21 #if (CRYPTOPP_SSSE3_AVAILABLE) 22 # include <pmmintrin.h> 23 # include <tmmintrin.h> 26 #if (CRYPTOPP_SSE41_AVAILABLE) 27 # include <smmintrin.h> 30 #if defined(__AVX512F__) && defined(__AVX512VL__) 31 # define CRYPTOPP_AVX512_ROTATE 1 32 # include <immintrin.h> 35 #if (CRYPTOPP_ARM_NEON_AVAILABLE) 36 # include <arm_neon.h> 41 #if defined(CRYPTOPP_ARM_ACLE_AVAILABLE) 43 # include <arm_acle.h> 48 #if (CRYPTOPP_GCC_VERSION >= 40900) 49 # define GCC_NO_UBSAN __attribute__ ((no_sanitize_undefined)) 54 ANONYMOUS_NAMESPACE_BEGIN
57 using CryptoPP::word32;
58 using CryptoPP::word64;
65 #if (CRYPTOPP_ARM_NEON_AVAILABLE) 68 inline T UnpackHigh32(
const T& a,
const T& b)
70 const uint32x2_t x(vget_high_u32((uint32x4_t)a));
71 const uint32x2_t y(vget_high_u32((uint32x4_t)b));
72 const uint32x2x2_t r = vzip_u32(x, y);
73 return (T)vcombine_u32(r.val[0], r.val[1]);
77 inline T UnpackLow32(
const T& a,
const T& b)
79 const uint32x2_t x(vget_low_u32((uint32x4_t)a));
80 const uint32x2_t y(vget_low_u32((uint32x4_t)b));
81 const uint32x2x2_t r = vzip_u32(x, y);
82 return (T)vcombine_u32(r.val[0], r.val[1]);
85 template <
unsigned int R>
86 inline uint32x4_t RotateLeft32(
const uint32x4_t& val)
88 const uint32x4_t a(vshlq_n_u32(val, R));
89 const uint32x4_t b(vshrq_n_u32(val, 32 - R));
90 return vorrq_u32(a, b);
93 template <
unsigned int R>
94 inline uint32x4_t RotateRight32(
const uint32x4_t& val)
96 const uint32x4_t a(vshlq_n_u32(val, 32 - R));
97 const uint32x4_t b(vshrq_n_u32(val, R));
98 return vorrq_u32(a, b);
101 #if defined(__aarch32__) || defined(__aarch64__) 104 inline uint32x4_t RotateLeft32<8>(
const uint32x4_t& val)
106 #if defined(CRYPTOPP_BIG_ENDIAN) 107 const uint8_t maskb[16] = { 14,13,12,15, 10,9,8,11, 6,5,4,7, 2,1,0,3 };
108 const uint8x16_t mask = vld1q_u8(maskb);
110 const uint8_t maskb[16] = { 3,0,1,2, 7,4,5,6, 11,8,9,10, 15,12,13,14 };
111 const uint8x16_t mask = vld1q_u8(maskb);
114 return vreinterpretq_u32_u8(
115 vqtbl1q_u8(vreinterpretq_u8_u32(val), mask));
120 inline uint32x4_t RotateRight32<8>(
const uint32x4_t& val)
122 #if defined(CRYPTOPP_BIG_ENDIAN) 123 const uint8_t maskb[16] = { 12,15,14,13, 8,11,10,9, 4,7,6,5, 0,3,2,1 };
124 const uint8x16_t mask = vld1q_u8(maskb);
126 const uint8_t maskb[16] = { 1,2,3,0, 5,6,7,4, 9,10,11,8, 13,14,14,12 };
127 const uint8x16_t mask = vld1q_u8(maskb);
130 return vreinterpretq_u32_u8(
131 vqtbl1q_u8(vreinterpretq_u8_u32(val), mask));
135 inline uint32x4_t SIMON64_f(
const uint32x4_t& val)
137 return veorq_u32(RotateLeft32<2>(val),
138 vandq_u32(RotateLeft32<1>(val), RotateLeft32<8>(val)));
141 inline void SIMON64_Enc_Block(uint32x4_t &block1, uint32x4_t &block0,
142 const word32 *subkeys,
unsigned int rounds)
149 uint32x4_t x1 = vuzpq_u32(block0, block1).val[1];
150 uint32x4_t y1 = vuzpq_u32(block0, block1).val[0];
152 for (
int i = 0; i < static_cast<int>(rounds & ~1)-1; i += 2)
154 const uint32x4_t rk1 = vld1q_dup_u32(subkeys+i);
155 y1 = veorq_u32(veorq_u32(y1, SIMON64_f(x1)), rk1);
157 const uint32x4_t rk2 = vld1q_dup_u32(subkeys+i+1);
158 x1 = veorq_u32(veorq_u32(x1, SIMON64_f(y1)), rk2);
163 const uint32x4_t rk = vld1q_dup_u32(subkeys+rounds-1);
165 y1 = veorq_u32(veorq_u32(y1, SIMON64_f(x1)), rk);
170 block0 = UnpackLow32(y1, x1);
171 block1 = UnpackHigh32(y1, x1);
174 inline void SIMON64_Dec_Block(uint32x4_t &block0, uint32x4_t &block1,
175 const word32 *subkeys,
unsigned int rounds)
182 uint32x4_t x1 = vuzpq_u32(block0, block1).val[1];
183 uint32x4_t y1 = vuzpq_u32(block0, block1).val[0];
188 const uint32x4_t rk = vld1q_dup_u32(subkeys + rounds - 1);
190 y1 = veorq_u32(veorq_u32(y1, rk), SIMON64_f(x1));
194 for (
int i = static_cast<int>(rounds-2); i >= 0; i -= 2)
196 const uint32x4_t rk1 = vld1q_dup_u32(subkeys+i+1);
197 x1 = veorq_u32(veorq_u32(x1, SIMON64_f(y1)), rk1);
