Coverage Report

Created: 2020-11-21 08:34

/src/botan/src/lib/utils/ghash/ghash_cpu/ghash_cpu.cpp
Line
Count
Source
1
/*
2
* Hook for CLMUL/PMULL/VPMSUM
3
* (C) 2013,2017,2019,2020 Jack Lloyd
4
*
5
* Botan is released under the Simplified BSD License (see license.txt)
6
*/
7
8
#include <botan/internal/ghash.h>
9
#include <botan/internal/simd_32.h>
10
11
#if defined(BOTAN_SIMD_USE_SSE2)
12
  #include <immintrin.h>
13
  #include <wmmintrin.h>
14
#endif
15
16
namespace Botan {
17
18
namespace {
19
20
BOTAN_FORCE_INLINE SIMD_4x32 BOTAN_FUNC_ISA(BOTAN_VPERM_ISA) reverse_vector(const SIMD_4x32& in)
21
38.8k
   {
22
38.8k
#if defined(BOTAN_SIMD_USE_SSE2)
23
38.8k
   const __m128i BSWAP_MASK = _mm_set_epi8(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15);
24
38.8k
   return SIMD_4x32(_mm_shuffle_epi8(in.raw(), BSWAP_MASK));
25
#elif defined(BOTAN_SIMD_USE_NEON)
26
   const uint8_t maskb[16] = { 15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0 };
27
   const uint8x16_t mask = vld1q_u8(maskb);
28
   return SIMD_4x32(vreinterpretq_u32_u8(vqtbl1q_u8(vreinterpretq_u8_u32(in.raw()), mask)));
29
#elif defined(BOTAN_SIMD_USE_ALTIVEC)
30
   const __vector unsigned char mask = {15,14,13,12, 11,10,9,8, 7,6,5,4, 3,2,1,0};
31
   return SIMD_4x32(vec_perm(in.raw(), in.raw(), mask));
32
#endif
33
38.8k
   }
34
35
template<int M>
36
BOTAN_FORCE_INLINE SIMD_4x32 BOTAN_FUNC_ISA(BOTAN_CLMUL_ISA) clmul(const SIMD_4x32& H, const SIMD_4x32& x)
37
104k
   {
38
104k
   static_assert(M == 0x00 || M == 0x01 || M == 0x10 || M == 0x11, "Valid clmul mode");
39
40
104k
#if defined(BOTAN_SIMD_USE_SSE2)
41
104k
   return SIMD_4x32(_mm_clmulepi64_si128(x.raw(), H.raw(), M));
42
#elif defined(BOTAN_SIMD_USE_NEON)
43
   const uint64_t a = vgetq_lane_u64(vreinterpretq_u64_u32(x.raw()), M & 0x01);
44
   const uint64_t b = vgetq_lane_u64(vreinterpretq_u64_u32(H.raw()), (M & 0x10) >> 4);
45
   return SIMD_4x32(reinterpret_cast<uint32x4_t>(vmull_p64(a, b)));
46
#elif defined(BOTAN_SIMD_USE_ALTIVEC)
47
   const SIMD_4x32 mask_lo = SIMD_4x32(0, 0, 0xFFFFFFFF, 0xFFFFFFFF);
48
49
   SIMD_4x32 i1 = x;
50
   SIMD_4x32 i2 = H;
51
52
   if(M == 0x11)
53
      {
54
      i1 &= mask_lo;
55
      i2 &= mask_lo;
56
      }
57
   else if(M == 0x10)
58
      {
59
      i1 = i1.shift_elems_left<2>();
60
      }
61
   else if(M == 0x01)
62
      {
63
      i2 = i2.shift_elems_left<2>();
64
      }
65
   else if(M == 0x00)
66
      {
67
      i1 = mask_lo.andc(i1);
68
      i2 = mask_lo.andc(i2);
69
      }
70
71
   return SIMD_4x32(
72
      reinterpret_cast<__vector unsigned int>(
73
         __builtin_crypto_vpmsumd(
74
            reinterpret_cast<__vector unsigned long>(i1.raw()),
