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
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
|
/*===------------- avx512vlvnniintrin.h - VNNI intrinsics ------------------===
*
*
* Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
* See https://llvm.org/LICENSE.txt for license information.
* SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
*
*===-----------------------------------------------------------------------===
*/
#ifndef __IMMINTRIN_H
#error "Never use <avx512vlvnniintrin.h> directly; include <immintrin.h> instead."
#endif
#ifndef __AVX512VLVNNIINTRIN_H
#define __AVX512VLVNNIINTRIN_H
/* Define the default attributes for the functions in this file. */
#define __DEFAULT_FN_ATTRS128 __attribute__((__always_inline__, __nodebug__, __target__("avx512vl,avx512vnni"), __min_vector_width__(128)))
#define __DEFAULT_FN_ATTRS256 __attribute__((__always_inline__, __nodebug__, __target__("avx512vl,avx512vnni"), __min_vector_width__(256)))
/// 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
#define _mm256_dpbusd_epi32(S, A, B) \
((__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
#define _mm256_dpbusds_epi32(S, A, B) \
((__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
#define _mm256_dpwssd_epi32(S, A, B) \
((__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
#define _mm256_dpwssds_epi32(S, A, B) \
((__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
#define _mm_dpbusd_epi32(S, A, B) \
((__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
#define _mm_dpbusds_epi32(S, A, B) \
((__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
#define _mm_dpwssd_epi32(S, A, B) \
((__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
#define _mm_dpwssds_epi32(S, A, B) \
((__m128i)__builtin_ia32_vpdpwssds128((__v4si)(S), (__v4si)(A), (__v4si)(B)))
static __inline__ __m256i __DEFAULT_FN_ATTRS256
_mm256_mask_dpbusd_epi32(__m256i __S, __mmask8 __U, __m256i __A, __m256i __B)
{
return (__m256i)__builtin_ia32_selectd_256(__U,
(__v8si)_mm256_dpbusd_epi32(__S, __A, __B),
(__v8si)__S);
}
static __inline__ __m256i __DEFAULT_FN_ATTRS256
_mm256_maskz_dpbusd_epi32(__mmask8 __U, __m256i __S, __m256i __A, __m256i __B)
{
return (__m256i)__builtin_ia32_selectd_256(__U,
(__v8si)_mm256_dpbusd_epi32(__S, __A, __B),
(__v8si)_mm256_setzero_si256());
}
static __inline__ __m256i __DEFAULT_FN_ATTRS256
_mm256_mask_dpbusds_epi32(__m256i __S, __mmask8 __U, __m256i __A, __m256i __B)
{
return (__m256i)__builtin_ia32_selectd_256(__U,
(__v8si)_mm256_dpbusds_epi32(__S, __A, __B),
(__v8si)__S);
}
static __inline__ __m256i __DEFAULT_FN_ATTRS256
_mm256_maskz_dpbusds_epi32(__mmask8 __U, __m256i __S, __m256i __A, __m256i __B)
{
return (__m256i)__builtin_ia32_selectd_256(__U,
(__v8si)_mm256_dpbusds_epi32(__S, __A, __B),
(__v8si)_mm256_setzero_si256());
}
static __inline__ __m256i __DEFAULT_FN_ATTRS256