199 const uint32x4_t rk2 = vld1q_dup_u32(subkeys+i);
200 y1 = veorq_u32(veorq_u32(y1, SIMON64_f(x1)), rk2);
204 block0 = UnpackLow32(y1, x1);
205 block1 = UnpackHigh32(y1, x1);
208 inline void SIMON64_Enc_6_Blocks(uint32x4_t &block0, uint32x4_t &block1,
209 uint32x4_t &block2, uint32x4_t &block3, uint32x4_t &block4, uint32x4_t &block5,
210 const word32 *subkeys,
unsigned int rounds)
217 uint32x4_t x1 = vuzpq_u32(block0, block1).val[1];
218 uint32x4_t y1 = vuzpq_u32(block0, block1).val[0];
219 uint32x4_t x2 = vuzpq_u32(block2, block3).val[1];
220 uint32x4_t y2 = vuzpq_u32(block2, block3).val[0];
221 uint32x4_t x3 = vuzpq_u32(block4, block5).val[1];
222 uint32x4_t y3 = vuzpq_u32(block4, block5).val[0];
224 for (
int i = 0; i < static_cast<int>(rounds & ~1) - 1; i += 2)
226 const uint32x4_t rk1 = vld1q_dup_u32(subkeys+i);
227 y1 = veorq_u32(veorq_u32(y1, SIMON64_f(x1)), rk1);
228 y2 = veorq_u32(veorq_u32(y2, SIMON64_f(x2)), rk1);
229 y3 = veorq_u32(veorq_u32(y3, SIMON64_f(x3)), rk1);
231 const uint32x4_t rk2 = vld1q_dup_u32(subkeys+i+1);
232 x1 = veorq_u32(veorq_u32(x1, SIMON64_f(y1)), rk2);
233 x2 = veorq_u32(veorq_u32(x2, SIMON64_f(y2)), rk2);
234 x3 = veorq_u32(veorq_u32(x3, SIMON64_f(y3)), rk2);
239 const uint32x4_t rk = vld1q_dup_u32(subkeys + rounds - 1);
241 y1 = veorq_u32(veorq_u32(y1, SIMON64_f(x1)), rk);
242 y2 = veorq_u32(veorq_u32(y2, SIMON64_f(x2)), rk);
243 y3 = veorq_u32(veorq_u32(y3, SIMON64_f(x3)), rk);
244 std::swap(x1, y1); std::swap(x2, y2); std::swap(x3, y3);
248 block0 = UnpackLow32(y1, x1);
249 block1 = UnpackHigh32(y1, x1);
250 block2 = UnpackLow32(y2, x2);
251 block3 = UnpackHigh32(y2, x2);
252 block4 = UnpackLow32(y3, x3);
253 block5 = UnpackHigh32(y3, x3);
256 inline void SIMON64_Dec_6_Blocks(uint32x4_t &block0, uint32x4_t &block1,
257 uint32x4_t &block2, uint32x4_t &block3, uint32x4_t &block4, uint32x4_t &block5,
258 const word32 *subkeys,
unsigned int rounds)
265 uint32x4_t x1 = vuzpq_u32(block0, block1).val[1];
266 uint32x4_t y1 = vuzpq_u32(block0, block1).val[0];
267 uint32x4_t x2 = vuzpq_u32(block2, block3).val[1];
268 uint32x4_t y2 = vuzpq_u32(block2, block3).val[0];
269 uint32x4_t x3 = vuzpq_u32(block4, block5).val[1];
270 uint32x4_t y3 = vuzpq_u32(block4, block5).val[0];
274 std::swap(x1, y1); std::swap(x2, y2); std::swap(x3, y3);
275 const uint32x4_t rk = vld1q_dup_u32(subkeys + rounds - 1);
277 y1 = veorq_u32(veorq_u32(y1, rk), SIMON64_f(x1));
278 y2 = veorq_u32(veorq_u32(y2, rk), SIMON64_f(x2));
279 y3 = veorq_u32(veorq_u32(y3, rk), SIMON64_f(x3));
283 for (
int i = static_cast<int>(rounds-2); i >= 0; i -= 2)
285 const uint32x4_t rk1 = vld1q_dup_u32(subkeys + i + 1);
286 x1 = veorq_u32(veorq_u32(x1, SIMON64_f(y1)), rk1);
287 x2 = veorq_u32(veorq_u32(x2, SIMON64_f(y2)), rk1);
288 x3 = veorq_u32(veorq_u32(x3, SIMON64_f(y3)), rk1);
290 const uint32x4_t rk2 = vld1q_dup_u32(subkeys + i);
291 y1 = veorq_u32(veorq_u32(y1, SIMON64_f(x1)), rk2);
292 y2 = veorq_u32(veorq_u32(y2, SIMON64_f(x2)), rk2);
293 y3 = veorq_u32(veorq_u32(y3, SIMON64_f(x3)), rk2);
297 block0 = UnpackLow32(y1, x1);
298 block1 = UnpackHigh32(y1, x1);
299 block2 = UnpackLow32(y2, x2);
300 block3 = UnpackHigh32(y2, x2);
301 block4 = UnpackLow32(y3, x3);
302 block5 = UnpackHigh32(y3, x3);
305 #endif // CRYPTOPP_ARM_NEON_AVAILABLE 307 #if (CRYPTOPP_ARM_NEON_AVAILABLE) 310 inline T UnpackHigh64(
const T& a,
const T& b)
312 const uint64x1_t x(vget_high_u64((uint64x2_t)a));
313 const uint64x1_t y(vget_high_u64((uint64x2_t)b));
314 return (T)vcombine_u64(x, y);
318 inline T UnpackLow64(
const T& a,
const T& b)
320 const uint64x1_t x(vget_low_u64((uint64x2_t)a));
321 const uint64x1_t y(vget_low_u64((uint64x2_t)b));
322 return (T)vcombine_u64(x, y);
325 template <
unsigned int R>
326 inline uint64x2_t RotateLeft64(
const uint64x2_t& val)
