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

package/include/avxvnniint8intrin.h

@@ -14,6 +14,7 @@
 #ifndef __AVXVNNIINT8INTRIN_H
 #define __AVXVNNIINT8INTRIN_H
 
+// clang-format off
 /// Multiply groups of 4 adjacent pairs of signed 8-bit integers in \a __A with
 ///    corresponding signed 8-bit integers in \a __B, producing 4 intermediate
 ///    signed 16-bit results. Sum these 4 results with the corresponding
@@ -44,10 +45,12 @@
 /// ENDFOR
 /// dst[MAX:128] := 0
 /// \endcode
+// clang-format on
 #define _mm_dpbssd_epi32(__W, __A, __B)                                        \
-  ((__m128i)__builtin_ia32_vpdpbssd128((__v4si)(__W), (__v4si)(__A),           \
-                                       (__v4si)(__B)))
+  ((__m128i)__builtin_ia32_vpdpbssd128((__v4si)(__W), (__v16qi)(__A),          \
+                                       (__v16qi)(__B)))
 
+// clang-format off
 /// Multiply groups of 4 adjacent pairs of signed 8-bit integers in \a __A with
 ///    corresponding signed 8-bit integers in \a __B, producing 4 intermediate
 ///    signed 16-bit results. Sum these 4 results with the corresponding
@@ -78,10 +81,12 @@
 /// ENDFOR
 /// dst[MAX:256] := 0
 /// \endcode
+// clang-format on
 #define _mm256_dpbssd_epi32(__W, __A, __B)                                     \
-  ((__m256i)__builtin_ia32_vpdpbssd256((__v8si)(__W), (__v8si)(__A),           \
-                                       (__v8si)(__B)))
+  ((__m256i)__builtin_ia32_vpdpbssd256((__v8si)(__W), (__v32qi)(__A),          \
+                                       (__v32qi)(__B)))
 
+// clang-format off
 /// Multiply groups of 4 adjacent pairs of signed 8-bit integers in \a __A with
 ///    corresponding signed 8-bit integers in \a __B, producing 4 intermediate
 ///    signed 16-bit results. Sum these 4 results with the corresponding
@@ -94,7 +99,7 @@
 /// _mm_dpbssds_epi32( __m128i __W, __m128i __A, __m128i __B);
 /// \endcode
 ///
-/// This intrinsic corresponds to the \c VPDPBSSD instruction.
+/// This intrinsic corresponds to the \c VPDPBSSDS instruction.
 ///
 /// \param __A
 ///    A 128-bit vector of [16 x char].
@@ -113,10 +118,12 @@
 /// ENDFOR
 /// dst[MAX:128] := 0
 /// \endcode
+// clang-format on
 #define _mm_dpbssds_epi32(__W, __A, __B)                                       \
-  ((__m128i)__builtin_ia32_vpdpbssds128((__v4si)(__W), (__v4si)(__A),          \
-                                        (__v4si)(__B)))
+  ((__m128i)__builtin_ia32_vpdpbssds128((__v4si)(__W), (__v16qi)(__A),         \
+                                        (__v16qi)(__B)))
 
+// clang-format off
 /// Multiply groups of 4 adjacent pairs of signed 8-bit integers in \a __A with
 ///    corresponding signed 8-bit integers in \a __B, producing 4 intermediate
 ///    signed 16-bit results. Sum these 4 results with the corresponding
@@ -129,7 +136,7 @@
 /// _mm256_dpbssds_epi32(__m256i __W, __m256i __A, __m256i __B);
 /// \endcode
 ///
-/// This intrinsic corresponds to the \c VPDPBSSD instruction.
+/// This intrinsic corresponds to the \c VPDPBSSDS instruction.
 ///
 /// \param __A
 ///    A 256-bit vector of [32 x char].
@@ -148,10 +155,12 @@
 /// ENDFOR
 /// dst[MAX:256] := 0
 /// \endcode
+// clang-format on
 #define _mm256_dpbssds_epi32(__W, __A, __B)                                    \
-  ((__m256i)__builtin_ia32_vpdpbssds256((__v8si)(__W), (__v8si)(__A),          \
-                                        (__v8si)(__B)))
+  ((__m256i)__builtin_ia32_vpdpbssds256((__v8si)(__W), (__v32qi)(__A),         \
+                                        (__v32qi)(__B)))
 
