aboutsummaryrefslogtreecommitdiff
path: root/clang/lib/Headers
diff options
context:
space:
mode:
Diffstat (limited to 'clang/lib/Headers')
-rw-r--r--clang/lib/Headers/__clang_hip_runtime_wrapper.h2
-rw-r--r--clang/lib/Headers/avx2intrin.h13
-rw-r--r--clang/lib/Headers/avx512bwintrin.h30
-rw-r--r--clang/lib/Headers/avx512cdintrin.h86
-rw-r--r--clang/lib/Headers/avx512dqintrin.h60
-rw-r--r--clang/lib/Headers/avx512fintrin.h126
-rw-r--r--clang/lib/Headers/avx512vlbwintrin.h28
-rw-r--r--clang/lib/Headers/avx512vlcdintrin.h154
-rw-r--r--clang/lib/Headers/avx512vldqintrin.h50
-rw-r--r--clang/lib/Headers/avx512vlintrin.h361
-rw-r--r--clang/lib/Headers/tmmintrin.h33
11 files changed, 384 insertions, 559 deletions
diff --git a/clang/lib/Headers/__clang_hip_runtime_wrapper.h b/clang/lib/Headers/__clang_hip_runtime_wrapper.h
index da1e39a..fb0ece9 100644
--- a/clang/lib/Headers/__clang_hip_runtime_wrapper.h
+++ b/clang/lib/Headers/__clang_hip_runtime_wrapper.h
@@ -25,6 +25,8 @@
#define __constant__ __attribute__((constant))
#define __managed__ __attribute__((managed))
+#define __cluster_dims__(...) __attribute__((cluster_dims(__VA_ARGS__)))
+
#if !defined(__cplusplus) || __cplusplus < 201103L
#define nullptr NULL;
#endif
diff --git a/clang/lib/Headers/avx2intrin.h b/clang/lib/Headers/avx2intrin.h
index e150aa6..d35bc0e 100644
--- a/clang/lib/Headers/avx2intrin.h
+++ b/clang/lib/Headers/avx2intrin.h
@@ -1650,9 +1650,8 @@ _mm256_mul_epi32(__m256i __a, __m256i __b) {
/// \param __b
/// A 256-bit vector of [16 x i16] containing one of the source operands.
/// \returns A 256-bit vector of [16 x i16] containing the rounded products.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_mulhrs_epi16(__m256i __a, __m256i __b)
-{
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_mulhrs_epi16(__m256i __a, __m256i __b) {
return (__m256i)__builtin_ia32_pmulhrsw256((__v16hi)__a, (__v16hi)__b);
}
@@ -1670,8 +1669,7 @@ _mm256_mulhrs_epi16(__m256i __a, __m256i __b)
/// A 256-bit vector of [16 x i16] containing one of the source operands.
/// \returns A 256-bit vector of [16 x i16] containing the products.
static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
-_mm256_mulhi_epu16(__m256i __a, __m256i __b)
-{
+_mm256_mulhi_epu16(__m256i __a, __m256i __b) {
return (__m256i)__builtin_ia32_pmulhuw256((__v16hu)__a, (__v16hu)__b);
}
@@ -1852,9 +1850,8 @@ _mm256_sad_epu8(__m256i __a, __m256i __b)
/// control byte specify the index (within the same 128-bit half) of \a __a
/// to copy to the result byte.
/// \returns A 256-bit integer vector containing the result.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_shuffle_epi8(__m256i __a, __m256i __b)
-{
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_shuffle_epi8(__m256i __a, __m256i __b) {
return (__m256i)__builtin_ia32_pshufb256((__v32qi)__a, (__v32qi)__b);
}
diff --git a/clang/lib/Headers/avx512bwintrin.h b/clang/lib/Headers/avx512bwintrin.h
index 473fe94..ac75b6c 100644
--- a/clang/lib/Headers/avx512bwintrin.h
+++ b/clang/lib/Headers/avx512bwintrin.h
@@ -866,23 +866,20 @@ _mm512_mask_min_epu16(__m512i __W, __mmask32 __M, __m512i __A, __m512i __B) {
(__v32hi)__W);
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
-_mm512_shuffle_epi8(__m512i __A, __m512i __B)
-{
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_shuffle_epi8(__m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_pshufb512((__v64qi)__A,(__v64qi)__B);
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
-_mm512_mask_shuffle_epi8(__m512i __W, __mmask64 __U, __m512i __A, __m512i __B)
-{
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_mask_shuffle_epi8(__m512i __W, __mmask64 __U, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectb_512((__mmask64)__U,
(__v64qi)_mm512_shuffle_epi8(__A, __B),
(__v64qi)__W);
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
-_mm512_maskz_shuffle_epi8(__mmask64 __U, __m512i __A, __m512i __B)
-{
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_maskz_shuffle_epi8(__mmask64 __U, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectb_512((__mmask64)__U,
(__v64qi)_mm512_shuffle_epi8(__A, __B),
(__v64qi)_mm512_setzero_si512());
@@ -1006,23 +1003,20 @@ _mm512_maskz_permutex2var_epi16(__mmask32 __U, __m512i __A, __m512i __I,
(__v32hi)_mm512_setzero_si512());
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
-_mm512_mulhrs_epi16(__m512i __A, __m512i __B)
-{
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_mulhrs_epi16(__m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_pmulhrsw512((__v32hi)__A, (__v32hi)__B);
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
-_mm512_mask_mulhrs_epi16(__m512i __W, __mmask32 __U, __m512i __A, __m512i __B)
-{
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_mask_mulhrs_epi16(__m512i __W, __mmask32 __U, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectw_512((__mmask32)__U,
(__v32hi)_mm512_mulhrs_epi16(__A, __B),
(__v32hi)__W);
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
-_mm512_maskz_mulhrs_epi16(__mmask32 __U, __m512i __A, __m512i __B)
-{
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_maskz_mulhrs_epi16(__mmask32 __U, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectw_512((__mmask32)__U,
(__v32hi)_mm512_mulhrs_epi16(__A, __B),
(__v32hi)_mm512_setzero_si512());
diff --git a/clang/lib/Headers/avx512cdintrin.h b/clang/lib/Headers/avx512cdintrin.h
index 8899298..fb6dcb6 100644
--- a/clang/lib/Headers/avx512cdintrin.h
+++ b/clang/lib/Headers/avx512cdintrin.h
@@ -15,94 +15,82 @@
#define __AVX512CDINTRIN_H
/* Define the default attributes for the functions in this file. */
+#if defined(__cplusplus) && (__cplusplus >= 201103L)
+#define __DEFAULT_FN_ATTRS \
+ constexpr __attribute__((__always_inline__, __nodebug__, \
+ __target__("avx512cd"), __min_vector_width__(512)))
+#else
#define __DEFAULT_FN_ATTRS \
__attribute__((__always_inline__, __nodebug__, __target__("avx512cd"), \
__min_vector_width__(512)))
-
-#if defined(__cplusplus) && (__cplusplus >= 201103L)
-#define __DEFAULT_FN_ATTRS_CONSTEXPR __DEFAULT_FN_ATTRS constexpr
-#else
-#define __DEFAULT_FN_ATTRS_CONSTEXPR __DEFAULT_FN_ATTRS
#endif
static __inline__ __m512i __DEFAULT_FN_ATTRS
-_mm512_conflict_epi64 (__m512i __A)
-{
- return (__m512i) __builtin_ia32_vpconflictdi_512 ((__v8di) __A);
+_mm512_conflict_epi64(__m512i __A) {
+ return (__m512i)__builtin_ia32_vpconflictdi_512((__v8di)__A);
}
static __inline__ __m512i __DEFAULT_FN_ATTRS
-_mm512_mask_conflict_epi64 (__m512i __W, __mmask8 __U, __m512i __A)
-{
- return (__m512i)__builtin_ia32_selectq_512((__mmask8)__U,
- (__v8di)_mm512_conflict_epi64(__A),
- (__v8di)__W);
+_mm512_mask_conflict_epi64(__m512i __W, __mmask8 __U, __m512i __A) {
+ return (__m512i)__builtin_ia32_selectq_512(
+ (__mmask8)__U, (__v8di)_mm512_conflict_epi64(__A), (__v8di)__W);
}
static __inline__ __m512i __DEFAULT_FN_ATTRS
-_mm512_maskz_conflict_epi64 (__mmask8 __U, __m512i __A)
-{
+_mm512_maskz_conflict_epi64(__mmask8 __U, __m512i __A) {
return (__m512i)__builtin_ia32_selectq_512((__mmask8)__U,
(__v8di)_mm512_conflict_epi64(__A),
- (__v8di)_mm512_setzero_si512 ());
+ (__v8di)_mm512_setzero_si512());
}
static __inline__ __m512i __DEFAULT_FN_ATTRS
-_mm512_conflict_epi32 (__m512i __A)
-{
- return (__m512i) __builtin_ia32_vpconflictsi_512 ((__v16si) __A);
+_mm512_conflict_epi32(__m512i __A) {
+ return (__m512i)__builtin_ia32_vpconflictsi_512((__v16si)__A);
}
static __inline__ __m512i __DEFAULT_FN_ATTRS
-_mm512_mask_conflict_epi32 (__m512i __W, __mmask16 __U, __m512i __A)
-{
- return (__m512i)__builtin_ia32_selectd_512((__mmask16)__U,
- (__v16si)_mm512_conflict_epi32(__A),
- (__v16si)__W);
+_mm512_mask_conflict_epi32(__m512i __W, __mmask16 __U, __m512i __A) {
+ return (__m512i)__builtin_ia32_selectd_512(
+ (__mmask16)__U, (__v16si)_mm512_conflict_epi32(__A), (__v16si)__W);
}
static __inline__ __m512i __DEFAULT_FN_ATTRS
-_mm512_maskz_conflict_epi32 (__mmask16 __U, __m512i __A)
-{
- return (__m512i)__builtin_ia32_selectd_512((__mmask16)__U,
- (__v16si)_mm512_conflict_epi32(__A),
- (__v16si)_mm512_setzero_si512());
+_mm512_maskz_conflict_epi32(__mmask16 __U, __m512i __A) {
+ return (__m512i)__builtin_ia32_selectd_512(
+ (__mmask16)__U, (__v16si)_mm512_conflict_epi32(__A),
+ (__v16si)_mm512_setzero_si512());
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS_CONSTEXPR
-_mm512_lzcnt_epi32(__m512i __A) {
+static __inline__ __m512i __DEFAULT_FN_ATTRS _mm512_lzcnt_epi32(__m512i __A) {
return (__m512i)__builtin_elementwise_clzg((__v16si)__A,
(__v16si)_mm512_set1_epi32(32));
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS_CONSTEXPR
+static __inline__ __m512i __DEFAULT_FN_ATTRS
_mm512_mask_lzcnt_epi32(__m512i __W, __mmask16 __U, __m512i __A) {
- return (__m512i)__builtin_ia32_selectd_512((__mmask16)__U,
- (__v16si)_mm512_lzcnt_epi32(__A),
- (__v16si)__W);
+ return (__m512i)__builtin_ia32_selectd_512(
+ (__mmask16)__U, (__v16si)_mm512_lzcnt_epi32(__A), (__v16si)__W);