328 const uint64x2_t a(vshlq_n_u64(val, R));
329 const uint64x2_t b(vshrq_n_u64(val, 64 - R));
330 return vorrq_u64(a, b);
333 template <
unsigned int R>
334 inline uint64x2_t RotateRight64(
const uint64x2_t& val)
336 const uint64x2_t a(vshlq_n_u64(val, 64 - R));
337 const uint64x2_t b(vshrq_n_u64(val, R));
338 return vorrq_u64(a, b);
341 #if defined(__aarch32__) || defined(__aarch64__) 344 inline uint64x2_t RotateLeft64<8>(
const uint64x2_t& val)
346 #if defined(CRYPTOPP_BIG_ENDIAN) 347 const uint8_t maskb[16] = { 14,13,12,11, 10,9,8,15, 6,5,4,3, 2,1,0,7 };
348 const uint8x16_t mask = vld1q_u8(maskb);
350 const uint8_t maskb[16] = { 7,0,1,2, 3,4,5,6, 15,8,9,10, 11,12,13,14 };
351 const uint8x16_t mask = vld1q_u8(maskb);
354 return vreinterpretq_u64_u8(
355 vqtbl1q_u8(vreinterpretq_u8_u64(val), mask));
360 inline uint64x2_t RotateRight64<8>(
const uint64x2_t& val)
362 #if defined(CRYPTOPP_BIG_ENDIAN) 363 const uint8_t maskb[16] = { 8,15,14,13, 12,11,10,9, 0,7,6,5, 4,3,2,1 };
364 const uint8x16_t mask = vld1q_u8(maskb);
366 const uint8_t maskb[16] = { 1,2,3,4, 5,6,7,0, 9,10,11,12, 13,14,15,8 };
367 const uint8x16_t mask = vld1q_u8(maskb);
370 return vreinterpretq_u64_u8(
371 vqtbl1q_u8(vreinterpretq_u8_u64(val), mask));
375 inline uint64x2_t SIMON128_f(
const uint64x2_t& val)
377 return veorq_u64(RotateLeft64<2>(val),
378 vandq_u64(RotateLeft64<1>(val), RotateLeft64<8>(val)));
381 inline void SIMON128_Enc_Block(uint64x2_t &block0, uint64x2_t &block1,
382 const word64 *subkeys,
unsigned int rounds)
388 uint64x2_t x1 = UnpackHigh64(block0, block1);
389 uint64x2_t y1 = UnpackLow64(block0, block1);
391 for (
int i = 0; i < static_cast<int>(rounds & ~1)-1; i += 2)
393 const uint64x2_t rk1 = vld1q_dup_u64(subkeys+i);
394 y1 = veorq_u64(veorq_u64(y1, SIMON128_f(x1)), rk1);
396 const uint64x2_t rk2 = vld1q_dup_u64(subkeys+i+1);
397 x1 = veorq_u64(veorq_u64(x1, SIMON128_f(y1)), rk2);
402 const uint64x2_t rk = vld1q_dup_u64(subkeys+rounds-1);
404 y1 = veorq_u64(veorq_u64(y1, SIMON128_f(x1)), rk);
409 block0 = UnpackLow64(y1, x1);
410 block1 = UnpackHigh64(y1, x1);
413 inline void SIMON128_Enc_6_Blocks(uint64x2_t &block0, uint64x2_t &block1,
414 uint64x2_t &block2, uint64x2_t &block3, uint64x2_t &block4, uint64x2_t &block5,
415 const word64 *subkeys,
unsigned int rounds)
421 uint64x2_t x1 = UnpackHigh64(block0, block1);
422 uint64x2_t y1 = UnpackLow64(block0, block1);
423 uint64x2_t x2 = UnpackHigh64(block2, block3);
424 uint64x2_t y2 = UnpackLow64(block2, block3);
425 uint64x2_t x3 = UnpackHigh64(block4, block5);
426 uint64x2_t y3 = UnpackLow64(block4, block5);
428 for (
int i = 0; i < static_cast<int>(rounds & ~1) - 1; i += 2)
430 const uint64x2_t rk1 = vld1q_dup_u64(subkeys+i);
431 y1 = veorq_u64(veorq_u64(y1, SIMON128_f(x1)), rk1);
432 y2 = veorq_u64(veorq_u64(y2, SIMON128_f(x2)), rk1);
433 y3 = veorq_u64(veorq_u64(y3, SIMON128_f(x3)), rk1);
435 const uint64x2_t rk2 = vld1q_dup_u64(subkeys+i+1);
436 x1 = veorq_u64(veorq_u64(x1, SIMON128_f(y1)), rk2);
437 x2 = veorq_u64(veorq_u64(x2, SIMON128_f(y2)), rk2);
438 x3 = veorq_u64(veorq_u64(x3, SIMON128_f(y3)), rk2);
443 const uint64x2_t rk = vld1q_dup_u64(subkeys + rounds - 1);
445 y1 = veorq_u64(veorq_u64(y1, SIMON128_f(x1)), rk);
446 y2 = veorq_u64(veorq_u64(y2, SIMON128_f(x2)), rk);
447 y3 = veorq_u64(veorq_u64(y3, SIMON128_f(x3)), rk);
448 std::swap(x1, y1); std::swap(x2, y2); std::swap(x3, y3);
452 block0 = UnpackLow64(y1, x1);
453 block1 = UnpackHigh64(y1, x1);
454 block2 = UnpackLow64(y2, x2);
455 block3 = UnpackHigh64(y2, x2);
456 block4 = UnpackLow64(y3, x3);
457 block5 = UnpackHigh64(y3, x3);
460 inline void SIMON128_Dec_Block(uint64x2_t &block0, uint64x2_t &block1,
461 const word64 *subkeys,
unsigned int rounds)
467 uint64x2_t x1 = UnpackHigh64(block0, block1);
468 uint64x2_t y1 = UnpackLow64(block0, block1);
