| 1 |
/* |
| 2 |
* Based on public domain code by Sean Gulley |
| 3 |
* |
| 4 |
* Adapted to Botan by Jeffrey Walton. |
| 5 |
* |
| 6 |
* Further changes |
| 7 |
* |
| 8 |
* (C) 2017,2025 Jack Lloyd |
| 9 |
* |
| 10 |
* Botan is released under the Simplified BSD License (see license.txt) |
| 11 |
*/ |
| 12 |
|
| 13 |
#include <botan/internal/sha1.h> |
| 14 |
|
| 15 |
#include <botan/internal/isa_extn.h> |
| 16 |
#include <botan/internal/simd_4x32.h> |
| 17 |
#include <immintrin.h> |
| 18 |
|
| 19 |
namespace Botan { |
| 20 |
|
| 21 |
namespace { |
| 22 |
|
| 23 |
// NOLINTBEGIN(portability-simd-intrinsics) |
| 24 |
|
| 25 |
BOTAN_FORCE_INLINE BOTAN_FN_ISA_SHANI SIMD_4x32 sha1_x86_nexte(const SIMD_4x32& x, const SIMD_4x32& y) { |
| 26 |
return SIMD_4x32(_mm_sha1nexte_epu32(x.raw(), y.raw())); |
| 27 |
} |
| 28 |
|
| 29 |
BOTAN_FORCE_INLINE BOTAN_FN_ISA_SHANI SIMD_4x32 sha1_x86_msg1(const SIMD_4x32& W0, const SIMD_4x32& W1) { |
| 30 |
return SIMD_4x32(_mm_sha1msg1_epu32(W0.raw(), W1.raw())); |
| 31 |
} |
| 32 |
|
| 33 |
BOTAN_FORCE_INLINE BOTAN_FN_ISA_SHANI void sha1_x86_next_msg(const SIMD_4x32& W0, |
| 34 |
SIMD_4x32& W1, |
| 35 |
SIMD_4x32& W2, |
| 36 |
SIMD_4x32& W3) { |
| 37 |
W3 = SIMD_4x32(_mm_sha1msg1_epu32(W3.raw(), W0.raw())); |
| 38 |
W1 = SIMD_4x32(_mm_sha1msg2_epu32(W1.raw(), W0.raw())); |
| 39 |
W2 ^= W0; |
| 40 |
} |
| 41 |
|
| 42 |
template <uint8_t R1, uint8_t R2 = R1> |
| 43 |
BOTAN_FORCE_INLINE BOTAN_FN_ISA_SHANI void sha1_x86_first8(SIMD_4x32& ABCD, |
| 44 |
SIMD_4x32& E, |
| 45 |
const SIMD_4x32& W0, |
| 46 |
const SIMD_4x32& W1) { |
| 47 |
auto TE = ABCD; |
| 48 |
ABCD = SIMD_4x32(_mm_sha1rnds4_epu32(ABCD.raw(), (E + W0).raw(), R1)); |
| 49 |
|
| 50 |
E = ABCD; |
| 51 |
ABCD = SIMD_4x32(_mm_sha1rnds4_epu32(ABCD.raw(), sha1_x86_nexte(TE, W1).raw(), R2)); |
| 52 |
} |
| 53 |
|
| 54 |
template <uint8_t R1, uint8_t R2 = R1> |
| 55 |
BOTAN_FORCE_INLINE BOTAN_FN_ISA_SHANI void sha1_x86_rnds8(SIMD_4x32& ABCD, |
| 56 |
SIMD_4x32& E, |
| 57 |
const SIMD_4x32& W0, |
| 58 |
const SIMD_4x32& W1) { |
| 59 |
auto TE = ABCD; |
| 60 |
ABCD = SIMD_4x32(_mm_sha1rnds4_epu32(ABCD.raw(), sha1_x86_nexte(E, W0).raw(), R1)); |
| 61 |
|
| 62 |
E = ABCD; |
| 63 |
ABCD = SIMD_4x32(_mm_sha1rnds4_epu32(ABCD.raw(), sha1_x86_nexte(TE, W1).raw(), R2)); |
| 64 |
} |
| 65 |
|
| 66 |
BOTAN_FORCE_INLINE BOTAN_FN_ISA_SHANI SIMD_4x32 rev_words(const SIMD_4x32& v) { |
| 67 |
return SIMD_4x32(_mm_shuffle_epi32(v.raw(), 0b00011011)); |
| 68 |
} |
| 69 |
|
| 70 |
// NOLINTEND(portability-simd-intrinsics) |
