e_rounds(SIMD_8x64& L, |
| 218 |
SIMD_8x64& R, |
| 219 |
std::span<const uint64_t> SK) { |
| 220 |
R ^= camellia_f(L ^ SIMD_8x64::splat(SK[0])); |
| 221 |
L ^= camellia_f(R ^ SIMD_8x64::splat(SK[1])); |
| 222 |
R ^= camellia_f(L ^ SIMD_8x64::splat(SK[2])); |
| 223 |
L ^= camellia_f(R ^ SIMD_8x64::splat(SK[3])); |
| 224 |
R ^= camellia_f(L ^ SIMD_8x64::splat(SK[4])); |
| 225 |
L ^= camellia_f(R ^ SIMD_8x64::splat(SK[5])); |
| 226 |
} |
| 227 |
|
| 228 |
BOTAN_FORCE_INLINE BOTAN_FN_ISA_AVX512_GFNI void six_d_rounds(SIMD_8x64& L, |
| 229 |
SIMD_8x64& R, |
| 230 |
std::span<const uint64_t> SK) { |
| 231 |
R ^= camellia_f(L ^ SIMD_8x64::splat(SK[5])); |
| 232 |
L ^= camellia_f(R ^ SIMD_8x64::splat(SK[4])); |
| 233 |
R ^= camellia_f(L ^ SIMD_8x64::splat(SK[3])); |
| 234 |
L ^= camellia_f(R ^ SIMD_8x64::splat(SK[2])); |
| 235 |
R ^= camellia_f(L ^ SIMD_8x64::splat(SK[1])); |
| 236 |
L ^= camellia_f(R ^ SIMD_8x64::splat(SK[0])); |
| 237 |
} |
| 238 |
|
| 239 |
BOTAN_FORCE_INLINE BOTAN_FN_ISA_AVX512_GFNI void six_e_rounds_x2( |
| 240 |
SIMD_8x64& L1, SIMD_8x64& R1, SIMD_8x64& L2, SIMD_8x64& R2, std::span<const uint64_t> SK) { |
| 241 |
const auto K0 = SIMD_8x64::splat(SK[0]); |
| 242 |
const auto K1 = SIMD_8x64::splat(SK[1]); |
| 243 |
const auto K2 = SIMD_8x64::splat(SK[2]); |
| 244 |
const auto K3 = SIMD_8x64::splat(SK[3]); |
| 245 |
const auto K4 = SIMD_8x64::splat(SK[4]); |
| 246 |
const auto K5 = SIMD_8x64::splat(SK[5]); |
| 247 |
|
| 248 |
R1 ^= camellia_f(L1 ^ K0); |
| 249 |
R2 ^= camellia_f(L2 ^ K0); |
| 250 |
L1 ^= camellia_f(R1 ^ K1); |
| 251 |
L2 ^= camellia_f(R2 ^ K1); |
| 252 |
R1 ^= camellia_f(L1 ^ K2); |
| 253 |
R2 ^= camellia_f(L2 ^ K2); |
| 254 |
L1 ^= camellia_f(R1 ^ K3); |
| 255 |
L2 ^= camellia_f(R2 ^ K3); |
| 256 |
R1 ^= camellia_f(L1 ^ K4); |
| 257 |
R2 ^= camellia_f(L2 ^ K4); |
| 258 |
L1 ^= camellia_f(R1 ^ K5); |
| 259 |
L2 ^= camellia_f(R2 ^ K5); |
| 260 |
} |
| 261 |
|
| 262 |
BOTAN_FORCE_INLINE BOTAN_FN_ISA_AVX512_GFNI void six_d_rounds_x2( |
| 263 |
SIMD_8x64& L1, SIMD_8x64& R1, SIMD_8x64& L2, SIMD_8x64& R2, std::span<const uint64_t> SK) { |
| 264 |
const auto K0 = SIMD_8x64::splat(SK[0]); |
| 265 |
const auto K1 = SIMD_8x64::splat(SK[1]); |
| 266 |
const auto K2 = SIMD_8x64::splat(SK[2]); |
| 267 |
const auto K3 = SIMD_8x64::splat(SK[3]); |
| 268 |
const auto K4 = SIMD_8x64::splat(SK[4]); |
| 269 |
const auto K5 = SIMD_8x64::splat(SK[5]); |
| 270 |
|
| 271 |
R1 ^= camellia_f(L1 ^ K5); |
| 272 |
R2 ^= camellia_f(L2 ^ K5); |
| 273 |
L1 ^= camellia_f(R1 ^ K4); |
