@zigc/lib 0.17.0-dev.27 → 0.17.0-dev.296

package/include/f16cintrin.h

@@ -15,10 +15,21 @@
 #define __F16CINTRIN_H
 
 /* Define the default attributes for the functions in this file. */
-#define __DEFAULT_FN_ATTRS128 \
-  __attribute__((__always_inline__, __nodebug__, __target__("f16c"), __min_vector_width__(128)))
-#define __DEFAULT_FN_ATTRS256 \
-  __attribute__((__always_inline__, __nodebug__, __target__("f16c"), __min_vector_width__(256)))
+#if defined(__cplusplus) && (__cplusplus >= 201103L)
+#define __DEFAULT_FN_ATTRS128                                                  \
+  __attribute__((__always_inline__, __nodebug__, __target__("f16c"),           \
+                 __min_vector_width__(128))) constexpr
+#define __DEFAULT_FN_ATTRS256                                                  \
+  __attribute__((__always_inline__, __nodebug__, __target__("f16c"),           \
+                 __min_vector_width__(256))) constexpr
+#else
+#define __DEFAULT_FN_ATTRS128                                                  \
+  __attribute__((__always_inline__, __nodebug__, __target__("f16c"),           \
+                 __min_vector_width__(128)))
+#define __DEFAULT_FN_ATTRS256                                                  \
+  __attribute__((__always_inline__, __nodebug__, __target__("f16c"),           \
+                 __min_vector_width__(256)))
+#endif
 
 /* NOTE: Intel documents the 128-bit versions of these as being in emmintrin.h,
  * but that's because icc can emulate these without f16c using a library call.
@@ -38,9 +49,7 @@
 static __inline float __DEFAULT_FN_ATTRS128
 _cvtsh_ss(unsigned short __a)
 {
-  __v8hi __v = {(short)__a, 0, 0, 0, 0, 0, 0, 0};
-  __v4sf __r = __builtin_ia32_vcvtph2ps(__v);
-  return __r[0];
+  return (float)__builtin_bit_cast(__fp16, __a);
 }
 
 /// Converts a 32-bit single-precision float value to a 16-bit
@@ -109,7 +118,10 @@ _cvtsh_ss(unsigned short __a)
 static __inline __m128 __DEFAULT_FN_ATTRS128
 _mm_cvtph_ps(__m128i __a)
 {
-  return (__m128)__builtin_ia32_vcvtph2ps((__v8hi)__a);
+  typedef __fp16 __v4fp16 __attribute__((__vector_size__(8)));
+
+  __v4hi __v = __builtin_shufflevector((__v8hi)__a, (__v8hi)__a, 0, 1, 2, 3);
+  return (__m128) __builtin_convertvector((__v4fp16)__v, __v4sf);
 }
 
 /// Converts a 256-bit vector of [8 x float] into a 128-bit vector
@@ -153,7 +165,9 @@ _mm_cvtph_ps(__m128i __a)
 static __inline __m256 __DEFAULT_FN_ATTRS256
 _mm256_cvtph_ps(__m128i __a)
 {
-  return (__m256)__builtin_ia32_vcvtph2ps256((__v8hi)__a);
+  typedef __fp16 __v8fp16 __attribute__((__vector_size__(16), __aligned__(16)));
+
+  return (__m256) __builtin_convertvector((__v8fp16)__a, __v8sf);
 }
 
 #undef __DEFAULT_FN_ATTRS128

package/include/float.h

@@ -7,13 +7,21 @@
  *===-----------------------------------------------------------------------===
  */
 
-#ifndef __CLANG_FLOAT_H
-#define __CLANG_FLOAT_H
-
 #if defined(__MVS__) && __has_include_next(<float.h>)
+#include <__float_header_macro.h>
 #include_next <float.h>
 #else
 
