avx512vnniintrin.h 5.1 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129
  1. /*===------------- avx512vnniintrin.h - VNNI 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 <avx512vnniintrin.h> directly; include <immintrin.h> instead."
  26. #endif
  27. #ifndef __AVX512VNNIINTRIN_H
  28. #define __AVX512VNNIINTRIN_H
  29. /* Define the default attributes for the functions in this file. */
  30. #define __DEFAULT_FN_ATTRS __attribute__((__always_inline__, __nodebug__, __target__("avx512vnni"), __min_vector_width__(512)))
  31. static __inline__ __m512i __DEFAULT_FN_ATTRS
  32. _mm512_dpbusd_epi32(__m512i __S, __m512i __A, __m512i __B)
  33. {
  34. return (__m512i)__builtin_ia32_vpdpbusd512((__v16si)__S, (__v16si)__A,
  35. (__v16si)__B);
  36. }
  37. static __inline__ __m512i __DEFAULT_FN_ATTRS
  38. _mm512_mask_dpbusd_epi32(__m512i __S, __mmask16 __U, __m512i __A, __m512i __B)
  39. {
  40. return (__m512i)__builtin_ia32_selectd_512(__U,
  41. (__v16si)_mm512_dpbusd_epi32(__S, __A, __B),
  42. (__v16si)__S);
  43. }
  44. static __inline__ __m512i __DEFAULT_FN_ATTRS
  45. _mm512_maskz_dpbusd_epi32(__mmask16 __U, __m512i __S, __m512i __A, __m512i __B)
  46. {
  47. return (__m512i)__builtin_ia32_selectd_512(__U,
  48. (__v16si)_mm512_dpbusd_epi32(__S, __A, __B),
  49. (__v16si)_mm512_setzero_si512());
  50. }
  51. static __inline__ __m512i __DEFAULT_FN_ATTRS
  52. _mm512_dpbusds_epi32(__m512i __S, __m512i __A, __m512i __B)
  53. {
  54. return (__m512i)__builtin_ia32_vpdpbusds512((__v16si)__S, (__v16si)__A,
  55. (__v16si)__B);
  56. }
  57. static __inline__ __m512i __DEFAULT_FN_ATTRS
  58. _mm512_mask_dpbusds_epi32(__m512i __S, __mmask16 __U, __m512i __A, __m512i __B)
  59. {
  60. return (__m512i)__builtin_ia32_selectd_512(__U,
  61. (__v16si)_mm512_dpbusds_epi32(__S, __A, __B),
  62. (__v16si)__S);
  63. }
  64. static __inline__ __m512i __DEFAULT_FN_ATTRS
  65. _mm512_maskz_dpbusds_epi32(__mmask16 __U, __m512i __S, __m512i __A, __m512i __B)
  66. {
  67. return (__m512i)__builtin_ia32_selectd_512(__U,
  68. (__v16si)_mm512_dpbusds_epi32(__S, __A, __B),
  69. (__v16si)_mm512_setzero_si512());
  70. }
  71. static __inline__ __m512i __DEFAULT_FN_ATTRS
  72. _mm512_dpwssd_epi32(__m512i __S, __m512i __A, __m512i __B)
  73. {
  74. return (__m512i)__builtin_ia32_vpdpwssd512((__v16si)__S, (__v16si)__A,
  75. (__v16si)__B);
  76. }
  77. static __inline__ __m512i __DEFAULT_FN_ATTRS
  78. _mm512_mask_dpwssd_epi32(__m512i __S, __mmask16 __U, __m512i __A, __m512i __B)
  79. {
  80. return (__m512i)__builtin_ia32_selectd_512(__U,
  81. (__v16si)_mm512_dpwssd_epi32(__S, __A, __B),
  82. (__v16si)__S);
  83. }
  84. static __inline__ __m512i __DEFAULT_FN_ATTRS
  85. _mm512_maskz_dpwssd_epi32(__mmask16 __U, __m512i __S, __m512i __A, __m512i __B)
  86. {
  87. return (__m512i)__builtin_ia32_selectd_512(__U,
  88. (__v16si)_mm512_dpwssd_epi32(__S, __A, __B),
  89. (__v16si)_mm512_setzero_si512());
  90. }
  91. static __inline__ __m512i __DEFAULT_FN_ATTRS
  92. _mm512_dpwssds_epi32(__m512i __S, __m512i __A, __m512i __B)
  93. {
  94. return (__m512i)__builtin_ia32_vpdpwssds512((__v16si)__S, (__v16si)__A,
  95. (__v16si)__B);
  96. }
  97. static __inline__ __m512i __DEFAULT_FN_ATTRS
  98. _mm512_mask_dpwssds_epi32(__m512i __S, __mmask16 __U, __m512i __A, __m512i __B)
  99. {
  100. return (__m512i)__builtin_ia32_selectd_512(__U,
  101. (__v16si)_mm512_dpwssds_epi32(__S, __A, __B),
  102. (__v16si)__S);
  103. }
  104. static __inline__ __m512i __DEFAULT_FN_ATTRS
  105. _mm512_maskz_dpwssds_epi32(__mmask16 __U, __m512i __S, __m512i __A, __m512i __B)
  106. {
  107. return (__m512i)__builtin_ia32_selectd_512(__U,
  108. (__v16si)_mm512_dpwssds_epi32(__S, __A, __B),
  109. (__v16si)_mm512_setzero_si512());
  110. }
  111. #undef __DEFAULT_FN_ATTRS
  112. #endif