1 /*===----------------- gfniintrin.h - GFNI intrinsics ----------------------===
4 * Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
5 * See https://llvm.org/LICENSE.txt for license information.
6 * SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
8 *===-----------------------------------------------------------------------===
11 #error "Never use <gfniintrin.h> directly; include <immintrin.h> instead."
14 #ifndef __GFNIINTRIN_H
15 #define __GFNIINTRIN_H
17 /* Default attributes for simple form (no masking). */
18 #define __DEFAULT_FN_ATTRS __attribute__((__always_inline__, __nodebug__, __target__("gfni"), __min_vector_width__(128)))
20 /* Default attributes for YMM unmasked form. */
21 #define __DEFAULT_FN_ATTRS_Y __attribute__((__always_inline__, __nodebug__, __target__("avx,gfni"), __min_vector_width__(256)))
23 /* Default attributes for ZMM unmasked forms. */
24 #define __DEFAULT_FN_ATTRS_Z __attribute__((__always_inline__, __nodebug__, __target__("avx512f,gfni"), __min_vector_width__(512)))
25 /* Default attributes for ZMM masked forms. */
26 #define __DEFAULT_FN_ATTRS_Z_MASK __attribute__((__always_inline__, __nodebug__, __target__("avx512bw,gfni"), __min_vector_width__(512)))
28 /* Default attributes for VLX masked forms. */
29 #define __DEFAULT_FN_ATTRS_VL128 __attribute__((__always_inline__, __nodebug__, __target__("avx512bw,avx512vl,gfni"), __min_vector_width__(128)))
30 #define __DEFAULT_FN_ATTRS_VL256 __attribute__((__always_inline__, __nodebug__, __target__("avx512bw,avx512vl,gfni"), __min_vector_width__(256)))
32 #define _mm_gf2p8affineinv_epi64_epi8(A, B, I) \
33 ((__m128i)__builtin_ia32_vgf2p8affineinvqb_v16qi((__v16qi)(__m128i)(A), \
34 (__v16qi)(__m128i)(B), \
37 #define _mm_gf2p8affine_epi64_epi8(A, B, I) \
38 ((__m128i)__builtin_ia32_vgf2p8affineqb_v16qi((__v16qi)(__m128i)(A), \
39 (__v16qi)(__m128i)(B), \
42 static __inline__ __m128i __DEFAULT_FN_ATTRS
43 _mm_gf2p8mul_epi8(__m128i __A
, __m128i __B
)
45 return (__m128i
) __builtin_ia32_vgf2p8mulb_v16qi((__v16qi
) __A
,
50 #define _mm256_gf2p8affineinv_epi64_epi8(A, B, I) \
51 ((__m256i)__builtin_ia32_vgf2p8affineinvqb_v32qi((__v32qi)(__m256i)(A), \
52 (__v32qi)(__m256i)(B), \
55 #define _mm256_gf2p8affine_epi64_epi8(A, B, I) \
56 ((__m256i)__builtin_ia32_vgf2p8affineqb_v32qi((__v32qi)(__m256i)(A), \
57 (__v32qi)(__m256i)(B), \
60 static __inline__ __m256i __DEFAULT_FN_ATTRS_Y
61 _mm256_gf2p8mul_epi8(__m256i __A
, __m256i __B
)
63 return (__m256i
) __builtin_ia32_vgf2p8mulb_v32qi((__v32qi
) __A
,
66 #endif /* __AVXINTRIN_H */
68 #ifdef __AVX512BWINTRIN_H
69 #define _mm512_gf2p8affineinv_epi64_epi8(A, B, I) \
70 ((__m512i)__builtin_ia32_vgf2p8affineinvqb_v64qi((__v64qi)(__m512i)(A), \
71 (__v64qi)(__m512i)(B), \
74 #define _mm512_mask_gf2p8affineinv_epi64_epi8(S, U, A, B, I) \
75 ((__m512i)__builtin_ia32_selectb_512((__mmask64)(U), \
76 (__v64qi)_mm512_gf2p8affineinv_epi64_epi8(A, B, I), \
77 (__v64qi)(__m512i)(S)))
79 #define _mm512_maskz_gf2p8affineinv_epi64_epi8(U, A, B, I) \
80 _mm512_mask_gf2p8affineinv_epi64_epi8((__m512i)_mm512_setzero_si512(), \
83 #define _mm512_gf2p8affine_epi64_epi8(A, B, I) \
84 ((__m512i)__builtin_ia32_vgf2p8affineqb_v64qi((__v64qi)(__m512i)(A), \
85 (__v64qi)(__m512i)(B), \
88 #define _mm512_mask_gf2p8affine_epi64_epi8(S, U, A, B, I) \
89 ((__m512i)__builtin_ia32_selectb_512((__mmask64)(U), \
