Home | History | Annotate | Download | only in include
      1 /*===------------- avx512ifmavlintrin.h - IFMA intrinsics ------------------===
      2  *
      3  *
      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:
     10  *
     11  * The above copyright notice and this permission notice shall be included in
     12  * all copies or substantial portions of the Software.
     13  *
     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
     20  * THE SOFTWARE.
     21  *
     22  *===-----------------------------------------------------------------------===
     23  */
     24 #ifndef __IMMINTRIN_H
     25 #error "Never use <avx512ifmavlintrin.h> directly; include <immintrin.h> instead."
     26 #endif
     27 
     28 #ifndef __IFMAVLINTRIN_H
     29 #define __IFMAVLINTRIN_H
     30 
     31 /* Define the default attributes for the functions in this file. */
     32 #define __DEFAULT_FN_ATTRS __attribute__((__always_inline__, __nodebug__, __target__("avx512ifma,avx512vl")))
     33 
     34 
     35 
     36 static __inline__ __m128i __DEFAULT_FN_ATTRS
     37 _mm_madd52hi_epu64 (__m128i __X, __m128i __Y, __m128i __Z)
     38 {
     39   return (__m128i) __builtin_ia32_vpmadd52huq128_mask ((__v2di) __X,
     40                    (__v2di) __Y,
     41                    (__v2di) __Z,
     42                    (__mmask8) -1);
     43 }
     44 
     45 static __inline__ __m128i __DEFAULT_FN_ATTRS
     46 _mm_mask_madd52hi_epu64 (__m128i __W, __mmask8 __M, __m128i __X, __m128i __Y)
     47 {
     48   return (__m128i) __builtin_ia32_vpmadd52huq128_mask ((__v2di) __W,
     49                    (__v2di) __X,
     50                    (__v2di) __Y,
     51                    (__mmask8) __M);
     52 }
     53 
     54 static __inline__ __m128i __DEFAULT_FN_ATTRS
     55 _mm_maskz_madd52hi_epu64 (__mmask8 __M, __m128i __X, __m128i __Y, __m128i __Z)
     56 {
     57   return (__m128i) __builtin_ia32_vpmadd52huq128_maskz ((__v2di) __X,
     58               (__v2di) __Y,
     59               (__v2di) __Z,
     60               (__mmask8) __M);
     61 }
     62 
     63 static __inline__ __m256i __DEFAULT_FN_ATTRS
     64 _mm256_madd52hi_epu64 (__m256i __X, __m256i __Y, __m256i __Z)
     65 {
     66   return (__m256i) __builtin_ia32_vpmadd52huq256_mask ((__v4di) __X,
     67                    (__v4di) __Y,
     68                    (__v4di) __Z,
     69                    (__mmask8) -1);
     70 }
     71 
     72 static __inline__ __m256i __DEFAULT_FN_ATTRS
     73 _mm256_mask_madd52hi_epu64 (__m256i __W, __mmask8 __M, __m256i __X,
     74           __m256i __Y)
     75 {
     76   return (__m256i) __builtin_ia32_vpmadd52huq256_mask ((__v4di) __W,
     77                    (__v4di) __X,
     78                    (__v4di) __Y,
     79                    (__mmask8) __M);
     80 }
     81 
     82 static __inline__ __m256i __DEFAULT_FN_ATTRS
     83 _mm256_maskz_madd52hi_epu64 (__mmask8 __M, __m256i __X, __m256i __Y, __m256i __Z)
     84 {
     85   return (__m256i) __builtin_ia32_vpmadd52huq256_maskz ((__v4di) __X,
     86               (__v4di) __Y,
     87               (__v4di) __Z,
     88               (__mmask8) __M);
     89 }
     90 
     91 static __inline__ __m128i __DEFAULT_FN_ATTRS
     92 _mm_madd52lo_epu64 (__m128i __X, __m128i __Y, __m128i __Z)
     93 {
     94   return (__m128i) __builtin_ia32_vpmadd52luq128_mask ((__v2di) __X,
     95                    (__v2di) __Y,
     96                    (__v2di) __Z,
     97                    (__mmask8) -1);
     98 }
     99 
    100 static __inline__ __m128i __DEFAULT_FN_ATTRS
    101 _mm_mask_madd52lo_epu64 (__m128i __W, __mmask8 __M, __m128i __X, __m128i __Y)
    102 {
    103   return (__m128i) __builtin_ia32_vpmadd52luq128_mask ((__v2di) __W,
    104                    (__v2di) __X,
    105                    (__v2di) __Y,
    106                    (__mmask8) __M);
    107 }
    108 
    109 static __inline__ __m128i __DEFAULT_FN_ATTRS
    110 _mm_maskz_madd52lo_epu64 (__mmask8 __M, __m128i __X, __m128i __Y, __m128i __Z)
    111 {
    112   return (__m128i) __builtin_ia32_vpmadd52luq128_maskz ((__v2di) __X,
    113               (__v2di) __Y,
    114               (__v2di) __Z,
    115               (__mmask8) __M);
    116 }
    117 
    118 static __inline__ __m256i __DEFAULT_FN_ATTRS
    119 _mm256_madd52lo_epu64 (__m256i __X, __m256i __Y, __m256i __Z)
    120 {
    121   return (__m256i) __builtin_ia32_vpmadd52luq256_mask ((__v4di) __X,
    122                    (__v4di) __Y,
    123                    (__v4di) __Z,
    124                    (__mmask8) -1);
    125 }
    126 
    127 static __inline__ __m256i __DEFAULT_FN_ATTRS
    128 _mm256_mask_madd52lo_epu64 (__m256i __W, __mmask8 __M, __m256i __X,
    129           __m256i __Y)
    130 {
    131   return (__m256i) __builtin_ia32_vpmadd52luq256_mask ((__v4di) __W,
    132                    (__v4di) __X,
    133                    (__v4di) __Y,
    134                    (__mmask8) __M);
    135 }
    136 
    137 static __inline__ __m256i __DEFAULT_FN_ATTRS
    138 _mm256_maskz_madd52lo_epu64 (__mmask8 __M, __m256i __X, __m256i __Y, __m256i __Z)
    139 {
    140   return (__m256i) __builtin_ia32_vpmadd52luq256_maskz ((__v4di) __X,
    141               (__v4di) __Y,
    142               (__v4di) __Z,
    143               (__mmask8) __M);
    144 }
    145 
    146 
    147 #undef __DEFAULT_FN_ATTRS
    148 
    149 #endif
    150