+#if !defined(__need_infinity_nan)
+#define __need_float_float
+#if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 202311L) ||              \
+    !defined(__STRICT_ANSI__)
+#define __need_infinity_nan
+#endif
+#include <__float_header_macro.h>
+#endif
+
+#ifdef __need_float_float
 /* If we're on MinGW, fall back to the system's float.h, which might have
  * additional definitions provided for Windows.
  * For more details see http://msdn.microsoft.com/en-us/library/y0ybw9fy.aspx
@@ -26,162 +34,15 @@
 
 #  include_next <float.h>
 
-/* Undefine anything that we'll be redefining below. */
-#  undef FLT_EVAL_METHOD
-#  undef FLT_ROUNDS
-#  undef FLT_RADIX
-#  undef FLT_MANT_DIG
-#  undef DBL_MANT_DIG
-#  undef LDBL_MANT_DIG
-#if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L) ||              \
-    !defined(__STRICT_ANSI__) ||                                               \
-    (defined(__cplusplus) && __cplusplus >= 201103L) ||                        \
-    (__STDC_HOSTED__ && defined(_AIX) && defined(_ALL_SOURCE))
-#    undef DECIMAL_DIG
-#  endif
-#  undef FLT_DIG
-#  undef DBL_DIG
-#  undef LDBL_DIG
-#  undef FLT_MIN_EXP
-#  undef DBL_MIN_EXP
-#  undef LDBL_MIN_EXP
-#  undef FLT_MIN_10_EXP
-#  undef DBL_MIN_10_EXP
-#  undef LDBL_MIN_10_EXP
-#  undef FLT_MAX_EXP
-#  undef DBL_MAX_EXP
-#  undef LDBL_MAX_EXP
-#  undef FLT_MAX_10_EXP
-#  undef DBL_MAX_10_EXP
-#  undef LDBL_MAX_10_EXP
-#  undef FLT_MAX
-#  undef DBL_MAX
-#  undef LDBL_MAX
-#  undef FLT_EPSILON
-#  undef DBL_EPSILON
-#  undef LDBL_EPSILON
-#  undef FLT_MIN
-#  undef DBL_MIN
-#  undef LDBL_MIN
-#if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 201112L) ||              \
-    !defined(__STRICT_ANSI__) ||                                               \
-    (defined(__cplusplus) && __cplusplus >= 201703L) ||                        \
-    (__STDC_HOSTED__ && defined(_AIX) && defined(_ALL_SOURCE))
-#    undef FLT_TRUE_MIN
-#    undef DBL_TRUE_MIN
-#    undef LDBL_TRUE_MIN
-#    undef FLT_DECIMAL_DIG
-#    undef DBL_DECIMAL_DIG
-#    undef LDBL_DECIMAL_DIG
-#    undef FLT_HAS_SUBNORM
-#    undef DBL_HAS_SUBNORM
-#    undef LDBL_HAS_SUBNORM
-#  endif
-#if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 202311L) ||              \
-    !defined(__STRICT_ANSI__)
-#    undef FLT_NORM_MAX
-#    undef DBL_NORM_MAX
-#    undef LDBL_NORM_MAX
-#endif
-#endif
-
-#if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 202311L) ||              \
-    !defined(__STRICT_ANSI__)
-#  undef INFINITY
-#  undef NAN
-#endif
-
-/* Characteristics of floating point types, C99 5.2.4.2.2 */
-
-#if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L) ||              \
-    (defined(__cplusplus) && __cplusplus >= 201103L)
-#define FLT_EVAL_METHOD __FLT_EVAL_METHOD__
 #endif
-#define FLT_ROUNDS (__builtin_flt_rounds())
-#define FLT_RADIX __FLT_RADIX__
-
-#define FLT_MANT_DIG __FLT_MANT_DIG__
-#define DBL_MANT_DIG __DBL_MANT_DIG__
-#define LDBL_MANT_DIG __LDBL_MANT_DIG__
 