| 274 |
L2 ^= camellia_f(R2 ^ K4); |
| 275 |
R1 ^= camellia_f(L1 ^ K3); |
| 276 |
R2 ^= camellia_f(L2 ^ K3); |
| 277 |
L1 ^= camellia_f(R1 ^ K2); |
| 278 |
L2 ^= camellia_f(R2 ^ K2); |
| 279 |
R1 ^= camellia_f(L1 ^ K1); |
| 280 |
R2 ^= camellia_f(L2 ^ K1); |
| 281 |
L1 ^= camellia_f(R1 ^ K0); |
| 282 |
L2 ^= camellia_f(R2 ^ K0); |
| 283 |
} |
| 284 |
|
| 285 |
BOTAN_FN_ISA_AVX512_GFNI |
| 286 |
void camellia_encrypt_x16_18r(const uint8_t in[], uint8_t out[], std::span<const uint64_t> SK) { |
| 287 |
SIMD_8x64 L1; |
| 288 |
SIMD_8x64 R1; |
| 289 |
SIMD_8x64 L2; |
| 290 |
SIMD_8x64 R2; |
| 291 |
load_and_deinterleave(in, L1, R1); |
| 292 |
load_and_deinterleave(in + 128, L2, R2); |
| 293 |
|
| 294 |
const auto K0 = SIMD_8x64::splat(SK[0]); |
| 295 |
const auto K1 = SIMD_8x64::splat(SK[1]); |
| 296 |
L1 ^= K0; |
| 297 |
L2 ^= K0; |
| 298 |
R1 ^= K1; |
| 299 |
R2 ^= K1; |
| 300 |
|
| 301 |
six_e_rounds_x2(L1, R1, L2, R2, SK.subspan(2)); |
| 302 |
|
| 303 |
L1 = FL_8(L1, SK[8]); |
| 304 |
L2 = FL_8(L2, SK[8]); |
| 305 |
R1 = FLINV_8(R1, SK[9]); |
| 306 |
R2 = FLINV_8(R2, SK[9]); |
| 307 |
|
| 308 |
six_e_rounds_x2(L1, R1, L2, R2, SK.subspan(10)); |
| 309 |
|
| 310 |
L1 = FL_8(L1, SK[16]); |
| 311 |
L2 = FL_8(L2, SK[16]); |
| 312 |
R1 = FLINV_8(R1, SK[17]); |
| 313 |
R2 = FLINV_8(R2, SK[17]); |
| 314 |
|
| 315 |
six_e_rounds_x2(L1, R1, L2, R2, SK.subspan(18)); |
| 316 |
|
| 317 |
const auto K24 = SIMD_8x64::splat(SK[24]); |
| 318 |
const auto K25 = SIMD_8x64::splat(SK[25]); |
| 319 |
R1 ^= K24; |
| 320 |
R2 ^= K24; |
| 321 |
L1 ^= K25; |
| 322 |
L2 ^= K25; |
| 323 |
|
| 324 |
interleave_and_store(out, L1, R1); |
| 325 |
interleave_and_store(out + 128, L2, R2); |
| 326 |
} |
| 327 |
|
| 328 |
BOTAN_FN_ISA_AVX512_GFNI |
| 329 |
void camellia_decrypt_x16_18r(const uint8_t in[], uint8_t out[], std::span<const uint64_t> SK) { |
| 330 |
SIMD_8x64 L1; |
| 331 |
SIMD_8x64 R1; |
| 332 |
SIMD_8x64 L2; |
| 333 |
SIMD_8x64 R2; |
| 334 |
load_and_deinterleave(in, L1, R1); |
| 335 |
load_and_deinterleave(in + 128, L2, R2); |
| 336 |
|
| 337 |
const auto K25 = SIMD_8x64::splat(SK[25]); |
| 338 |
const auto K24 = SIMD_8x64::splat(SK[24]); |
| 339 |
R1 ^= K25; |
| 340 |
R2 ^= K25; |
| 341 |
L1 ^= K24; |
| 342 |
L2 ^= K24; |
| 343 |
|
| 344 |
six_d_rounds_x2(L1, R1, L2, R2, SK.subspan(18)); |
| 345 |
|
| 346 |
L1 = FL_8(L1, SK[17]); |
| 347 |
L2 = FL_8(L2, SK[17]); |
| 348 |
R1 = FLINV_8(R1, SK[16]); |
| 349 |
R2 = FLINV_8(R2, SK[16]); |
| 350 |
|
| 351 |
six_d_rounds_x2(L1, R1, L2, R2, SK.subspan(10)); |
