diff --git a/renderscript/clang-include/avx2intrin.h b/renderscript/clang-include/avx2intrin.h
index 63b1efc..1887fc8 100644
--- a/renderscript/clang-include/avx2intrin.h
+++ b/renderscript/clang-include/avx2intrin.h
@@ -25,6 +25,9 @@
 #error "Never use <avx2intrin.h> directly; include <immintrin.h> instead."
 #endif
 
+#ifndef __AVX2INTRIN_H
+#define __AVX2INTRIN_H
+
 /* SSE4 Multiple Packed Sums of Absolute Difference.  */
 #define _mm256_mpsadbw_epu8(X, Y, M) __builtin_ia32_mpsadbw256((X), (Y), (M))
 
@@ -1199,3 +1202,5 @@
   (__m256i)__builtin_ia32_gatherq_q256((__v4di)_mm256_setzero_si256(), \
              (const __v4di *)__m, (__v4di)__i, \
              (__v4di)_mm256_set1_epi64x(-1), (s)); })
+
+#endif /* __AVX2INTRIN_H */
diff --git a/renderscript/clang-include/avxintrin.h b/renderscript/clang-include/avxintrin.h
index 412d284..50454f2 100644
--- a/renderscript/clang-include/avxintrin.h
+++ b/renderscript/clang-include/avxintrin.h
@@ -25,6 +25,9 @@
 #error "Never use <avxintrin.h> directly; include <immintrin.h> instead."
 #endif
 
+#ifndef __AVXINTRIN_H
+#define __AVXINTRIN_H
+
 typedef double __v4df __attribute__ ((__vector_size__ (32)));
 typedef float __v8sf __attribute__ ((__vector_size__ (32)));
 typedef long long __v4di __attribute__ ((__vector_size__ (32)));
@@ -1078,78 +1081,75 @@
 
 /* Cast between vector types */
 static __inline __m256 __attribute__((__always_inline__, __nodebug__))
-_mm256_castpd_ps(__m256d __in)
+_mm256_castpd_ps(__m256d __a)
 {
-  return (__m256)__in;
+  return (__m256)__a;
 }
 
 static __inline __m256i __attribute__((__always_inline__, __nodebug__))
-_mm256_castpd_si256(__m256d __in)
+_mm256_castpd_si256(__m256d __a)
 {
-  return (__m256i)__in;
+  return (__m256i)__a;
 }
 
 static __inline __m256d __attribute__((__always_inline__, __nodebug__))
-_mm256_castps_pd(__m256 __in)
+_mm256_castps_pd(__m256 __a)
 {
-  return (__m256d)__in;
+  return (__m256d)__a;
 }
 
 static __inline __m256i __attribute__((__always_inline__, __nodebug__))
-_mm256_castps_si256(__m256 __in)
+_mm256_castps_si256(__m256 __a)
 {
-  return (__m256i)__in;
+  return (__m256i)__a;
 }
 
 static __inline __m256 __attribute__((__always_inline__, __nodebug__))
-_mm256_castsi256_ps(__m256i __in)
+_mm256_castsi256_ps(__m256i __a)
 {
-  return (__m256)__in;
+  return (__m256)__a;
 }
 
 static __inline __m256d __attribute__((__always_inline__, __nodebug__))
-_mm256_castsi256_pd(__m256i __in)
+_mm256_castsi256_pd(__m256i __a)
 {
-  return (__m256d)__in;
+  return (__m256d)__a;
 }
 
 static __inline __m128d __attribute__((__always_inline__, __nodebug__))
-_mm256_castpd256_pd128(__m256d __in)
+_mm256_castpd256_pd128(__m256d __a)
 {
-  return __builtin_shufflevector(__in, __in, 0, 1);
+  return __builtin_shufflevector(__a, __a, 0, 1);
 }
 
 static __inline __m128 __attribute__((__always_inline__, __nodebug__))
-_mm256_castps256_ps128(__m256 __in)
+_mm256_castps256_ps128(__m256 __a)
 {
-  return __builtin_shufflevector(__in, __in, 0, 1, 2, 3);
+  return __builtin_shufflevector(__a, __a, 0, 1, 2, 3);
 }
 
 static __inline __m128i __attribute__((__always_inline__, __nodebug__))
-_mm256_castsi256_si128(__m256i __in)
+_mm256_castsi256_si128(__m256i __a)
 {
-  return __builtin_shufflevector(__in, __in, 0, 1);
+  return __builtin_shufflevector(__a, __a, 0, 1);
 }
 
 static __inline __m256d __attribute__((__always_inline__, __nodebug__))
-_mm256_castpd128_pd256(__m128d __in)
+_mm256_castpd128_pd256(__m128d __a)
 {
-  __m128d __zero = _mm_setzero_pd();
-  return __builtin_shufflevector(__in, __zero, 0, 1, 2, 2);
+  return __builtin_shufflevector(__a, __a, 0, 1, -1, -1);
 }
 
