11 "Never use <avxvnniint8intrin.h> directly; include <immintrin.h> instead."
14 #ifndef __AVXVNNIINT8INTRIN_H
15 #define __AVXVNNIINT8INTRIN_H
18 #define __DEFAULT_FN_ATTRS256 \
19 __attribute__((__always_inline__, __nodebug__, __target__("avxvnniint8"), \
20 __min_vector_width__(256)))
21 #define __DEFAULT_FN_ATTRS128 \
22 __attribute__((__always_inline__, __nodebug__, __target__("avxvnniint8"), \
23 __min_vector_width__(128)))
58 return (__m128i)__builtin_ia32_vpdpbssd128((__v4si)__W, (__v4si)__A,
94 return (__m256i)__builtin_ia32_vpdpbssd256((__v8si)__W, (__v8si)__A,
132 return (__m128i)__builtin_ia32_vpdpbssds128((__v4si)__W, (__v4si)__A,
169 return (__m256i)__builtin_ia32_vpdpbssds256((__v8si)__W, (__v8si)__A,
206 return (__m128i)__builtin_ia32_vpdpbsud128((__v4si)__W, (__v4si)__A,
242 return (__m256i)__builtin_ia32_vpdpbsud256((__v8si)__W, (__v8si)__A,
280 return (__m128i)__builtin_ia32_vpdpbsuds128((__v4si)__W, (__v4si)__A,
317 return (__m256i)__builtin_ia32_vpdpbsuds256((__v8si)__W, (__v8si)__A,
354 return (__m128i)__builtin_ia32_vpdpbuud128((__v4si)__W, (__v4si)__A,
390 return (__m256i)__builtin_ia32_vpdpbuud256((__v8si)__W, (__v8si)__A,
428 return (__m128i)__builtin_ia32_vpdpbuuds128((__v4si)__W, (__v4si)__A,
465 return (__m256i)__builtin_ia32_vpdpbuuds256((__v8si)__W, (__v8si)__A,
468 #undef __DEFAULT_FN_ATTRS128
469 #undef __DEFAULT_FN_ATTRS256
static __inline__ __m128i __DEFAULT_FN_ATTRS128 _mm_dpbssd_epi32(__m128i __W, __m128i __A, __m128i __B)
Multiply groups of 4 adjacent pairs of signed 8-bit integers in __A with corresponding signed 8-bit i...
static __inline__ __m256i __DEFAULT_FN_ATTRS256 _mm256_dpbsud_epi32(__m256i __W, __m256i __A, __m256i __B)
Multiply groups of 4 adjacent pairs of signed 8-bit integers in __A with corresponding unsigned 8-bit...
static __inline__ __m128i __DEFAULT_FN_ATTRS128 _mm_dpbsuds_epi32(__m128i __W, __m128i __A, __m128i __B)
Multiply groups of 4 adjacent pairs of signed 8-bit integers in __A with corresponding unsigned 8-bit...
static __inline__ __m128i __DEFAULT_FN_ATTRS128 _mm_dpbuuds_epi32(__m128i __W, __m128i __A, __m128i __B)
Multiply groups of 4 adjacent pairs of unsigned 8-bit integers in __A with corresponding unsigned 8-b...
static __inline__ __m256i __DEFAULT_FN_ATTRS256 _mm256_dpbssds_epi32(__m256i __W, __m256i __A, __m256i __B)
Multiply groups of 4 adjacent pairs of signed 8-bit integers in __A with corresponding signed 8-bit i...
static __inline__ __m256i __DEFAULT_FN_ATTRS256 _mm256_dpbuud_epi32(__m256i __W, __m256i __A, __m256i __B)
Multiply groups of 4 adjacent pairs of unsigned 8-bit integers in __A with corresponding unsigned 8-b...
#define __DEFAULT_FN_ATTRS256
static __inline__ __m256i __DEFAULT_FN_ATTRS256 _mm256_dpbuuds_epi32(__m256i __W, __m256i __A, __m256i __B)
Multiply groups of 4 adjacent pairs of signed 8-bit integers in __A with corresponding unsigned 8-bit...
#define __DEFAULT_FN_ATTRS128
static __inline__ __m256i __DEFAULT_FN_ATTRS256 _mm256_dpbsuds_epi32(__m256i __W, __m256i __A, __m256i __B)
Multiply groups of 4 adjacent pairs of signed 8-bit integers in __A with corresponding unsigned 8-bit...
static __inline__ __m256i __DEFAULT_FN_ATTRS256 _mm256_dpbssd_epi32(__m256i __W, __m256i __A, __m256i __B)
Multiply groups of 4 adjacent pairs of signed 8-bit integers in __A with corresponding signed 8-bit i...
static __inline__ __m128i __DEFAULT_FN_ATTRS128 _mm_dpbuud_epi32(__m128i __W, __m128i __A, __m128i __B)
Multiply groups of 4 adjacent pairs of unsigned 8-bit integers in __A with corresponding unsigned 8-b...
static __inline__ __m128i __DEFAULT_FN_ATTRS128 _mm_dpbsud_epi32(__m128i __W, __m128i __A, __m128i __B)
Multiply groups of 4 adjacent pairs of signed 8-bit integers in __A with corresponding unsigned 8-bit...
static __inline__ __m128i __DEFAULT_FN_ATTRS128 _mm_dpbssds_epi32(__m128i __W, __m128i __A, __m128i __B)
Multiply groups of 4 adjacent pairs of signed 8-bit integers in __A with corresponding signed 8-bit i...