75
            reinterpret_cast<__vector unsigned long>(i2.raw())))
76
      );
77
#endif
78
104k
   }
ghash_cpu.cpp:Botan::SIMD_4x32 Botan::(anonymous namespace)::clmul<17>(Botan::SIMD_4x32 const&, Botan::SIMD_4x32 const&)
Line
Count
Source
37
32.3k
   {
38
32.3k
   static_assert(M == 0x00 || M == 0x01 || M == 0x10 || M == 0x11, "Valid clmul mode");
39
40
32.3k
#if defined(BOTAN_SIMD_USE_SSE2)
41
32.3k
   return SIMD_4x32(_mm_clmulepi64_si128(x.raw(), H.raw(), M));
42
#elif defined(BOTAN_SIMD_USE_NEON)
43
   const uint64_t a = vgetq_lane_u64(vreinterpretq_u64_u32(x.raw()), M & 0x01);
44
   const uint64_t b = vgetq_lane_u64(vreinterpretq_u64_u32(H.raw()), (M & 0x10) >> 4);
45
   return SIMD_4x32(reinterpret_cast<uint32x4_t>(vmull_p64(a, b)));
46
#elif defined(BOTAN_SIMD_USE_ALTIVEC)
47
   const SIMD_4x32 mask_lo = SIMD_4x32(0, 0, 0xFFFFFFFF, 0xFFFFFFFF);
48
49
   SIMD_4x32 i1 = x;
50
   SIMD_4x32 i2 = H;
51
52
   if(M == 0x11)
53
      {
54
      i1 &= mask_lo;
55
      i2 &= mask_lo;
56
      }
57
   else if(M == 0x10)
58
      {
59
      i1 = i1.shift_elems_left<2>();
60
      }
61
   else if(M == 0x01)
62
      {
63
      i2 = i2.shift_elems_left<2>();
64
      }
65
   else if(M == 0x00)
66
      {
67
      i1 = mask_lo.andc(i1);
68
      i2 = mask_lo.andc(i2);
69
      }
70
71
   return SIMD_4x32(
72
      reinterpret_cast<__vector unsigned int>(
73
         __builtin_crypto_vpmsumd(
74
            reinterpret_cast<__vector unsigned long>(i1.raw()),
75
            reinterpret_cast<__vector unsigned long>(i2.raw())))
76
      );
77
#endif
78
32.3k
   }
ghash_cpu.cpp:Botan::SIMD_4x32 Botan::(anonymous namespace)::clmul<16>(Botan::SIMD_4x32 const&, Botan::SIMD_4x32 const&)
Line
Count
Source
37
7.12k
   {
38
7.12k
   static_assert(M == 0x00 || M == 0x01 || M == 0x10 || M == 0x11, "Valid clmul mode");
39
40
7.12k
#if defined(BOTAN_SIMD_USE_SSE2)
41
7.12k
   return SIMD_4x32(_mm_clmulepi64_si128(x.raw(), H.raw(), M));
42
#elif defined(BOTAN_SIMD_USE_NEON)
43
   const uint64_t a = vgetq_lane_u64(vreinterpretq_u64_u32(x.raw()), M & 0x01);
44
   const uint64_t b = vgetq_lane_u64(vreinterpretq_u64_u32(H.raw()), (M & 0x10) >> 4);
45
   return SIMD_4x32(reinterpret_cast<uint32x4_t>(vmull_p64(a, b)));
46
#elif defined(BOTAN_SIMD_USE_ALTIVEC)
47
   const SIMD_4x32 mask_lo = SIMD_4x32(0, 0, 0xFFFFFFFF, 0xFFFFFFFF);
48
49
   SIMD_4x32 i1 = x;
50
   SIMD_4x32 i2 = H;
51
52
   if(M == 0x11)
53
      {
54
      i1 &= mask_lo;
55
      i2 &= mask_lo;
56
      }
57
   else if(M == 0x10)
58
      {
59
      i1 = i1.shift_elems_left<2>();
60
      }
61
   else if(M == 0x01)
62
      {
63
      i2 = i2.shift_elems_left<2>();
64
      }
65