 static __inline __m256 __attribute__((__always_inline__, __nodebug__))
-_mm256_castps128_ps256(__m128 __in)
+_mm256_castps128_ps256(__m128 __a)
 {
-  __m128 __zero = _mm_setzero_ps();
-  return __builtin_shufflevector(__in, __zero, 0, 1, 2, 3, 4, 4, 4, 4);
+  return __builtin_shufflevector(__a, __a, 0, 1, 2, 3, -1, -1, -1, -1);
 }
 
 static __inline __m256i __attribute__((__always_inline__, __nodebug__))
-_mm256_castsi128_si256(__m128i __in)
+_mm256_castsi128_si256(__m128i __a)
 {
-  __m128i __zero = _mm_setzero_si128();
-  return __builtin_shufflevector(__in, __zero, 0, 1, 2, 2);
+  return __builtin_shufflevector(__a, __a, 0, 1, -1, -1);
 }
 
 /* SIMD load ops (unaligned) */
@@ -1220,3 +1220,5 @@
   __v128 = _mm256_extractf128_si256(__a, 1);
   __builtin_ia32_storedqu((char *)__addr_hi, (__v16qi)__v128);
 }
+
+#endif /* __AVXINTRIN_H */
diff --git a/renderscript/clang-include/cpuid.h b/renderscript/clang-include/cpuid.h
index 6d7d61d..8f12cae 100644
--- a/renderscript/clang-include/cpuid.h
+++ b/renderscript/clang-include/cpuid.h
@@ -25,10 +25,132 @@
 #error this header is for x86 only
 #endif
 