| 71 |
|
| 72 |
} // namespace |
| 73 |
|
| 74 |
void BOTAN_FN_ISA_SHANI SHA_1::sha1_compress_x86(digest_type& digest, |
| 75 |
std::span<const uint8_t> input_span, |
| 76 |
size_t blocks) { |
| 77 |
const uint8_t* input = input_span.data(); |
| 78 |
|
| 79 |
SIMD_4x32 ABCD = rev_words(SIMD_4x32::load_le(&digest[0])); // NOLINT(*-container-data-pointer) |
| 80 |
SIMD_4x32 E0 = SIMD_4x32(0, 0, 0, digest[4]); |
| 81 |
|
| 82 |
while(blocks > 0) { |
| 83 |
// Save current hash |
| 84 |
const auto ABCD_SAVE = ABCD; |
| 85 |
const auto E0_SAVE = E0; |
| 86 |
|
| 87 |
auto W0 = rev_words(SIMD_4x32::load_be(input)); |
| 88 |
auto W1 = rev_words(SIMD_4x32::load_be(input + 16)); |
| 89 |
auto W2 = rev_words(SIMD_4x32::load_be(input + 32)); |
| 90 |
auto W3 = rev_words(SIMD_4x32::load_be(input + 48)); |
| 91 |
|
| 92 |
sha1_x86_first8<0>(ABCD, E0, W0, W1); |
| 93 |
sha1_x86_rnds8<0>(ABCD, E0, W2, W3); |
| 94 |
|
| 95 |
W0 = sha1_x86_msg1(W0, W1); |
| 96 |
W1 = sha1_x86_msg1(W1, W2); |
| 97 |
W0 ^= W2; |
| 98 |
|
| 99 |
sha1_x86_next_msg(W3, W0, W1, W2); |
| 100 |
sha1_x86_next_msg(W0, W1, W2, W3); |
| 101 |
sha1_x86_rnds8<0, 1>(ABCD, E0, W0, W1); |
| 102 |
|
| 103 |
sha1_x86_next_msg(W1, W2, W3, W0); |
| 104 |
sha1_x86_next_msg(W2, W3, W0, W1); |
| 105 |
sha1_x86_rnds8<1>(ABCD, E0, W2, W3); |
| 106 |
|
| 107 |
sha1_x86_next_msg(W3, W0, W1, W2); |
| 108 |
sha1_x86_next_msg(W0, W1, W2, W3); |
| 109 |
sha1_x86_rnds8<1>(ABCD, E0, W0, W1); |
| 110 |
|
| 111 |
sha1_x86_next_msg(W1, W2, W3, W0); |
| 112 |
sha1_x86_next_msg(W2, W3, W0, W1); |
| 113 |
sha1_x86_rnds8<2>(ABCD, E0, W2, W3); |
| 114 |
|
| 115 |
sha1_x86_next_msg(W3, W0, W1, W2); |
| 116 |
sha1_x86_next_msg(W0, W1, W2, W3); |
| 117 |
sha1_x86_rnds8<2>(ABCD, E0, W0, W1); |
| 118 |
|
| 119 |
sha1_x86_next_msg(W1, W2, W3, W0); |
| 120 |
sha1_x86_next_msg(W2, W3, W0, W1); |
| 121 |
sha1_x86_rnds8<2, 3>(ABCD, E0, W2, W3); |
| 122 |
|
| 123 |
sha1_x86_next_msg(W3, W0, W1, W2); |
| 124 |
sha1_x86_next_msg(W0, W1, W2, W3); |
| 125 |
sha1_x86_rnds8<3>(ABCD, E0, W0, W1); |
| 126 |
|
| 127 |
sha1_x86_next_msg(W1, W2, W3, W0); |
| 128 |
sha1_x86_next_msg(W2, W3, W0, W1); |
| 129 |
sha1_x86_rnds8<3>(ABCD, E0, W2, W3); |
| 130 |
|
| 131 |
ABCD += ABCD_SAVE; |
| 132 |
E0 = sha1_x86_nexte(E0, E0_SAVE); |
| 133 |
|
| 134 |
input += 64; |
| 135 |
blocks--; |
| 136 |
} |
| 137 |
|
| 138 |
rev_words(ABCD).store_le(&digest[0]); // NOLINT(*-container-data-pointer) |
| 139 |
digest[4] = _mm_extract_epi32(E0.raw(), 3); |
| 140 |
} |
| 141 |
|
| 142 |
} // namespace Botan |