/* * (C) 2025,2026 Jack Lloyd * * Botan is released under the Simplified BSD License (see license.txt) */ #include #include #include #include namespace Botan { namespace Camellia_AVX2_GFNI { /* * This follows exactly the approach used in the AVX-512+GFNI implementation * with only minor complications due to missing rotate and masked operations. */ namespace { constexpr uint64_t pre123_a = gfni_matrix(R"( 1 1 1 0 1 1 0 1 0 0 1 1 0 0 1 0 1 1 0 1 0 0 0 0 1 0 1 1 0 0 1 1 0 0 0 0 1 1 0 0 1 0 1 0 0 1 0 0 0 0 1 0 1 1 0 0 1 0 0 0 0 1 1 0)"); constexpr uint64_t pre4_a = gfni_matrix(R"( 1 1 0 1 1 0 1 1 0 1 1 0 0 1 0 0 1 0 1 0 0 0 0 1 0 1 1 0 0 1 1 1 0 0 0 1 1 0 0 0 0 1 0 0 1 0 0 1 0 1 0 1 1 0 0 0 0 0 0 0 1 1 0 1)"); constexpr uint8_t pre_c = 0b01000101; constexpr uint64_t post2_a = gfni_matrix(R"( 0 0 0 1 1 1 0 0 0 0 0 0 0 0 0 1 0 1 1 0 0 1 1 0 1 0 1 1 1 1 1 0 0 0 0 1 1 0 1 1 1 0 0 0 1 1 1 0 0 1 0 1 1 1 1 0 0 1 1 1 1 1 1 1)"); constexpr uint64_t post3_a = gfni_matrix(R"( 0 1 1 0 0 1 1 0 1 0 1 1 1 1 1 0 0 0 0 1 1 0 1 1 1 0 0 0 1 1 1 0 0 1 0 1 1 1 1 0 0 1 1 1 1 1 1 1 0 0 0 1 1 1 0 0 0 0 0 0 0 0 0 1)"); constexpr uint64_t post14_a = gfni_matrix(R"( 0 0 0 0 0 0 0 1 0 1 1 0 0 1 1 0 1 0 1 1 1 1 1 0 0 0 0 1 1 0 1 1 1 0 0 0 1 1 1 0 0 1 0 1 1 1 1 0 0 1 1 1 1 1 1 1 0 0 0 1 1 1 0 0)"); // NOLINTBEGIN(portability-simd-intrinsics) BOTAN_FORCE_INLINE BOTAN_FN_ISA_AVX2_GFNI SIMD_4x64 camellia_f(SIMD_4x64 x) { const __m256i xr = x.raw(); // Compute S1(x), S2(x), S3(x), S4(x) for all bytes const auto y123 = _mm256_gf2p8affine_epi64_epi8(xr, _mm256_set1_epi64x(pre123_a), pre_c); const auto y4 = _mm256_gf2p8affine_epi64_epi8(xr, _mm256_set1_epi64x(pre4_a), pre_c); const auto s1 = _mm256_gf2p8affineinv_epi64_epi8(y123, _mm256_set1_epi64x(post14_a), 0x6E); const auto s2 = _mm256_gf2p8affineinv_epi64_epi8(y123, _mm256_set1_epi64x(post2_a), 0xDC); const auto s3 = _mm256_gf2p8affineinv_epi64_epi8(y123, _mm256_set1_epi64x(post3_a), 0x37); const auto s4 = _mm256_gf2p8affineinv_epi64_epi8(y4, _mm256_set1_epi64x(post14_a), 0x6E); // Blend to find correct S(x) for each byte position const auto mask_s2 = _mm256_set1_epi64x(0x00FF0000FF000000); const auto mask_s3 = _mm256_set1_epi64x(0x0000FF0000FF0000); const auto mask_s4 = _mm256_set1_epi64x(0x000000FF0000FF00); auto sx = s1; sx = _mm256_blendv_epi8(sx, s2, mask_s2); sx = _mm256_blendv_epi8(sx, s3, mask_s3); sx = _mm256_blendv_epi8(sx, s4, mask_s4); // Linear mixing layer const auto P1 = _mm256_set_epi64x(0x0808080908080809, 0x0000000100000001, 0x0808080908080809, 0x0000000100000001); const auto P2 = _mm256_set_epi64x(0x09090A0A09090A0A, 0x0101020201010202, 0x09090A0A09090A0A, 0x0101020201010202); const auto P3 = _mm256_set_epi64x(0x0A0B0B0B0A0B0B0B, 0x0203030302030303, 0x0A0B0B0B0A0B0B0B, 0x0203030302030303); const auto P4 = _mm256_set_epi64x(0x0C0C0D0C0E0D0C0C, 0x0404050406050404, 