avx512ifmavlintrin.h 5.3 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133
  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. #ifndef __IFMAVLINTRIN_H
  28. #define __IFMAVLINTRIN_H
  29. /* Define the default attributes for the functions in this file. */
  30. #define __DEFAULT_FN_ATTRS128 __attribute__((__always_inline__, __nodebug__, __target__("avx512ifma,avx512vl"), __min_vector_width__(128)))
  31. #define __DEFAULT_FN_ATTRS256 __attribute__((__always_inline__, __nodebug__, __target__("avx512ifma,avx512vl"), __min_vector_width__(256)))
  32. static __inline__ __m128i __DEFAULT_FN_ATTRS128
  33. _mm_madd52hi_epu64 (__m128i __X, __m128i __Y, __m128i __Z)
  34. {
  35. return (__m128i)__builtin_ia32_vpmadd52huq128((__v2di) __X, (__v2di) __Y,
  36. (__v2di) __Z);
  37. }
  38. static __inline__ __m128i __DEFAULT_FN_ATTRS128
  39. _mm_mask_madd52hi_epu64 (__m128i __W, __mmask8 __M, __m128i __X, __m128i __Y)
  40. {
  41. return (__m128i)__builtin_ia32_selectq_128(__M,
  42. (__v2di)_mm_madd52hi_epu64(__W, __X, __Y),
  43. (__v2di)__W);
  44. }
  45. static __inline__ __m128i __DEFAULT_FN_ATTRS128
  46. _mm_maskz_madd52hi_epu64 (__mmask8 __M, __m128i __X, __m128i __Y, __m128i __Z)
  47. {
  48. return (__m128i)__builtin_ia32_selectq_128(__M,
  49. (__v2di)_mm_madd52hi_epu64(__X, __Y, __Z),
  50. (__v2di)_mm_setzero_si128());
  51. }
  52. static __inline__ __m256i __DEFAULT_FN_ATTRS256
  53. _mm256_madd52hi_epu64 (__m256i __X, __m256i __Y, __m256i __Z)
  54. {
  55. return (__m256i)__builtin_ia32_vpmadd52huq256((__v4di)__X, (__v4di)__Y,
  56. (__v4di)__Z);
  57. }
  58. static __inline__ __m256i __DEFAULT_FN_ATTRS256
  59. _mm256_mask_madd52hi_epu64 (__m256i __W, __mmask8 __M, __m256i __X, __m256i __Y)
  60. {
  61. return (__m256i)__builtin_ia32_selectq_256(__M,
  62. (__v4di)_mm256_madd52hi_epu64(__W, __X, __Y),
  63. (__v4di)__W);
  64. }
  65. static __inline__ __m256i __DEFAULT_FN_ATTRS256
  66. _mm256_maskz_madd52hi_epu64 (__mmask8 __M, __m256i __X, __m256i __Y, __m256i __Z)
  67. {
  68. return (__m256i)__builtin_ia32_selectq_256(__M,
  69. (__v4di)_mm256_madd52hi_epu64(__X, __Y, __Z),
  70. (__v4di)_mm256_setzero_si256());
  71. }
  72. static __inline__ __m128i __DEFAULT_FN_ATTRS128
  73. _mm_madd52lo_epu64 (__m128i __X, __m128i __Y, __m128i __Z)
  74. {
  75. return (__m128i)__builtin_ia32_vpmadd52luq128((__v2di)__X, (__v2di)__Y,
  76. (__v2di)__Z);
  77. }
  78. static __inline__ __m128i __DEFAULT_FN_ATTRS128
  79. _mm_mask_madd52lo_epu64 (__m128i __W, __mmask8 __M, __m128i __X, __m128i __Y)
  80. {
  81. return (__m128i)__builtin_ia32_selectq_128(__M,
  82. (__v2di)_mm_madd52lo_epu64(__W, __X, __Y),
  83. (__v2di)__W);
  84. }
  85. static __inline__ __m128i __DEFAULT_FN_ATTRS128
  86. _mm_maskz_madd52lo_epu64 (__mmask8 __M, __m128i __X, __m128i __Y, __m128i __Z)
  87. {
  88. return (__m128i)__builtin_ia32_selectq_128(__M,
  89. (__v2di)_mm_madd52lo_epu64(__X, __Y, __Z),
  90. (__v2di)_mm_setzero_si128());
  91. }
  92. static __inline__ __m256i __DEFAULT_FN_ATTRS256
  93. _mm256_madd52lo_epu64 (__m256i __X, __m256i __Y, __m256i __Z)
  94. {
  95. return (__m256i)__builtin_ia32_vpmadd52luq256((__v4di)__X, (__v4di)__Y,
  96. (__v4di)__Z);
  97. }
  98. static __inline__ __m256i __DEFAULT_FN_ATTRS256
  99. _mm256_mask_madd52lo_epu64 (__m256i __W, __mmask8 __M, __m256i __X, __m256i __Y)
  100. {
  101. return (__m256i)__builtin_ia32_selectq_256(__M,
  102. (__v4di)_mm256_madd52lo_epu64(__W, __X, __Y),
  103. (__v4di)__W);
  104. }
  105. static __inline__ __m256i __DEFAULT_FN_ATTRS256
  106. _mm256_maskz_madd52lo_epu64 (__mmask8 __M, __m256i __X, __m256i __Y, __m256i __Z)
  107. {
  108. return (__m256i)__builtin_ia32_selectq_256(__M,
  109. (__v4di)_mm256_madd52lo_epu64(__X, __Y, __Z),
  110. (__v4di)_mm256_setzero_si256());
  111. }
  112. #undef __DEFAULT_FN_ATTRS128
  113. #undef __DEFAULT_FN_ATTRS256
  114. #endif