+// clang-format off
 /// Multiply groups of 4 adjacent pairs of signed 8-bit integers in \a __A with
 ///    corresponding unsigned 8-bit integers in \a __B, producing 4 intermediate
 ///    signed 16-bit results. Sum these 4 results with the corresponding
@@ -163,7 +172,7 @@
 /// _mm_dpbsud_epi32(__m128i __W, __m128i __A, __m128i __B);
 /// \endcode
 ///
-/// This intrinsic corresponds to the \c VPDPBSSD instruction.
+/// This intrinsic corresponds to the \c VPDPBSUD instruction.
 ///
 /// \param __A
 ///    A 128-bit vector of [16 x char].
@@ -182,10 +191,12 @@
 /// ENDFOR
 /// dst[MAX:128] := 0
 /// \endcode
+// clang-format on
 #define _mm_dpbsud_epi32(__W, __A, __B)                                        \
-  ((__m128i)__builtin_ia32_vpdpbsud128((__v4si)(__W), (__v4si)(__A),           \
-                                       (__v4si)(__B)))
+  ((__m128i)__builtin_ia32_vpdpbsud128((__v4si)(__W), (__v16qi)(__A),          \
+                                       (__v16qu)(__B)))
 
+// clang-format off
 /// Multiply groups of 4 adjacent pairs of signed 8-bit integers in \a __A with
 ///    corresponding unsigned 8-bit integers in \a __B, producing 4 intermediate
 ///    signed 16-bit results. Sum these 4 results with the corresponding
@@ -197,7 +208,7 @@
 /// _mm256_dpbsud_epi32(__m256i __W, __m256i __A, __m256i __B);
 /// \endcode
 ///
-/// This intrinsic corresponds to the \c VPDPBSSD instruction.
+/// This intrinsic corresponds to the \c VPDPBSUD instruction.
 ///
 /// \param __A
 ///    A 256-bit vector of [32 x char].
@@ -216,10 +227,12 @@
 /// ENDFOR
 /// dst[MAX:256] := 0
 /// \endcode
+// clang-format on
 #define _mm256_dpbsud_epi32(__W, __A, __B)                                     \
-  ((__m256i)__builtin_ia32_vpdpbsud256((__v8si)(__W), (__v8si)(__A),           \
-                                       (__v8si)(__B)))
+  ((__m256i)__builtin_ia32_vpdpbsud256((__v8si)(__W), (__v32qi)(__A),          \
+                                       (__v32qu)(__B)))
 
+// clang-format off
 /// Multiply groups of 4 adjacent pairs of signed 8-bit integers in \a __A with
 ///    corresponding unsigned 8-bit integers in \a __B, producing 4 intermediate
 ///    signed 16-bit results. Sum these 4 results with the corresponding
@@ -232,7 +245,7 @@
 /// _mm_dpbsuds_epi32( __m128i __W, __m128i __A, __m128i __B);
 /// \endcode
 ///
-/// This intrinsic corresponds to the \c VPDPBSSD instruction.
+/// This intrinsic corresponds to the \c VPDPBSUDS instruction.
 ///
 /// \param __A
 ///    A 128-bit vector of [16 x char].
@@ -251,10 +264,12 @@
 /// ENDFOR
 /// dst[MAX:128] := 0
 /// \endcode
+// clang-format on
 #define _mm_dpbsuds_epi32(__W, __A, __B)                                       \
-  ((__m128i)__builtin_ia32_vpdpbsuds128((__v4si)(__W), (__v4si)(__A),          \
-                                        (__v4si)(__B)))
+  ((__m128i)__builtin_ia32_vpdpbsuds128((__v4si)(__W), (__v16qi)(__A),         \
+                                        (__v16qu)(__B)))
 
