Training courses

Kernel and Embedded Linux

Bootlin training courses

Embedded Linux, kernel,
Yocto Project, Buildroot, real-time,
graphics, boot time, debugging...

Bootlin logo

Elixir Cross Referencer

Open Menu /external/apache2/llvm/dist/clang/lib/Headers/avxvnniintrin.h 
  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
/*===--------------- avxvnniintrin.h - VNNI intrinsics --------------------===
 *
 *
 * Permission is hereby granted, free of charge, to any person obtaining a copy
 * of this software and associated documentation files (the "Software"), to deal
 * in the Software without restriction, including without limitation the rights
 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
 * copies of the Software, and to permit persons to whom the Software is
 * furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be included in
 * all copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
 * THE SOFTWARE.
 *
 *===-----------------------------------------------------------------------===
 */
#ifndef __IMMINTRIN_H
#error "Never use <avxvnniintrin.h> directly; include <immintrin.h> instead."
#endif

#ifndef __AVXVNNIINTRIN_H
#define __AVXVNNIINTRIN_H

/* Below intrinsics defined in avx512vlvnniintrin.h can be used for AVXVNNI */
/// \fn __m256i _mm256_dpbusd_epi32(__m256i __S, __m256i __A, __m256i __B)
/// \fn __m256i _mm256_dpbusds_epi32(__m256i __S, __m256i __A, __m256i __B)
/// \fn __m256i _mm256_dpwssd_epi32(__m256i __S, __m256i __A, __m256i __B)
/// \fn __m256i _mm256_dpwssds_epi32(__m256i __S, __m256i __A, __m256i __B)
/// \fn __m128i _mm_dpbusd_epi32(__m128i __S, __m128i __A, __m128i __B)
/// \fn __m128i _mm_dpbusds_epi32(__m128i __S, __m128i __A, __m128i __B)
/// \fn __m128i _mm_dpwssd_epi32(__m128i __S, __m128i __A, __m128i __B)
/// \fn __m128i _mm_dpwssds_epi32(__m128i __S, __m128i __A, __m128i __B)

/* Intrinsics with _avx_ prefix are for compatibility with msvc. */
/* Define the default attributes for the functions in this file. */
#define __DEFAULT_FN_ATTRS256 __attribute__((__always_inline__, __nodebug__, __target__("avxvnni"), __min_vector_width__(256)))
#define __DEFAULT_FN_ATTRS128 __attribute__((__always_inline__, __nodebug__, __target__("avxvnni"), __min_vector_width__(128)))

/// Multiply groups of 4 adjacent pairs of unsigned 8-bit integers in \a __A with
/// corresponding signed 8-bit integers in \a __B, producing 4 intermediate signed
/// 16-bit results. Sum these 4 results with the corresponding 32-bit integer
/// in \a __S, and store the packed 32-bit results in DST.
///
/// This intrinsic corresponds to the <c> VPDPBUSD </c> instructions.
///
/// \operation
///    FOR j := 0 to 7
///      tmp1.word := Signed(ZeroExtend16(__A.byte[4*j]) * SignExtend16(__B.byte[4*j]))
///      tmp2.word := Signed(ZeroExtend16(__A.byte[4*j+1]) * SignExtend16(__B.byte[4*j+1]))
///      tmp3.word := Signed(ZeroExtend16(__A.byte[4*j+2]) * SignExtend16(__B.byte[4*j+2]))
///      tmp4.word := Signed(ZeroExtend16(__A.byte[4*j+3]) * SignExtend16(__B.byte[4*j+3]))
///      DST.dword[j] := __S.dword[j] + tmp1 + tmp2 + tmp3 + tmp4
///    ENDFOR
///    DST[MAX:256] := 0
/// \endoperation
static __inline__ __m256i __DEFAULT_FN_ATTRS256
_mm256_dpbusd_avx_epi32(__m256i __S, __m256i __A, __m256i __B)
{
  return (__m256i)__builtin_ia32_vpdpbusd256((__v8si)__S, (__v8si)__A, (__v8si)__B);
}