| 352 |
|
| 353 |
L1 = FL_8(L1, SK[9]); |
| 354 |
L2 = FL_8(L2, SK[9]); |
| 355 |
R1 = FLINV_8(R1, SK[8]); |
| 356 |
R2 = FLINV_8(R2, SK[8]); |
| 357 |
|
| 358 |
six_d_rounds_x2(L1, R1, L2, R2, SK.subspan(2)); |
| 359 |
|
| 360 |
const auto K1 = SIMD_8x64::splat(SK[1]); |
| 361 |
const auto K0 = SIMD_8x64::splat(SK[0]); |
| 362 |
L1 ^= K1; |
| 363 |
L2 ^= K1; |
| 364 |
R1 ^= K0; |
| 365 |
R2 ^= K0; |
| 366 |
|
| 367 |
interleave_and_store(out, L1, R1); |
| 368 |
interleave_and_store(out + 128, L2, R2); |
| 369 |
} |
| 370 |
|
| 371 |
BOTAN_FN_ISA_AVX512_GFNI |
| 372 |
void camellia_encrypt_x16_24r(const uint8_t in[], uint8_t out[], std::span<const uint64_t> SK) { |
| 373 |
SIMD_8x64 L1; |
| 374 |
SIMD_8x64 R1; |
| 375 |
SIMD_8x64 L2; |
| 376 |
SIMD_8x64 R2; |
| 377 |
load_and_deinterleave(in, L1, R1); |
| 378 |
load_and_deinterleave(in + 128, L2, R2); |
| 379 |
|
| 380 |
const auto K0 = SIMD_8x64::splat(SK[0]); |
| 381 |
const auto K1 = SIMD_8x64::splat(SK[1]); |
| 382 |
L1 ^= K0; |
| 383 |
L2 ^= K0; |
| 384 |
R1 ^= K1; |
| 385 |
R2 ^= K1; |
| 386 |
|
| 387 |
six_e_rounds_x2(L1, R1, L2, R2, SK.subspan(2)); |
| 388 |
|
| 389 |
L1 = FL_8(L1, SK[8]); |
| 390 |
L2 = FL_8(L2, SK[8]); |
| 391 |
R1 = FLINV_8(R1, SK[9]); |
| 392 |
R2 = FLINV_8(R2, SK[9]); |
| 393 |
|
| 394 |
six_e_rounds_x2(L1, R1, L2, R2, SK.subspan(10)); |
| 395 |
|
| 396 |
L1 = FL_8(L1, SK[16]); |
| 397 |
L2 = FL_8(L2, SK[16]); |
| 398 |
R1 = FLINV_8(R1, SK[17]); |
| 399 |
R2 = FLINV_8(R2, SK[17]); |
| 400 |
|
| 401 |
six_e_rounds_x2(L1, R1, L2, R2, SK.subspan(18)); |
| 402 |
|
| 403 |
L1 = FL_8(L1, SK[24]); |
| 404 |
L2 = FL_8(L2, SK[24]); |
| 405 |
R1 = FLINV_8(R1, SK[25]); |
| 406 |
R2 = FLINV_8(R2, SK[25]); |
| 407 |
|
| 408 |
six_e_rounds_x2(L1, R1, L2, R2, SK.subspan(26)); |
| 409 |
|
| 410 |
const auto K32 = SIMD_8x64::splat(SK[32]); |
| 411 |
const auto K33 = SIMD_8x64::splat(SK[33]); |
| 412 |
R1 ^= K32; |
| 413 |
R2 ^= K32; |
| 414 |
L1 ^= K33; |
| 415 |
L2 ^= K33; |
| 416 |
|
| 417 |
interleave_and_store(out, L1, R1); |
| 418 |
interleave_and_store(out + 128, L2, R2); |
| 419 |
} |
| 420 |
|
| 421 |
BOTAN_FN_ISA_AVX512_GFNI |
| 422 |
void camellia_decrypt_x16_24r(const uint8_t in[], uint8_t out[], std::span<const uint64_t> SK) { |
| 423 |
SIMD_8x64 L1; |
| 424 |
SIMD_8x64 R1; |
| 425 |
SIMD_8x64 L2; |
| 426 |
SIMD_8x64 R2; |
| 427 |
load_and_deinterleave(in, L1, R1); |
| 428 |
load_and_deinterleave(in + 128, L2, R2); |
| 429 |
|
| 430 |
const auto K33 = SIMD_8x64::splat(SK[33]); |