+// clang-format off
 /// Multiply groups of 4 adjacent pairs of signed 8-bit integers in \a __A with
 ///    corresponding unsigned 8-bit integers in \a __B, producing 4 intermediate
 ///    signed 16-bit results. Sum these 4 results with the corresponding
@@ -267,7 +282,7 @@
 /// _mm256_dpbsuds_epi32(__m256i __W, __m256i __A, __m256i __B);
 /// \endcode
 ///
-/// This intrinsic corresponds to the \c VPDPBSSD instruction.
+/// This intrinsic corresponds to the \c VPDPBSUDS instruction.
 ///
 /// \param __A
 ///    A 256-bit vector of [32 x char].
@@ -286,10 +301,12 @@
 /// ENDFOR
 /// dst[MAX:256] := 0
 /// \endcode
+// clang-format on
 #define _mm256_dpbsuds_epi32(__W, __A, __B)                                    \
-  ((__m256i)__builtin_ia32_vpdpbsuds256((__v8si)(__W), (__v8si)(__A),          \
-                                        (__v8si)(__B)))
+  ((__m256i)__builtin_ia32_vpdpbsuds256((__v8si)(__W), (__v32qi)(__A),         \
+                                        (__v32qu)(__B)))
 
+// clang-format off
 /// Multiply groups of 4 adjacent pairs of unsigned 8-bit integers in \a __A with
 ///    corresponding unsigned 8-bit integers in \a __B, producing 4 intermediate
 ///    signed 16-bit results. Sum these 4 results with the corresponding
@@ -301,7 +318,7 @@
 /// _mm_dpbuud_epi32(__m128i __W, __m128i __A, __m128i __B);
 /// \endcode
 ///
-/// This intrinsic corresponds to the \c VPDPBSSD instruction.
+/// This intrinsic corresponds to the \c VPDPBUUD instruction.
 ///
 /// \param __A
 ///    A 128-bit vector of [16 x unsigned char].
@@ -320,10 +337,12 @@
 /// ENDFOR
 /// dst[MAX:128] := 0
 /// \endcode
+// clang-format on
 #define _mm_dpbuud_epi32(__W, __A, __B)                                        \
-  ((__m128i)__builtin_ia32_vpdpbuud128((__v4si)(__W), (__v4si)(__A),           \
-                                       (__v4si)(__B)))
+  ((__m128i)__builtin_ia32_vpdpbuud128((__v4si)(__W), (__v16qu)(__A),          \
+                                       (__v16qu)(__B)))
 
+// clang-format off
 /// Multiply groups of 4 adjacent pairs of unsigned 8-bit integers in \a __A with
 ///    corresponding unsigned 8-bit integers in \a __B, producing 4 intermediate
 ///    signed 16-bit results. Sum these 4 results with the corresponding
@@ -335,7 +354,7 @@
 /// _mm256_dpbuud_epi32(__m256i __W, __m256i __A, __m256i __B);
 /// \endcode
 ///
-/// This intrinsic corresponds to the \c VPDPBSSD instruction.
+/// This intrinsic corresponds to the \c VPDPBUUD instruction.
 ///
 /// \param __A
 ///    A 256-bit vector of [32 x unsigned char].
@@ -354,10 +373,12 @@
 /// ENDFOR
 /// dst[MAX:256] := 0
 /// \endcode
+// clang-format on
 #define _mm256_dpbuud_epi32(__W, __A, __B)                                     \
-  ((__m256i)__builtin_ia32_vpdpbuud256((__v8si)(__W), (__v8si)(__A),           \
-                                       (__v8si)(__B)))
+  ((__m256i)__builtin_ia32_vpdpbuud256((__v8si)(__W), (__v32qu)(__A),          \
+                                       (__v32qu)(__B)))
 
+// clang-format off
 /// Multiply groups of 4 adjacent pairs of unsigned 8-bit integers in \a __A with
 ///    corresponding unsigned 8-bit integers in \a __B, producing 4 intermediate
 ///    signed 16-bit results. Sum these 4 results with the corresponding
@@ -389,10 +410,12 @@
 /// ENDFOR
 /// dst[MAX:128] := 0
 /// \endcode
+// clang-format on
 #define _mm_dpbuuds_epi32(__W, __A, __B)                                       \
-  ((__m128i)__builtin_ia32_vpdpbuuds128((__v4si)(__W), (__v4si)(__A),          \
-                                        (__v4si)(__B)))
+  ((__m128i)__builtin_ia32_vpdpbuuds128((__v4si)(__W), (__v16qu)(__A),         \
+                                        (__v16qu)(__B)))
 