/// Multiply groups of 4 adjacent pairs of unsigned 8-bit integers in \a __A with
/// corresponding signed 8-bit integers in \a __B, producing 4 intermediate signed
/// 16-bit results. Sum these 4 results with the corresponding 32-bit integer
/// in \a __S using signed saturation, and store the packed 32-bit results in DST.
///
/// This intrinsic corresponds to the <c> VPDPBUSDS </c> instructions.
///
/// \operation
///    FOR j := 0 to 7
///      tmp1.word := Signed(ZeroExtend16(__A.byte[4*j]) * SignExtend16(__B.byte[4*j]))
///      tmp2.word := Signed(ZeroExtend16(__A.byte[4*j+1]) * SignExtend16(__B.byte[4*j+1]))
///      tmp3.word := Signed(ZeroExtend16(__A.byte[4*j+2]) * SignExtend16(__B.byte[4*j+2]))
///      tmp4.word := Signed(ZeroExtend16(__A.byte[4*j+3]) * SignExtend16(__B.byte[4*j+3]))
///      DST.dword[j] := Saturate32(__S.dword[j] + tmp1 + tmp2 + tmp3 + tmp4)
///    ENDFOR
///    DST[MAX:256] := 0
/// \endoperation
static __inline__ __m256i __DEFAULT_FN_ATTRS256
_mm256_dpbusds_avx_epi32(__m256i __S, __m256i __A, __m256i __B)
{
  return (__m256i)__builtin_ia32_vpdpbusds256((__v8si)__S, (__v8si)__A, (__v8si)__B);
}

/// Multiply groups of 2 adjacent pairs of signed 16-bit integers in \a __A with
/// corresponding 16-bit integers in \a __B, producing 2 intermediate signed 32-bit
/// results. Sum these 2 results with the corresponding 32-bit integer in \a __S,
///  and store the packed 32-bit results in DST.
///
/// This intrinsic corresponds to the <c> VPDPWSSD </c> instructions.
///
/// \operation
///    FOR j := 0 to 7
///      tmp1.dword := SignExtend32(__A.word[2*j]) * SignExtend32(__B.word[2*j])
///      tmp2.dword := SignExtend32(__A.word[2*j+1]) * SignExtend32(__B.word[2*j+1])
///      DST.dword[j] := __S.dword[j] + tmp1 + tmp2
///    ENDFOR
///    DST[MAX:256] := 0
/// \endoperation
static __inline__ __m256i __DEFAULT_FN_ATTRS256
_mm256_dpwssd_avx_epi32(__m256i __S, __m256i __A, __m256i __B)
{
  return (__m256i)__builtin_ia32_vpdpwssd256((__v8si)__S, (__v8si)__A, (__v8si)__B);
}

/// Multiply groups of 2 adjacent pairs of signed 16-bit integers in \a __A with
/// corresponding 16-bit integers in \a __B, producing 2 intermediate signed 32-bit
/// results. Sum these 2 results with the corresponding 32-bit integer in \a __S
/// using signed saturation, and store the packed 32-bit results in DST.
///
/// This intrinsic corresponds to the <c> VPDPWSSDS </c> instructions.
///
/// \operation
///    FOR j := 0 to 7
///      tmp1.dword := SignExtend32(__A.word[2*j]) * SignExtend32(__B.word[2*j])
///      tmp2.dword := SignExtend32(__A.word[2*j+1]) * SignExtend32(__B.word[2*j+1])
///      DST.dword[j] := Saturate32(__S.dword[j] + tmp1 + tmp2)
///    ENDFOR
///    DST[MAX:256] := 0
/// \endoperation
static __inline__ __m256i __DEFAULT_FN_ATTRS256
_mm256_dpwssds_avx_epi32(__m256i __S, __m256i __A, __m256i __B)
{
  return (__m256i)__builtin_ia32_vpdpwssds256((__v8si)__S, (__v8si)__A, (__v8si)__B);
}

/// Multiply groups of 4 adjacent pairs of unsigned 8-bit integers in \a __A with
/// corresponding signed 8-bit integers in \a __B, producing 4 intermediate signed
/// 16-bit results. Sum these 4 results with the corresponding 32-bit integer
/// in \a __S, and store the packed 32-bit results in DST.
///
/// This intrinsic corresponds to the <c> VPDPBUSD </c> instructions.
///
/// \operation
///    FOR j := 0 to 3
///      tmp1.word := Signed(ZeroExtend16(__A.byte[4*j]) * SignExtend16(__B.byte[4*j]))
///      tmp2.word := Signed(ZeroExtend16(__A.byte[4*j+1]) * SignExtend16(__B.byte[4*j+1]))
///      tmp3.word := Signed(ZeroExtend16(__A.byte[4*j+2]) * SignExtend16(__B.byte[4*j+2]))
///      tmp4.word := Signed(ZeroExtend16(__A.byte[4*j+3]) * SignExtend16(__B.byte[4*j+3]))
///      DST.dword[j] := __S.dword[j] + tmp1 + tmp2 + tmp3 + tmp4
///    ENDFOR
///    DST[MAX:128] := 0
/// \endoperation
static __inline__ __m128i __DEFAULT_FN_ATTRS128
_mm_dpbusd_avx_epi32(__m128i __S, __m128i __A, __m128i __B)
{
  return (__m128i)__builtin_ia32_vpdpbusd128((__v4si)__S, (__v4si)__A, (__v4si)__B);
}