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS_CONSTEXPR
+static __inline__ __m512i __DEFAULT_FN_ATTRS
_mm512_maskz_lzcnt_epi32(__mmask16 __U, __m512i __A) {
return (__m512i)__builtin_ia32_selectd_512((__mmask16)__U,
(__v16si)_mm512_lzcnt_epi32(__A),
(__v16si)_mm512_setzero_si512());
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS_CONSTEXPR
-_mm512_lzcnt_epi64(__m512i __A) {
+static __inline__ __m512i __DEFAULT_FN_ATTRS _mm512_lzcnt_epi64(__m512i __A) {
return (__m512i)__builtin_elementwise_clzg(
(__v8di)__A, (__v8di)_mm512_set1_epi64((long long)64));
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS_CONSTEXPR
+static __inline__ __m512i __DEFAULT_FN_ATTRS
_mm512_mask_lzcnt_epi64(__m512i __W, __mmask8 __U, __m512i __A) {
- return (__m512i)__builtin_ia32_selectq_512((__mmask8)__U,
- (__v8di)_mm512_lzcnt_epi64(__A),
- (__v8di)__W);
+ return (__m512i)__builtin_ia32_selectq_512(
+ (__mmask8)__U, (__v8di)_mm512_lzcnt_epi64(__A), (__v8di)__W);
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS_CONSTEXPR
+static __inline__ __m512i __DEFAULT_FN_ATTRS
_mm512_maskz_lzcnt_epi64(__mmask8 __U, __m512i __A) {
return (__m512i)__builtin_ia32_selectq_512((__mmask8)__U,
(__v8di)_mm512_lzcnt_epi64(__A),
@@ -110,19 +98,15 @@ _mm512_maskz_lzcnt_epi64(__mmask8 __U, __m512i __A) {
}
static __inline__ __m512i __DEFAULT_FN_ATTRS
-_mm512_broadcastmb_epi64 (__mmask8 __A)
-{
- return (__m512i) _mm512_set1_epi64((long long) __A);
+_mm512_broadcastmb_epi64(__mmask8 __A) {
+ return (__m512i)_mm512_set1_epi64((long long)__A);
}
static __inline__ __m512i __DEFAULT_FN_ATTRS
-_mm512_broadcastmw_epi32 (__mmask16 __A)
-{
- return (__m512i) _mm512_set1_epi32((int) __A);
-
+_mm512_broadcastmw_epi32(__mmask16 __A) {
+ return (__m512i)_mm512_set1_epi32((int)__A);
}
#undef __DEFAULT_FN_ATTRS
-#undef __DEFAULT_FN_ATTRS_CONSTEXPR
#endif
diff --git a/clang/lib/Headers/avx512dqintrin.h b/clang/lib/Headers/avx512dqintrin.h
index fb65bf9..3681cca 100644
--- a/clang/lib/Headers/avx512dqintrin.h
+++ b/clang/lib/Headers/avx512dqintrin.h
@@ -1083,17 +1083,15 @@ _mm512_broadcast_f32x2(__m128 __A) {
0, 1, 0, 1, 0, 1, 0, 1);
}
-static __inline__ __m512 __DEFAULT_FN_ATTRS512
-_mm512_mask_broadcast_f32x2 (__m512 __O, __mmask16 __M, __m128 __A)
-{
+static __inline__ __m512 __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_mask_broadcast_f32x2(__m512 __O, __mmask16 __M, __m128 __A) {
return (__m512)__builtin_ia32_selectps_512((__mmask16)__M,
(__v16sf)_mm512_broadcast_f32x2(__A),
(__v16sf)__O);
}
-static __inline__ __m512 __DEFAULT_FN_ATTRS512
-_mm512_maskz_broadcast_f32x2 (__mmask16 __M, __m128 __A)
-{
+static __inline__ __m512 __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_maskz_broadcast_f32x2(__mmask16 __M, __m128 __A) {
return (__m512)__builtin_ia32_selectps_512((__mmask16)__M,
(__v16sf)_mm512_broadcast_f32x2(__A),
(__v16sf)_mm512_setzero_ps());
@@ -1106,17 +1104,15 @@ _mm512_broadcast_f32x8(__m256 __A) {
0, 1, 2, 3, 4, 5, 6, 7);
}
-static __inline__ __m512 __DEFAULT_FN_ATTRS512
-_mm512_mask_broadcast_f32x8(__m512 __O, __mmask16 __M, __m256 __A)
-{
+static __inline__ __m512 __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_mask_broadcast_f32x8(__m512 __O, __mmask16 __M, __m256 __A) {
return (__m512)__builtin_ia32_selectps_512((__mmask16)__M,
(__v16sf)_mm512_broadcast_f32x8(__A),
(__v16sf)__O);
}
-static __inline__ __m512 __DEFAULT_FN_ATTRS512
-_mm512_maskz_broadcast_f32x8(__mmask16 __M, __m256 __A)
-{
+static __inline__ __m512 __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_maskz_broadcast_f32x8(__mmask16 __M, __m256 __A) {
return (__m512)__builtin_ia32_selectps_512((__mmask16)__M,
(__v16sf)_mm512_broadcast_f32x8(__A),
(__v16sf)_mm512_setzero_ps());
@@ -1128,17 +1124,15 @@ _mm512_broadcast_f64x2(__m128d __A) {
0, 1, 0, 1, 0, 1, 0, 1);
}
-static __inline__ __m512d __DEFAULT_FN_ATTRS512
-_mm512_mask_broadcast_f64x2(__m512d __O, __mmask8 __M, __m128d __A)
-{
+static __inline__ __m512d __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_mask_broadcast_f64x2(__m512d __O, __mmask8 __M, __m128d __A) {
return (__m512d)__builtin_ia32_selectpd_512((__mmask8)__M,
(__v8df)_mm512_broadcast_f64x2(__A),
(__v8df)__O);
}
-static __inline__ __m512d __DEFAULT_FN_ATTRS512
-_mm512_maskz_broadcast_f64x2(__mmask8 __M, __m128d __A)
-{
+static __inline__ __m512d __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_maskz_broadcast_f64x2(__mmask8 __M, __m128d __A) {
return (__m512d)__builtin_ia32_selectpd_512((__mmask8)__M,
(__v8df)_mm512_broadcast_f64x2(__A),
(__v8df)_mm512_setzero_pd());
@@ -1151,17 +1145,15 @@ _mm512_broadcast_i32x2(__m128i __A) {
0, 1, 0, 1, 0, 1, 0, 1);
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
-_mm512_mask_broadcast_i32x2 (__m512i __O, __mmask16 __M, __m128i __A)
-{
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_mask_broadcast_i32x2(__m512i __O, __mmask16 __M, __m128i __A) {
return (__m512i)__builtin_ia32_selectd_512((__mmask16)__M,
(__v16si)_mm512_broadcast_i32x2(__A),
(__v16si)__O);
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
-_mm512_maskz_broadcast_i32x2 (__mmask16 __M, __m128i __A)
-{
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_maskz_broadcast_i32x2(__mmask16 __M, __m128i __A) {
return (__m512i)__builtin_ia32_selectd_512((__mmask16)__M,
(__v16si)_mm512_broadcast_i32x2(__A),
(__v16si)_mm512_setzero_si512());
@@ -1174,17 +1166,15 @@ _mm512_broadcast_i32x8(__m256i __A) {
0, 1, 2, 3, 4, 5, 6, 7);
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
-_mm512_mask_broadcast_i32x8(__m512i __O, __mmask16 __M, __m256i __A)
-{
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_mask_broadcast_i32x8(__m512i __O, __mmask16 __M, __m256i __A) {
return (__m512i)__builtin_ia32_selectd_512((__mmask16)__M,
(__v16si)_mm512_broadcast_i32x8(__A),
(__v16si)__O);
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
-_mm512_maskz_broadcast_i32x8(__mmask16 __M, __m256i __A)
-{
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_maskz_broadcast_i32x8(__mmask16 __M, __m256i __A) {
return (__m512i)__builtin_ia32_selectd_512((__mmask16)__M,
(__v16si)_mm512_broadcast_i32x8(__A),
(__v16si)_mm512_setzero_si512());
@@ -1196,17 +1186,15 @@ _mm512_broadcast_i64x2(__m128i __A) {
0, 1, 0, 1, 0, 1, 0, 1);
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
-_mm512_mask_broadcast_i64x2(__m512i __O, __mmask8 __M, __m128i __A)
-{
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_mask_broadcast_i64x2(__m512i __O, __mmask8 __M, __m128i __A) {
return (__m512i)__builtin_ia32_selectq_512((__mmask8)__M,
(__v8di)_mm512_broadcast_i64x2(__A),
(__v8di)__O);
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
-_mm512_maskz_broadcast_i64x2(__mmask8 __M, __m128i __A)
-{
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_maskz_broadcast_i64x2(__mmask8 __M, __m128i __A) {
return (__m512i)__builtin_ia32_selectq_512((__mmask8)__M,
(__v8di)_mm512_broadcast_i64x2(__A),
(__v8di)_mm512_setzero_si512());
diff --git a/clang/lib/Headers/avx512fintrin.h b/clang/lib/Headers/avx512fintrin.h
index 80e5842..07de036 100644
--- a/clang/lib/Headers/avx512fintrin.h
+++ b/clang/lib/Headers/avx512fintrin.h
@@ -225,17 +225,15 @@ _mm512_broadcastd_epi32(__m128i __A) {
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0);
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
-_mm512_mask_broadcastd_epi32 (__m512i __O, __mmask16 __M, __m128i __A)
-{
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_mask_broadcastd_epi32(__m512i __O, __mmask16 __M, __m128i __A) {
return (__m512i)__builtin_ia32_selectd_512(__M,
(__v16si) _mm512_broadcastd_epi32(__A),
(__v16si) __O);
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
-_mm512_maskz_broadcastd_epi32 (__mmask16 __M, __m128i __A)
-{
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_maskz_broadcastd_epi32(__mmask16 __M, __m128i __A) {
return (__m512i)__builtin_ia32_selectd_512(__M,
(__v16si) _mm512_broadcastd_epi32(__A),
(__v16si) _mm512_setzero_si512());
@@ -247,18 +245,14 @@ _mm512_broadcastq_epi64(__m128i __A) {
0, 0, 0, 0, 0, 0, 0, 0);
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
-_mm512_mask_broadcastq_epi64 (__m512i __O, __mmask8 __M, __m128i __A)
-{
- return (__m512i)__builtin_ia32_selectq_512(__M,
- (__v8di) _mm512_broadcastq_epi64(__A),
- (__v8di) __O);
-
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_mask_broadcastq_epi64(__m512i __O, __mmask8 __M, __m128i __A) {
+ return (__m512i)__builtin_ia32_selectq_512(
+ __M, (__v8di)_mm512_broadcastq_epi64(__A), (__v8di)__O);
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
-_mm512_maskz_broadcastq_epi64 (__mmask8 __M, __m128i __A)
-{
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_maskz_broadcastq_epi64(__mmask8 __M, __m128i __A) {
return (__m512i)__builtin_ia32_selectq_512(__M,
(__v8di) _mm512_broadcastq_epi64(__A),
(__v8di) _mm512_setzero_si512());
@@ -321,9 +315,8 @@ _mm512_set1_epi32(int __s)
__s, __s, __s, __s, __s, __s, __s, __s };
}