+// clang-format off
 ///    corresponding unsigned 8-bit integers in \a __B, producing 4 intermediate
 ///    signed 16-bit results. Sum these 4 results with the corresponding
 ///    32-bit integer in \a __W with signed saturation, and store the packed
@@ -423,8 +446,9 @@
 /// ENDFOR
 /// dst[MAX:256] := 0
 /// \endcode
+// clang-format on
 #define _mm256_dpbuuds_epi32(__W, __A, __B)                                    \
-  ((__m256i)__builtin_ia32_vpdpbuuds256((__v8si)(__W), (__v8si)(__A),          \
-                                        (__v8si)(__B)))
+  ((__m256i)__builtin_ia32_vpdpbuuds256((__v8si)(__W), (__v32qu)(__A),         \
+                                        (__v32qu)(__B)))
 
 #endif // __AVXVNNIINT8INTRIN_H

package/include/avxvnniintrin.h

@@ -63,7 +63,8 @@
 static __inline__ __m256i __DEFAULT_FN_ATTRS256
 _mm256_dpbusd_avx_epi32(__m256i __S, __m256i __A, __m256i __B)
 {
-  return (__m256i)__builtin_ia32_vpdpbusd256((__v8si)__S, (__v8si)__A, (__v8si)__B);
+  return (__m256i)__builtin_ia32_vpdpbusd256((__v8si)__S, (__v32qu)__A,
+                                             (__v32qi)__B);
 }
 
 /// Multiply groups of 4 adjacent pairs of unsigned 8-bit integers in \a __A with
@@ -86,7 +87,8 @@ _mm256_dpbusd_avx_epi32(__m256i __S, __m256i __A, __m256i __B)
 static __inline__ __m256i __DEFAULT_FN_ATTRS256
 _mm256_dpbusds_avx_epi32(__m256i __S, __m256i __A, __m256i __B)
 {
-  return (__m256i)__builtin_ia32_vpdpbusds256((__v8si)__S, (__v8si)__A, (__v8si)__B);
+  return (__m256i)__builtin_ia32_vpdpbusds256((__v8si)__S, (__v32qu)__A,
+                                              (__v32qi)__B);
 }
 
 /// Multiply groups of 2 adjacent pairs of signed 16-bit integers in \a __A with
@@ -107,7 +109,8 @@ _mm256_dpbusds_avx_epi32(__m256i __S, __m256i __A, __m256i __B)
 static __inline__ __m256i __DEFAULT_FN_ATTRS256
 _mm256_dpwssd_avx_epi32(__m256i __S, __m256i __A, __m256i __B)
 {
-  return (__m256i)__builtin_ia32_vpdpwssd256((__v8si)__S, (__v8si)__A, (__v8si)__B);
+  return (__m256i)__builtin_ia32_vpdpwssd256((__v8si)__S, (__v16hi)__A,
+                                             (__v16hi)__B);
 }
 
 /// Multiply groups of 2 adjacent pairs of signed 16-bit integers in \a __A with
@@ -128,7 +131,8 @@ _mm256_dpwssd_avx_epi32(__m256i __S, __m256i __A, __m256i __B)
 static __inline__ __m256i __DEFAULT_FN_ATTRS256
 _mm256_dpwssds_avx_epi32(__m256i __S, __m256i __A, __m256i __B)
 {
-  return (__m256i)__builtin_ia32_vpdpwssds256((__v8si)__S, (__v8si)__A, (__v8si)__B);
+  return (__m256i)__builtin_ia32_vpdpwssds256((__v8si)__S, (__v16hi)__A,
+                                              (__v16hi)__B);
 }
 
 /// Multiply groups of 4 adjacent pairs of unsigned 8-bit integers in \a __A with
@@ -151,7 +155,8 @@ _mm256_dpwssds_avx_epi32(__m256i __S, __m256i __A, __m256i __B)
 static __inline__ __m128i __DEFAULT_FN_ATTRS128
 _mm_dpbusd_avx_epi32(__m128i __S, __m128i __A, __m128i __B)
 {
-  return (__m128i)__builtin_ia32_vpdpbusd128((__v4si)__S, (__v4si)__A, (__v4si)__B);
+  return (__m128i)__builtin_ia32_vpdpbusd128((__v4si)__S, (__v16qu)__A,
+                                             (__v16qi)__B);
 }
 
 /// Multiply groups of 4 adjacent pairs of unsigned 8-bit integers in \a __A with
@@ -174,7 +179,8 @@ _mm_dpbusd_avx_epi32(__m128i __S, __m128i __A, __m128i __B)
 static __inline__ __m128i __DEFAULT_FN_ATTRS128
 _mm_dpbusds_avx_epi32(__m128i __S, __m128i __A, __m128i __B)
 {
-  return (__m128i)__builtin_ia32_vpdpbusds128((__v4si)__S, (__v4si)__A, (__v4si)__B);
+  return (__m128i)__builtin_ia32_vpdpbusds128((__v4si)__S, (__v16qu)__A,
+                                              (__v16qi)__B);
 }
 