-#if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L) ||              \
-    !defined(__STRICT_ANSI__) ||                                               \
-    (defined(__cplusplus) && __cplusplus >= 201103L) ||                        \
-    (__STDC_HOSTED__ && defined(_AIX) && defined(_ALL_SOURCE))
-#  define DECIMAL_DIG __DECIMAL_DIG__
+#include <__float_float.h>
+#undef __need_float_float
 #endif
 
-#define FLT_DIG __FLT_DIG__
-#define DBL_DIG __DBL_DIG__
-#define LDBL_DIG __LDBL_DIG__
-
-#define FLT_MIN_EXP __FLT_MIN_EXP__
-#define DBL_MIN_EXP __DBL_MIN_EXP__
-#define LDBL_MIN_EXP __LDBL_MIN_EXP__
-
-#define FLT_MIN_10_EXP __FLT_MIN_10_EXP__
-#define DBL_MIN_10_EXP __DBL_MIN_10_EXP__
-#define LDBL_MIN_10_EXP __LDBL_MIN_10_EXP__
-
-#define FLT_MAX_EXP __FLT_MAX_EXP__
-#define DBL_MAX_EXP __DBL_MAX_EXP__
-#define LDBL_MAX_EXP __LDBL_MAX_EXP__
-
-#define FLT_MAX_10_EXP __FLT_MAX_10_EXP__
-#define DBL_MAX_10_EXP __DBL_MAX_10_EXP__
-#define LDBL_MAX_10_EXP __LDBL_MAX_10_EXP__
-
-#define FLT_MAX __FLT_MAX__
-#define DBL_MAX __DBL_MAX__
-#define LDBL_MAX __LDBL_MAX__
-
-#define FLT_EPSILON __FLT_EPSILON__
-#define DBL_EPSILON __DBL_EPSILON__
-#define LDBL_EPSILON __LDBL_EPSILON__
-
-#define FLT_MIN __FLT_MIN__
-#define DBL_MIN __DBL_MIN__
-#define LDBL_MIN __LDBL_MIN__
-
-#if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 201112L) ||              \
-    !defined(__STRICT_ANSI__) ||                                               \
-    (defined(__cplusplus) && __cplusplus >= 201703L) ||                        \
-    (__STDC_HOSTED__ && defined(_AIX) && defined(_ALL_SOURCE))
-#  define FLT_TRUE_MIN __FLT_DENORM_MIN__
-#  define DBL_TRUE_MIN __DBL_DENORM_MIN__
-#  define LDBL_TRUE_MIN __LDBL_DENORM_MIN__
-#  define FLT_DECIMAL_DIG __FLT_DECIMAL_DIG__
-#  define DBL_DECIMAL_DIG __DBL_DECIMAL_DIG__
-#  define LDBL_DECIMAL_DIG __LDBL_DECIMAL_DIG__
-#  define FLT_HAS_SUBNORM __FLT_HAS_DENORM__
-#  define DBL_HAS_SUBNORM __DBL_HAS_DENORM__
-#  define LDBL_HAS_SUBNORM __LDBL_HAS_DENORM__
+#ifdef __need_infinity_nan
+#include <__float_infinity_nan.h>
+#undef __need_infinity_nan
 #endif
 
-#if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 202311L) ||              \
-    !defined(__STRICT_ANSI__)
-   /* C23 5.2.5.3.3p29-30 */
-#  define INFINITY (__builtin_inff())
-#  define NAN (__builtin_nanf(""))
-   /* C23 5.2.5.3.3p32 */
-#  define FLT_NORM_MAX __FLT_NORM_MAX__
-#  define DBL_NORM_MAX __DBL_NORM_MAX__
-#  define LDBL_NORM_MAX __LDBL_NORM_MAX__
-#endif
-
-#ifdef __STDC_WANT_IEC_60559_TYPES_EXT__
-#  define FLT16_MANT_DIG    __FLT16_MANT_DIG__
-#  define FLT16_DECIMAL_DIG __FLT16_DECIMAL_DIG__
-#  define FLT16_DIG         __FLT16_DIG__
-#  define FLT16_MIN_EXP     __FLT16_MIN_EXP__
-#  define FLT16_MIN_10_EXP  __FLT16_MIN_10_EXP__
-#  define FLT16_MAX_EXP     __FLT16_MAX_EXP__
-#  define FLT16_MAX_10_EXP  __FLT16_MAX_10_EXP__
-#  define FLT16_MAX         __FLT16_MAX__
-#  define FLT16_EPSILON     __FLT16_EPSILON__
-#  define FLT16_MIN         __FLT16_MIN__
-#  define FLT16_TRUE_MIN    __FLT16_TRUE_MIN__
-#endif /* __STDC_WANT_IEC_60559_TYPES_EXT__ */
-
 #endif /* __MVS__ */
-#endif /* __CLANG_FLOAT_H */