| 431 |
const auto K32 = SIMD_8x64::splat(SK[32]); |
| 432 |
R1 ^= K33; |
| 433 |
R2 ^= K33; |
| 434 |
L1 ^= K32; |
| 435 |
L2 ^= K32; |
| 436 |
|
| 437 |
six_d_rounds_x2(L1, R1, L2, R2, SK.subspan(26)); |
| 438 |
|
| 439 |
L1 = FL_8(L1, SK[25]); |
| 440 |
L2 = FL_8(L2, SK[25]); |
| 441 |
R1 = FLINV_8(R1, SK[24]); |
| 442 |
R2 = FLINV_8(R2, SK[24]); |
| 443 |
|
| 444 |
six_d_rounds_x2(L1, R1, L2, R2, SK.subspan(18)); |
| 445 |
|
| 446 |
L1 = FL_8(L1, SK[17]); |
| 447 |
L2 = FL_8(L2, SK[17]); |
| 448 |
R1 = FLINV_8(R1, SK[16]); |
| 449 |
R2 = FLINV_8(R2, SK[16]); |
| 450 |
|
| 451 |
six_d_rounds_x2(L1, R1, L2, R2, SK.subspan(10)); |
| 452 |
|
| 453 |
L1 = FL_8(L1, SK[9]); |
| 454 |
L2 = FL_8(L2, SK[9]); |
| 455 |
R1 = FLINV_8(R1, SK[8]); |
| 456 |
R2 = FLINV_8(R2, SK[8]); |
| 457 |
|
| 458 |
six_d_rounds_x2(L1, R1, L2, R2, SK.subspan(2)); |
| 459 |
|
| 460 |
const auto K1 = SIMD_8x64::splat(SK[1]); |
| 461 |
const auto K0 = SIMD_8x64::splat(SK[0]); |
| 462 |
L1 ^= K1; |
| 463 |
L2 ^= K1; |
| 464 |
R1 ^= K0; |
| 465 |
R2 ^= K0; |
| 466 |
|
| 467 |
interleave_and_store(out, L1, R1); |
| 468 |
interleave_and_store(out + 128, L2, R2); |
| 469 |
} |
| 470 |
|
| 471 |
BOTAN_FN_ISA_AVX512_GFNI |
| 472 |
void camellia_encrypt_x8_18r(const uint8_t in[], uint8_t out[], std::span<const uint64_t> SK) { |
| 473 |
SIMD_8x64 L; |
| 474 |
SIMD_8x64 R; |
| 475 |
load_and_deinterleave(in, L, R); |
| 476 |
|
| 477 |
L ^= SIMD_8x64::splat(SK[0]); |
| 478 |
R ^= SIMD_8x64::splat(SK[1]); |
| 479 |
|
| 480 |
six_e_rounds(L, R, SK.subspan(2)); |
| 481 |
|
| 482 |
L = FL_8(L, SK[8]); |
| 483 |
R = FLINV_8(R, SK[9]); |
| 484 |
|
| 485 |
six_e_rounds(L, R, SK.subspan(10)); |
| 486 |
|
| 487 |
L = FL_8(L, SK[16]); |
| 488 |
R = FLINV_8(R, SK[17]); |
| 489 |
|
| 490 |
six_e_rounds(L, R, SK.subspan(18)); |
| 491 |
|
| 492 |
R ^= SIMD_8x64::splat(SK[24]); |
| 493 |
L ^= SIMD_8x64::splat(SK[25]); |
| 494 |
|
| 495 |
interleave_and_store(out, L, R); |
| 496 |
} |
| 497 |
|
| 498 |
BOTAN_FN_ISA_AVX512_GFNI |
| 499 |
void camellia_decrypt_x8_18r(const uint8_t in[], uint8_t out[], std::span<const uint64_t> SK) { |
| 500 |
SIMD_8x64 L; |
| 501 |
SIMD_8x64 R; |
| 502 |
load_and_deinterleave(in, L, R); |
| 503 |
|
| 504 |
R ^= SIMD_8x64::splat(SK[25]); |
| 505 |
L ^= SIMD_8x64::splat(SK[24]); |
| 506 |
|
| 507 |
six_d_rounds(L, R, SK.subspan(18)); |
| 508 |
|
| 509 |
L = FL_8(L, SK[17]); |
| 510 |
R = FLINV_8(R, SK[16]); |
| 511 |
|
| 512 |
six_d_rounds(L, R, SK.subspan(10)); |
| 513 |
|
| 514 |