-static __inline __m512i __DEFAULT_FN_ATTRS512
-_mm512_maskz_set1_epi32(__mmask16 __M, int __A)
-{
+static __inline __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_maskz_set1_epi32(__mmask16 __M, int __A) {
return (__m512i)__builtin_ia32_selectd_512(__M,
(__v16si)_mm512_set1_epi32(__A),
(__v16si)_mm512_setzero_si512());
@@ -335,9 +328,8 @@ _mm512_set1_epi64(long long __d)
return __extension__(__m512i)(__v8di){ __d, __d, __d, __d, __d, __d, __d, __d };
}
-static __inline __m512i __DEFAULT_FN_ATTRS512
-_mm512_maskz_set1_epi64(__mmask8 __M, long long __A)
-{
+static __inline __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_maskz_set1_epi64(__mmask8 __M, long long __A) {
return (__m512i)__builtin_ia32_selectq_512(__M,
(__v8di)_mm512_set1_epi64(__A),
(__v8di)_mm512_setzero_si512());
@@ -6552,17 +6544,15 @@ _mm512_broadcast_f32x4(__m128 __A) {
0, 1, 2, 3, 0, 1, 2, 3);
}
-static __inline__ __m512 __DEFAULT_FN_ATTRS512
-_mm512_mask_broadcast_f32x4(__m512 __O, __mmask16 __M, __m128 __A)
-{
+static __inline__ __m512 __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_mask_broadcast_f32x4(__m512 __O, __mmask16 __M, __m128 __A) {
return (__m512)__builtin_ia32_selectps_512((__mmask16)__M,
(__v16sf)_mm512_broadcast_f32x4(__A),
(__v16sf)__O);
}
-static __inline__ __m512 __DEFAULT_FN_ATTRS512
-_mm512_maskz_broadcast_f32x4(__mmask16 __M, __m128 __A)
-{
+static __inline__ __m512 __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_maskz_broadcast_f32x4(__mmask16 __M, __m128 __A) {
return (__m512)__builtin_ia32_selectps_512((__mmask16)__M,
(__v16sf)_mm512_broadcast_f32x4(__A),
(__v16sf)_mm512_setzero_ps());
@@ -6597,17 +6587,15 @@ _mm512_broadcast_i32x4(__m128i __A) {
0, 1, 2, 3, 0, 1, 2, 3);
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
-_mm512_mask_broadcast_i32x4(__m512i __O, __mmask16 __M, __m128i __A)
-{
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_mask_broadcast_i32x4(__m512i __O, __mmask16 __M, __m128i __A) {
return (__m512i)__builtin_ia32_selectd_512((__mmask16)__M,
(__v16si)_mm512_broadcast_i32x4(__A),
(__v16si)__O);
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
-_mm512_maskz_broadcast_i32x4(__mmask16 __M, __m128i __A)
-{
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_maskz_broadcast_i32x4(__mmask16 __M, __m128i __A) {
return (__m512i)__builtin_ia32_selectd_512((__mmask16)__M,
(__v16si)_mm512_broadcast_i32x4(__A),
(__v16si)_mm512_setzero_si512());
@@ -6635,33 +6623,29 @@ _mm512_maskz_broadcast_i64x4(__mmask8 __M, __m256i __A)
(__v8di)_mm512_setzero_si512());
}
-static __inline__ __m512d __DEFAULT_FN_ATTRS512
-_mm512_mask_broadcastsd_pd (__m512d __O, __mmask8 __M, __m128d __A)
-{
+static __inline__ __m512d __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_mask_broadcastsd_pd(__m512d __O, __mmask8 __M, __m128d __A) {
return (__m512d)__builtin_ia32_selectpd_512(__M,
(__v8df) _mm512_broadcastsd_pd(__A),
(__v8df) __O);
}
-static __inline__ __m512d __DEFAULT_FN_ATTRS512
-_mm512_maskz_broadcastsd_pd (__mmask8 __M, __m128d __A)
-{
+static __inline__ __m512d __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_maskz_broadcastsd_pd(__mmask8 __M, __m128d __A) {
return (__m512d)__builtin_ia32_selectpd_512(__M,
(__v8df) _mm512_broadcastsd_pd(__A),
(__v8df) _mm512_setzero_pd());
}
-static __inline__ __m512 __DEFAULT_FN_ATTRS512
-_mm512_mask_broadcastss_ps (__m512 __O, __mmask16 __M, __m128 __A)
-{
+static __inline__ __m512 __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_mask_broadcastss_ps(__m512 __O, __mmask16 __M, __m128 __A) {
return (__m512)__builtin_ia32_selectps_512(__M,
(__v16sf) _mm512_broadcastss_ps(__A),
(__v16sf) __O);
}
-static __inline__ __m512 __DEFAULT_FN_ATTRS512
-_mm512_maskz_broadcastss_ps (__mmask16 __M, __m128 __A)
-{
+static __inline__ __m512 __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_maskz_broadcastss_ps(__mmask16 __M, __m128 __A) {
return (__m512)__builtin_ia32_selectps_512(__M,
(__v16sf) _mm512_broadcastss_ps(__A),
(__v16sf) _mm512_setzero_ps());
@@ -8381,17 +8365,15 @@ _mm512_movehdup_ps (__m512 __A)
1, 1, 3, 3, 5, 5, 7, 7, 9, 9, 11, 11, 13, 13, 15, 15);
}
-static __inline__ __m512 __DEFAULT_FN_ATTRS512
-_mm512_mask_movehdup_ps (__m512 __W, __mmask16 __U, __m512 __A)
-{
+static __inline__ __m512 __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_mask_movehdup_ps(__m512 __W, __mmask16 __U, __m512 __A) {
return (__m512)__builtin_ia32_selectps_512((__mmask16)__U,
(__v16sf)_mm512_movehdup_ps(__A),
(__v16sf)__W);
}
-static __inline__ __m512 __DEFAULT_FN_ATTRS512
-_mm512_maskz_movehdup_ps (__mmask16 __U, __m512 __A)
-{
+static __inline__ __m512 __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_maskz_movehdup_ps(__mmask16 __U, __m512 __A) {
return (__m512)__builtin_ia32_selectps_512((__mmask16)__U,
(__v16sf)_mm512_movehdup_ps(__A),
(__v16sf)_mm512_setzero_ps());
@@ -8404,44 +8386,38 @@ _mm512_moveldup_ps (__m512 __A)
0, 0, 2, 2, 4, 4, 6, 6, 8, 8, 10, 10, 12, 12, 14, 14);
}
-static __inline__ __m512 __DEFAULT_FN_ATTRS512
-_mm512_mask_moveldup_ps (__m512 __W, __mmask16 __U, __m512 __A)
-{
+static __inline__ __m512 __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_mask_moveldup_ps(__m512 __W, __mmask16 __U, __m512 __A) {
return (__m512)__builtin_ia32_selectps_512((__mmask16)__U,
(__v16sf)_mm512_moveldup_ps(__A),
(__v16sf)__W);
}
-static __inline__ __m512 __DEFAULT_FN_ATTRS512
-_mm512_maskz_moveldup_ps (__mmask16 __U, __m512 __A)
-{
+static __inline__ __m512 __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_maskz_moveldup_ps(__mmask16 __U, __m512 __A) {
return (__m512)__builtin_ia32_selectps_512((__mmask16)__U,
(__v16sf)_mm512_moveldup_ps(__A),
(__v16sf)_mm512_setzero_ps());
}
-static __inline__ __m128 __DEFAULT_FN_ATTRS128
-_mm_mask_move_ss (__m128 __W, __mmask8 __U, __m128 __A, __m128 __B)
-{
+static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_mask_move_ss(__m128 __W, __mmask8 __U, __m128 __A, __m128 __B) {
return __builtin_ia32_selectss_128(__U, _mm_move_ss(__A, __B), __W);
}
-static __inline__ __m128 __DEFAULT_FN_ATTRS128
-_mm_maskz_move_ss (__mmask8 __U, __m128 __A, __m128 __B)
-{
+static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_maskz_move_ss(__mmask8 __U, __m128 __A, __m128 __B) {
return __builtin_ia32_selectss_128(__U, _mm_move_ss(__A, __B),
_mm_setzero_ps());
}
-static __inline__ __m128d __DEFAULT_FN_ATTRS128
-_mm_mask_move_sd (__m128d __W, __mmask8 __U, __m128d __A, __m128d __B)
-{
+static __inline__ __m128d __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_mask_move_sd(__m128d __W, __mmask8 __U, __m128d __A, __m128d __B) {
return __builtin_ia32_selectsd_128(__U, _mm_move_sd(__A, __B), __W);
}
-static __inline__ __m128d __DEFAULT_FN_ATTRS128
-_mm_maskz_move_sd (__mmask8 __U, __m128d __A, __m128d __B)
-{
+static __inline__ __m128d __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_maskz_move_sd(__mmask8 __U, __m128d __A, __m128d __B) {
return __builtin_ia32_selectsd_128(__U, _mm_move_sd(__A, __B),
_mm_setzero_pd());
}
@@ -8884,17 +8860,15 @@ _mm_cvtu64_ss (__m128 __A, unsigned long long __B)
}
#endif
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
-_mm512_mask_set1_epi32 (__m512i __O, __mmask16 __M, int __A)
-{
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_mask_set1_epi32(__m512i __O, __mmask16 __M, int __A) {
return (__m512i) __builtin_ia32_selectd_512(__M,
(__v16si) _mm512_set1_epi32(__A),
(__v16si) __O);
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
-_mm512_mask_set1_epi64 (__m512i __O, __mmask8 __M, long long __A)
-{
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_mask_set1_epi64(__m512i __O, __mmask8 __M, long long __A) {
return (__m512i) __builtin_ia32_selectq_512(__M,
(__v8di) _mm512_set1_epi64(__A),
(__v8di) __O);
diff --git a/clang/lib/Headers/avx512vlbwintrin.h b/clang/lib/Headers/avx512vlbwintrin.h
index 81e4cbb9..0fcfe37 100644
--- a/clang/lib/Headers/avx512vlbwintrin.h
+++ b/clang/lib/Headers/avx512vlbwintrin.h
@@ -1067,33 +1067,29 @@ _mm256_mask_min_epu16(__m256i __W, __mmask16 __M, __m256i __A, __m256i __B) {
(__v16hi)__W);
}
-static __inline__ __m128i __DEFAULT_FN_ATTRS128
-_mm_mask_shuffle_epi8(__m128i __W, __mmask16 __U, __m128i __A, __m128i __B)
-{
+static __inline__ __m128i __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_mask_shuffle_epi8(__m128i __W, __mmask16 __U, __m128i __A, __m128i __B) {
return (__m128i)__builtin_ia32_selectb_128((__mmask16)__U,
(__v16qi)_mm_shuffle_epi8(__A, __B),
(__v16qi)__W);
}
-static __inline__ __m128i __DEFAULT_FN_ATTRS128
-_mm_maskz_shuffle_epi8(__mmask16 __U, __m128i __A, __m128i __B)
-{
+static __inline__ __m128i __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_maskz_shuffle_epi8(__mmask16 __U, __m128i __A, __m128i __B) {
return (__m128i)__builtin_ia32_selectb_128((__mmask16)__U,
(__v16qi)_mm_shuffle_epi8(__A, __B),
(__v16qi)_mm_setzero_si128());
}
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_mask_shuffle_epi8(__m256i __W, __mmask32 __U, __m256i __A, __m256i __B)