package/include/fma4intrin.h

@@ -20,100 +20,100 @@
 #define __DEFAULT_FN_ATTRS128 __attribute__((__always_inline__, __nodebug__, __target__("fma4"), __min_vector_width__(128)))
 #define __DEFAULT_FN_ATTRS256 __attribute__((__always_inline__, __nodebug__, __target__("fma4"), __min_vector_width__(256)))
 
-static __inline__ __m128 __DEFAULT_FN_ATTRS128
-_mm_macc_ps(__m128 __A, __m128 __B, __m128 __C)
-{
-  return (__m128)__builtin_ia32_vfmaddps((__v4sf)__A, (__v4sf)__B, (__v4sf)__C);
+#if defined(__cplusplus) && (__cplusplus >= 201103L)
+#define __DEFAULT_FN_ATTRS128_CONSTEXPR __DEFAULT_FN_ATTRS128 constexpr
+#define __DEFAULT_FN_ATTRS256_CONSTEXPR __DEFAULT_FN_ATTRS256 constexpr
+#else
+#define __DEFAULT_FN_ATTRS128_CONSTEXPR __DEFAULT_FN_ATTRS128
+#define __DEFAULT_FN_ATTRS256_CONSTEXPR __DEFAULT_FN_ATTRS256
+#endif
+
+static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_macc_ps(__m128 __A, __m128 __B, __m128 __C) {
+  return (__m128)__builtin_elementwise_fma((__v4sf)__A, (__v4sf)__B,
+                                           (__v4sf)__C);
 }
 
-static __inline__ __m128d __DEFAULT_FN_ATTRS128
-_mm_macc_pd(__m128d __A, __m128d __B, __m128d __C)
-{
-  return (__m128d)__builtin_ia32_vfmaddpd((__v2df)__A, (__v2df)__B, (__v2df)__C);
+static __inline__ __m128d __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_macc_pd(__m128d __A, __m128d __B, __m128d __C) {
+  return (__m128d)__builtin_elementwise_fma((__v2df)__A, (__v2df)__B,
+                                            (__v2df)__C);
 }
 
-static __inline__ __m128 __DEFAULT_FN_ATTRS128
-_mm_macc_ss(__m128 __A, __m128 __B, __m128 __C)
-{
-  return (__m128)__builtin_ia32_vfmaddss((__v4sf)__A, (__v4sf)__B, (__v4sf)__C);
+static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_macc_ss(__m128 __A, __m128 __B, __m128 __C) {
+  return _mm_set_ss(__builtin_elementwise_fma(__A[0], __B[0], __C[0]));
 }
 
-static __inline__ __m128d __DEFAULT_FN_ATTRS128
-_mm_macc_sd(__m128d __A, __m128d __B, __m128d __C)
-{
-  return (__m128d)__builtin_ia32_vfmaddsd((__v2df)__A, (__v2df)__B, (__v2df)__C);
+static __inline__ __m128d __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_macc_sd(__m128d __A, __m128d __B, __m128d __C) {
+  return _mm_set_sd(__builtin_elementwise_fma(__A[0], __B[0], __C[0]));
 }
 