 /// Multiply groups of 2 adjacent pairs of signed 16-bit integers in \a __A with
@@ -195,7 +201,8 @@ _mm_dpbusds_avx_epi32(__m128i __S, __m128i __A, __m128i __B)
 static __inline__ __m128i __DEFAULT_FN_ATTRS128
 _mm_dpwssd_avx_epi32(__m128i __S, __m128i __A, __m128i __B)
 {
-  return (__m128i)__builtin_ia32_vpdpwssd128((__v4si)__S, (__v4si)__A, (__v4si)__B);
+  return (__m128i)__builtin_ia32_vpdpwssd128((__v4si)__S, (__v8hi)__A,
+                                             (__v8hi)__B);
 }
 
 /// Multiply groups of 2 adjacent pairs of signed 16-bit integers in \a __A with
@@ -216,7 +223,8 @@ _mm_dpwssd_avx_epi32(__m128i __S, __m128i __A, __m128i __B)
 static __inline__ __m128i __DEFAULT_FN_ATTRS128
 _mm_dpwssds_avx_epi32(__m128i __S, __m128i __A, __m128i __B)
 {
-  return (__m128i)__builtin_ia32_vpdpwssds128((__v4si)__S, (__v4si)__A, (__v4si)__B);
+  return (__m128i)__builtin_ia32_vpdpwssds128((__v4si)__S, (__v8hi)__A,
+                                              (__v8hi)__B);
 }
 
 #undef __DEFAULT_FN_ATTRS128

package/include/cpuid.h

@@ -253,10 +253,6 @@
 #define bit_RDPRU       0x00000010
 #define bit_WBNOINVD    0x00000200
 
-/* Features in %ebx for leaf 0x24 */
-#define bit_AVX10_256   0x00020000
-#define bit_AVX10_512   0x00040000
-
 #ifdef __i386__
 #define __cpuid(__leaf, __eax, __ebx, __ecx, __edx) \
     __asm("cpuid" : "=a"(__eax), "=b" (__ebx), "=c"(__ecx), "=d"(__edx) \
@@ -282,6 +278,24 @@
         : "0"(__leaf), "2"(__count))
 #endif
 
+/// Queries the processor to determine the highest supported \c CPUID leaf.
+/// This intrinsic is only available on x86 and x64.
+///
+/// \headerfile <cpuid.h>
+///
+/// This intrinsic corresponds to the <c> CPUID </c> instruction.
+///
+/// \param __leaf
+///    \a __leaf can be either 0x0 or 0x8000000. If \a __leaf == 0x0, the
+///    highest supported value for basic \c CPUID information is returned.
+///    If \a __leaf == 0x8000000, the highest supported value for extended
+///    \c CPUID information is returned.
+/// \param __sig
+///    If the \a __sig pointer is non-null, the first four bytes of the
+///    signature (as found in the \c EBX register) are returned in the
+///    location pointed to by \a __sig.
+/// \returns Returns 0 if \c CPUID is supported; otherwise returns the value
+///    that \c CPUID returns in the \c EAX register.
 static __inline unsigned int __get_cpuid_max (unsigned int __leaf,
                                               unsigned int *__sig)
 {
@@ -315,6 +329,32 @@ static __inline unsigned int __get_cpuid_max (unsigned int __leaf,
     return __eax;
 }
 