-{
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_mask_shuffle_epi8(__m256i __W, __mmask32 __U, __m256i __A, __m256i __B) {
return (__m256i)__builtin_ia32_selectb_256((__mmask32)__U,
(__v32qi)_mm256_shuffle_epi8(__A, __B),
(__v32qi)__W);
}
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_maskz_shuffle_epi8(__mmask32 __U, __m256i __A, __m256i __B)
-{
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_maskz_shuffle_epi8(__mmask32 __U, __m256i __A, __m256i __B) {
return (__m256i)__builtin_ia32_selectb_256((__mmask32)__U,
(__v32qi)_mm256_shuffle_epi8(__A, __B),
(__v32qi)_mm256_setzero_si256());
@@ -1514,28 +1510,28 @@ _mm256_mask_cvtusepi16_storeu_epi8 (void * __P, __mmask16 __M, __m256i __A)
__builtin_ia32_pmovuswb256mem_mask ((__v16qi*) __P, (__v16hi) __A, __M);
}
-static __inline__ __m128i __DEFAULT_FN_ATTRS128
+static __inline__ __m128i __DEFAULT_FN_ATTRS128_CONSTEXPR
_mm_mask_mulhrs_epi16(__m128i __W, __mmask8 __U, __m128i __X, __m128i __Y) {
return (__m128i)__builtin_ia32_selectw_128((__mmask8)__U,
(__v8hi)_mm_mulhrs_epi16(__X, __Y),
(__v8hi)__W);
}
-static __inline__ __m128i __DEFAULT_FN_ATTRS128
+static __inline__ __m128i __DEFAULT_FN_ATTRS128_CONSTEXPR
_mm_maskz_mulhrs_epi16(__mmask8 __U, __m128i __X, __m128i __Y) {
return (__m128i)__builtin_ia32_selectw_128((__mmask8)__U,
(__v8hi)_mm_mulhrs_epi16(__X, __Y),
(__v8hi)_mm_setzero_si128());
}
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
_mm256_mask_mulhrs_epi16(__m256i __W, __mmask16 __U, __m256i __X, __m256i __Y) {
return (__m256i)__builtin_ia32_selectw_256((__mmask16)__U,
(__v16hi)_mm256_mulhrs_epi16(__X, __Y),
(__v16hi)__W);
}
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
_mm256_maskz_mulhrs_epi16(__mmask16 __U, __m256i __X, __m256i __Y) {
return (__m256i)__builtin_ia32_selectw_256((__mmask16)__U,
(__v16hi)_mm256_mulhrs_epi16(__X, __Y),
diff --git a/clang/lib/Headers/avx512vlcdintrin.h b/clang/lib/Headers/avx512vlcdintrin.h
index 30c9f90..7719680f 100644
--- a/clang/lib/Headers/avx512vlcdintrin.h
+++ b/clang/lib/Headers/avx512vlcdintrin.h
@@ -14,208 +14,182 @@
#define __AVX512VLCDINTRIN_H
/* Define the default attributes for the functions in this file. */
+#if defined(__cplusplus) && (__cplusplus >= 201103L)
+#define __DEFAULT_FN_ATTRS128 \
+ constexpr __attribute__((__always_inline__, __nodebug__, \
+ __target__("avx512vl,avx512cd"), \
+ __min_vector_width__(128)))
+#define __DEFAULT_FN_ATTRS256 \
+ constexpr __attribute__((__always_inline__, __nodebug__, \
+ __target__("avx512vl,avx512cd"), \
+ __min_vector_width__(256)))
+#else
#define __DEFAULT_FN_ATTRS128 \
__attribute__((__always_inline__, __nodebug__, \
__target__("avx512vl,avx512cd"), __min_vector_width__(128)))
#define __DEFAULT_FN_ATTRS256 \
__attribute__((__always_inline__, __nodebug__, \
__target__("avx512vl,avx512cd"), __min_vector_width__(256)))
-
-#if defined(__cplusplus) && (__cplusplus >= 201103L)
-#define __DEFAULT_FN_ATTRS256_CONSTEXPR __DEFAULT_FN_ATTRS256 constexpr
-#define __DEFAULT_FN_ATTRS128_CONSTEXPR __DEFAULT_FN_ATTRS128 constexpr
-#else
-#define __DEFAULT_FN_ATTRS256_CONSTEXPR __DEFAULT_FN_ATTRS256
-#define __DEFAULT_FN_ATTRS128_CONSTEXPR __DEFAULT_FN_ATTRS128
#endif
static __inline__ __m128i __DEFAULT_FN_ATTRS128
-_mm_broadcastmb_epi64 (__mmask8 __A)
-{
+_mm_broadcastmb_epi64(__mmask8 __A) {
return (__m128i) _mm_set1_epi64x((long long) __A);
}
static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_broadcastmb_epi64 (__mmask8 __A)
-{
- return (__m256i) _mm256_set1_epi64x((long long)__A);
+_mm256_broadcastmb_epi64(__mmask8 __A) {
+ return (__m256i)_mm256_set1_epi64x((long long)__A);
}
static __inline__ __m128i __DEFAULT_FN_ATTRS128
-_mm_broadcastmw_epi32 (__mmask16 __A)
-{
+_mm_broadcastmw_epi32(__mmask16 __A) {
return (__m128i) _mm_set1_epi32((int)__A);
}
static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_broadcastmw_epi32 (__mmask16 __A)
-{
+_mm256_broadcastmw_epi32(__mmask16 __A) {
return (__m256i) _mm256_set1_epi32((int)__A);
}
-
static __inline__ __m128i __DEFAULT_FN_ATTRS128
-_mm_conflict_epi64 (__m128i __A)
-{
- return (__m128i) __builtin_ia32_vpconflictdi_128 ((__v2di) __A);
+_mm_conflict_epi64(__m128i __A) {
+ return (__m128i)__builtin_ia32_vpconflictdi_128((__v2di)__A);
}
static __inline__ __m128i __DEFAULT_FN_ATTRS128
-_mm_mask_conflict_epi64 (__m128i __W, __mmask8 __U, __m128i __A)
-{
- return (__m128i)__builtin_ia32_selectq_128((__mmask8)__U,
- (__v2di)_mm_conflict_epi64(__A),
- (__v2di)__W);
+_mm_mask_conflict_epi64(__m128i __W, __mmask8 __U, __m128i __A) {
+ return (__m128i)__builtin_ia32_selectq_128(
+ (__mmask8)__U, (__v2di)_mm_conflict_epi64(__A), (__v2di)__W);
}
static __inline__ __m128i __DEFAULT_FN_ATTRS128
-_mm_maskz_conflict_epi64 (__mmask8 __U, __m128i __A)
-{
+_mm_maskz_conflict_epi64(__mmask8 __U, __m128i __A) {
return (__m128i)__builtin_ia32_selectq_128((__mmask8)__U,
(__v2di)_mm_conflict_epi64(__A),
(__v2di)_mm_setzero_si128());
}
static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_conflict_epi64 (__m256i __A)
-{
- return (__m256i) __builtin_ia32_vpconflictdi_256 ((__v4di) __A);
+_mm256_conflict_epi64(__m256i __A) {
+ return (__m256i)__builtin_ia32_vpconflictdi_256((__v4di)__A);
}
static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_mask_conflict_epi64 (__m256i __W, __mmask8 __U, __m256i __A)
-{
- return (__m256i)__builtin_ia32_selectq_256((__mmask8)__U,
- (__v4di)_mm256_conflict_epi64(__A),
- (__v4di)__W);
+_mm256_mask_conflict_epi64(__m256i __W, __mmask8 __U, __m256i __A) {
+ return (__m256i)__builtin_ia32_selectq_256(
+ (__mmask8)__U, (__v4di)_mm256_conflict_epi64(__A), (__v4di)__W);
}
static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_maskz_conflict_epi64 (__mmask8 __U, __m256i __A)
-{
+_mm256_maskz_conflict_epi64(__mmask8 __U, __m256i __A) {
return (__m256i)__builtin_ia32_selectq_256((__mmask8)__U,
(__v4di)_mm256_conflict_epi64(__A),
(__v4di)_mm256_setzero_si256());
}
static __inline__ __m128i __DEFAULT_FN_ATTRS128
-_mm_conflict_epi32 (__m128i __A)
-{
- return (__m128i) __builtin_ia32_vpconflictsi_128 ((__v4si) __A);
+_mm_conflict_epi32(__m128i __A) {
+ return (__m128i)__builtin_ia32_vpconflictsi_128((__v4si)__A);
}
static __inline__ __m128i __DEFAULT_FN_ATTRS128
-_mm_mask_conflict_epi32 (__m128i __W, __mmask8 __U, __m128i __A)
-{
- return (__m128i)__builtin_ia32_selectd_128((__mmask8)__U,
- (__v4si)_mm_conflict_epi32(__A),
- (__v4si)__W);
+_mm_mask_conflict_epi32(__m128i __W, __mmask8 __U, __m128i __A) {
+ return (__m128i)__builtin_ia32_selectd_128(
+ (__mmask8)__U, (__v4si)_mm_conflict_epi32(__A), (__v4si)__W);
}
static __inline__ __m128i __DEFAULT_FN_ATTRS128
-_mm_maskz_conflict_epi32 (__mmask8 __U, __m128i __A)
-{
+_mm_maskz_conflict_epi32(__mmask8 __U, __m128i __A) {
return (__m128i)__builtin_ia32_selectd_128((__mmask8)__U,
(__v4si)_mm_conflict_epi32(__A),
(__v4si)_mm_setzero_si128());
}
static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_conflict_epi32 (__m256i __A)
-{
- return (__m256i) __builtin_ia32_vpconflictsi_256 ((__v8si) __A);
+_mm256_conflict_epi32(__m256i __A) {
+ return (__m256i)__builtin_ia32_vpconflictsi_256((__v8si)__A);
}
static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_mask_conflict_epi32 (__m256i __W, __mmask8 __U, __m256i __A)
-{
- return (__m256i)__builtin_ia32_selectd_256((__mmask8)__U,
- (__v8si)_mm256_conflict_epi32(__A),
- (__v8si)__W);
+_mm256_mask_conflict_epi32(__m256i __W, __mmask8 __U, __m256i __A) {
+ return (__m256i)__builtin_ia32_selectd_256(
+ (__mmask8)__U, (__v8si)_mm256_conflict_epi32(__A), (__v8si)__W);
}
static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_maskz_conflict_epi32 (__mmask8 __U, __m256i __A)
-{
+_mm256_maskz_conflict_epi32(__mmask8 __U, __m256i __A) {
return (__m256i)__builtin_ia32_selectd_256((__mmask8)__U,
(__v8si)_mm256_conflict_epi32(__A),
(__v8si)_mm256_setzero_si256());
}
-static __inline__ __m128i __DEFAULT_FN_ATTRS128_CONSTEXPR
-_mm_lzcnt_epi32(__m128i __A) {
+static __inline__ __m128i __DEFAULT_FN_ATTRS128 _mm_lzcnt_epi32(__m128i __A) {
return (__m128i)__builtin_elementwise_clzg((__v4si)__A,
(__v4si)_mm_set1_epi32(32));
}
-static __inline__ __m128i __DEFAULT_FN_ATTRS128_CONSTEXPR
+static __inline__ __m128i __DEFAULT_FN_ATTRS128
_mm_mask_lzcnt_epi32(__m128i __W, __mmask8 __U, __m128i __A) {
- return (__m128i)__builtin_ia32_selectd_128((__mmask8)__U,
- (__v4si)_mm_lzcnt_epi32(__A),
- (__v4si)__W);
+ return (__m128i)__builtin_ia32_selectd_128(
+ (__mmask8)__U, (__v4si)_mm_lzcnt_epi32(__A), (__v4si)__W);
}
-static __inline__ __m128i __DEFAULT_FN_ATTRS128_CONSTEXPR
+static __inline__ __m128i __DEFAULT_FN_ATTRS128
_mm_maskz_lzcnt_epi32(__mmask8 __U, __m128i __A) {
- return (__m128i)__builtin_ia32_selectd_128((__mmask8)__U,
- (__v4si)_mm_lzcnt_epi32(__A),
- (__v4si)_mm_setzero_si128());
+ return (__m128i)__builtin_ia32_selectd_128(
+ (__mmask8)__U, (__v4si)_mm_lzcnt_epi32(__A), (__v4si)_mm_setzero_si128());
}