L = FL_8(L, SK[9]); |
| 515 |
R = FLINV_8(R, SK[8]); |
| 516 |
|
| 517 |
six_d_rounds(L, R, SK.subspan(2)); |
| 518 |
|
| 519 |
L ^= SIMD_8x64::splat(SK[1]); |
| 520 |
R ^= SIMD_8x64::splat(SK[0]); |
| 521 |
|
| 522 |
interleave_and_store(out, L, R); |
| 523 |
} |
| 524 |
|
| 525 |
BOTAN_FN_ISA_AVX512_GFNI |
| 526 |
void camellia_encrypt_x8_24r(const uint8_t in[], uint8_t out[], std::span<const uint64_t> SK) { |
| 527 |
SIMD_8x64 L; |
| 528 |
SIMD_8x64 R; |
| 529 |
load_and_deinterleave(in, L, R); |
| 530 |
|
| 531 |
L ^= SIMD_8x64::splat(SK[0]); |
| 532 |
R ^= SIMD_8x64::splat(SK[1]); |
| 533 |
|
| 534 |
six_e_rounds(L, R, SK.subspan(2)); |
| 535 |
|
| 536 |
L = FL_8(L, SK[8]); |
| 537 |
R = FLINV_8(R, SK[9]); |
| 538 |
|
| 539 |
six_e_rounds(L, R, SK.subspan(10)); |
| 540 |
|
| 541 |
L = FL_8(L, SK[16]); |
| 542 |
R = FLINV_8(R, SK[17]); |
| 543 |
|
| 544 |
six_e_rounds(L, R, SK.subspan(18)); |
| 545 |
|
| 546 |
L = FL_8(L, SK[24]); |
| 547 |
R = FLINV_8(R, SK[25]); |
| 548 |
|
| 549 |
six_e_rounds(L, R, SK.subspan(26)); |
| 550 |
|
| 551 |
R ^= SIMD_8x64::splat(SK[32]); |
| 552 |
L ^= SIMD_8x64::splat(SK[33]); |
| 553 |
|
| 554 |
interleave_and_store(out, L, R); |
| 555 |
} |
| 556 |
|
| 557 |
BOTAN_FN_ISA_AVX512_GFNI |
| 558 |
void camellia_decrypt_x8_24r(const uint8_t in[], uint8_t out[], std::span<const uint64_t> SK) { |
| 559 |
SIMD_8x64 L; |
| 560 |
SIMD_8x64 R; |
| 561 |
load_and_deinterleave(in, L, R); |
| 562 |
|
| 563 |
R ^= SIMD_8x64::splat(SK[33]); |
| 564 |
L ^= SIMD_8x64::splat(SK[32]); |
| 565 |
|
| 566 |
six_d_rounds(L, R, SK.subspan(26)); |
| 567 |
|
| 568 |
L = FL_8(L, SK[25]); |
| 569 |
R = FLINV_8(R, SK[24]); |
| 570 |
|
| 571 |
six_d_rounds(L, R, SK.subspan(18)); |
| 572 |
|
| 573 |
L = FL_8(L, SK[17]); |
| 574 |
R = FLINV_8(R, SK[16]); |
| 575 |
|
| 576 |
six_d_rounds(L, R, SK.subspan(10)); |
| 577 |
|
| 578 |
L = FL_8(L, SK[9]); |
| 579 |
R = FLINV_8(R, SK[8]); |
| 580 |
|
| 581 |
six_d_rounds(L, R, SK.subspan(2)); |
| 582 |
|
| 583 |
L ^= SIMD_8x64::splat(SK[1]); |
| 584 |
R ^= SIMD_8x64::splat(SK[0]); |
| 585 |
|
| 586 |
interleave_and_store(out, L, R); |
| 587 |
} |
| 588 |
|
| 589 |
} // namespace |
| 590 |
|
| 591 |
} // namespace Camellia_AVX512 |
| 592 |
|
| 593 |
// static |
| 594 |
void BOTAN_FN_ISA_AVX512_GFNI Camellia_128::avx512_gfni_encrypt(const uint8_t in[], |
| 595 |
uint8_t out[], |
| 596 |
size_t blocks, |
| 597 |
std::span<const uint64_t> SK) { |
| 598 |
while(blocks >= 16) { |
| 599 |
Camellia_AVX512::camellia_encrypt_x16_18r(in, out, SK); |