/// Multiply groups of 4 adjacent pairs of unsigned 8-bit integers in \a __A with
/// corresponding signed 8-bit integers in \a __B, producing 4 intermediate signed
/// 16-bit results. Sum these 4 results with the corresponding 32-bit integer
/// in \a __S using signed saturation, and store the packed 32-bit results in DST.
///
/// This intrinsic corresponds to the <c> VPDPBUSDS </c> instructions.
///
/// \operation
///    FOR j := 0 to 3
///      tmp1.word := Signed(ZeroExtend16(__A.byte[4*j]) * SignExtend16(__B.byte[4*j]))
///      tmp2.word := Signed(ZeroExtend16(__A.byte[4*j+1]) * SignExtend16(__B.byte[4*j+1]))
///      tmp3.word := Signed(ZeroExtend16(__A.byte[4*j+2]) * SignExtend16(__B.byte[4*j+2]))
///      tmp4.word := Signed(ZeroExtend16(__A.byte[4*j+3]) * SignExtend16(__B.byte[4*j+3]))
///      DST.dword[j] := Saturate32(__S.dword[j] + tmp1 + tmp2 + tmp3 + tmp4)
///    ENDFOR
///    DST[MAX:128] := 0
/// \endoperation
static __inline__ __m128i __DEFAULT_FN_ATTRS128
_mm_dpbusds_avx_epi32(__m128i __S, __m128i __A, __m128i __B)
{
  return (__m128i)__builtin_ia32_vpdpbusds128((__v4si)__S, (__v4si)__A, (__v4si)__B);
}

/// Multiply groups of 2 adjacent pairs of signed 16-bit integers in \a __A with
/// corresponding 16-bit integers in \a __B, producing 2 intermediate signed 32-bit
/// results. Sum these 2 results with the corresponding 32-bit integer in \a __S,
/// and store the packed 32-bit results in DST.
///
/// This intrinsic corresponds to the <c> VPDPWSSD </c> instructions.
///
/// \operation
///    FOR j := 0 to 3
///      tmp1.dword := SignExtend32(__A.word[2*j]) * SignExtend32(__B.word[2*j])
///      tmp2.dword := SignExtend32(__A.word[2*j+1]) * SignExtend32(__B.word[2*j+1])
///      DST.dword[j] := __S.dword[j] + tmp1 + tmp2
///    ENDFOR
///    DST[MAX:128] := 0
/// \endoperation
static __inline__ __m128i __DEFAULT_FN_ATTRS128
_mm_dpwssd_avx_epi32(__m128i __S, __m128i __A, __m128i __B)
{
  return (__m128i)__builtin_ia32_vpdpwssd128((__v4si)__S, (__v4si)__A, (__v4si)__B);
}

/// Multiply groups of 2 adjacent pairs of signed 16-bit integers in \a __A with
/// corresponding 16-bit integers in \a __B, producing 2 intermediate signed 32-bit
/// results. Sum these 2 results with the corresponding 32-bit integer in \a __S
/// using signed saturation, and store the packed 32-bit results in DST.
///
/// This intrinsic corresponds to the <c> VPDPWSSDS </c> instructions.
///
/// \operation
///    FOR j := 0 to 3
///      tmp1.dword := SignExtend32(__A.word[2*j]) * SignExtend32(__B.word[2*j])
///      tmp2.dword := SignExtend32(__A.word[2*j+1]) * SignExtend32(__B.word[2*j+1])
///      DST.dword[j] := Saturate32(__S.dword[j] + tmp1 + tmp2)
///    ENDFOR
///    DST[MAX:128] := 0
/// \endoperation
static __inline__ __m128i __DEFAULT_FN_ATTRS128
_mm_dpwssds_avx_epi32(__m128i __S, __m128i __A, __m128i __B)
{
  return (__m128i)__builtin_ia32_vpdpwssds128((__v4si)__S, (__v4si)__A, (__v4si)__B);
}

#undef __DEFAULT_FN_ATTRS128
#undef __DEFAULT_FN_ATTRS256

#endif // __AVXVNNIINTRIN_H