473 const uint64x2_t rk = vld1q_dup_u64(subkeys + rounds - 1);
475 y1 = veorq_u64(veorq_u64(y1, rk), SIMON128_f(x1));
479 for (
int i = static_cast<int>(rounds-2); i >= 0; i -= 2)
481 const uint64x2_t rk1 = vld1q_dup_u64(subkeys+i+1);
482 x1 = veorq_u64(veorq_u64(x1, SIMON128_f(y1)), rk1);
484 const uint64x2_t rk2 = vld1q_dup_u64(subkeys+i);
485 y1 = veorq_u64(veorq_u64(y1, SIMON128_f(x1)), rk2);
489 block0 = UnpackLow64(y1, x1);
490 block1 = UnpackHigh64(y1, x1);
493 inline void SIMON128_Dec_6_Blocks(uint64x2_t &block0, uint64x2_t &block1,
494 uint64x2_t &block2, uint64x2_t &block3, uint64x2_t &block4, uint64x2_t &block5,
495 const word64 *subkeys,
unsigned int rounds)
501 uint64x2_t x1 = UnpackHigh64(block0, block1);
502 uint64x2_t y1 = UnpackLow64(block0, block1);
503 uint64x2_t x2 = UnpackHigh64(block2, block3);
504 uint64x2_t y2 = UnpackLow64(block2, block3);
505 uint64x2_t x3 = UnpackHigh64(block4, block5);
506 uint64x2_t y3 = UnpackLow64(block4, block5);
510 std::swap(x1, y1); std::swap(x2, y2); std::swap(x3, y3);
511 const uint64x2_t rk = vld1q_dup_u64(subkeys + rounds - 1);
513 y1 = veorq_u64(veorq_u64(y1, rk), SIMON128_f(x1));
514 y2 = veorq_u64(veorq_u64(y2, rk), SIMON128_f(x2));
515 y3 = veorq_u64(veorq_u64(y3, rk), SIMON128_f(x3));
519 for (
int i = static_cast<int>(rounds-2); i >= 0; i -= 2)
521 const uint64x2_t rk1 = vld1q_dup_u64(subkeys + i + 1);
522 x1 = veorq_u64(veorq_u64(x1, SIMON128_f(y1)), rk1);
523 x2 = veorq_u64(veorq_u64(x2, SIMON128_f(y2)), rk1);
524 x3 = veorq_u64(veorq_u64(x3, SIMON128_f(y3)), rk1);
526 const uint64x2_t rk2 = vld1q_dup_u64(subkeys + i);
527 y1 = veorq_u64(veorq_u64(y1, SIMON128_f(x1)), rk2);
528 y2 = veorq_u64(veorq_u64(y2, SIMON128_f(x2)), rk2);
529 y3 = veorq_u64(veorq_u64(y3, SIMON128_f(x3)), rk2);
533 block0 = UnpackLow64(y1, x1);
534 block1 = UnpackHigh64(y1, x1);
535 block2 = UnpackLow64(y2, x2);
536 block3 = UnpackHigh64(y2, x2);
537 block4 = UnpackLow64(y3, x3);
538 block5 = UnpackHigh64(y3, x3);
541 #endif // CRYPTOPP_ARM_NEON_AVAILABLE 545 #if defined(CRYPTOPP_SSSE3_AVAILABLE) 549 # define M128_CAST(x) ((__m128i *)(void *)(x)) 551 #ifndef CONST_M128_CAST 552 # define CONST_M128_CAST(x) ((const __m128i *)(const void *)(x)) 557 # define DOUBLE_CAST(x) ((double *)(void *)(x)) 559 #ifndef CONST_DOUBLE_CAST 560 # define CONST_DOUBLE_CAST(x) ((const double *)(const void *)(x)) 563 inline void Swap128(__m128i& a,__m128i& b)
565 #if defined(__SUNPRO_CC) && (__SUNPRO_CC <= 0x5120) 574 #if defined(CRYPTOPP_AVX512_ROTATE) 575 template <
unsigned int R>
576 inline __m128i RotateLeft64(
const __m128i& val)
578 return _mm_rol_epi64(val, R);
581 template <
unsigned int R>
582 inline __m128i RotateRight64(
const __m128i& val)
584 return _mm_ror_epi64(val, R);
587 template <
unsigned int R>
588 inline __m128i RotateLeft64(
const __m128i& val)
591 _mm_slli_epi64(val, R), _mm_srli_epi64(val, 64-R));
594 template <
unsigned int R>
595 inline __m128i RotateRight64(
const __m128i& val)
598 _mm_slli_epi64(val, 64-R), _mm_srli_epi64(val, R));
603 inline __m128i RotateLeft64<8>(
const __m128i& val)
605 const __m128i mask = _mm_set_epi8(14,13,12,11, 10,9,8,15, 6,5,4,3, 2,1,0,7);
606 return _mm_shuffle_epi8(val, mask);
611 inline __m128i RotateRight64<8>(
const __m128i& val)
613 const __m128i mask = _mm_set_epi8(8,15,14,13, 12,11,10,9, 0,7,6,5, 4,3,2,1);
614 return _mm_shuffle_epi8(val, mask);
616 #endif // CRYPTOPP_AVX512_ROTATE 618 inline __m128i SIMON128_f(
const __m128i& v)
620 return _mm_xor_si128(RotateLeft64<2>(v),
621 _mm_and_si128(RotateLeft64<1>(v), RotateLeft64<8>(v)));
624 inline void GCC_NO_UBSAN SIMON128_Enc_Block(__m128i &block0, __m128i &block1,
625 const word64 *subkeys,
unsigned int rounds)
631 __m128i x1 = _mm_unpackhi_epi64(block0, block1);
632 __m128i y1 = _mm_unpacklo_epi64(block0, block1);