| 600 |
in += 16 * 16; |
| 601 |
out += 16 * 16; |
| 602 |
blocks -= 16; |
| 603 |
} |
| 604 |
|
| 605 |
while(blocks >= 8) { |
| 606 |
Camellia_AVX512::camellia_encrypt_x8_18r(in, out, SK); |
| 607 |
in += 8 * 16; |
| 608 |
out += 8 * 16; |
| 609 |
blocks -= 8; |
| 610 |
} |
| 611 |
|
| 612 |
if(blocks > 0) { |
| 613 |
uint8_t ibuf[8 * 16] = {0}; |
| 614 |
uint8_t obuf[8 * 16] = {0}; |
| 615 |
copy_mem(ibuf, in, blocks * 16); |
| 616 |
Camellia_AVX512::camellia_encrypt_x8_18r(ibuf, obuf, SK); |
| 617 |
copy_mem(out, obuf, blocks * 16); |
| 618 |
} |
| 619 |
} |
| 620 |
|
| 621 |
// static |
| 622 |
void BOTAN_FN_ISA_AVX512_GFNI Camellia_128::avx512_gfni_decrypt(const uint8_t in[], |
| 623 |
uint8_t out[], |
| 624 |
size_t blocks, |
| 625 |
std::span<const uint64_t> SK) { |
| 626 |
while(blocks >= 16) { |
| 627 |
Camellia_AVX512::camellia_decrypt_x16_18r(in, out, SK); |
| 628 |
in += 16 * 16; |
| 629 |
out += 16 * 16; |
| 630 |
blocks -= 16; |
| 631 |
} |
| 632 |
|
| 633 |
while(blocks >= 8) { |
| 634 |
Camellia_AVX512::camellia_decrypt_x8_18r(in, out, SK); |
| 635 |
in += 8 * 16; |
| 636 |
out += 8 * 16; |
| 637 |
blocks -= 8; |
| 638 |
} |
| 639 |
|
| 640 |
if(blocks > 0) { |
| 641 |
uint8_t ibuf[8 * 16] = {0}; |
| 642 |
uint8_t obuf[8 * 16] = {0}; |
| 643 |
copy_mem(ibuf, in, blocks * 16); |
| 644 |
Camellia_AVX512::camellia_decrypt_x8_18r(ibuf, obuf, SK); |
| 645 |
copy_mem(out, obuf, blocks * 16); |
| 646 |
} |
| 647 |
} |
| 648 |
|
| 649 |
// static |
| 650 |
void BOTAN_FN_ISA_AVX512_GFNI Camellia_192::avx512_gfni_encrypt(const uint8_t in[], |
| 651 |
uint8_t out[], |
| 652 |
size_t blocks, |
| 653 |
std::span<const uint64_t> SK) { |
| 654 |
while(blocks >= 16) { |
| 655 |
Camellia_AVX512::camellia_encrypt_x16_24r(in, out, SK); |
| 656 |
in += 16 * 16; |
| 657 |
out += 16 * 16; |
| 658 |
blocks -= 16; |
| 659 |
} |
| 660 |
|
| 661 |
while(blocks >= 8) { |
| 662 |
Camellia_AVX512::camellia_encrypt_x8_24r(in, out, SK); |
| 663 |
in += 8 * 16; |
| 664 |
out += 8 * 16; |
| 665 |
blocks -= 8; |
| 666 |
} |
| 667 |
|
| 668 |
if(blocks > 0) { |
| 669 |
uint8_t ibuf[8 * 16] = {0}; |
| 670 |
uint8_t obuf[8 * 16] = {0}; |
| 671 |
copy_mem(ibuf, in, blocks * 16); |
| 672 |
Camellia_AVX512::camellia_encrypt_x8_24r(ibuf, obuf, SK); |
| 673 |
copy_mem(out, obuf, blocks * 16); |
| 674 |
} |
| 675 |
} |
| 676 |
|
| 677 |
// static |
| 678 |
void BOTAN_FN_ISA_AVX512_GFNI Camellia_192::avx512_gfni_decrypt(const uint8_t in[], |
| 679 |
uint8_t out[], |
| 680 |
size_t blocks, |