90 (__v64qi)_mm512_gf2p8affine_epi64_epi8((A), (B), (I)), \
91 (__v64qi)(__m512i)(S)))
93 #define _mm512_maskz_gf2p8affine_epi64_epi8(U, A, B, I) \
94 _mm512_mask_gf2p8affine_epi64_epi8((__m512i)_mm512_setzero_si512(), \
97 static __inline__ __m512i __DEFAULT_FN_ATTRS_Z
98 _mm512_gf2p8mul_epi8(__m512i __A
, __m512i __B
)
100 return (__m512i
) __builtin_ia32_vgf2p8mulb_v64qi((__v64qi
) __A
,
104 static __inline__ __m512i __DEFAULT_FN_ATTRS_Z_MASK
105 _mm512_mask_gf2p8mul_epi8(__m512i __S
, __mmask64 __U
, __m512i __A
, __m512i __B
)
107 return (__m512i
) __builtin_ia32_selectb_512(__U
,
108 (__v64qi
) _mm512_gf2p8mul_epi8(__A
, __B
),
112 static __inline__ __m512i __DEFAULT_FN_ATTRS_Z_MASK
113 _mm512_maskz_gf2p8mul_epi8(__mmask64 __U
, __m512i __A
, __m512i __B
)
115 return _mm512_mask_gf2p8mul_epi8((__m512i
)_mm512_setzero_si512(),
118 #endif /* __AVX512BWINTRIN_H */
120 #ifdef __AVX512VLBWINTRIN_H
121 #define _mm_mask_gf2p8affineinv_epi64_epi8(S, U, A, B, I) \
122 ((__m128i)__builtin_ia32_selectb_128((__mmask16)(U), \
123 (__v16qi)_mm_gf2p8affineinv_epi64_epi8(A, B, I), \
124 (__v16qi)(__m128i)(S)))
126 #define _mm_maskz_gf2p8affineinv_epi64_epi8(U, A, B, I) \
127 _mm_mask_gf2p8affineinv_epi64_epi8((__m128i)_mm_setzero_si128(), \
130 #define _mm256_mask_gf2p8affineinv_epi64_epi8(S, U, A, B, I) \
131 ((__m256i)__builtin_ia32_selectb_256((__mmask32)(U), \
132 (__v32qi)_mm256_gf2p8affineinv_epi64_epi8(A, B, I), \
133 (__v32qi)(__m256i)(S)))
135 #define _mm256_maskz_gf2p8affineinv_epi64_epi8(U, A, B, I) \
136 _mm256_mask_gf2p8affineinv_epi64_epi8((__m256i)_mm256_setzero_si256(), \
139 #define _mm_mask_gf2p8affine_epi64_epi8(S, U, A, B, I) \
140 ((__m128i)__builtin_ia32_selectb_128((__mmask16)(U), \
141 (__v16qi)_mm_gf2p8affine_epi64_epi8(A, B, I), \
142 (__v16qi)(__m128i)(S)))
144 #define _mm_maskz_gf2p8affine_epi64_epi8(U, A, B, I) \
145 _mm_mask_gf2p8affine_epi64_epi8((__m128i)_mm_setzero_si128(), U, A, B, I)
147 #define _mm256_mask_gf2p8affine_epi64_epi8(S, U, A, B, I) \
148 ((__m256i)__builtin_ia32_selectb_256((__mmask32)(U), \
149 (__v32qi)_mm256_gf2p8affine_epi64_epi8(A, B, I), \
150 (__v32qi)(__m256i)(S)))
152 #define _mm256_maskz_gf2p8affine_epi64_epi8(U, A, B, I) \
153 _mm256_mask_gf2p8affine_epi64_epi8((__m256i)_mm256_setzero_si256(), \
156 static __inline__ __m128i __DEFAULT_FN_ATTRS_VL128
157 _mm_mask_gf2p8mul_epi8(__m128i __S
, __mmask16 __U
, __m128i __A
, __m128i __B
)
159 return (__m128i
) __builtin_ia32_selectb_128(__U
,
160 (__v16qi
) _mm_gf2p8mul_epi8(__A
, __B
),
164 static __inline__ __m128i __DEFAULT_FN_ATTRS_VL128
165 _mm_maskz_gf2p8mul_epi8(__mmask16 __U
, __m128i __A
, __m128i __B
)
167 return _mm_mask_gf2p8mul_epi8((__m128i
)_mm_setzero_si128(),
171 static __inline__ __m256i __DEFAULT_FN_ATTRS_VL256
172 _mm256_mask_gf2p8mul_epi8(__m256i __S
, __mmask32 __U
, __m256i __A
, __m256i __B
)
174 return (__m256i
) __builtin_ia32_selectb_256(__U
,
175 (__v32qi
) _mm256_gf2p8mul_epi8(__A
, __B
),
179 static __inline__ __m256i __DEFAULT_FN_ATTRS_VL256
180 _mm256_maskz_gf2p8mul_epi8(__mmask32 __U
, __m256i __A
, __m256i __B
)
182 return _mm256_mask_gf2p8mul_epi8((__m256i
)_mm256_setzero_si256(),
185 #endif /* __AVX512VLBWINTRIN_H */
187 #undef __DEFAULT_FN_ATTRS
188 #undef __DEFAULT_FN_ATTRS_Y
189 #undef __DEFAULT_FN_ATTRS_Z
190 #undef __DEFAULT_FN_ATTRS_VL128
191 #undef __DEFAULT_FN_ATTRS_VL256
193 #endif /* __GFNIINTRIN_H */