-static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+static __inline__ __m256i __DEFAULT_FN_ATTRS256
_mm256_lzcnt_epi32(__m256i __A) {
return (__m256i)__builtin_elementwise_clzg((__v8si)__A,
(__v8si)_mm256_set1_epi32(32));
}
-static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+static __inline__ __m256i __DEFAULT_FN_ATTRS256
_mm256_mask_lzcnt_epi32(__m256i __W, __mmask8 __U, __m256i __A) {
- return (__m256i)__builtin_ia32_selectd_256((__mmask8)__U,
- (__v8si)_mm256_lzcnt_epi32(__A),
- (__v8si)__W);
+ return (__m256i)__builtin_ia32_selectd_256(
+ (__mmask8)__U, (__v8si)_mm256_lzcnt_epi32(__A), (__v8si)__W);
}
-static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+static __inline__ __m256i __DEFAULT_FN_ATTRS256
_mm256_maskz_lzcnt_epi32(__mmask8 __U, __m256i __A) {
return (__m256i)__builtin_ia32_selectd_256((__mmask8)__U,
(__v8si)_mm256_lzcnt_epi32(__A),
(__v8si)_mm256_setzero_si256());
}
-static __inline__ __m128i __DEFAULT_FN_ATTRS128_CONSTEXPR
-_mm_lzcnt_epi64(__m128i __A) {
+static __inline__ __m128i __DEFAULT_FN_ATTRS128 _mm_lzcnt_epi64(__m128i __A) {
return (__m128i)__builtin_elementwise_clzg(
(__v2di)__A, (__v2di)_mm_set1_epi64x((long long)64));
}
-static __inline__ __m128i __DEFAULT_FN_ATTRS128_CONSTEXPR
+static __inline__ __m128i __DEFAULT_FN_ATTRS128
_mm_mask_lzcnt_epi64(__m128i __W, __mmask8 __U, __m128i __A) {
- return (__m128i)__builtin_ia32_selectq_128((__mmask8)__U,
- (__v2di)_mm_lzcnt_epi64(__A),
- (__v2di)__W);
+ return (__m128i)__builtin_ia32_selectq_128(
+ (__mmask8)__U, (__v2di)_mm_lzcnt_epi64(__A), (__v2di)__W);
}
-static __inline__ __m128i __DEFAULT_FN_ATTRS128_CONSTEXPR
+static __inline__ __m128i __DEFAULT_FN_ATTRS128
_mm_maskz_lzcnt_epi64(__mmask8 __U, __m128i __A) {
- return (__m128i)__builtin_ia32_selectq_128((__mmask8)__U,
- (__v2di)_mm_lzcnt_epi64(__A),
- (__v2di)_mm_setzero_si128());
+ return (__m128i)__builtin_ia32_selectq_128(
+ (__mmask8)__U, (__v2di)_mm_lzcnt_epi64(__A), (__v2di)_mm_setzero_si128());
}
-static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+static __inline__ __m256i __DEFAULT_FN_ATTRS256
_mm256_lzcnt_epi64(__m256i __A) {
return (__m256i)__builtin_elementwise_clzg(
(__v4di)__A, (__v4di)_mm256_set1_epi64x((long long)64));
}
-static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+static __inline__ __m256i __DEFAULT_FN_ATTRS256
_mm256_mask_lzcnt_epi64(__m256i __W, __mmask8 __U, __m256i __A) {
- return (__m256i)__builtin_ia32_selectq_256((__mmask8)__U,
- (__v4di)_mm256_lzcnt_epi64(__A),
- (__v4di)__W);
+ return (__m256i)__builtin_ia32_selectq_256(
+ (__mmask8)__U, (__v4di)_mm256_lzcnt_epi64(__A), (__v4di)__W);
}
-static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+static __inline__ __m256i __DEFAULT_FN_ATTRS256
_mm256_maskz_lzcnt_epi64(__mmask8 __U, __m256i __A) {
return (__m256i)__builtin_ia32_selectq_256((__mmask8)__U,
(__v4di)_mm256_lzcnt_epi64(__A),
@@ -224,7 +198,5 @@ _mm256_maskz_lzcnt_epi64(__mmask8 __U, __m256i __A) {
#undef __DEFAULT_FN_ATTRS128
#undef __DEFAULT_FN_ATTRS256
-#undef __DEFAULT_FN_ATTRS128_CONSTEXPR
-#undef __DEFAULT_FN_ATTRS256_CONSTEXPR
#endif /* __AVX512VLCDINTRIN_H */
diff --git a/clang/lib/Headers/avx512vldqintrin.h b/clang/lib/Headers/avx512vldqintrin.h
index 68bd52e..ee7974e 100644
--- a/clang/lib/Headers/avx512vldqintrin.h
+++ b/clang/lib/Headers/avx512vldqintrin.h
@@ -968,17 +968,15 @@ _mm256_broadcast_f32x2(__m128 __A) {
0, 1, 0, 1, 0, 1, 0, 1);
}
-static __inline__ __m256 __DEFAULT_FN_ATTRS256
-_mm256_mask_broadcast_f32x2 (__m256 __O, __mmask8 __M, __m128 __A)
-{
+static __inline__ __m256 __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_mask_broadcast_f32x2(__m256 __O, __mmask8 __M, __m128 __A) {
return (__m256)__builtin_ia32_selectps_256((__mmask8)__M,
(__v8sf)_mm256_broadcast_f32x2(__A),
(__v8sf)__O);
}
-static __inline__ __m256 __DEFAULT_FN_ATTRS256
-_mm256_maskz_broadcast_f32x2 (__mmask8 __M, __m128 __A)
-{
+static __inline__ __m256 __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_maskz_broadcast_f32x2(__mmask8 __M, __m128 __A) {
return (__m256)__builtin_ia32_selectps_256((__mmask8)__M,
(__v8sf)_mm256_broadcast_f32x2(__A),
(__v8sf)_mm256_setzero_ps());
@@ -990,17 +988,15 @@ _mm256_broadcast_f64x2(__m128d __A) {
0, 1, 0, 1);
}
-static __inline__ __m256d __DEFAULT_FN_ATTRS256
-_mm256_mask_broadcast_f64x2(__m256d __O, __mmask8 __M, __m128d __A)
-{
+static __inline__ __m256d __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_mask_broadcast_f64x2(__m256d __O, __mmask8 __M, __m128d __A) {
return (__m256d)__builtin_ia32_selectpd_256((__mmask8)__M,
(__v4df)_mm256_broadcast_f64x2(__A),
(__v4df)__O);
}
-static __inline__ __m256d __DEFAULT_FN_ATTRS256
-_mm256_maskz_broadcast_f64x2 (__mmask8 __M, __m128d __A)
-{
+static __inline__ __m256d __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_maskz_broadcast_f64x2(__mmask8 __M, __m128d __A) {
return (__m256d)__builtin_ia32_selectpd_256((__mmask8)__M,
(__v4df)_mm256_broadcast_f64x2(__A),
(__v4df)_mm256_setzero_pd());
@@ -1012,17 +1008,15 @@ _mm_broadcast_i32x2(__m128i __A) {
0, 1, 0, 1);
}
-static __inline__ __m128i __DEFAULT_FN_ATTRS128
-_mm_mask_broadcast_i32x2 (__m128i __O, __mmask8 __M, __m128i __A)
-{
+static __inline__ __m128i __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_mask_broadcast_i32x2(__m128i __O, __mmask8 __M, __m128i __A) {
return (__m128i)__builtin_ia32_selectd_128((__mmask8)__M,
(__v4si)_mm_broadcast_i32x2(__A),
(__v4si)__O);
}
-static __inline__ __m128i __DEFAULT_FN_ATTRS128
-_mm_maskz_broadcast_i32x2 (__mmask8 __M, __m128i __A)
-{
+static __inline__ __m128i __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_maskz_broadcast_i32x2(__mmask8 __M, __m128i __A) {
return (__m128i)__builtin_ia32_selectd_128((__mmask8)__M,
(__v4si)_mm_broadcast_i32x2(__A),
(__v4si)_mm_setzero_si128());
@@ -1034,17 +1028,15 @@ _mm256_broadcast_i32x2(__m128i __A) {
0, 1, 0, 1, 0, 1, 0, 1);
}
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_mask_broadcast_i32x2 (__m256i __O, __mmask8 __M, __m128i __A)
-{
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_mask_broadcast_i32x2(__m256i __O, __mmask8 __M, __m128i __A) {
return (__m256i)__builtin_ia32_selectd_256((__mmask8)__M,
(__v8si)_mm256_broadcast_i32x2(__A),
(__v8si)__O);
}
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_maskz_broadcast_i32x2 (__mmask8 __M, __m128i __A)
-{
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_maskz_broadcast_i32x2(__mmask8 __M, __m128i __A) {
return (__m256i)__builtin_ia32_selectd_256((__mmask8)__M,
(__v8si)_mm256_broadcast_i32x2(__A),
(__v8si)_mm256_setzero_si256());
@@ -1056,17 +1048,15 @@ _mm256_broadcast_i64x2(__m128i __A) {
0, 1, 0, 1);
}
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_mask_broadcast_i64x2(__m256i __O, __mmask8 __M, __m128i __A)
-{
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_mask_broadcast_i64x2(__m256i __O, __mmask8 __M, __m128i __A) {
return (__m256i)__builtin_ia32_selectq_256((__mmask8)__M,
(__v4di)_mm256_broadcast_i64x2(__A),
(__v4di)__O);
}
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_maskz_broadcast_i64x2 (__mmask8 __M, __m128i __A)
-{
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_maskz_broadcast_i64x2(__mmask8 __M, __m128i __A) {
return (__m256i)__builtin_ia32_selectq_256((__mmask8)__M,
(__v4di)_mm256_broadcast_i64x2(__A),
(__v4di)_mm256_setzero_si256());
diff --git a/clang/lib/Headers/avx512vlintrin.h b/clang/lib/Headers/avx512vlintrin.h
index 965741f..676b5a0 100644
--- a/clang/lib/Headers/avx512vlintrin.h
+++ b/clang/lib/Headers/avx512vlintrin.h
@@ -5101,69 +5101,55 @@ _mm256_maskz_movedup_pd (__mmask8 __U, __m256d __A)
(__v4df)_mm256_setzero_pd());
}
-static __inline__ __m128i __DEFAULT_FN_ATTRS128
-_mm_mask_set1_epi32(__m128i __O, __mmask8 __M, int __A)
-{
- return (__m128i)__builtin_ia32_selectd_128(__M,
- (__v4si) _mm_set1_epi32(__A),
- (__v4si)__O);
+static __inline__ __m128i __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_mask_set1_epi32(__m128i __O, __mmask8 __M, int __A) {
+ return (__m128i)__builtin_ia32_selectd_128(__M, (__v4si)_mm_set1_epi32(__A),
+ (__v4si)__O);
}
-static __inline__ __m128i __DEFAULT_FN_ATTRS128
-_mm_maskz_set1_epi32( __mmask8 __M, int __A)
-{
- return (__m128i)__builtin_ia32_selectd_128(__M,
- (__v4si) _mm_set1_epi32(__A),
- (__v4si)_mm_setzero_si128());
+static __inline__ __m128i __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_maskz_set1_epi32(__mmask8 __M, int __A) {
+ return (__m128i)__builtin_ia32_selectd_128(__M, (__v4si)_mm_set1_epi32(__A),
+ (__v4si)_mm_setzero_si128());
}
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_mask_set1_epi32(__m256i __O, __mmask8 __M, int __A)
-{
- return (__m256i)__builtin_ia32_selectd_256(__M,
- (__v8si) _mm256_set1_epi32(__A),
- (__v8si)__O);
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_mask_set1_epi32(__m256i __O, __mmask8 __M, int __A) {
+ return (__m256i)__builtin_ia32_selectd_256(
+ __M, (__v8si)_mm256_set1_epi32(__A), (__v8si)__O);
}
