avx512pfintrin.h 5.7 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111
  1. /*===------------- avx512pfintrin.h - PF 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 <avx512pfintrin.h> directly; include <immintrin.h> instead."
  26. #endif
  27. #ifndef __AVX512PFINTRIN_H
  28. #define __AVX512PFINTRIN_H
  29. /* Define the default attributes for the functions in this file. */
  30. #define __DEFAULT_FN_ATTRS __attribute__((__always_inline__, __nodebug__, __target__("avx512pf")))
  31. #define _mm512_mask_prefetch_i32gather_pd(index, mask, addr, scale, hint) __extension__ ({\
  32. __builtin_ia32_gatherpfdpd((__mmask8)(mask), (__v8si)(__m256i)(index), \
  33. (long long const *)(addr), (int)(scale), \
  34. (int)(hint)); })
  35. #define _mm512_prefetch_i32gather_pd(index, addr, scale, hint) __extension__ ({\
  36. __builtin_ia32_gatherpfdpd((__mmask8) -1, (__v8si)(__m256i)(index), \
  37. (long long const *)(addr), (int)(scale), \
  38. (int)(hint)); })
  39. #define _mm512_mask_prefetch_i32gather_ps(index, mask, addr, scale, hint) ({\
  40. __builtin_ia32_gatherpfdps((__mmask16)(mask), \
  41. (__v16si)(__m512i)(index), (int const *)(addr), \
  42. (int)(scale), (int)(hint)); })
  43. #define _mm512_prefetch_i32gather_ps(index, addr, scale, hint) ({\
  44. __builtin_ia32_gatherpfdps((__mmask16) -1, \
  45. (__v16si)(__m512i)(index), (int const *)(addr), \
  46. (int)(scale), (int)(hint)); })
  47. #define _mm512_mask_prefetch_i64gather_pd(index, mask, addr, scale, hint) __extension__ ({\
  48. __builtin_ia32_gatherpfqpd((__mmask8)(mask), (__v8di)(__m512i)(index), \
  49. (long long const *)(addr), (int)(scale), \
  50. (int)(hint)); })
  51. #define _mm512_prefetch_i64gather_pd(index, addr, scale, hint) __extension__ ({\
  52. __builtin_ia32_gatherpfqpd((__mmask8) -1, (__v8di)(__m512i)(index), \
  53. (long long const *)(addr), (int)(scale), \
  54. (int)(hint)); })
  55. #define _mm512_mask_prefetch_i64gather_ps(index, mask, addr, scale, hint) ({\
  56. __builtin_ia32_gatherpfqps((__mmask8)(mask), (__v8di)(__m512i)(index), \
  57. (int const *)(addr), (int)(scale), (int)(hint)); })
  58. #define _mm512_prefetch_i64gather_ps(index, addr, scale, hint) ({\
  59. __builtin_ia32_gatherpfqps((__mmask8) -1, (__v8di)(__m512i)(index), \
  60. (int const *)(addr), (int)(scale), (int)(hint)); })
  61. #define _mm512_prefetch_i32scatter_pd(addr, index, scale, hint) __extension__ ({\
  62. __builtin_ia32_scatterpfdpd((__mmask8)-1, (__v8si)(__m256i)(index), \
  63. (long long *)(addr), (int)(scale), \
  64. (int)(hint)); })
  65. #define _mm512_mask_prefetch_i32scatter_pd(addr, mask, index, scale, hint) __extension__ ({\
  66. __builtin_ia32_scatterpfdpd((__mmask8)(mask), (__v8si)(__m256i)(index), \
  67. (long long *)(addr), (int)(scale), \
  68. (int)(hint)); })
  69. #define _mm512_prefetch_i32scatter_ps(addr, index, scale, hint) __extension__ ({\
  70. __builtin_ia32_scatterpfdps((__mmask16)-1, (__v16si)(__m512i)(index), \
  71. (int *)(addr), (int)(scale), (int)(hint)); })
  72. #define _mm512_mask_prefetch_i32scatter_ps(addr, mask, index, scale, hint) __extension__ ({\
  73. __builtin_ia32_scatterpfdps((__mmask16)(mask), \
  74. (__v16si)(__m512i)(index), (int *)(addr), \
  75. (int)(scale), (int)(hint)); })
  76. #define _mm512_prefetch_i64scatter_pd(addr, index, scale, hint) __extension__ ({\
  77. __builtin_ia32_scatterpfqpd((__mmask8)-1, (__v8di)(__m512i)(index), \
  78. (long long *)(addr), (int)(scale), \
  79. (int)(hint)); })
  80. #define _mm512_mask_prefetch_i64scatter_pd(addr, mask, index, scale, hint) __extension__ ({\
  81. __builtin_ia32_scatterpfqpd((__mmask8)(mask), (__v8di)(__m512i)(index), \
  82. (long long *)(addr), (int)(scale), \
  83. (int)(hint)); })
  84. #define _mm512_prefetch_i64scatter_ps(addr, index, scale, hint) __extension__ ({\
  85. __builtin_ia32_scatterpfqps((__mmask8)-1, (__v8di)(__m512i)(index), \
  86. (int *)(addr), (int)(scale), (int)(hint)); })
  87. #define _mm512_mask_prefetch_i64scatter_ps(addr, mask, index, scale, hint) __extension__ ({\
  88. __builtin_ia32_scatterpfqps((__mmask8)(mask), (__v8di)(__m512i)(index), \
  89. (int *)(addr), (int)(scale), (int)(hint)); })
  90. #undef __DEFAULT_FN_ATTRS
  91. #endif