   else if(M == 0x00)
66
      {
67
      i1 = mask_lo.andc(i1);
68
      i2 = mask_lo.andc(i2);
69
      }
70
71
   return SIMD_4x32(
72
      reinterpret_cast<__vector unsigned int>(
73
         __builtin_crypto_vpmsumd(
74
            reinterpret_cast<__vector unsigned long>(i1.raw()),
75
            reinterpret_cast<__vector unsigned long>(i2.raw())))
76
      );
77
#endif
78
7.12k
   }
ghash_cpu.cpp:Botan::SIMD_4x32 Botan::(anonymous namespace)::clmul<1>(Botan::SIMD_4x32 const&, Botan::SIMD_4x32 const&)
Line
Count
Source
37
7.12k
   {
38
7.12k
   static_assert(M == 0x00 || M == 0x01 || M == 0x10 || M == 0x11, "Valid clmul mode");
39
40
7.12k
#if defined(BOTAN_SIMD_USE_SSE2)
41
7.12k
   return SIMD_4x32(_mm_clmulepi64_si128(x.raw(), H.raw(), M));
42
#elif defined(BOTAN_SIMD_USE_NEON)
43
   const uint64_t a = vgetq_lane_u64(vreinterpretq_u64_u32(x.raw()), M & 0x01);
44
   const uint64_t b = vgetq_lane_u64(vreinterpretq_u64_u32(H.raw()), (M & 0x10) >> 4);
45
   return SIMD_4x32(reinterpret_cast<uint32x4_t>(vmull_p64(a, b)));
46
#elif defined(BOTAN_SIMD_USE_ALTIVEC)
47
   const SIMD_4x32 mask_lo = SIMD_4x32(0, 0, 0xFFFFFFFF, 0xFFFFFFFF);
48
49
   SIMD_4x32 i1 = x;
50
   SIMD_4x32 i2 = H;
51
52
   if(M == 0x11)
53
      {
54
      i1 &= mask_lo;
55
      i2 &= mask_lo;
56
      }
57
   else if(M == 0x10)
58
      {
59
      i1 = i1.shift_elems_left<2>();
60
      }
61
   else if(M == 0x01)
62
      {
63
      i2 = i2.shift_elems_left<2>();
64
      }
65
   else if(M == 0x00)
66
      {
67
      i1 = mask_lo.andc(i1);
68
      i2 = mask_lo.andc(i2);
69
      }
70
71
   return SIMD_4x32(
72
      reinterpret_cast<__vector unsigned int>(
73
         __builtin_crypto_vpmsumd(
74
            reinterpret_cast<__vector unsigned long>(i1.raw()),
75
            reinterpret_cast<__vector unsigned long>(i2.raw())))
76
      );
77
#endif
78
7.12k
   }
ghash_cpu.cpp:Botan::SIMD_4x32 Botan::(anonymous namespace)::clmul<0>(Botan::SIMD_4x32 const&, Botan::SIMD_4x32 const&)
Line
Count
Source
37
57.5k
   {
38
57.5k
   static_assert(M == 0x00 || M == 0x01 || M == 0x10 || M == 0x11, "Valid clmul mode");
39
40
57.5k
#if defined(BOTAN_SIMD_USE_SSE2)
41
57.5k
   return SIMD_4x32(_mm_clmulepi64_si128(x.raw(), H.raw(), M));
42
#elif defined(BOTAN_SIMD_USE_NEON)
43
   const uint64_t a = vgetq_lane_u64(vreinterpretq_u64_u32(x.raw()), M & 0x01);
44
   const uint64_t b = vgetq_lane_u64(vreinterpretq_u64_u32(H.raw()), (M & 0x10) >> 4);
45
   return SIMD_4x32(reinterpret_cast<uint32x4_t>(vmull_p64(a, b)));
46
#elif defined(BOTAN_SIMD_USE_ALTIVEC)
47
   const SIMD_4x32 mask_lo = SIMD_4x32(0, 0, 0xFFFFFFFF, 0xFFFFFFFF);
48
49
   SIMD_4x32 i1 = x;
50
   SIMD_4x32 i2 = H;