-static __inline__ __m128 __DEFAULT_FN_ATTRS128
-_mm_msub_ps(__m128 __A, __m128 __B, __m128 __C)
-{
-  return (__m128)__builtin_ia32_vfmaddps((__v4sf)__A, (__v4sf)__B, -(__v4sf)__C);
+static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_msub_ps(__m128 __A, __m128 __B, __m128 __C) {
+  return (__m128)__builtin_elementwise_fma((__v4sf)__A, (__v4sf)__B,
+                                           -(__v4sf)__C);
 }
 
-static __inline__ __m128d __DEFAULT_FN_ATTRS128
-_mm_msub_pd(__m128d __A, __m128d __B, __m128d __C)
-{
-  return (__m128d)__builtin_ia32_vfmaddpd((__v2df)__A, (__v2df)__B, -(__v2df)__C);
+static __inline__ __m128d __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_msub_pd(__m128d __A, __m128d __B, __m128d __C) {
+  return (__m128d)__builtin_elementwise_fma((__v2df)__A, (__v2df)__B,
+                                            -(__v2df)__C);
 }
 
-static __inline__ __m128 __DEFAULT_FN_ATTRS128
-_mm_msub_ss(__m128 __A, __m128 __B, __m128 __C)
-{
-  return (__m128)__builtin_ia32_vfmaddss((__v4sf)__A, (__v4sf)__B, -(__v4sf)__C);
+static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_msub_ss(__m128 __A, __m128 __B, __m128 __C) {
+  return _mm_set_ss(__builtin_elementwise_fma(__A[0], __B[0], -__C[0]));
 }
 
-static __inline__ __m128d __DEFAULT_FN_ATTRS128
-_mm_msub_sd(__m128d __A, __m128d __B, __m128d __C)
-{
-  return (__m128d)__builtin_ia32_vfmaddsd((__v2df)__A, (__v2df)__B, -(__v2df)__C);
+static __inline__ __m128d __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_msub_sd(__m128d __A, __m128d __B, __m128d __C) {
+  return _mm_set_sd(__builtin_elementwise_fma(__A[0], __B[0], -__C[0]));
 }
 
-static __inline__ __m128 __DEFAULT_FN_ATTRS128
-_mm_nmacc_ps(__m128 __A, __m128 __B, __m128 __C)
-{
-  return (__m128)__builtin_ia32_vfmaddps(-(__v4sf)__A, (__v4sf)__B, (__v4sf)__C);
+static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_nmacc_ps(__m128 __A, __m128 __B, __m128 __C) {
+  return (__m128)__builtin_elementwise_fma(-(__v4sf)__A, (__v4sf)__B,
+                                           (__v4sf)__C);
 }
 
-static __inline__ __m128d __DEFAULT_FN_ATTRS128
-_mm_nmacc_pd(__m128d __A, __m128d __B, __m128d __C)
-{
-  return (__m128d)__builtin_ia32_vfmaddpd(-(__v2df)__A, (__v2df)__B, (__v2df)__C);
+static __inline__ __m128d __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_nmacc_pd(__m128d __A, __m128d __B, __m128d __C) {
+  return (__m128d)__builtin_elementwise_fma(-(__v2df)__A, (__v2df)__B,
+                                            (__v2df)__C);
 }
 
-static __inline__ __m128 __DEFAULT_FN_ATTRS128
-_mm_nmacc_ss(__m128 __A, __m128 __B, __m128 __C)
-{
-  return (__m128)__builtin_ia32_vfmaddss(-(__v4sf)__A, (__v4sf)__B, (__v4sf)__C);
+static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_nmacc_ss(__m128 __A, __m128 __B, __m128 __C) {
+  return _mm_set_ss(__builtin_elementwise_fma(-__A[0], __B[0], __C[0]));
 }
 
-static __inline__ __m128d __DEFAULT_FN_ATTRS128
-_mm_nmacc_sd(__m128d __A, __m128d __B, __m128d __C)
-{
-  return (__m128d)__builtin_ia32_vfmaddsd(-(__v2df)__A, (__v2df)__B, (__v2df)__C);
+static __inline__ __m128d __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_nmacc_sd(__m128d __A, __m128d __B, __m128d __C) {
+  return _mm_set_sd(__builtin_elementwise_fma(-__A[0], __B[0], __C[0]));
 }
 
