diff options
Diffstat (limited to 'lib/Headers/avx512vbmivlintrin.h')
-rw-r--r-- | lib/Headers/avx512vbmivlintrin.h | 183 |
1 files changed, 75 insertions, 108 deletions
diff --git a/lib/Headers/avx512vbmivlintrin.h b/lib/Headers/avx512vbmivlintrin.h index 105c6d142fa6f..9a0400b2b5d5b 100644 --- a/lib/Headers/avx512vbmivlintrin.h +++ b/lib/Headers/avx512vbmivlintrin.h @@ -29,161 +29,127 @@ #define __VBMIVLINTRIN_H /* Define the default attributes for the functions in this file. */ -#define __DEFAULT_FN_ATTRS __attribute__((__always_inline__, __nodebug__, __target__("avx512vbmi,avx512vl"))) +#define __DEFAULT_FN_ATTRS128 __attribute__((__always_inline__, __nodebug__, __target__("avx512vbmi,avx512vl"), __min_vector_width__(128))) +#define __DEFAULT_FN_ATTRS256 __attribute__((__always_inline__, __nodebug__, __target__("avx512vbmi,avx512vl"), __min_vector_width__(256))) -static __inline__ __m128i __DEFAULT_FN_ATTRS -_mm_mask2_permutex2var_epi8 (__m128i __A, __m128i __I, __mmask16 __U, - __m128i __B) +static __inline__ __m128i __DEFAULT_FN_ATTRS128 +_mm_permutex2var_epi8(__m128i __A, __m128i __I, __m128i __B) { - return (__m128i) __builtin_ia32_vpermi2varqi128_mask ((__v16qi) __A, - (__v16qi) __I - /* idx */ , - (__v16qi) __B, - (__mmask16) - __U); + return (__m128i)__builtin_ia32_vpermi2varqi128((__v16qi)__A, + (__v16qi)__I, + (__v16qi)__B); } -static __inline__ __m256i __DEFAULT_FN_ATTRS -_mm256_mask2_permutex2var_epi8 (__m256i __A, __m256i __I, - __mmask32 __U, __m256i __B) +static __inline__ __m128i __DEFAULT_FN_ATTRS128 +_mm_mask_permutex2var_epi8(__m128i __A, __mmask16 __U, __m128i __I, + __m128i __B) { - return (__m256i) __builtin_ia32_vpermi2varqi256_mask ((__v32qi) __A, - (__v32qi) __I - /* idx */ , - (__v32qi) __B, - (__mmask32) - __U); + return (__m128i)__builtin_ia32_selectb_128(__U, + (__v16qi)_mm_permutex2var_epi8(__A, __I, __B), + (__v16qi)__A); } -static __inline__ __m128i __DEFAULT_FN_ATTRS -_mm_permutex2var_epi8 (__m128i __A, __m128i __I, __m128i __B) +static __inline__ __m128i __DEFAULT_FN_ATTRS128 +_mm_mask2_permutex2var_epi8(__m128i __A, __m128i __I, __mmask16 __U, + __m128i __B) { - return (__m128i) __builtin_ia32_vpermt2varqi128_mask ((__v16qi) __I - /* idx */ , - (__v16qi) __A, - (__v16qi) __B, - (__mmask16) - - 1); + return (__m128i)__builtin_ia32_selectb_128(__U, + (__v16qi)_mm_permutex2var_epi8(__A, __I, __B), + (__v16qi)__I); } -static __inline__ __m128i __DEFAULT_FN_ATTRS -_mm_mask_permutex2var_epi8 (__m128i __A, __mmask16 __U, __m128i __I, - __m128i __B) +static __inline__ __m128i __DEFAULT_FN_ATTRS128 +_mm_maskz_permutex2var_epi8(__mmask16 __U, __m128i __A, __m128i __I, + __m128i __B) { - return (__m128i) __builtin_ia32_vpermt2varqi128_mask ((__v16qi) __I - /* idx */ , - (__v16qi) __A, - (__v16qi) __B, - (__mmask16) - __U); + return (__m128i)__builtin_ia32_selectb_128(__U, + (__v16qi)_mm_permutex2var_epi8(__A, __I, __B), + (__v16qi)_mm_setzero_si128()); } -static __inline__ __m128i __DEFAULT_FN_ATTRS -_mm_maskz_permutex2var_epi8 (__mmask16 __U, __m128i __A, __m128i __I, - __m128i __B) +static __inline__ __m256i __DEFAULT_FN_ATTRS256 +_mm256_permutex2var_epi8(__m256i __A, __m256i __I, __m256i __B) { - return (__m128i) __builtin_ia32_vpermt2varqi128_maskz ((__v16qi) __I - /* idx */ , - (__v16qi) __A, - (__v16qi) __B, - (__mmask16) - __U); + return (__m256i)__builtin_ia32_vpermi2varqi256((__v32qi)__A, (__v32qi)__I, + (__v32qi)__B); } -static __inline__ __m256i __DEFAULT_FN_ATTRS -_mm256_permutex2var_epi8 (__m256i __A, __m256i __I, __m256i __B) +static __inline__ __m256i __DEFAULT_FN_ATTRS256 +_mm256_mask_permutex2var_epi8(__m256i __A, __mmask32 __U, __m256i __I, + __m256i __B) { - return (__m256i) __builtin_ia32_vpermt2varqi256_mask ((__v32qi) __I - /* idx */ , - (__v32qi) __A, - (__v32qi) __B, - (__mmask32) - - 1); + return (__m256i)__builtin_ia32_selectb_256(__U, + (__v32qi)_mm256_permutex2var_epi8(__A, __I, __B), + (__v32qi)__A); } -static __inline__ __m256i __DEFAULT_FN_ATTRS -_mm256_mask_permutex2var_epi8 (__m256i __A, __mmask32 __U, - __m256i __I, __m256i __B) +static __inline__ __m256i __DEFAULT_FN_ATTRS256 +_mm256_mask2_permutex2var_epi8(__m256i __A, __m256i __I, __mmask32 __U, + __m256i __B) { - return (__m256i) __builtin_ia32_vpermt2varqi256_mask ((__v32qi) __I - /* idx */ , - (__v32qi) __A, - (__v32qi) __B, - (__mmask32) - __U); + return (__m256i)__builtin_ia32_selectb_256(__U, + (__v32qi)_mm256_permutex2var_epi8(__A, __I, __B), + (__v32qi)__I); } -static __inline__ __m256i __DEFAULT_FN_ATTRS -_mm256_maskz_permutex2var_epi8 (__mmask32 __U, __m256i __A, - __m256i __I, __m256i __B) +static __inline__ __m256i __DEFAULT_FN_ATTRS256 +_mm256_maskz_permutex2var_epi8(__mmask32 __U, __m256i __A, __m256i __I, + __m256i __B) { - return (__m256i) __builtin_ia32_vpermt2varqi256_maskz ((__v32qi) __I - /* idx */ , - (__v32qi) __A, - (__v32qi) __B, - (__mmask32) - __U); + return (__m256i)__builtin_ia32_selectb_256(__U, + (__v32qi)_mm256_permutex2var_epi8(__A, __I, __B), + (__v32qi)_mm256_setzero_si256()); } -static __inline__ __m128i __DEFAULT_FN_ATTRS +static __inline__ __m128i __DEFAULT_FN_ATTRS128 _mm_permutexvar_epi8 (__m128i __A, __m128i __B) { - return (__m128i) __builtin_ia32_permvarqi128_mask ((__v16qi) __B, - (__v16qi) __A, - (__v16qi) _mm_undefined_si128 (), - (__mmask16) -1); + return (__m128i)__builtin_ia32_permvarqi128((__v16qi)__B, (__v16qi)__A); } -static __inline__ __m128i __DEFAULT_FN_ATTRS +static __inline__ __m128i __DEFAULT_FN_ATTRS128 _mm_maskz_permutexvar_epi8 (__mmask16 __M, __m128i __A, __m128i __B) { - return (__m128i) __builtin_ia32_permvarqi128_mask ((__v16qi) __B, - (__v16qi) __A, - (__v16qi) _mm_setzero_si128 (), - (__mmask16) __M); + return (__m128i)__builtin_ia32_selectb_128((__mmask16)__M, + (__v16qi)_mm_permutexvar_epi8(__A, __B), + (__v16qi)_mm_setzero_si128()); } -static __inline__ __m128i __DEFAULT_FN_ATTRS +static __inline__ __m128i __DEFAULT_FN_ATTRS128 _mm_mask_permutexvar_epi8 (__m128i __W, __mmask16 __M, __m128i __A, __m128i __B) { - return (__m128i) __builtin_ia32_permvarqi128_mask ((__v16qi) __B, - (__v16qi) __A, - (__v16qi) __W, - (__mmask16) __M); + return (__m128i)__builtin_ia32_selectb_128((__mmask16)__M, + (__v16qi)_mm_permutexvar_epi8(__A, __B), + (__v16qi)__W); } -static __inline__ __m256i __DEFAULT_FN_ATTRS +static __inline__ __m256i __DEFAULT_FN_ATTRS256 _mm256_permutexvar_epi8 (__m256i __A, __m256i __B) { - return (__m256i) __builtin_ia32_permvarqi256_mask ((__v32qi) __B, - (__v32qi) __A, - (__v32qi) _mm256_undefined_si256 (), - (__mmask32) -1); + return (__m256i)__builtin_ia32_permvarqi256((__v32qi) __B, (__v32qi) __A); } -static __inline__ __m256i __DEFAULT_FN_ATTRS +static __inline__ __m256i __DEFAULT_FN_ATTRS256 _mm256_maskz_permutexvar_epi8 (__mmask32 __M, __m256i __A, __m256i __B) { - return (__m256i) __builtin_ia32_permvarqi256_mask ((__v32qi) __B, - (__v32qi) __A, - (__v32qi) _mm256_setzero_si256 (), - (__mmask32) __M); + return (__m256i)__builtin_ia32_selectb_256((__mmask32)__M, + (__v32qi)_mm256_permutexvar_epi8(__A, __B), + (__v32qi)_mm256_setzero_si256()); } -static __inline__ __m256i __DEFAULT_FN_ATTRS +static __inline__ __m256i __DEFAULT_FN_ATTRS256 _mm256_mask_permutexvar_epi8 (__m256i __W, __mmask32 __M, __m256i __A, __m256i __B) { - return (__m256i) __builtin_ia32_permvarqi256_mask ((__v32qi) __B, - (__v32qi) __A, - (__v32qi) __W, - (__mmask32) __M); + return (__m256i)__builtin_ia32_selectb_256((__mmask32)__M, + (__v32qi)_mm256_permutexvar_epi8(__A, __B), + (__v32qi)__W); } -static __inline__ __m128i __DEFAULT_FN_ATTRS +static __inline__ __m128i __DEFAULT_FN_ATTRS128 _mm_mask_multishift_epi64_epi8 (__m128i __W, __mmask16 __M, __m128i __X, __m128i __Y) { return (__m128i) __builtin_ia32_vpmultishiftqb128_mask ((__v16qi) __X, @@ -192,7 +158,7 @@ _mm_mask_multishift_epi64_epi8 (__m128i __W, __mmask16 __M, __m128i __X, __m128i (__mmask16) __M); } -static __inline__ __m128i __DEFAULT_FN_ATTRS +static __inline__ __m128i __DEFAULT_FN_ATTRS128 _mm_maskz_multishift_epi64_epi8 (__mmask16 __M, __m128i __X, __m128i __Y) { return (__m128i) __builtin_ia32_vpmultishiftqb128_mask ((__v16qi) __X, @@ -202,7 +168,7 @@ _mm_maskz_multishift_epi64_epi8 (__mmask16 __M, __m128i __X, __m128i __Y) (__mmask16) __M); } -static __inline__ __m128i __DEFAULT_FN_ATTRS +static __inline__ __m128i __DEFAULT_FN_ATTRS128 _mm_multishift_epi64_epi8 (__m128i __X, __m128i __Y) { return (__m128i) __builtin_ia32_vpmultishiftqb128_mask ((__v16qi) __X, @@ -212,7 +178,7 @@ _mm_multishift_epi64_epi8 (__m128i __X, __m128i __Y) (__mmask16) -1); } -static __inline__ __m256i __DEFAULT_FN_ATTRS +static __inline__ __m256i __DEFAULT_FN_ATTRS256 _mm256_mask_multishift_epi64_epi8 (__m256i __W, __mmask32 __M, __m256i __X, __m256i __Y) { return (__m256i) __builtin_ia32_vpmultishiftqb256_mask ((__v32qi) __X, @@ -221,7 +187,7 @@ _mm256_mask_multishift_epi64_epi8 (__m256i __W, __mmask32 __M, __m256i __X, __m2 (__mmask32) __M); } -static __inline__ __m256i __DEFAULT_FN_ATTRS +static __inline__ __m256i __DEFAULT_FN_ATTRS256 _mm256_maskz_multishift_epi64_epi8 (__mmask32 __M, __m256i __X, __m256i __Y) { return (__m256i) __builtin_ia32_vpmultishiftqb256_mask ((__v32qi) __X, @@ -231,7 +197,7 @@ _mm256_maskz_multishift_epi64_epi8 (__mmask32 __M, __m256i __X, __m256i __Y) (__mmask32) __M); } -static __inline__ __m256i __DEFAULT_FN_ATTRS +static __inline__ __m256i __DEFAULT_FN_ATTRS256 _mm256_multishift_epi64_epi8 (__m256i __X, __m256i __Y) { return (__m256i) __builtin_ia32_vpmultishiftqb256_mask ((__v32qi) __X, @@ -242,6 +208,7 @@ _mm256_multishift_epi64_epi8 (__m256i __X, __m256i __Y) } -#undef __DEFAULT_FN_ATTRS +#undef __DEFAULT_FN_ATTRS128 +#undef __DEFAULT_FN_ATTRS256 #endif |