51
52
   if(M == 0x11)
53
      {
54
      i1 &= mask_lo;
55
      i2 &= mask_lo;
56
      }
57
   else if(M == 0x10)
58
      {
59
      i1 = i1.shift_elems_left<2>();
60
      }
61
   else if(M == 0x01)
62
      {
63
      i2 = i2.shift_elems_left<2>();
64
      }
65
   else if(M == 0x00)
66
      {
67
      i1 = mask_lo.andc(i1);
68
      i2 = mask_lo.andc(i2);
69
      }
70
71
   return SIMD_4x32(
72
      reinterpret_cast<__vector unsigned int>(
73
         __builtin_crypto_vpmsumd(
74
            reinterpret_cast<__vector unsigned long>(i1.raw()),
75
            reinterpret_cast<__vector unsigned long>(i2.raw())))
76
      );
77
#endif
78
57.5k
   }
79
80
inline SIMD_4x32 gcm_reduce(const SIMD_4x32& B0, const SIMD_4x32& B1)
81
13.4k
   {
82
13.4k
   SIMD_4x32 X0 = B1.shr<31>();
83
13.4k
   SIMD_4x32 X1 = B1.shl<1>();
84
13.4k
   SIMD_4x32 X2 = B0.shr<31>();
85
13.4k
   SIMD_4x32 X3 = B0.shl<1>();
86
87
13.4k
   X3 |= X0.shift_elems_right<3>();
88
13.4k
   X3 |= X2.shift_elems_left<1>();
89
13.4k
   X1 |= X0.shift_elems_left<1>();
90
91
13.4k
   X0 = X1.shl<31>() ^ X1.shl<30>() ^ X1.shl<25>();
92
93
13.4k
   X1 ^= X0.shift_elems_left<3>();
94
95
13.4k
   X0 = X1 ^ X3 ^ X0.shift_elems_right<1>();
96
13.4k
   X0 ^= X1.shr<7>() ^ X1.shr<2>() ^ X1.shr<1>();
97
13.4k
   return X0;
98
13.4k
   }
99
100
inline SIMD_4x32 BOTAN_FUNC_ISA(BOTAN_CLMUL_ISA) gcm_multiply(const SIMD_4x32& H, const SIMD_4x32& x)
101
7.12k
   {
102
7.12k
   SIMD_4x32 T0 = clmul<0x11>(H, x);
103
7.12k
   SIMD_4x32 T1 = clmul<0x10>(H, x);
104
7.12k
   SIMD_4x32 T2 = clmul<0x01>(H, x);
105
7.12k
   SIMD_4x32 T3 = clmul<0x00>(H, x);
106
107
7.12k
   T1 ^= T2;
108
7.12k
   T0 ^= T1.shift_elems_right<2>();
109
7.12k
   T3 ^= T1.shift_elems_left<2>();
110
111
7.12k
   return gcm_reduce(T0, T3);
112
7.12k
   }
113
114
inline SIMD_4x32 BOTAN_FUNC_ISA(BOTAN_CLMUL_ISA)
115
   gcm_multiply_x4(const SIMD_4x32& H1, const SIMD_4x32& H2, const SIMD_4x32& H3, const SIMD_4x32& H4,
116
                   const SIMD_4x32& X1, const SIMD_4x32& X2, const SIMD_4x32& X3, const SIMD_4x32& X4)
117
6.30k
   {
118
   /*
119
   * Mutiply with delayed reduction, algorithm by Krzysztof Jankowski
120
   * and Pierre Laurent of Intel
121
   */
122
123
6.30k
   const SIMD_4x32 lo = (clmul<0x00>(H1, X1) ^ clmul<0x00>(H2, X2)) ^
124
6.30k
                        (clmul<0x00>(H3, X3) ^ clmul<0x00>(H4, X4));
125
126
6.30k
   const SIMD_4x32 hi = (clmul<0x11>(H1, X1) ^ clmul<0x11>(H2, X2)) ^
127
6.30k
                        (clmul<0x11>(H3, X3) ^ clmul<0x11>(H4, X4));
128
129
6.30k
   SIMD_4x32 T;
130
131
6.30k
   T ^= clmul<0x00>(H1 ^ H1.shift_elems_right<2>(), X1 ^ X1.shift_elems_right<2>());