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_maskz_set1_epi32( __mmask8 __M, int __A)
-{
- return (__m256i)__builtin_ia32_selectd_256(__M,
- (__v8si) _mm256_set1_epi32(__A),
- (__v8si)_mm256_setzero_si256());
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_maskz_set1_epi32(__mmask8 __M, int __A) {
+ return (__m256i)__builtin_ia32_selectd_256(
+ __M, (__v8si)_mm256_set1_epi32(__A), (__v8si)_mm256_setzero_si256());
}
-
-static __inline__ __m128i __DEFAULT_FN_ATTRS128
-_mm_mask_set1_epi64 (__m128i __O, __mmask8 __M, long long __A)
-{
+static __inline__ __m128i __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_mask_set1_epi64(__m128i __O, __mmask8 __M, long long __A) {
return (__m128i) __builtin_ia32_selectq_128(__M,
(__v2di) _mm_set1_epi64x(__A),
(__v2di) __O);
}
-static __inline__ __m128i __DEFAULT_FN_ATTRS128
-_mm_maskz_set1_epi64 (__mmask8 __M, long long __A)
-{
+static __inline__ __m128i __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_maskz_set1_epi64(__mmask8 __M, long long __A) {
return (__m128i) __builtin_ia32_selectq_128(__M,
(__v2di) _mm_set1_epi64x(__A),
(__v2di) _mm_setzero_si128());
}
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_mask_set1_epi64 (__m256i __O, __mmask8 __M, long long __A)
-{
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_mask_set1_epi64(__m256i __O, __mmask8 __M, long long __A) {
return (__m256i) __builtin_ia32_selectq_256(__M,
(__v4di) _mm256_set1_epi64x(__A),
(__v4di) __O) ;
}
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_maskz_set1_epi64 (__mmask8 __M, long long __A)
-{
- return (__m256i) __builtin_ia32_selectq_256(__M,
- (__v4di) _mm256_set1_epi64x(__A),
- (__v4di) _mm256_setzero_si256());
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_maskz_set1_epi64(__mmask8 __M, long long __A) {
+ return (__m256i)__builtin_ia32_selectq_256(
+ __M, (__v4di)_mm256_set1_epi64x(__A), (__v4di)_mm256_setzero_si256());
}
#define _mm_fixupimm_pd(A, B, C, imm) \
@@ -5610,130 +5596,113 @@ _mm256_mask_storeu_ps (void *__P, __mmask8 __U, __m256 __A)
(__mmask8) __U);
}
-
-static __inline__ __m128d __DEFAULT_FN_ATTRS128
-_mm_mask_unpackhi_pd(__m128d __W, __mmask8 __U, __m128d __A, __m128d __B)
-{
+static __inline__ __m128d __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_mask_unpackhi_pd(__m128d __W, __mmask8 __U, __m128d __A, __m128d __B) {
return (__m128d)__builtin_ia32_selectpd_128((__mmask8)__U,
(__v2df)_mm_unpackhi_pd(__A, __B),
(__v2df)__W);
}
-static __inline__ __m128d __DEFAULT_FN_ATTRS128
-_mm_maskz_unpackhi_pd(__mmask8 __U, __m128d __A, __m128d __B)
-{
+static __inline__ __m128d __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_maskz_unpackhi_pd(__mmask8 __U, __m128d __A, __m128d __B) {
return (__m128d)__builtin_ia32_selectpd_128((__mmask8)__U,
(__v2df)_mm_unpackhi_pd(__A, __B),
(__v2df)_mm_setzero_pd());
}
-static __inline__ __m256d __DEFAULT_FN_ATTRS256
-_mm256_mask_unpackhi_pd(__m256d __W, __mmask8 __U, __m256d __A, __m256d __B)
-{
+static __inline__ __m256d __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_mask_unpackhi_pd(__m256d __W, __mmask8 __U, __m256d __A, __m256d __B) {
return (__m256d)__builtin_ia32_selectpd_256((__mmask8)__U,
(__v4df)_mm256_unpackhi_pd(__A, __B),
(__v4df)__W);
}
-static __inline__ __m256d __DEFAULT_FN_ATTRS256
-_mm256_maskz_unpackhi_pd(__mmask8 __U, __m256d __A, __m256d __B)
-{
+static __inline__ __m256d __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_maskz_unpackhi_pd(__mmask8 __U, __m256d __A, __m256d __B) {
return (__m256d)__builtin_ia32_selectpd_256((__mmask8)__U,
(__v4df)_mm256_unpackhi_pd(__A, __B),
(__v4df)_mm256_setzero_pd());
}
-static __inline__ __m128 __DEFAULT_FN_ATTRS128
-_mm_mask_unpackhi_ps(__m128 __W, __mmask8 __U, __m128 __A, __m128 __B)
-{
+static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_mask_unpackhi_ps(__m128 __W, __mmask8 __U, __m128 __A, __m128 __B) {
return (__m128)__builtin_ia32_selectps_128((__mmask8)__U,
(__v4sf)_mm_unpackhi_ps(__A, __B),
(__v4sf)__W);
}
-static __inline__ __m128 __DEFAULT_FN_ATTRS128
-_mm_maskz_unpackhi_ps(__mmask8 __U, __m128 __A, __m128 __B)
-{
+static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_maskz_unpackhi_ps(__mmask8 __U, __m128 __A, __m128 __B) {
return (__m128)__builtin_ia32_selectps_128((__mmask8)__U,
(__v4sf)_mm_unpackhi_ps(__A, __B),
(__v4sf)_mm_setzero_ps());
}
-static __inline__ __m256 __DEFAULT_FN_ATTRS256
-_mm256_mask_unpackhi_ps(__m256 __W, __mmask8 __U, __m256 __A, __m256 __B)
-{
+static __inline__ __m256 __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_mask_unpackhi_ps(__m256 __W, __mmask8 __U, __m256 __A, __m256 __B) {
return (__m256)__builtin_ia32_selectps_256((__mmask8)__U,
(__v8sf)_mm256_unpackhi_ps(__A, __B),
(__v8sf)__W);
}
-static __inline__ __m256 __DEFAULT_FN_ATTRS256
-_mm256_maskz_unpackhi_ps(__mmask8 __U, __m256 __A, __m256 __B)
-{
+static __inline__ __m256 __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_maskz_unpackhi_ps(__mmask8 __U, __m256 __A, __m256 __B) {
return (__m256)__builtin_ia32_selectps_256((__mmask8)__U,
(__v8sf)_mm256_unpackhi_ps(__A, __B),
(__v8sf)_mm256_setzero_ps());
}
-static __inline__ __m128d __DEFAULT_FN_ATTRS128
-_mm_mask_unpacklo_pd(__m128d __W, __mmask8 __U, __m128d __A, __m128d __B)
-{
+static __inline__ __m128d __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_mask_unpacklo_pd(__m128d __W, __mmask8 __U, __m128d __A, __m128d __B) {
return (__m128d)__builtin_ia32_selectpd_128((__mmask8)__U,
(__v2df)_mm_unpacklo_pd(__A, __B),
(__v2df)__W);
}
-static __inline__ __m128d __DEFAULT_FN_ATTRS128
-_mm_maskz_unpacklo_pd(__mmask8 __U, __m128d __A, __m128d __B)
-{
+static __inline__ __m128d __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_maskz_unpacklo_pd(__mmask8 __U, __m128d __A, __m128d __B) {
return (__m128d)__builtin_ia32_selectpd_128((__mmask8)__U,
(__v2df)_mm_unpacklo_pd(__A, __B),
(__v2df)_mm_setzero_pd());
}
-static __inline__ __m256d __DEFAULT_FN_ATTRS256
-_mm256_mask_unpacklo_pd(__m256d __W, __mmask8 __U, __m256d __A, __m256d __B)
-{
+static __inline__ __m256d __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_mask_unpacklo_pd(__m256d __W, __mmask8 __U, __m256d __A, __m256d __B) {
return (__m256d)__builtin_ia32_selectpd_256((__mmask8)__U,
(__v4df)_mm256_unpacklo_pd(__A, __B),
(__v4df)__W);
}
-static __inline__ __m256d __DEFAULT_FN_ATTRS256
-_mm256_maskz_unpacklo_pd(__mmask8 __U, __m256d __A, __m256d __B)
-{
+static __inline__ __m256d __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_maskz_unpacklo_pd(__mmask8 __U, __m256d __A, __m256d __B) {
return (__m256d)__builtin_ia32_selectpd_256((__mmask8)__U,
(__v4df)_mm256_unpacklo_pd(__A, __B),
(__v4df)_mm256_setzero_pd());
}
-static __inline__ __m128 __DEFAULT_FN_ATTRS128
-_mm_mask_unpacklo_ps(__m128 __W, __mmask8 __U, __m128 __A, __m128 __B)
-{
+static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_mask_unpacklo_ps(__m128 __W, __mmask8 __U, __m128 __A, __m128 __B) {
return (__m128)__builtin_ia32_selectps_128((__mmask8)__U,
(__v4sf)_mm_unpacklo_ps(__A, __B),
(__v4sf)__W);
}
-static __inline__ __m128 __DEFAULT_FN_ATTRS128
-_mm_maskz_unpacklo_ps(__mmask8 __U, __m128 __A, __m128 __B)
-{
+static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_maskz_unpacklo_ps(__mmask8 __U, __m128 __A, __m128 __B) {
return (__m128)__builtin_ia32_selectps_128((__mmask8)__U,
(__v4sf)_mm_unpacklo_ps(__A, __B),
(__v4sf)_mm_setzero_ps());
}
-static __inline__ __m256 __DEFAULT_FN_ATTRS256
-_mm256_mask_unpacklo_ps(__m256 __W, __mmask8 __U, __m256 __A, __m256 __B)
-{
+static __inline__ __m256 __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_mask_unpacklo_ps(__m256 __W, __mmask8 __U, __m256 __A, __m256 __B) {
return (__m256)__builtin_ia32_selectps_256((__mmask8)__U,
(__v8sf)_mm256_unpacklo_ps(__A, __B),
(__v8sf)__W);
}
-static __inline__ __m256 __DEFAULT_FN_ATTRS256
-_mm256_maskz_unpacklo_ps(__mmask8 __U, __m256 __A, __m256 __B)
-{
+static __inline__ __m256 __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_maskz_unpacklo_ps(__mmask8 __U, __m256 __A, __m256 __B) {
return (__m256)__builtin_ia32_selectps_256((__mmask8)__U,
(__v8sf)_mm256_unpacklo_ps(__A, __B),
(__v8sf)_mm256_setzero_ps());
@@ -6055,129 +6024,117 @@ _mm256_mask_testn_epi64_mask (__mmask8 __U, __m256i __A, __m256i __B)
_mm256_setzero_si256());
}
-static __inline__ __m128i __DEFAULT_FN_ATTRS128
-_mm_mask_unpackhi_epi32(__m128i __W, __mmask8 __U, __m128i __A, __m128i __B)
-{
+static __inline__ __m128i __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_mask_unpackhi_epi32(__m128i __W, __mmask8 __U, __m128i __A, __m128i __B) {
return (__m128i)__builtin_ia32_selectd_128((__mmask8)__U,
(__v4si)_mm_unpackhi_epi32(__A, __B),
(__v4si)__W);
}
-static __inline__ __m128i __DEFAULT_FN_ATTRS128
-_mm_maskz_unpackhi_epi32(__mmask8 __U, __m128i __A, __m128i __B)
-{
+static __inline__ __m128i __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_maskz_unpackhi_epi32(__mmask8 __U, __m128i __A, __m128i __B) {