0x0C0C0D0C0E0D0C0C, 0x0404050406050404); const auto P5 = _mm256_set_epi64x(0x0D0E0E0D0F0E0D0F, 0x0506060507060507, 0x0D0E0E0D0F0E0D0F, 0x0506060507060507); const auto P6 = _mm256_set_epi64x(0x0F0F0F0EFFFFFFFF, 0x07070706FFFFFFFF, 0x0F0F0F0EFFFFFFFF, 0x07070706FFFFFFFF); const auto t1 = SIMD_4x64(_mm256_shuffle_epi8(sx, P1)); const auto t2 = SIMD_4x64(_mm256_shuffle_epi8(sx, P2)); const auto t3 = SIMD_4x64(_mm256_shuffle_epi8(sx, P3)); const auto t4 = SIMD_4x64(_mm256_shuffle_epi8(sx, P4)); const auto t5 = SIMD_4x64(_mm256_shuffle_epi8(sx, P5)); const auto t6 = SIMD_4x64(_mm256_shuffle_epi8(sx, P6)); return (t1 ^ t2 ^ t3 ^ t4 ^ t5 ^ t6); } BOTAN_FORCE_INLINE BOTAN_FN_ISA_AVX2 void load_and_deinterleave(const uint8_t in[], SIMD_4x64& L, SIMD_4x64& R) { auto A = SIMD_4x64::load_be(in); auto B = SIMD_4x64::load_be(in + 32); auto Ap = _mm256_permute4x64_epi64(A.raw(), 0b11'01'10'00); // [L[0], L[1], R[0], R[1]] auto Bp = _mm256_permute4x64_epi64(B.raw(), 0b11'01'10'00); // [L[2], L[3], R[2], R[3]] L = SIMD_4x64(_mm256_permute2x128_si256(Ap, Bp, 0x20)); // [L[0], L[1], L[2], L[3]] R = SIMD_4x64(_mm256_permute2x128_si256(Ap, Bp, 0x31)); // [R[0], R[1], R[2], R[3]] } BOTAN_FORCE_INLINE BOTAN_FN_ISA_AVX2 void interleave_and_store(uint8_t out[], SIMD_4x64 L, SIMD_4x64 R) { auto T1 = _mm256_permute2x128_si256(R.raw(), L.raw(), 0x20); // [R[0], R[1], L[0], L[1]] auto T2 = _mm256_permute2x128_si256(R.raw(), L.raw(), 0x31); // [R[2], R[3], L[2], L[3]] auto A = SIMD_4x64(_mm256_permute4x64_epi64(T1, 0b11'01'10'00)); // [R[0], L[0], R[1], L[1]] auto B = SIMD_4x64(_mm256_permute4x64_epi64(T2, 0b11'01'10'00)); // [R[2], L[2], R[3], L[3]] A.store_be(out); B.store_be(out + 32); } /* * 32-bit rotate on SIMD_4x64 helper for FL/FLINV */ BOTAN_FORCE_INLINE BOTAN_FN_ISA_AVX2 SIMD_4x64 rotl32_1(SIMD_4x64 t) { return SIMD_4x64(_mm256_or_si256(_mm256_slli_epi32(t.raw(), 1), _mm256_srli_epi32(t.raw(), 31))); } // NOLINTEND(portability-simd-intrinsics) BOTAN_FORCE_INLINE BOTAN_FN_ISA_AVX2 SIMD_4x64 FL_4(SIMD_4x64 v, uint64_t K) { const uint32_t k1 = static_cast(K >> 32); const uint32_t k2 = static_cast(K & 0xFFFFFFFF); auto x1 = v.shr<32>(); auto x2 = v & SIMD_4x64::splat(0xFFFFFFFF); x2 ^= rotl32_1(x1 & SIMD_4x64::splat(k1)); x1 ^= (x2 | SIMD_4x64::splat(k2)); return x1.shl<32>() | x2; } BOTAN_FORCE_INLINE BOTAN_FN_ISA_AVX2 SIMD_4x64 FLINV_4(SIMD_4x64 v, uint64_t K) { const uint32_t k1 = static_cast(K >> 32); const uint32_t k2 = static_cast(K & 0xFFFFFFFF); auto x1 = v.shr<32>(); auto x2 = v & SIMD_4x64::splat(0xFFFFFFFF); x1 ^= (x2 | SIMD_4x64::splat(k2)); x2 ^= rotl32_1(x1 & SIMD_4x64::splat(k1)); return x1.shl<32>() | x2; } // Helpers for 6 round iterations BOTAN_FORCE_INLINE BOTAN_FN_ISA_AVX2_GFNI void six_e_rounds(SIMD_4x64& L, SIMD_4x64& R, std::span SK) { R ^= camellia_f(L ^ SIMD_4x64::splat(SK[0])); L ^= camellia_f(R ^ SIMD_4x64::splat(SK[1])); R ^= camellia_f(L ^ SIMD_4x64::splat(SK[2])); L ^= camellia_f(R ^ SIMD_4x64::splat(SK[3])); R ^= camellia_f(L ^ SIMD_4x64::splat(SK[4])); L ^= camellia_f(R ^ SIMD_4x64::splat(SK[5])); } BOTAN_FORCE_INLINE BOTAN_FN_ISA_AVX2_GFNI void six_d_rounds(SIMD_4x64& L, SIMD_4x64& R, std::span SK) { R ^= camellia_f(L ^ SIMD_4x64::splat(SK[5])); L ^= camellia_f(R ^ SIMD_4x64::splat(SK[4])); R ^= camellia_f(L ^ SIMD_4x64::splat(SK[3])); L ^= camellia_f(R ^ SIMD_4x64::splat(SK[2])); R ^= camellia_f(L ^ SIMD_4x64::splat(SK[1])); L ^= camellia_f(R ^ SIMD_4x64::splat(SK[0])); } BOTAN_FN_ISA_AVX2_GFNI void camellia_encrypt_x4_18r(const uint8_t in[], uint8_t out[], std::span SK) { SIMD_4x64 L; SIMD_4x64 R; load_and_deinterleave(in, L, R); L ^= SIMD_4x64::splat(SK[0]); R ^= SIMD_4x64::splat(SK[1]); six_e_rounds(L, R, SK.subspan(2)); L = FL_4(L, SK[8]); R = FLINV_4(R, SK[9]); six_e_rounds(L, R, SK.subspan(10)); L = FL_4(L, SK[16]); R = FLINV_4(R, SK[17]); six_e_rounds(L, R, SK.subspan(18)); R ^= SIMD_4x64::splat(SK[24]); L ^= SIMD_4x64::splat(SK[25]); interleave_and_store(out, L, R); } BOTAN_FN_ISA_AVX2_GFNI void camellia_decrypt_x4_18r(const uint8_t in[], uint8_t out[], std::span SK) { SIMD_4x64 L; SIMD_4x64 R; load_and_deinterleave(in, L, R); R ^= SIMD_4x64::splat(SK[25]); L ^= SIMD_4x64::splat(SK[24]); six_d_rounds(L, R, SK.subspan(18)); L = FL_4(L, SK[17]); R = FLINV_4(R, SK[16]); six_d_rounds(L, R, SK.subspan(10)); L = FL_4(L, SK[9]); R = FLINV_4(R, SK[8]); six_d_rounds(L, R, SK.subspan(2)); L ^= SIMD_4x64::splat(SK[1]); R ^= SIMD_4x64::splat(SK[0]); interleave_and_store(out, L, R); } BOTAN_FN_ISA_AVX2_GFNI void camellia_encrypt_x4_24r(const uint8_t in[], uint8_t out[], std::span SK) { SIMD_4x64 L; SIMD_4x64 R; load_and_deinterleave(in, L, R); L ^= SIMD_4x64::splat(SK[0]); R ^= SIMD_4x64::splat(SK[1]); six_e_rounds(L, R, SK.subspan(2)); L = FL_4(L, SK[8]); R = FLINV_4(R, SK[9]); six_e_rounds(L, R, SK.subspan(10)); L = FL_4(L, SK[16]); R = FLINV_4(R, SK[17]); six_e_rounds(L, R, SK.subspan(18)); L = FL_4(L, SK[24]); R = FLINV_4(R, SK[25]); six_e_rounds(L, R, SK.subspan(26)); R ^= SIMD_4x64::splat(SK[32]); L ^= SIMD_4x64::splat(SK[33]); interleave_and_store(out, L, R); } BOTAN_FN_ISA_AVX2_GFNI void camellia_decrypt_x4_24r(const uint8_t in[], uint8_t out[], std::span SK) { SIMD_4x64 L; SIMD_4x64 R; load_and_deinterleave(in, L, R); R ^= SIMD_4x64::splat(SK[33]); L ^= SIMD_4x64::splat(SK[32]); six_d_rounds(L, R, SK.subspan(26)); L = FL_4(L, SK[25]); R = FLINV_4(R, SK[24]); six_d_rounds(L, R, SK.subspan(18)); L = FL_4(L, SK[17]); R = FLINV_4(R, SK[16]); six_d_rounds(L, R, SK.subspan(10)); L = FL_4(L, SK[9]); R = FLINV_4(R, SK[8]); six_d_rounds(L, R, SK.subspan(2)); L ^= SIMD_4x64::splat(SK[1]); R ^= SIMD_4x64::splat(SK[0]); interleave_and_store(out, L, R); } } // namespace } // namespace Camellia_AVX2_GFNI // static void BOTAN_FN_ISA_AVX2_GFNI Camellia_128::avx2_gfni_encrypt(const uint8_t in[], uint8_t out[], size_t blocks, std::span SK) { while(blocks >= 4) { Camellia_AVX2_GFNI::camellia_encrypt_x4_18r(in, out, SK); in += 4 * 16; out += 4 * 16; blocks -= 4; } if(blocks > 0) { uint8_t ibuf[4 * 16] = {0}; uint8_t obuf[4 * 16] = {0}; copy_mem(ibuf, in, blocks * 16); Camellia_AVX2_GFNI::camellia_encrypt_x4_18r(ibuf, obuf, SK); copy_mem(out, obuf, blocks * 16); } } // static void BOTAN_FN_ISA_AVX2_GFNI Camellia_128::avx2_gfni_decrypt(const uint8_t in[], uint8_t out[], size_t blocks, std::span SK) { while(blocks >= 4) { Camellia_AVX2_GFNI::camellia_decrypt_x4_18r(in, out, SK); in += 4 * 16; out += 4 * 16; blocks -= 4; } if(blocks > 0) { uint8_t ibuf[4 * 16] = {0}; uint8_t obuf[4 * 16] = {0}; copy_mem(ibuf, in, blocks * 16); Camellia_AVX2_GFNI::camellia_decrypt_x4_18r(ibuf, obuf, SK); copy_mem(out, obuf, blocks * 16); } } // static void BOTAN_FN_ISA_AVX2_GFNI Camellia_192::avx2_gfni_encrypt(const uint8_t in[], uint8_t out[], size_t blocks, std::span SK) { while(blocks >= 4) { Camellia_AVX2_GFNI::camellia_encrypt_x4_24r(in, out, SK); in += 4 * 16; out += 4 * 16; blocks -= 4; } if(blocks > 0) { uint8_t ibuf[4 * 16] = {0}; uint8_t obuf[4 * 16] = {0}; copy_mem(ibuf, in, blocks * 16); Camellia_AVX2_GFNI::camellia_encrypt_x4_24r(ibuf, obuf, SK); copy_mem(out, obuf, blocks * 16); } } // static void BOTAN_FN_ISA_AVX2_GFNI Camellia_192::avx2_gfni_decrypt(const uint8_t in[], uint8_t out[], size_t blocks, std::span SK) { while(blocks >= 4) { Camellia_AVX2_GFNI::camellia_decrypt_x4_24r(in, out, SK); in += 4 * 16; out += 4 * 16; blocks -= 4; } if(blocks > 0) { uint8_t ibuf[4 * 16] = {0}; uint8_t obuf[4 * 16] = {0}; copy_mem(ibuf, in, blocks * 16); Camellia_AVX2_GFNI::camellia_decrypt_x4_24r(ibuf, obuf, SK); copy_mem(out, obuf, blocks * 16); } } // static void BOTAN_FN_ISA_AVX2_GFNI Camellia_256::avx2_gfni_encrypt(const uint8_t in[], uint8_t out[], size_t blocks, std::span SK) { while(blocks >= 4) { Camellia_AVX2_GFNI::camellia_encrypt_x4_24r(in, out, SK); in += 4 * 16; out += 4 * 16; blocks -= 4; } if(blocks > 0) { uint8_t ibuf[4 * 16] = {0}; uint8_t obuf[4 * 16] = {0}; copy_mem(ibuf, in, blocks * 16); Camellia_AVX2_GFNI::camellia_encrypt_x4_24r(ibuf, obuf, SK); copy_mem(out, obuf, blocks * 16); } } // static void BOTAN_FN_ISA_AVX2_GFNI Camellia_256::avx2_gfni_decrypt(const uint8_t in[], uint8_t out[], size_t blocks, std::span SK) { while(blocks >= 4) { Camellia_AVX2_GFNI::camellia_decrypt_x4_24r(in, out, SK); in += 4 * 16; out += 4 * 16; blocks -= 4; } if(blocks > 0) { uint8_t ibuf[4 * 16] = {0}; uint8_t obuf[4 * 16] = {0}; copy_mem(ibuf, in, blocks * 16); Camellia_AVX2_GFNI::camellia_decrypt_x4_24r(ibuf, obuf, SK); copy_mem(out, obuf, blocks * 16); } } } // namespace Botan