+/// For the requested \c CPUID leaf, queries the processor for information
+/// about the CPU type and CPU features (such as processor vendor, supported
+/// instruction sets, CPU capabilities, cache sizes, CPU model and family, and
+/// other hardware details). This intrinsic is only available on x86 and x64.
+///
+/// \headerfile <cpuid.h>
+///
+/// This intrinsic corresponds to the <c> CPUID </c> instruction.
+///
+/// \param __leaf
+///    An unsigned integer that identifies the level (also called "leaf") at
+///    which the \c CPUID instruction will be executed.
+/// \param __eax
+///    A pointer to an integer that corresponds to the \c EAX register where
+///    \c CPUID stores output results.
+/// \param __ebx
+///    A pointer to an integer that corresponds to the \c EBX register where
+///    \c CPUID stores output results.
+/// \param __ecx
+///    A pointer to an integer that corresponds to the \c ECX register where
+///    \c CPUID stores output results.
+/// \param __edx
+///    A pointer to an integer that corresponds to the \c EDX register where
+///    \c CPUID stores output results.
+/// \returns Returns 1 if the requested \c CPUID leaf is supported; otherwise
+///    returns 0.
 static __inline int __get_cpuid (unsigned int __leaf, unsigned int *__eax,
                                  unsigned int *__ebx, unsigned int *__ecx,
                                  unsigned int *__edx)
@@ -328,6 +368,36 @@ static __inline int __get_cpuid (unsigned int __leaf, unsigned int *__eax,
     return 1;
 }
 
+/// For the requested \c CPUID leaf and subleaf, queries the processor for
+/// information about the CPU type and CPU features (such as processor vendor,
+/// supported instruction sets, CPU capabilities, cache sizes, CPU model and
+/// family, and other hardware details). This intrinsic is only available on
+/// x86 and x64.
+///
+/// \headerfile <cpuid.h>
+///
+/// This intrinsic corresponds to the <c> CPUID </c> instruction.
+///
+/// \param __leaf
+///    An unsigned integer that identifies the level (also called "leaf") at
+///    which the \c CPUID instruction will be executed.
+/// \param __subleaf
+///    An unsigned integer that identifies the sublevel (also called
+///    "subleaf") at which the \c CPUID instruction will be executed.
+/// \param __eax
+///    A pointer to an integer that corresponds to the \c EAX register where
+///    \c CPUID stores output results.
+/// \param __ebx
+///    A pointer to an integer that corresponds to the \c EBX register where
+///    \c CPUID stores output results.
+/// \param __ecx
+///    A pointer to an integer that corresponds to the \c ECX register where
+///    \c CPUID stores output results.
+/// \param __edx
+///    A pointer to an integer that corresponds to the \c EDX register where
+///    \c CPUID stores output results.
+/// \returns Returns 1 if the requested \c CPUID leaf is supported; otherwise
+///    returns 0.
 static __inline int __get_cpuid_count (unsigned int __leaf,
                                        unsigned int __subleaf,
                                        unsigned int *__eax, unsigned int *__ebx,
@@ -345,10 +415,37 @@ static __inline int __get_cpuid_count (unsigned int __leaf,
 // In some configurations, __cpuidex is defined as a builtin (primarily
 // -fms-extensions) which will conflict with the __cpuidex definition below.
 #if !(__has_builtin(__cpuidex))
+// In some cases, offloading will set the host as the aux triple and define the
+// builtin. Given __has_builtin does not detect builtins on aux triples, we need
+// to explicitly check for some offloading cases.
+#if !defined(__NVPTX__) && !defined(__AMDGPU__) && !defined(__SPIRV__)
+/// Executes the \c CPUID instruction with the specified leaf and subleaf
+/// values, and returns the results from the CPU's registers. This intrinsic
+/// is only available on x86 and x64.
+///
+/// \headerfile <cpuid.h>
+///
+/// This intrinsic corresponds to the <c> CPUID </c> instruction.
+///
+/// \param __cpu_info
+///    An output array of four integers:
+///    <ul>
+///    <li>\a __cpuInfo[0] receives the value of the \c EAX register.</li>
+///    <li>\a __cpuInfo[1] receives the value of the \c EBX register.</li>
+///    <li>\a __cpuInfo[2] receives the value of the \c ECX register.</li>
+///    <li>\a __cpuInfo[3] receives the value of the \c EDX register.</li>
+///    </ul>
+/// \param __leaf
+///    An unsigned integer that identifies the level (also called the "leaf")
+///    at which the \c CPUID instruction will be executed.
+/// \param __subleaf
+///    An unsigned integer that identifies the sublevel (also called the
+///    "subleaf") at which the \c CPUID instruction will be executed.
 static __inline void __cpuidex(int __cpu_info[4], int __leaf, int __subleaf) {
   __cpuid_count(__leaf, __subleaf, __cpu_info[0], __cpu_info[1], __cpu_info[2],
                 __cpu_info[3]);
 }
 #endif
+#endif
 
 #endif /* __CPUID_H */