return (__m128i)__builtin_ia32_selectd_128((__mmask8)__U,
(__v4si)_mm_unpackhi_epi32(__A, __B),
(__v4si)_mm_setzero_si128());
}
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_mask_unpackhi_epi32(__m256i __W, __mmask8 __U, __m256i __A, __m256i __B)
-{
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_mask_unpackhi_epi32(__m256i __W, __mmask8 __U, __m256i __A,
+ __m256i __B) {
return (__m256i)__builtin_ia32_selectd_256((__mmask8)__U,
(__v8si)_mm256_unpackhi_epi32(__A, __B),
(__v8si)__W);
}
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_maskz_unpackhi_epi32(__mmask8 __U, __m256i __A, __m256i __B)
-{
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_maskz_unpackhi_epi32(__mmask8 __U, __m256i __A, __m256i __B) {
return (__m256i)__builtin_ia32_selectd_256((__mmask8)__U,
(__v8si)_mm256_unpackhi_epi32(__A, __B),
(__v8si)_mm256_setzero_si256());
}
-static __inline__ __m128i __DEFAULT_FN_ATTRS128
-_mm_mask_unpackhi_epi64(__m128i __W, __mmask8 __U, __m128i __A, __m128i __B)
-{
+static __inline__ __m128i __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_mask_unpackhi_epi64(__m128i __W, __mmask8 __U, __m128i __A, __m128i __B) {
return (__m128i)__builtin_ia32_selectq_128((__mmask8)__U,
(__v2di)_mm_unpackhi_epi64(__A, __B),
(__v2di)__W);
}
-static __inline__ __m128i __DEFAULT_FN_ATTRS128
-_mm_maskz_unpackhi_epi64(__mmask8 __U, __m128i __A, __m128i __B)
-{
+static __inline__ __m128i __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_maskz_unpackhi_epi64(__mmask8 __U, __m128i __A, __m128i __B) {
return (__m128i)__builtin_ia32_selectq_128((__mmask8)__U,
(__v2di)_mm_unpackhi_epi64(__A, __B),
(__v2di)_mm_setzero_si128());
}
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_mask_unpackhi_epi64(__m256i __W, __mmask8 __U, __m256i __A, __m256i __B)
-{
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_mask_unpackhi_epi64(__m256i __W, __mmask8 __U, __m256i __A,
+ __m256i __B) {
return (__m256i)__builtin_ia32_selectq_256((__mmask8)__U,
(__v4di)_mm256_unpackhi_epi64(__A, __B),
(__v4di)__W);
}
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_maskz_unpackhi_epi64(__mmask8 __U, __m256i __A, __m256i __B)
-{
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_maskz_unpackhi_epi64(__mmask8 __U, __m256i __A, __m256i __B) {
return (__m256i)__builtin_ia32_selectq_256((__mmask8)__U,
(__v4di)_mm256_unpackhi_epi64(__A, __B),
(__v4di)_mm256_setzero_si256());
}
-static __inline__ __m128i __DEFAULT_FN_ATTRS128
-_mm_mask_unpacklo_epi32(__m128i __W, __mmask8 __U, __m128i __A, __m128i __B)
-{
+static __inline__ __m128i __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_mask_unpacklo_epi32(__m128i __W, __mmask8 __U, __m128i __A, __m128i __B) {
return (__m128i)__builtin_ia32_selectd_128((__mmask8)__U,
(__v4si)_mm_unpacklo_epi32(__A, __B),
(__v4si)__W);
}
-static __inline__ __m128i __DEFAULT_FN_ATTRS128
-_mm_maskz_unpacklo_epi32(__mmask8 __U, __m128i __A, __m128i __B)
-{
+static __inline__ __m128i __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_maskz_unpacklo_epi32(__mmask8 __U, __m128i __A, __m128i __B) {
return (__m128i)__builtin_ia32_selectd_128((__mmask8)__U,
(__v4si)_mm_unpacklo_epi32(__A, __B),
(__v4si)_mm_setzero_si128());
}
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_mask_unpacklo_epi32(__m256i __W, __mmask8 __U, __m256i __A, __m256i __B)
-{
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_mask_unpacklo_epi32(__m256i __W, __mmask8 __U, __m256i __A,
+ __m256i __B) {
return (__m256i)__builtin_ia32_selectd_256((__mmask8)__U,
(__v8si)_mm256_unpacklo_epi32(__A, __B),
(__v8si)__W);
}
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_maskz_unpacklo_epi32(__mmask8 __U, __m256i __A, __m256i __B)
-{
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_maskz_unpacklo_epi32(__mmask8 __U, __m256i __A, __m256i __B) {
return (__m256i)__builtin_ia32_selectd_256((__mmask8)__U,
(__v8si)_mm256_unpacklo_epi32(__A, __B),
(__v8si)_mm256_setzero_si256());
}
-static __inline__ __m128i __DEFAULT_FN_ATTRS128
-_mm_mask_unpacklo_epi64(__m128i __W, __mmask8 __U, __m128i __A, __m128i __B)
-{
+static __inline__ __m128i __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_mask_unpacklo_epi64(__m128i __W, __mmask8 __U, __m128i __A, __m128i __B) {
return (__m128i)__builtin_ia32_selectq_128((__mmask8)__U,
(__v2di)_mm_unpacklo_epi64(__A, __B),
(__v2di)__W);
}
-static __inline__ __m128i __DEFAULT_FN_ATTRS128
-_mm_maskz_unpacklo_epi64(__mmask8 __U, __m128i __A, __m128i __B)
-{
+static __inline__ __m128i __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_maskz_unpacklo_epi64(__mmask8 __U, __m128i __A, __m128i __B) {
return (__m128i)__builtin_ia32_selectq_128((__mmask8)__U,
(__v2di)_mm_unpacklo_epi64(__A, __B),
(__v2di)_mm_setzero_si128());
}
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_mask_unpacklo_epi64(__m256i __W, __mmask8 __U, __m256i __A, __m256i __B)
-{
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_mask_unpacklo_epi64(__m256i __W, __mmask8 __U, __m256i __A,
+ __m256i __B) {
return (__m256i)__builtin_ia32_selectq_256((__mmask8)__U,
(__v4di)_mm256_unpacklo_epi64(__A, __B),
(__v4di)__W);
}
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_maskz_unpacklo_epi64(__mmask8 __U, __m256i __A, __m256i __B)
-{
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_maskz_unpacklo_epi64(__mmask8 __U, __m256i __A, __m256i __B) {
return (__m256i)__builtin_ia32_selectq_256((__mmask8)__U,
(__v4di)_mm256_unpacklo_epi64(__A, __B),
(__v4di)_mm256_setzero_si256());
@@ -6594,17 +6551,15 @@ _mm256_broadcast_f32x4(__m128 __A) {
0, 1, 2, 3, 0, 1, 2, 3);
}
-static __inline__ __m256 __DEFAULT_FN_ATTRS256
-_mm256_mask_broadcast_f32x4(__m256 __O, __mmask8 __M, __m128 __A)
-{
+static __inline__ __m256 __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_mask_broadcast_f32x4(__m256 __O, __mmask8 __M, __m128 __A) {
return (__m256)__builtin_ia32_selectps_256((__mmask8)__M,
(__v8sf)_mm256_broadcast_f32x4(__A),
(__v8sf)__O);
}
-static __inline__ __m256 __DEFAULT_FN_ATTRS256
-_mm256_maskz_broadcast_f32x4 (__mmask8 __M, __m128 __A)
-{
+static __inline__ __m256 __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_maskz_broadcast_f32x4(__mmask8 __M, __m128 __A) {
return (__m256)__builtin_ia32_selectps_256((__mmask8)__M,
(__v8sf)_mm256_broadcast_f32x4(__A),
(__v8sf)_mm256_setzero_ps());
@@ -6616,129 +6571,113 @@ _mm256_broadcast_i32x4(__m128i __A) {
0, 1, 2, 3, 0, 1, 2, 3);
}
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_mask_broadcast_i32x4(__m256i __O, __mmask8 __M, __m128i __A)
-{
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_mask_broadcast_i32x4(__m256i __O, __mmask8 __M, __m128i __A) {
return (__m256i)__builtin_ia32_selectd_256((__mmask8)__M,
(__v8si)_mm256_broadcast_i32x4(__A),
(__v8si)__O);
}
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_maskz_broadcast_i32x4(__mmask8 __M, __m128i __A)
-{
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_maskz_broadcast_i32x4(__mmask8 __M, __m128i __A) {
return (__m256i)__builtin_ia32_selectd_256((__mmask8)__M,
(__v8si)_mm256_broadcast_i32x4(__A),
(__v8si)_mm256_setzero_si256());
}
-static __inline__ __m256d __DEFAULT_FN_ATTRS256
-_mm256_mask_broadcastsd_pd (__m256d __O, __mmask8 __M, __m128d __A)
-{
+static __inline__ __m256d __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_mask_broadcastsd_pd(__m256d __O, __mmask8 __M, __m128d __A) {
return (__m256d)__builtin_ia32_selectpd_256(__M,
(__v4df) _mm256_broadcastsd_pd(__A),
(__v4df) __O);
}
-static __inline__ __m256d __DEFAULT_FN_ATTRS256
-_mm256_maskz_broadcastsd_pd (__mmask8 __M, __m128d __A)
-{
+static __inline__ __m256d __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_maskz_broadcastsd_pd(__mmask8 __M, __m128d __A) {
return (__m256d)__builtin_ia32_selectpd_256(__M,
(__v4df) _mm256_broadcastsd_pd(__A),
(__v4df) _mm256_setzero_pd());
}
-static __inline__ __m128 __DEFAULT_FN_ATTRS128
-_mm_mask_broadcastss_ps (__m128 __O, __mmask8 __M, __m128 __A)
-{
+static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_mask_broadcastss_ps(__m128 __O, __mmask8 __M, __m128 __A) {
return (__m128)__builtin_ia32_selectps_128(__M,
(__v4sf) _mm_broadcastss_ps(__A),
(__v4sf) __O);
}
-static __inline__ __m128 __DEFAULT_FN_ATTRS128
-_mm_maskz_broadcastss_ps (__mmask8 __M, __m128 __A)
-{
+static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_maskz_broadcastss_ps(__mmask8 __M, __m128 __A) {
return (__m128)__builtin_ia32_selectps_128(__M,
(__v4sf) _mm_broadcastss_ps(__A),
(__v4sf) _mm_setzero_ps());
}
-static __inline__ __m256 __DEFAULT_FN_ATTRS256
-_mm256_mask_broadcastss_ps (__m256 __O, __mmask8 __M, __m128 __A)
-{
+static __inline__ __m256 __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_mask_broadcastss_ps(__m256 __O, __mmask8 __M, __m128 __A) {
return (__m256)__builtin_ia32_selectps_256(__M,
(__v8sf) _mm256_broadcastss_ps(__A),
(__v8sf) __O);
}
-static __inline__ __m256 __DEFAULT_FN_ATTRS256
-_mm256_maskz_broadcastss_ps (__mmask8 __M, __m128 __A)