634 for (
int i = 0; i < static_cast<int>(rounds & ~1)-1; i += 2)
636 const __m128i rk1 = _mm_castpd_si128(
637 _mm_loaddup_pd(CONST_DOUBLE_CAST(subkeys+i)));
638 y1 = _mm_xor_si128(_mm_xor_si128(y1, SIMON128_f(x1)), rk1);
640 const __m128i rk2 = _mm_castpd_si128(
641 _mm_loaddup_pd(CONST_DOUBLE_CAST(subkeys+i+1)));
642 x1 = _mm_xor_si128(_mm_xor_si128(x1, SIMON128_f(y1)), rk2);
647 const __m128i rk = _mm_castpd_si128(
648 _mm_loaddup_pd(CONST_DOUBLE_CAST(subkeys+rounds-1)));
650 y1 = _mm_xor_si128(_mm_xor_si128(y1, SIMON128_f(x1)), rk);
655 block0 = _mm_unpacklo_epi64(y1, x1);
656 block1 = _mm_unpackhi_epi64(y1, x1);
659 inline void GCC_NO_UBSAN SIMON128_Enc_6_Blocks(__m128i &block0, __m128i &block1,
660 __m128i &block2, __m128i &block3, __m128i &block4, __m128i &block5,
661 const word64 *subkeys,
unsigned int rounds)
667 __m128i x1 = _mm_unpackhi_epi64(block0, block1);
668 __m128i y1 = _mm_unpacklo_epi64(block0, block1);
669 __m128i x2 = _mm_unpackhi_epi64(block2, block3);
670 __m128i y2 = _mm_unpacklo_epi64(block2, block3);
671 __m128i x3 = _mm_unpackhi_epi64(block4, block5);
672 __m128i y3 = _mm_unpacklo_epi64(block4, block5);
674 for (
int i = 0; i < static_cast<int>(rounds & ~1) - 1; i += 2)
676 const __m128i rk1 = _mm_castpd_si128(
677 _mm_loaddup_pd(CONST_DOUBLE_CAST(subkeys + i)));
678 y1 = _mm_xor_si128(_mm_xor_si128(y1, SIMON128_f(x1)), rk1);
679 y2 = _mm_xor_si128(_mm_xor_si128(y2, SIMON128_f(x2)), rk1);
680 y3 = _mm_xor_si128(_mm_xor_si128(y3, SIMON128_f(x3)), rk1);
682 const __m128i rk2 = _mm_castpd_si128(
683 _mm_loaddup_pd(CONST_DOUBLE_CAST(subkeys + i + 1)));
684 x1 = _mm_xor_si128(_mm_xor_si128(x1, SIMON128_f(y1)), rk2);
685 x2 = _mm_xor_si128(_mm_xor_si128(x2, SIMON128_f(y2)), rk2);
686 x3 = _mm_xor_si128(_mm_xor_si128(x3, SIMON128_f(y3)), rk2);
691 const __m128i rk = _mm_castpd_si128(
692 _mm_loaddup_pd(CONST_DOUBLE_CAST(subkeys + rounds - 1)));
693 y1 = _mm_xor_si128(_mm_xor_si128(y1, SIMON128_f(x1)), rk);
694 y2 = _mm_xor_si128(_mm_xor_si128(y2, SIMON128_f(x2)), rk);
695 y3 = _mm_xor_si128(_mm_xor_si128(y3, SIMON128_f(x3)), rk);
696 Swap128(x1, y1); Swap128(x2, y2); Swap128(x3, y3);
700 block0 = _mm_unpacklo_epi64(y1, x1);
701 block1 = _mm_unpackhi_epi64(y1, x1);
702 block2 = _mm_unpacklo_epi64(y2, x2);
703 block3 = _mm_unpackhi_epi64(y2, x2);
704 block4 = _mm_unpacklo_epi64(y3, x3);
705 block5 = _mm_unpackhi_epi64(y3, x3);
708 inline void GCC_NO_UBSAN SIMON128_Dec_Block(__m128i &block0, __m128i &block1,
709 const word64 *subkeys,
unsigned int rounds)
715 __m128i x1 = _mm_unpackhi_epi64(block0, block1);
716 __m128i y1 = _mm_unpacklo_epi64(block0, block1);
720 const __m128i rk = _mm_castpd_si128(
721 _mm_loaddup_pd(CONST_DOUBLE_CAST(subkeys + rounds - 1)));
724 y1 = _mm_xor_si128(_mm_xor_si128(y1, rk), SIMON128_f(x1));
728 for (
int i = static_cast<int>(rounds-2); i >= 0; i -= 2)
730 const __m128i rk1 = _mm_castpd_si128(
731 _mm_loaddup_pd(CONST_DOUBLE_CAST(subkeys+i+1)));
732 x1 = _mm_xor_si128(_mm_xor_si128(x1, SIMON128_f(y1)), rk1);
734 const __m128i rk2 = _mm_castpd_si128(
735 _mm_loaddup_pd(CONST_DOUBLE_CAST(subkeys+i)));
736 y1 = _mm_xor_si128(_mm_xor_si128(y1, SIMON128_f(x1)), rk2);
740 block0 = _mm_unpacklo_epi64(y1, x1);
741 block1 = _mm_unpackhi_epi64(y1, x1);
744 inline void GCC_NO_UBSAN SIMON128_Dec_6_Blocks(__m128i &block0, __m128i &block1,
745 __m128i &block2, __m128i &block3, __m128i &block4, __m128i &block5,
746 const word64 *subkeys,
unsigned int rounds)
752 __m128i x1 = _mm_unpackhi_epi64(block0, block1);
753 __m128i y1 = _mm_unpacklo_epi64(block0, block1);
754 __m128i x2 = _mm_unpackhi_epi64(block2, block3);
755 __m128i y2 = _mm_unpacklo_epi64(block2, block3);
756 __m128i x3 = _mm_unpackhi_epi64(block4, block5);
757 __m128i y3 = _mm_unpacklo_epi64(block4, block5);
761 const __m128i rk = _mm_castpd_si128(