132
6.30k
   T ^= clmul<0x00>(H2 ^ H2.shift_elems_right<2>(), X2 ^ X2.shift_elems_right<2>());
133
6.30k
   T ^= clmul<0x00>(H3 ^ H3.shift_elems_right<2>(), X3 ^ X3.shift_elems_right<2>());
134
6.30k
   T ^= clmul<0x00>(H4 ^ H4.shift_elems_right<2>(), X4 ^ X4.shift_elems_right<2>());
135
6.30k
   T ^= lo;
136
6.30k
   T ^= hi;
137
138
6.30k
   return gcm_reduce(hi ^ T.shift_elems_right<2>(),
139
6.30k
                     lo ^ T.shift_elems_left<2>());
140
6.30k
   }
141
142
}
143
144
BOTAN_FUNC_ISA(BOTAN_VPERM_ISA)
145
void GHASH::ghash_precompute_cpu(const uint8_t H_bytes[16], uint64_t H_pow[4*2])
146
929
   {
147
929
   const SIMD_4x32 H1 = reverse_vector(SIMD_4x32::load_le(H_bytes));
148
929
   const SIMD_4x32 H2 = gcm_multiply(H1, H1);
149
929
   const SIMD_4x32 H3 = gcm_multiply(H1, H2);
150
929
   const SIMD_4x32 H4 = gcm_multiply(H2, H2);
151
152
929
   H1.store_le(H_pow);
153
929
   H2.store_le(H_pow + 2);
154
929
   H3.store_le(H_pow + 4);
155
929
   H4.store_le(H_pow + 6);
156
929
   }
157
158
BOTAN_FUNC_ISA(BOTAN_VPERM_ISA)
159
void GHASH::ghash_multiply_cpu(uint8_t x[16],
160
                               const uint64_t H_pow[8],
161
                               const uint8_t input[], size_t blocks)
162
4.16k
   {
163
   /*
164
   * Algorithms 1 and 5 from Intel's CLMUL guide
165
   */
166
4.16k
   const SIMD_4x32 H1 = SIMD_4x32::load_le(H_pow);
167
168
4.16k
   SIMD_4x32 a = reverse_vector(SIMD_4x32::load_le(x));
169
170
4.16k
   if(blocks >= 4)
171
528
      {
172
528
      const SIMD_4x32 H2 = SIMD_4x32::load_le(H_pow + 2);
173
528
      const SIMD_4x32 H3 = SIMD_4x32::load_le(H_pow + 4);
174
528
      const SIMD_4x32 H4 = SIMD_4x32::load_le(H_pow + 6);
175
176
6.83k
      while(blocks >= 4)
177
6.30k
         {
178
6.30k
         const SIMD_4x32 m0 = reverse_vector(SIMD_4x32::load_le(input       ));
179
6.30k
         const SIMD_4x32 m1 = reverse_vector(SIMD_4x32::load_le(input + 16*1));
180
6.30k
         const SIMD_4x32 m2 = reverse_vector(SIMD_4x32::load_le(input + 16*2));
181
6.30k
         const SIMD_4x32 m3 = reverse_vector(SIMD_4x32::load_le(input + 16*3));
182
183
6.30k
         a ^= m0;
184
6.30k
         a = gcm_multiply_x4(H1, H2, H3, H4, m3, m2, m1, a);
185
186
6.30k
         input += 4*16;
187
6.30k
         blocks -= 4;
188
6.30k
         }
189
528
      }
190
191
8.49k
   for(size_t i = 0; i != blocks; ++i)
192
4.33k
      {
193
4.33k
      const SIMD_4x32 m = reverse_vector(SIMD_4x32::load_le(input + 16*i));
194
195
4.33k
      a ^= m;
196
4.33k
      a = gcm_multiply(H1, a);
197
4.33k
      }
198
199
4.16k
   a = reverse_vector(a);
200
4.16k
   a.store_le(x);
201
4.16k
   }
202
203
}