-{
+static __inline__ __m256 __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_maskz_broadcastss_ps(__mmask8 __M, __m128 __A) {
return (__m256)__builtin_ia32_selectps_256(__M,
(__v8sf) _mm256_broadcastss_ps(__A),
(__v8sf) _mm256_setzero_ps());
}
-static __inline__ __m128i __DEFAULT_FN_ATTRS128
-_mm_mask_broadcastd_epi32 (__m128i __O, __mmask8 __M, __m128i __A)
-{
+static __inline__ __m128i __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_mask_broadcastd_epi32(__m128i __O, __mmask8 __M, __m128i __A) {
return (__m128i)__builtin_ia32_selectd_128(__M,
(__v4si) _mm_broadcastd_epi32(__A),
(__v4si) __O);
}
-static __inline__ __m128i __DEFAULT_FN_ATTRS128
-_mm_maskz_broadcastd_epi32 (__mmask8 __M, __m128i __A)
-{
+static __inline__ __m128i __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_maskz_broadcastd_epi32(__mmask8 __M, __m128i __A) {
return (__m128i)__builtin_ia32_selectd_128(__M,
(__v4si) _mm_broadcastd_epi32(__A),
(__v4si) _mm_setzero_si128());
}
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_mask_broadcastd_epi32 (__m256i __O, __mmask8 __M, __m128i __A)
-{
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_mask_broadcastd_epi32(__m256i __O, __mmask8 __M, __m128i __A) {
return (__m256i)__builtin_ia32_selectd_256(__M,
(__v8si) _mm256_broadcastd_epi32(__A),
(__v8si) __O);
}
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_maskz_broadcastd_epi32 (__mmask8 __M, __m128i __A)
-{
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_maskz_broadcastd_epi32(__mmask8 __M, __m128i __A) {
return (__m256i)__builtin_ia32_selectd_256(__M,
(__v8si) _mm256_broadcastd_epi32(__A),
(__v8si) _mm256_setzero_si256());
}
-static __inline__ __m128i __DEFAULT_FN_ATTRS128
-_mm_mask_broadcastq_epi64 (__m128i __O, __mmask8 __M, __m128i __A)
-{
+static __inline__ __m128i __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_mask_broadcastq_epi64(__m128i __O, __mmask8 __M, __m128i __A) {
return (__m128i)__builtin_ia32_selectq_128(__M,
(__v2di) _mm_broadcastq_epi64(__A),
(__v2di) __O);
}
-static __inline__ __m128i __DEFAULT_FN_ATTRS128
-_mm_maskz_broadcastq_epi64 (__mmask8 __M, __m128i __A)
-{
+static __inline__ __m128i __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_maskz_broadcastq_epi64(__mmask8 __M, __m128i __A) {
return (__m128i)__builtin_ia32_selectq_128(__M,
(__v2di) _mm_broadcastq_epi64(__A),
(__v2di) _mm_setzero_si128());
}
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_mask_broadcastq_epi64 (__m256i __O, __mmask8 __M, __m128i __A)
-{
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_mask_broadcastq_epi64(__m256i __O, __mmask8 __M, __m128i __A) {
return (__m256i)__builtin_ia32_selectq_256(__M,
(__v4di) _mm256_broadcastq_epi64(__A),
(__v4di) __O);
}
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_maskz_broadcastq_epi64 (__mmask8 __M, __m128i __A)
-{
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_maskz_broadcastq_epi64(__mmask8 __M, __m128i __A) {
return (__m256i)__builtin_ia32_selectq_256(__M,
(__v4di) _mm256_broadcastq_epi64(__A),
(__v4di) _mm256_setzero_si256());
@@ -8003,65 +7942,57 @@ _mm256_maskz_permutexvar_epi32(__mmask8 __M, __m256i __X, __m256i __Y)
(__v4di)_mm256_alignr_epi64((A), (B), (imm)), \
(__v4di)_mm256_setzero_si256()))
-static __inline__ __m128 __DEFAULT_FN_ATTRS128
-_mm_mask_movehdup_ps (__m128 __W, __mmask8 __U, __m128 __A)
-{
+static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_mask_movehdup_ps(__m128 __W, __mmask8 __U, __m128 __A) {
return (__m128)__builtin_ia32_selectps_128((__mmask8)__U,
(__v4sf)_mm_movehdup_ps(__A),
(__v4sf)__W);
}
-static __inline__ __m128 __DEFAULT_FN_ATTRS128
-_mm_maskz_movehdup_ps (__mmask8 __U, __m128 __A)
-{
+static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_maskz_movehdup_ps(__mmask8 __U, __m128 __A) {
return (__m128)__builtin_ia32_selectps_128((__mmask8)__U,
(__v4sf)_mm_movehdup_ps(__A),
(__v4sf)_mm_setzero_ps());
}
-static __inline__ __m256 __DEFAULT_FN_ATTRS256
-_mm256_mask_movehdup_ps (__m256 __W, __mmask8 __U, __m256 __A)
-{
+static __inline__ __m256 __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_mask_movehdup_ps(__m256 __W, __mmask8 __U, __m256 __A) {
return (__m256)__builtin_ia32_selectps_256((__mmask8)__U,
(__v8sf)_mm256_movehdup_ps(__A),
(__v8sf)__W);
}
-static __inline__ __m256 __DEFAULT_FN_ATTRS256
-_mm256_maskz_movehdup_ps (__mmask8 __U, __m256 __A)
-{
+static __inline__ __m256 __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_maskz_movehdup_ps(__mmask8 __U, __m256 __A) {
return (__m256)__builtin_ia32_selectps_256((__mmask8)__U,
(__v8sf)_mm256_movehdup_ps(__A),
(__v8sf)_mm256_setzero_ps());
}
-static __inline__ __m128 __DEFAULT_FN_ATTRS128
-_mm_mask_moveldup_ps (__m128 __W, __mmask8 __U, __m128 __A)
-{
+static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_mask_moveldup_ps(__m128 __W, __mmask8 __U, __m128 __A) {
return (__m128)__builtin_ia32_selectps_128((__mmask8)__U,
(__v4sf)_mm_moveldup_ps(__A),
(__v4sf)__W);
}
-static __inline__ __m128 __DEFAULT_FN_ATTRS128
-_mm_maskz_moveldup_ps (__mmask8 __U, __m128 __A)
-{
+static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_maskz_moveldup_ps(__mmask8 __U, __m128 __A) {
return (__m128)__builtin_ia32_selectps_128((__mmask8)__U,
(__v4sf)_mm_moveldup_ps(__A),
(__v4sf)_mm_setzero_ps());
}
-static __inline__ __m256 __DEFAULT_FN_ATTRS256
-_mm256_mask_moveldup_ps (__m256 __W, __mmask8 __U, __m256 __A)
-{
+static __inline__ __m256 __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_mask_moveldup_ps(__m256 __W, __mmask8 __U, __m256 __A) {
return (__m256)__builtin_ia32_selectps_256((__mmask8)__U,
(__v8sf)_mm256_moveldup_ps(__A),
(__v8sf)__W);
}
-static __inline__ __m256 __DEFAULT_FN_ATTRS256
-_mm256_maskz_moveldup_ps (__mmask8 __U, __m256 __A)
-{
+static __inline__ __m256 __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_maskz_moveldup_ps(__mmask8 __U, __m256 __A) {
return (__m256)__builtin_ia32_selectps_256((__mmask8)__U,
(__v8sf)_mm256_moveldup_ps(__A),
(__v8sf)_mm256_setzero_ps());
diff --git a/clang/lib/Headers/tmmintrin.h b/clang/lib/Headers/tmmintrin.h
index 9d007c8..5d0f20f 100644
--- a/clang/lib/Headers/tmmintrin.h
+++ b/clang/lib/Headers/tmmintrin.h
@@ -544,8 +544,8 @@ _mm_maddubs_pi16(__m64 __a, __m64 __b) {
/// A 128-bit vector of [8 x i16] containing one of the source operands.
/// \returns A 128-bit vector of [8 x i16] containing the rounded and scaled
/// products of both operands.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_mulhrs_epi16(__m128i __a,
- __m128i __b) {
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_mulhrs_epi16(__m128i __a, __m128i __b) {
return (__m128i)__builtin_ia32_pmulhrsw128((__v8hi)__a, (__v8hi)__b);
}
@@ -563,11 +563,10 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_mulhrs_epi16(__m128i __a,
/// A 64-bit vector of [4 x i16] containing one of the source operands.
/// \returns A 64-bit vector of [4 x i16] containing the rounded and scaled
/// products of both operands.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
-_mm_mulhrs_pi16(__m64 __a, __m64 __b)
-{
- return __trunc64(__builtin_ia32_pmulhrsw128((__v8hi)__anyext128(__a),
- (__v8hi)__anyext128(__b)));
+static __inline__ __m64 __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_mulhrs_pi16(__m64 __a, __m64 __b) {
+ return __trunc64(__builtin_ia32_pmulhrsw128((__v8hi)__zext128(__a),
+ (__v8hi)__zext128(__b)));
}
/// Copies the 8-bit integers from a 128-bit integer vector to the
@@ -590,10 +589,9 @@ _mm_mulhrs_pi16(__m64 __a, __m64 __b)
/// Bits [6:4] Reserved. \n
/// Bits [3:0] select the source byte to be copied.
/// \returns A 128-bit integer vector containing the copied or cleared values.
-static __inline__ __m128i __DEFAULT_FN_ATTRS
-_mm_shuffle_epi8(__m128i __a, __m128i __b)
-{
- return (__m128i)__builtin_ia32_pshufb128((__v16qi)__a, (__v16qi)__b);
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_shuffle_epi8(__m128i __a, __m128i __b) {
+ return (__m128i)__builtin_ia32_pshufb128((__v16qi)__a, (__v16qi)__b);
}
/// Copies the 8-bit integers from a 64-bit integer vector to the
@@ -615,13 +613,12 @@ _mm_shuffle_epi8(__m128i __a, __m128i __b)
/// destination. \n
/// Bits [2:0] select the source byte to be copied.
/// \returns A 64-bit integer vector containing the copied or cleared values.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
-_mm_shuffle_pi8(__m64 __a, __m64 __b)
-{
- return __trunc64(__builtin_ia32_pshufb128(
- (__v16qi)__builtin_shufflevector(
- (__v2si)(__a), __extension__ (__v2si){}, 0, 1, 0, 1),
- (__v16qi)__anyext128(__b)));
+static __inline__ __m64 __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_shuffle_pi8(__m64 __a, __m64 __b) {
+ return __trunc64(__builtin_ia32_pshufb128(
+ (__v16qi)__builtin_shufflevector((__v2si)(__a), __extension__(__v2si){},
+ 0, 1, 0, 1),
+ (__v16qi)__zext128(__b)));
}
/// For each 8-bit integer in the first source operand, perform one of
generated by cgit v1.1 at 2025-10-18 22:27:19 +0000