762 _mm_loaddup_pd(CONST_DOUBLE_CAST(subkeys + rounds - 1)));
764 Swap128(x1, y1); Swap128(x2, y2); Swap128(x3, y3);
765 y1 = _mm_xor_si128(_mm_xor_si128(y1, rk), SIMON128_f(x1));
766 y2 = _mm_xor_si128(_mm_xor_si128(y2, rk), SIMON128_f(x2));
767 y3 = _mm_xor_si128(_mm_xor_si128(y3, rk), SIMON128_f(x3));
771 for (
int i = static_cast<int>(rounds-2); i >= 0; i -= 2)
773 const __m128i rk1 = _mm_castpd_si128(
774 _mm_loaddup_pd(CONST_DOUBLE_CAST(subkeys + i + 1)));
775 x1 = _mm_xor_si128(_mm_xor_si128(x1, SIMON128_f(y1)), rk1);
776 x2 = _mm_xor_si128(_mm_xor_si128(x2, SIMON128_f(y2)), rk1);
777 x3 = _mm_xor_si128(_mm_xor_si128(x3, SIMON128_f(y3)), rk1);
779 const __m128i rk2 = _mm_castpd_si128(
780 _mm_loaddup_pd(CONST_DOUBLE_CAST(subkeys + i)));
781 y1 = _mm_xor_si128(_mm_xor_si128(y1, SIMON128_f(x1)), rk2);
782 y2 = _mm_xor_si128(_mm_xor_si128(y2, SIMON128_f(x2)), rk2);
783 y3 = _mm_xor_si128(_mm_xor_si128(y3, SIMON128_f(x3)), rk2);
787 block0 = _mm_unpacklo_epi64(y1, x1);
788 block1 = _mm_unpackhi_epi64(y1, x1);
789 block2 = _mm_unpacklo_epi64(y2, x2);
790 block3 = _mm_unpackhi_epi64(y2, x2);
791 block4 = _mm_unpacklo_epi64(y3, x3);
792 block5 = _mm_unpackhi_epi64(y3, x3);
795 #endif // CRYPTOPP_SSSE3_AVAILABLE 797 #if defined(CRYPTOPP_SSE41_AVAILABLE) 799 template <
unsigned int R>
800 inline __m128i RotateLeft32(
const __m128i& val)
803 _mm_slli_epi32(val, R), _mm_srli_epi32(val, 32-R));
806 template <
unsigned int R>
807 inline __m128i RotateRight32(
const __m128i& val)
810 _mm_slli_epi32(val, 32-R), _mm_srli_epi32(val, R));
815 inline __m128i RotateLeft32<8>(
const __m128i& val)
817 const __m128i mask = _mm_set_epi8(14,13,12,15, 10,9,8,11, 6,5,4,7, 2,1,0,3);
818 return _mm_shuffle_epi8(val, mask);
823 inline __m128i RotateRight32<8>(
const __m128i& val)
825 const __m128i mask = _mm_set_epi8(12,15,14,13, 8,11,10,9, 4,7,6,5, 0,3,2,1);
826 return _mm_shuffle_epi8(val, mask);
829 inline __m128i SIMON64_f(
const __m128i& v)
831 return _mm_xor_si128(RotateLeft32<2>(v),
832 _mm_and_si128(RotateLeft32<1>(v), RotateLeft32<8>(v)));
835 inline void GCC_NO_UBSAN SIMON64_Enc_Block(__m128i &block0, __m128i &block1,
836 const word32 *subkeys,
unsigned int rounds)
843 const __m128 t0 = _mm_castsi128_ps(block0);
844 const __m128 t1 = _mm_castsi128_ps(block1);
845 __m128i x1 = _mm_castps_si128(_mm_shuffle_ps(t0, t1, _MM_SHUFFLE(3,1,3,1)));
846 __m128i y1 = _mm_castps_si128(_mm_shuffle_ps(t0, t1, _MM_SHUFFLE(2,0,2,0)));
848 for (
int i = 0; i < static_cast<int>(rounds & ~1)-1; i += 2)
850 const __m128i rk1 = _mm_set1_epi32(subkeys[i]);
851 y1 = _mm_xor_si128(_mm_xor_si128(y1, SIMON64_f(x1)), rk1);
853 const __m128i rk2 = _mm_set1_epi32(subkeys[i+1]);
854 x1 = _mm_xor_si128(_mm_xor_si128(x1, SIMON64_f(y1)), rk2);
859 const __m128i rk = _mm_set1_epi32(subkeys[rounds-1]);
860 y1 = _mm_xor_si128(_mm_xor_si128(y1, SIMON64_f(x1)), rk);
866 block0 = _mm_unpacklo_epi32(y1, x1);
867 block1 = _mm_unpackhi_epi32(y1, x1);
870 inline void GCC_NO_UBSAN SIMON64_Dec_Block(__m128i &block0, __m128i &block1,
871 const word32 *subkeys,
unsigned int rounds)
878 const __m128 t0 = _mm_castsi128_ps(block0);
879 const __m128 t1 = _mm_castsi128_ps(block1);
880 __m128i x1 = _mm_castps_si128(_mm_shuffle_ps(t0, t1, _MM_SHUFFLE(3,1,3,1)));
881 __m128i y1 = _mm_castps_si128(_mm_shuffle_ps(t0, t1, _MM_SHUFFLE(2,0,2,0)));
886 const __m128i rk = _mm_set1_epi32(subkeys[rounds-1]);
887 y1 = _mm_xor_si128(_mm_xor_si128(y1, rk), SIMON64_f(x1));
891 for (
int i = static_cast<int>(rounds-2); i >= 0; i -= 2)
893 const __m128i rk1 = _mm_set1_epi32(subkeys[i+1]);
894 x1 = _mm_xor_si128(_mm_xor_si128(x1, SIMON64_f(y1)), rk1);
896 const __m128i rk2 = _mm_set1_epi32(subkeys[i]);
897 y1 = _mm_xor_si128(_mm_xor_si128(y1, SIMON64_f(x1)), rk2);
902 block0 = _mm_unpacklo_epi32(y1, x1);