| 681 |
std::span<const uint64_t> SK) { |
| 682 |
while(blocks >= 16) { |
| 683 |
Camellia_AVX512::camellia_decrypt_x16_24r(in, out, SK); |
| 684 |
in += 16 * 16; |
| 685 |
out += 16 * 16; |
| 686 |
blocks -= 16; |
| 687 |
} |
| 688 |
|
| 689 |
while(blocks >= 8) { |
| 690 |
Camellia_AVX512::camellia_decrypt_x8_24r(in, out, SK); |
| 691 |
in += 8 * 16; |
| 692 |
out += 8 * 16; |
| 693 |
blocks -= 8; |
| 694 |
} |
| 695 |
|
| 696 |
if(blocks > 0) { |
| 697 |
uint8_t ibuf[8 * 16] = {0}; |
| 698 |
uint8_t obuf[8 * 16] = {0}; |
| 699 |
copy_mem(ibuf, in, blocks * 16); |
| 700 |
Camellia_AVX512::camellia_decrypt_x8_24r(ibuf, obuf, SK); |
| 701 |
copy_mem(out, obuf, blocks * 16); |
| 702 |
} |
| 703 |
} |
| 704 |
|
| 705 |
// static |
| 706 |
void BOTAN_FN_ISA_AVX512_GFNI Camellia_256::avx512_gfni_encrypt(const uint8_t in[], |
| 707 |
uint8_t out[], |
| 708 |
size_t blocks, |
| 709 |
std::span<const uint64_t> SK) { |
| 710 |
while(blocks >= 16) { |
| 711 |
Camellia_AVX512::camellia_encrypt_x16_24r(in, out, SK); |
| 712 |
in += 16 * 16; |
| 713 |
out += 16 * 16; |
| 714 |
blocks -= 16; |
| 715 |
} |
| 716 |
|
| 717 |
while(blocks >= 8) { |
| 718 |
Camellia_AVX512::camellia_encrypt_x8_24r(in, out, SK); |
| 719 |
in += 8 * 16; |
| 720 |
out += 8 * 16; |
| 721 |
blocks -= 8; |
| 722 |
} |
| 723 |
|
| 724 |
if(blocks > 0) { |
| 725 |
uint8_t ibuf[8 * 16] = {0}; |
| 726 |
uint8_t obuf[8 * 16] = {0}; |
| 727 |
copy_mem(ibuf, in, blocks * 16); |
| 728 |
Camellia_AVX512::camellia_encrypt_x8_24r(ibuf, obuf, SK); |
| 729 |
copy_mem(out, obuf, blocks * 16); |
| 730 |
} |
| 731 |
} |
| 732 |
|
| 733 |
// static |
| 734 |
void BOTAN_FN_ISA_AVX512_GFNI Camellia_256::avx512_gfni_decrypt(const uint8_t in[], |
| 735 |
uint8_t out[], |
| 736 |
size_t blocks, |
| 737 |
std::span<const uint64_t> SK) { |
| 738 |
while(blocks >= 16) { |
| 739 |
Camellia_AVX512::camellia_decrypt_x16_24r(in, out, SK); |
| 740 |
in += 16 * 16; |
| 741 |
out += 16 * 16; |
| 742 |
blocks -= 16; |
| 743 |
} |
| 744 |
|
| 745 |
while(blocks >= 8) { |
| 746 |
Camellia_AVX512::camellia_decrypt_x8_24r(in, out, SK); |
| 747 |
in += 8 * 16; |
| 748 |
out += 8 * 16; |
| 749 |
blocks -= 8; |
| 750 |
} |
| 751 |
|
| 752 |
if(blocks > 0) { |
| 753 |
uint8_t ibuf[8 * 16] = {0}; |
| 754 |
uint8_t obuf[8 * 16] = {0}; |
| 755 |
copy_mem(ibuf, in, blocks * 16); |
| 756 |
Camellia_AVX512::camellia_decrypt_x8_24r(ibuf, obuf, SK); |
| 757 |
copy_mem(out, obuf, blocks * 16); |
| 758 |
} |
| 759 |
} |
| 760 |
|
| 761 |
} // namespace Botan |