-static __inline__ __m128 __DEFAULT_FN_ATTRS128
-_mm_nmsub_ps(__m128 __A, __m128 __B, __m128 __C)
-{
-  return (__m128)__builtin_ia32_vfmaddps(-(__v4sf)__A, (__v4sf)__B, -(__v4sf)__C);
+static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_nmsub_ps(__m128 __A, __m128 __B, __m128 __C) {
+  return (__m128)__builtin_elementwise_fma(-(__v4sf)__A, (__v4sf)__B,
+                                           -(__v4sf)__C);
 }
 
-static __inline__ __m128d __DEFAULT_FN_ATTRS128
-_mm_nmsub_pd(__m128d __A, __m128d __B, __m128d __C)
-{
-  return (__m128d)__builtin_ia32_vfmaddpd(-(__v2df)__A, (__v2df)__B, -(__v2df)__C);
+static __inline__ __m128d __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_nmsub_pd(__m128d __A, __m128d __B, __m128d __C) {
+  return (__m128d)__builtin_elementwise_fma(-(__v2df)__A, (__v2df)__B,
+                                            -(__v2df)__C);
 }
 
-static __inline__ __m128 __DEFAULT_FN_ATTRS128
-_mm_nmsub_ss(__m128 __A, __m128 __B, __m128 __C)
-{
-  return (__m128)__builtin_ia32_vfmaddss(-(__v4sf)__A, (__v4sf)__B, -(__v4sf)__C);
+static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_nmsub_ss(__m128 __A, __m128 __B, __m128 __C) {
+  return _mm_set_ss(__builtin_elementwise_fma(-__A[0], __B[0], -__C[0]));
 }
 
-static __inline__ __m128d __DEFAULT_FN_ATTRS128
-_mm_nmsub_sd(__m128d __A, __m128d __B, __m128d __C)
-{
-  return (__m128d)__builtin_ia32_vfmaddsd(-(__v2df)__A, (__v2df)__B, -(__v2df)__C);
+static __inline__ __m128d __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_nmsub_sd(__m128d __A, __m128d __B, __m128d __C) {
+  return _mm_set_sd(__builtin_elementwise_fma(-__A[0], __B[0], -__C[0]));
 }
 
 static __inline__ __m128 __DEFAULT_FN_ATTRS128
@@ -140,52 +140,52 @@ _mm_msubadd_pd(__m128d __A, __m128d __B, __m128d __C)
   return (__m128d)__builtin_ia32_vfmaddsubpd((__v2df)__A, (__v2df)__B, -(__v2df)__C);
 }
 
-static __inline__ __m256 __DEFAULT_FN_ATTRS256
-_mm256_macc_ps(__m256 __A, __m256 __B, __m256 __C)
-{
-  return (__m256)__builtin_ia32_vfmaddps256((__v8sf)__A, (__v8sf)__B, (__v8sf)__C);
+static __inline__ __m256 __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_macc_ps(__m256 __A, __m256 __B, __m256 __C) {
+  return (__m256)__builtin_elementwise_fma((__v8sf)__A, (__v8sf)__B,
+                                           (__v8sf)__C);
 }
 
-static __inline__ __m256d __DEFAULT_FN_ATTRS256
-_mm256_macc_pd(__m256d __A, __m256d __B, __m256d __C)
-{
-  return (__m256d)__builtin_ia32_vfmaddpd256((__v4df)__A, (__v4df)__B, (__v4df)__C);
+static __inline__ __m256d __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_macc_pd(__m256d __A, __m256d __B, __m256d __C) {
+  return (__m256d)__builtin_elementwise_fma((__v4df)__A, (__v4df)__B,
+                                            (__v4df)__C);
 }
 
