1 /*===------------- avx512vlbitalgintrin.h - BITALG intrinsics ------------------===
4 * Permission is hereby granted, free of charge, to any person obtaining a copy
5 * of this software and associated documentation files (the "Software"), to deal
6 * in the Software without restriction, including without limitation the rights
7 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
8 * copies of the Software, and to permit persons to whom the Software is
9 * furnished to do so, subject to the following conditions:
11 * The above copyright notice and this permission notice shall be included in
12 * all copies or substantial portions of the Software.
14 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
15 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
16 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
17 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
18 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
19 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
22 *===-----------------------------------------------------------------------===
25 #error "Never use <avx512vlbitalgintrin.h> directly; include <immintrin.h> instead."
28 #ifndef __AVX512VLBITALGINTRIN_H
29 #define __AVX512VLBITALGINTRIN_H
31 /* Define the default attributes for the functions in this file. */
32 #define __DEFAULT_FN_ATTRS __attribute__((__always_inline__, __nodebug__, __target__("avx512vl,avx512bitalg")))
34 static __inline__ __m256i __DEFAULT_FN_ATTRS
35 _mm256_popcnt_epi16(__m256i __A)
37 return (__m256i) __builtin_ia32_vpopcntw_256((__v16hi) __A);
40 static __inline__ __m256i __DEFAULT_FN_ATTRS
41 _mm256_mask_popcnt_epi16(__m256i __A, __mmask16 __U, __m256i __B)
43 return (__m256i) __builtin_ia32_selectw_256((__mmask16) __U,
44 (__v16hi) _mm256_popcnt_epi16(__B),
48 static __inline__ __m256i __DEFAULT_FN_ATTRS
49 _mm256_maskz_popcnt_epi16(__mmask16 __U, __m256i __B)
51 return _mm256_mask_popcnt_epi16((__m256i) _mm256_setzero_si256(),
56 static __inline__ __m128i __DEFAULT_FN_ATTRS
57 _mm128_popcnt_epi16(__m128i __A)
59 return (__m128i) __builtin_ia32_vpopcntw_128((__v8hi) __A);
62 static __inline__ __m128i __DEFAULT_FN_ATTRS
63 _mm128_mask_popcnt_epi16(__m128i __A, __mmask8 __U, __m128i __B)
65 return (__m128i) __builtin_ia32_selectw_128((__mmask8) __U,
66 (__v8hi) _mm128_popcnt_epi16(__B),
70 static __inline__ __m128i __DEFAULT_FN_ATTRS
71 _mm128_maskz_popcnt_epi16(__mmask8 __U, __m128i __B)
73 return _mm128_mask_popcnt_epi16((__m128i) _mm_setzero_si128(),
78 static __inline__ __m256i __DEFAULT_FN_ATTRS
79 _mm256_popcnt_epi8(__m256i __A)
81 return (__m256i) __builtin_ia32_vpopcntb_256((__v32qi) __A);
84 static __inline__ __m256i __DEFAULT_FN_ATTRS
85 _mm256_mask_popcnt_epi8(__m256i __A, __mmask32 __U, __m256i __B)
87 return (__m256i) __builtin_ia32_selectb_256((__mmask32) __U,
88 (__v32qi) _mm256_popcnt_epi8(__B),
92 static __inline__ __m256i __DEFAULT_FN_ATTRS
93 _mm256_maskz_popcnt_epi8(__mmask32 __U, __m256i __B)
95 return _mm256_mask_popcnt_epi8((__m256i) _mm256_setzero_si256(),
100 static __inline__ __m128i __DEFAULT_FN_ATTRS
101 _mm128_popcnt_epi8(__m128i __A)
103 return (__m128i) __builtin_ia32_vpopcntb_128((__v16qi) __A);
106 static __inline__ __m128i __DEFAULT_FN_ATTRS
107 _mm128_mask_popcnt_epi8(__m128i __A, __mmask16 __U, __m128i __B)
109 return (__m128i) __builtin_ia32_selectb_128((__mmask16) __U,
110 (__v16qi) _mm128_popcnt_epi8(__B),
114 static __inline__ __m128i __DEFAULT_FN_ATTRS
115 _mm128_maskz_popcnt_epi8(__mmask16 __U, __m128i __B)
117 return _mm128_mask_popcnt_epi8((__m128i) _mm_setzero_si128(),
122 static __inline__ __mmask32 __DEFAULT_FN_ATTRS
123 _mm256_mask_bitshuffle_epi32_mask(__mmask32 __U, __m256i __A, __m256i __B)
125 return (__mmask32) __builtin_ia32_vpshufbitqmb256_mask((__v32qi) __A,
130 static __inline__ __mmask32 __DEFAULT_FN_ATTRS
131 _mm256_bitshuffle_epi32_mask(__m256i __A, __m256i __B)
133 return _mm256_mask_bitshuffle_epi32_mask((__mmask32) -1,
138 static __inline__ __mmask16 __DEFAULT_FN_ATTRS
139 _mm128_mask_bitshuffle_epi16_mask(__mmask16 __U, __m128i __A, __m128i __B)
141 return (__mmask16) __builtin_ia32_vpshufbitqmb128_mask((__v16qi) __A,
146 static __inline__ __mmask16 __DEFAULT_FN_ATTRS
147 _mm128_bitshuffle_epi16_mask(__m128i __A, __m128i __B)
149 return _mm128_mask_bitshuffle_epi16_mask((__mmask16) -1,
155 #undef __DEFAULT_FN_ATTRS