903 block1 = _mm_unpackhi_epi32(y1, x1);
906 inline void GCC_NO_UBSAN SIMON64_Enc_6_Blocks(__m128i &block0, __m128i &block1,
907 __m128i &block2, __m128i &block3, __m128i &block4, __m128i &block5,
908 const word32 *subkeys,
unsigned int rounds)
915 const __m128 t0 = _mm_castsi128_ps(block0);
916 const __m128 t1 = _mm_castsi128_ps(block1);
917 __m128i x1 = _mm_castps_si128(_mm_shuffle_ps(t0, t1, _MM_SHUFFLE(3,1,3,1)));
918 __m128i y1 = _mm_castps_si128(_mm_shuffle_ps(t0, t1, _MM_SHUFFLE(2,0,2,0)));
920 const __m128 t2 = _mm_castsi128_ps(block2);
921 const __m128 t3 = _mm_castsi128_ps(block3);
922 __m128i x2 = _mm_castps_si128(_mm_shuffle_ps(t2, t3, _MM_SHUFFLE(3,1,3,1)));
923 __m128i y2 = _mm_castps_si128(_mm_shuffle_ps(t2, t3, _MM_SHUFFLE(2,0,2,0)));
925 const __m128 t4 = _mm_castsi128_ps(block4);
926 const __m128 t5 = _mm_castsi128_ps(block5);
927 __m128i x3 = _mm_castps_si128(_mm_shuffle_ps(t4, t5, _MM_SHUFFLE(3,1,3,1)));
928 __m128i y3 = _mm_castps_si128(_mm_shuffle_ps(t4, t5, _MM_SHUFFLE(2,0,2,0)));
930 for (
int i = 0; i < static_cast<int>(rounds & ~1)-1; i += 2)
932 const __m128i rk1 = _mm_set1_epi32(subkeys[i]);
933 y1 = _mm_xor_si128(_mm_xor_si128(y1, SIMON64_f(x1)), rk1);
934 y2 = _mm_xor_si128(_mm_xor_si128(y2, SIMON64_f(x2)), rk1);
935 y3 = _mm_xor_si128(_mm_xor_si128(y3, SIMON64_f(x3)), rk1);
937 const __m128i rk2 = _mm_set1_epi32(subkeys[i+1]);
938 x1 = _mm_xor_si128(_mm_xor_si128(x1, SIMON64_f(y1)), rk2);
939 x2 = _mm_xor_si128(_mm_xor_si128(x2, SIMON64_f(y2)), rk2);
940 x3 = _mm_xor_si128(_mm_xor_si128(x3, SIMON64_f(y3)), rk2);
945 const __m128i rk = _mm_set1_epi32(subkeys[rounds-1]);
946 y1 = _mm_xor_si128(_mm_xor_si128(y1, SIMON64_f(x1)), rk);
947 y2 = _mm_xor_si128(_mm_xor_si128(y2, SIMON64_f(x2)), rk);
948 y3 = _mm_xor_si128(_mm_xor_si128(y3, SIMON64_f(x3)), rk);
949 Swap128(x1, y1); Swap128(x2, y2); Swap128(x3, y3);
954 block0 = _mm_unpacklo_epi32(y1, x1);
955 block1 = _mm_unpackhi_epi32(y1, x1);
956 block2 = _mm_unpacklo_epi32(y2, x2);
957 block3 = _mm_unpackhi_epi32(y2, x2);
958 block4 = _mm_unpacklo_epi32(y3, x3);
959 block5 = _mm_unpackhi_epi32(y3, x3);
962 inline void GCC_NO_UBSAN SIMON64_Dec_6_Blocks(__m128i &block0, __m128i &block1,
963 __m128i &block2, __m128i &block3, __m128i &block4, __m128i &block5,
964 const word32 *subkeys,
unsigned int rounds)
971 const __m128 t0 = _mm_castsi128_ps(block0);
972 const __m128 t1 = _mm_castsi128_ps(block1);
973 __m128i x1 = _mm_castps_si128(_mm_shuffle_ps(t0, t1, _MM_SHUFFLE(3,1,3,1)));
974 __m128i y1 = _mm_castps_si128(_mm_shuffle_ps(t0, t1, _MM_SHUFFLE(2,0,2,0)));
976 const __m128 t2 = _mm_castsi128_ps(block2);
977 const __m128 t3 = _mm_castsi128_ps(block3);
978 __m128i x2 = _mm_castps_si128(_mm_shuffle_ps(t2, t3, _MM_SHUFFLE(3,1,3,1)));
979 __m128i y2 = _mm_castps_si128(_mm_shuffle_ps(t2, t3, _MM_SHUFFLE(2,0,2,0)));
981 const __m128 t4 = _mm_castsi128_ps(block4);
982 const __m128 t5 = _mm_castsi128_ps(block5);
983 __m128i x3 = _mm_castps_si128(_mm_shuffle_ps(t4, t5, _MM_SHUFFLE(3,1,3,1)));
984 __m128i y3 = _mm_castps_si128(_mm_shuffle_ps(t4, t5, _MM_SHUFFLE(2,0,2,0)));
988 Swap128(x1, y1); Swap128(x2, y2); Swap128(x3, y3);
989 const __m128i rk = _mm_set1_epi32(subkeys[rounds-1]);
990 y1 = _mm_xor_si128(_mm_xor_si128(y1, rk), SIMON64_f(x1));
991 y2 = _mm_xor_si128(_mm_xor_si128(y2, rk), SIMON64_f(x2));
992 y3 = _mm_xor_si128(_mm_xor_si128(y3, rk), SIMON64_f(x3));
996 for (
int i = static_cast<int>(rounds-2); i >= 0; i -= 2)
998 const __m128i rk1 = _mm_set1_epi32(subkeys[i+1]);
999 x1 = _mm_xor_si128(_mm_xor_si128(x1, SIMON64_f(y1)), rk1);
1000 x2 = _mm_xor_si128(_mm_xor_si128(x2, SIMON64_f(y2)), rk1);
1001 x3 = _mm_xor_si128(_mm_xor_si128(x3, SIMON64_f(y3)), rk1);
1003 const __m128i rk2 = _mm_set1_epi32(subkeys[i]);
1004 y1 = _mm_xor_si128(_mm_xor_si128(y1, SIMON64_f(x1)), rk2);