-static __inline__ __m256 __DEFAULT_FN_ATTRS256
-_mm256_msub_ps(__m256 __A, __m256 __B, __m256 __C)
-{
-  return (__m256)__builtin_ia32_vfmaddps256((__v8sf)__A, (__v8sf)__B, -(__v8sf)__C);
+static __inline__ __m256 __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_msub_ps(__m256 __A, __m256 __B, __m256 __C) {
+  return (__m256)__builtin_elementwise_fma((__v8sf)__A, (__v8sf)__B,
+                                           -(__v8sf)__C);
 }
 
-static __inline__ __m256d __DEFAULT_FN_ATTRS256
-_mm256_msub_pd(__m256d __A, __m256d __B, __m256d __C)
-{
-  return (__m256d)__builtin_ia32_vfmaddpd256((__v4df)__A, (__v4df)__B, -(__v4df)__C);
+static __inline__ __m256d __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_msub_pd(__m256d __A, __m256d __B, __m256d __C) {
+  return (__m256d)__builtin_elementwise_fma((__v4df)__A, (__v4df)__B,
+                                            -(__v4df)__C);
 }
 
-static __inline__ __m256 __DEFAULT_FN_ATTRS256
-_mm256_nmacc_ps(__m256 __A, __m256 __B, __m256 __C)
-{
-  return (__m256)__builtin_ia32_vfmaddps256(-(__v8sf)__A, (__v8sf)__B, (__v8sf)__C);
+static __inline__ __m256 __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_nmacc_ps(__m256 __A, __m256 __B, __m256 __C) {
+  return (__m256)__builtin_elementwise_fma(-(__v8sf)__A, (__v8sf)__B,
+                                           (__v8sf)__C);
 }
 
-static __inline__ __m256d __DEFAULT_FN_ATTRS256
-_mm256_nmacc_pd(__m256d __A, __m256d __B, __m256d __C)
-{
-  return (__m256d)__builtin_ia32_vfmaddpd256(-(__v4df)__A, (__v4df)__B, (__v4df)__C);
+static __inline__ __m256d __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_nmacc_pd(__m256d __A, __m256d __B, __m256d __C) {
+  return (__m256d)__builtin_elementwise_fma(-(__v4df)__A, (__v4df)__B,
+                                            (__v4df)__C);
 }
 
-static __inline__ __m256 __DEFAULT_FN_ATTRS256
-_mm256_nmsub_ps(__m256 __A, __m256 __B, __m256 __C)
-{
-  return (__m256)__builtin_ia32_vfmaddps256(-(__v8sf)__A, (__v8sf)__B, -(__v8sf)__C);
+static __inline__ __m256 __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_nmsub_ps(__m256 __A, __m256 __B, __m256 __C) {
+  return (__m256)__builtin_elementwise_fma(-(__v8sf)__A, (__v8sf)__B,
+                                           -(__v8sf)__C);
 }
 
-static __inline__ __m256d __DEFAULT_FN_ATTRS256
-_mm256_nmsub_pd(__m256d __A, __m256d __B, __m256d __C)
-{
-  return (__m256d)__builtin_ia32_vfmaddpd256(-(__v4df)__A, (__v4df)__B, -(__v4df)__C);
+static __inline__ __m256d __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_nmsub_pd(__m256d __A, __m256d __B, __m256d __C) {
+  return (__m256d)__builtin_elementwise_fma(-(__v4df)__A, (__v4df)__B,
+                                            -(__v4df)__C);
 }
 
 static __inline__ __m256 __DEFAULT_FN_ATTRS256
@@ -214,5 +214,7 @@ _mm256_msubadd_pd(__m256d __A, __m256d __B, __m256d __C)
 
 #undef __DEFAULT_FN_ATTRS128
 #undef __DEFAULT_FN_ATTRS256
+#undef __DEFAULT_FN_ATTRS128_CONSTEXPR
+#undef __DEFAULT_FN_ATTRS256_CONSTEXPR
 
 #endif /* __FMA4INTRIN_H */