_mm256_mask_dpwssd_epi32(__m256i __S, __mmask8 __U, __m256i __A, __m256i __B)
{
return (__m256i)__builtin_ia32_selectd_256(__U,
(__v8si)_mm256_dpwssd_epi32(__S, __A, __B),
(__v8si)__S);
}
static __inline__ __m256i __DEFAULT_FN_ATTRS256
_mm256_maskz_dpwssd_epi32(__mmask8 __U, __m256i __S, __m256i __A, __m256i __B)
{
return (__m256i)__builtin_ia32_selectd_256(__U,
(__v8si)_mm256_dpwssd_epi32(__S, __A, __B),
(__v8si)_mm256_setzero_si256());
}
static __inline__ __m256i __DEFAULT_FN_ATTRS256
_mm256_mask_dpwssds_epi32(__m256i __S, __mmask8 __U, __m256i __A, __m256i __B)
{
return (__m256i)__builtin_ia32_selectd_256(__U,
(__v8si)_mm256_dpwssds_epi32(__S, __A, __B),
(__v8si)__S);
}
static __inline__ __m256i __DEFAULT_FN_ATTRS256
_mm256_maskz_dpwssds_epi32(__mmask8 __U, __m256i __S, __m256i __A, __m256i __B)
{
return (__m256i)__builtin_ia32_selectd_256(__U,
(__v8si)_mm256_dpwssds_epi32(__S, __A, __B),
(__v8si)_mm256_setzero_si256());
}
static __inline__ __m128i __DEFAULT_FN_ATTRS128
_mm_mask_dpbusd_epi32(__m128i __S, __mmask8 __U, __m128i __A, __m128i __B)
{
return (__m128i)__builtin_ia32_selectd_128(__U,
(__v4si)_mm_dpbusd_epi32(__S, __A, __B),
(__v4si)__S);
}
static __inline__ __m128i __DEFAULT_FN_ATTRS128
_mm_maskz_dpbusd_epi32(__mmask8 __U, __m128i __S, __m128i __A, __m128i __B)
{
return (__m128i)__builtin_ia32_selectd_128(__U,
(__v4si)_mm_dpbusd_epi32(__S, __A, __B),
(__v4si)_mm_setzero_si128());
}
static __inline__ __m128i __DEFAULT_FN_ATTRS128
_mm_mask_dpbusds_epi32(__m128i __S, __mmask8 __U, __m128i __A, __m128i __B)
{
return (__m128i)__builtin_ia32_selectd_128(__U,
(__v4si)_mm_dpbusds_epi32(__S, __A, __B),
(__v4si)__S);
}
static __inline__ __m128i __DEFAULT_FN_ATTRS128
_mm_maskz_dpbusds_epi32(__mmask8 __U, __m128i __S, __m128i __A, __m128i __B)
{
return (__m128i)__builtin_ia32_selectd_128(__U,
(__v4si)_mm_dpbusds_epi32(__S, __A, __B),
(__v4si)_mm_setzero_si128());
}
static __inline__ __m128i __DEFAULT_FN_ATTRS128
_mm_mask_dpwssd_epi32(__m128i __S, __mmask8 __U, __m128i __A, __m128i __B)
{
return (__m128i)__builtin_ia32_selectd_128(__U,
(__v4si)_mm_dpwssd_epi32(__S, __A, __B),
(__v4si)__S);
}
static __inline__ __m128i __DEFAULT_FN_ATTRS128
_mm_maskz_dpwssd_epi32(__mmask8 __U, __m128i __S, __m128i __A, __m128i __B)
{
return (__m128i)__builtin_ia32_selectd_128(__U,
(__v4si)_mm_dpwssd_epi32(__S, __A, __B),
(__v4si)_mm_setzero_si128());
}
static __inline__ __m128i __DEFAULT_FN_ATTRS128
_mm_mask_dpwssds_epi32(__m128i __S, __mmask8 __U, __m128i __A, __m128i __B)
{
return (__m128i)__builtin_ia32_selectd_128(__U,
(__v4si)_mm_dpwssds_epi32(__S, __A, __B),
(__v4si)__S);
}
static __inline__ __m128i __DEFAULT_FN_ATTRS128
_mm_maskz_dpwssds_epi32(__mmask8 __U, __m128i __S, __m128i __A, __m128i __B)
{
return (__m128i)__builtin_ia32_selectd_128(__U,
(__v4si)_mm_dpwssds_epi32(__S, __A, __B),
(__v4si)_mm_setzero_si128());
}
#undef __DEFAULT_FN_ATTRS128
#undef __DEFAULT_FN_ATTRS256
#endif
|