1005 y2 = _mm_xor_si128(_mm_xor_si128(y2, SIMON64_f(x2)), rk2);
1006 y3 = _mm_xor_si128(_mm_xor_si128(y3, SIMON64_f(x3)), rk2);
1011 block0 = _mm_unpacklo_epi32(y1, x1);
1012 block1 = _mm_unpackhi_epi32(y1, x1);
1013 block2 = _mm_unpacklo_epi32(y2, x2);
1014 block3 = _mm_unpackhi_epi32(y2, x2);
1015 block4 = _mm_unpacklo_epi32(y3, x3);
1016 block5 = _mm_unpackhi_epi32(y3, x3);
1019 #endif // CRYPTOPP_SSE41_AVAILABLE 1021 ANONYMOUS_NAMESPACE_END
1029 #if (CRYPTOPP_ARM_NEON_AVAILABLE) 1030 size_t SIMON64_Enc_AdvancedProcessBlocks_NEON(
const word32* subKeys,
size_t rounds,
1031 const byte *inBlocks,
const byte *xorBlocks, byte *outBlocks,
size_t length, word32 flags)
1033 return AdvancedProcessBlocks64_6x2_NEON(SIMON64_Enc_Block, SIMON64_Enc_6_Blocks,
1034 subKeys, rounds, inBlocks, xorBlocks, outBlocks, length, flags);
1037 size_t SIMON64_Dec_AdvancedProcessBlocks_NEON(
const word32* subKeys,
size_t rounds,
1038 const byte *inBlocks,
const byte *xorBlocks, byte *outBlocks,
size_t length, word32 flags)
1040 return AdvancedProcessBlocks64_6x2_NEON(SIMON64_Dec_Block, SIMON64_Dec_6_Blocks,
1041 subKeys, rounds, inBlocks, xorBlocks, outBlocks, length, flags);
1043 #endif // CRYPTOPP_ARM_NEON_AVAILABLE 1045 #if (CRYPTOPP_ARM_NEON_AVAILABLE) 1046 size_t SIMON128_Enc_AdvancedProcessBlocks_NEON(
const word64* subKeys,
size_t rounds,
1047 const byte *inBlocks,
const byte *xorBlocks, byte *outBlocks,
size_t length, word32 flags)
1049 return AdvancedProcessBlocks128_6x2_NEON(SIMON128_Enc_Block, SIMON128_Enc_6_Blocks,
1050 subKeys, rounds, inBlocks, xorBlocks, outBlocks, length, flags);
1053 size_t SIMON128_Dec_AdvancedProcessBlocks_NEON(
const word64* subKeys,
size_t rounds,
1054 const byte *inBlocks,
const byte *xorBlocks, byte *outBlocks,
size_t length, word32 flags)
1056 return AdvancedProcessBlocks128_6x2_NEON(SIMON128_Dec_Block, SIMON128_Dec_6_Blocks,
1057 subKeys, rounds, inBlocks, xorBlocks, outBlocks, length, flags);
1059 #endif // CRYPTOPP_ARM_NEON_AVAILABLE 1063 #if defined(CRYPTOPP_SSE41_AVAILABLE) 1064 size_t SIMON64_Enc_AdvancedProcessBlocks_SSE41(
const word32* subKeys,
size_t rounds,
1065 const byte *inBlocks,
const byte *xorBlocks, byte *outBlocks,
size_t length, word32 flags)
1067 return AdvancedProcessBlocks64_6x2_SSE(SIMON64_Enc_Block, SIMON64_Enc_6_Blocks,
1068 subKeys, rounds, inBlocks, xorBlocks, outBlocks, length, flags);
1071 size_t SIMON64_Dec_AdvancedProcessBlocks_SSE41(
const word32* subKeys,
size_t rounds,
1072 const byte *inBlocks,
const byte *xorBlocks, byte *outBlocks,
size_t length, word32 flags)
1074 return AdvancedProcessBlocks64_6x2_SSE(SIMON64_Dec_Block, SIMON64_Dec_6_Blocks,
1075 subKeys, rounds, inBlocks, xorBlocks, outBlocks, length, flags);
1079 #if defined(CRYPTOPP_SSSE3_AVAILABLE) 1080 size_t SIMON128_Enc_AdvancedProcessBlocks_SSSE3(
const word64* subKeys,
size_t rounds,
1081 const byte *inBlocks,
const byte *xorBlocks, byte *outBlocks,
size_t length, word32 flags)
1083 return AdvancedProcessBlocks128_6x2_SSE(SIMON128_Enc_Block, SIMON128_Enc_6_Blocks,
1084 subKeys, rounds, inBlocks, xorBlocks, outBlocks, length, flags);
1087 size_t SIMON128_Dec_AdvancedProcessBlocks_SSSE3(
const word64* subKeys,
size_t rounds,
1088 const byte *inBlocks,
const byte *xorBlocks, byte *outBlocks,
size_t length, word32 flags)
1090 return AdvancedProcessBlocks128_6x2_SSE(SIMON128_Dec_Block, SIMON128_Dec_6_Blocks,
1091 subKeys, rounds, inBlocks, xorBlocks, outBlocks, length, flags);
1093 #endif // CRYPTOPP_SSSE3_AVAILABLE Utility functions for the Crypto++ library.
T rotlFixed(T x, unsigned int y)
Performs a left rotate.
Library configuration file.
Classes for the Simon block cipher.
Crypto++ library namespace.
T rotrFixed(T x, unsigned int y)
Performs a right rotate.
void vec_swap(T &a, T &b)
Swaps two variables which are arrays.