-static inline int __get_cpuid (unsigned int __level, unsigned int *__eax,
-                               unsigned int *__ebx, unsigned int *__ecx,
-                               unsigned int *__edx) {
-    __asm("cpuid" : "=a"(*__eax), "=b" (*__ebx), "=c"(*__ecx), "=d"(*__edx)
-                  : "0"(__level));
+/* Features in %ecx for level 1 */
+#define bit_SSE3        0x00000001
+#define bit_PCLMULQDQ   0x00000002
+#define bit_DTES64      0x00000004
+#define bit_MONITOR     0x00000008
+#define bit_DSCPL       0x00000010
+#define bit_VMX         0x00000020
+#define bit_SMX         0x00000040
+#define bit_EIST        0x00000080
+#define bit_TM2         0x00000100
+#define bit_SSSE3       0x00000200
+#define bit_CNXTID      0x00000400
+#define bit_FMA         0x00001000
+#define bit_CMPXCHG16B  0x00002000
+#define bit_xTPR        0x00004000
+#define bit_PDCM        0x00008000
+#define bit_PCID        0x00020000
+#define bit_DCA         0x00040000
+#define bit_SSE41       0x00080000
+#define bit_SSE42       0x00100000
+#define bit_x2APIC      0x00200000
+#define bit_MOVBE       0x00400000
+#define bit_POPCNT      0x00800000
+#define bit_TSCDeadline 0x01000000
+#define bit_AESNI       0x02000000
+#define bit_XSAVE       0x04000000
+#define bit_OSXSAVE     0x08000000
+#define bit_AVX         0x10000000
+#define bit_RDRAND      0x40000000
+
+/* Features in %edx for level 1 */
+#define bit_FPU         0x00000001
+#define bit_VME         0x00000002
+#define bit_DE          0x00000004
+#define bit_PSE         0x00000008
+#define bit_TSC         0x00000010
+#define bit_MSR         0x00000020
+#define bit_PAE         0x00000040
+#define bit_MCE         0x00000080
+#define bit_CX8         0x00000100
+#define bit_APIC        0x00000200
+#define bit_SEP         0x00000800
+#define bit_MTRR        0x00001000
+#define bit_PGE         0x00002000
+#define bit_MCA         0x00004000
+#define bit_CMOV        0x00008000
+#define bit_PAT         0x00010000
+#define bit_PSE36       0x00020000
+#define bit_PSN         0x00040000
+#define bit_CLFSH       0x00080000
+#define bit_DS          0x00200000
+#define bit_ACPI        0x00400000
+#define bit_MMX         0x00800000
+#define bit_FXSR        0x01000000
+#define bit_SSE         0x02000000
+#define bit_SSE2        0x04000000
+#define bit_SS          0x08000000
+#define bit_HTT         0x10000000
+#define bit_TM          0x20000000
+#define bit_PBE         0x80000000
+
+/* Features in %ebx for level 7 sub-leaf 0 */
+#define bit_FSGSBASE    0x00000001
+#define bit_SMEP        0x00000080
+#define bit_ENH_MOVSB   0x00000200
+
+/* PIC on i386 uses %ebx, so preserve it. */
+#if __i386__
+#define __cpuid(__level, __eax, __ebx, __ecx, __edx) \
+    __asm("  pushl  %%ebx\n" \
+          "  cpuid\n" \
+          "  mov    %%ebx,%1\n" \
+          "  popl   %%ebx" \
+        : "=a"(__eax), "=r" (__ebx), "=c"(__ecx), "=d"(__edx) \
+        : "0"(__level))
+
+#define __cpuid_count(__level, __count, __eax, __ebx, __ecx, __edx) \
+    __asm("  pushl  %%ebx\n" \
+          "  cpuid\n" \
+          "  mov    %%ebx,%1\n" \
+          "  popl   %%ebx" \
+        : "=a"(__eax), "=r" (__ebx), "=c"(__ecx), "=d"(__edx) \
+        : "0"(__level), "2"(__count))
+#else
+#define __cpuid(__level, __eax, __ebx, __ecx, __edx) \
+    __asm("cpuid" : "=a"(__eax), "=b" (__ebx), "=c"(__ecx), "=d"(__edx) \
+                  : "0"(__level))
+
+#define __cpuid_count(__level, __count, __eax, __ebx, __ecx, __edx) \
+    __asm("cpuid" : "=a"(__eax), "=b" (__ebx), "=c"(__ecx), "=d"(__edx) \
+                  : "0"(__level), "2"(__count))
+#endif
+
+static __inline int __get_cpuid (unsigned int __level, unsigned int *__eax,
+                                 unsigned int *__ebx, unsigned int *__ecx,
+                                 unsigned int *__edx) {
+    __cpuid(__level, *__eax, *__ebx, *__ecx, *__edx);
     return 1;
 }
+
+static __inline int __get_cpuid_max (unsigned int __level, unsigned int *__sig)
+{
+    unsigned int __eax, __ebx, __ecx, __edx;
+#if __i386__
+    int __cpuid_supported;
+
+    __asm("  pushfl\n"
+          "  popl   %%eax\n"
+          "  movl   %%eax,%%ecx\n"
+          "  xorl   $0x00200000,%%eax\n"
+          "  pushl  %%eax\n"
+          "  popfl\n"
+          "  pushfl\n"
+          "  popl   %%eax\n"
+          "  movl   $0,%0\n"
+          "  cmpl   %%eax,%%ecx\n"
+          "  je     1f\n"
+          "  movl   $1,%0\n"
+          "1:"
+        : "=r" (__cpuid_supported) : : "eax", "ecx");
+    if (!__cpuid_supported)
+        return 0;
+#endif
+
+    __cpuid(__level, __eax, __ebx, __ecx, __edx);
+    if (__sig)
+        *__sig = __ebx;
+    return __eax;
+}
diff --git a/renderscript/clang-include/emmintrin.h b/renderscript/clang-include/emmintrin.h
index e18fae4..f965dce 100644
--- a/renderscript/clang-include/emmintrin.h
+++ b/renderscript/clang-include/emmintrin.h
@@ -245,13 +245,15 @@
 static __inline__ __m128d __attribute__((__always_inline__, __nodebug__))
 _mm_cmpgt_sd(__m128d __a, __m128d __b)
 {
-  return (__m128d)__builtin_ia32_cmpsd(__b, __a, 1);
+  __m128d __c = __builtin_ia32_cmpsd(__b, __a, 1);
+  return (__m128d) { __c[0], __a[1] };
 }
 
 static __inline__ __m128d __attribute__((__always_inline__, __nodebug__))
 _mm_cmpge_sd(__m128d __a, __m128d __b)
 {
-  return (__m128d)__builtin_ia32_cmpsd(__b, __a, 2);
+  __m128d __c = __builtin_ia32_cmpsd(__b, __a, 2);
+  return (__m128d) { __c[0], __a[1] };
 }
 
 static __inline__ __m128d __attribute__((__always_inline__, __nodebug__))
@@ -287,13 +289,15 @@
 static __inline__ __m128d __attribute__((__always_inline__, __nodebug__))
 _mm_cmpngt_sd(__m128d __a, __m128d __b)
 {
-  return (__m128d)__builtin_ia32_cmpsd(__b, __a, 5);
+  __m128d __c = __builtin_ia32_cmpsd(__b, __a, 5);
+  return (__m128d) { __c[0], __a[1] };
 }
 
 static __inline__ __m128d __attribute__((__always_inline__, __nodebug__))
 _mm_cmpnge_sd(__m128d __a, __m128d __b)
 {
-  return (__m128d)__builtin_ia32_cmpsd(__b, __a, 6);
+  __m128d __c = __builtin_ia32_cmpsd(__b, __a, 6);
+  return (__m128d) { __c[0], __a[1] };
 }
 
 static __inline__ int __attribute__((__always_inline__, __nodebug__))
@@ -1379,39 +1383,39 @@
   __builtin_shufflevector(__a, __b, (i) & 1, (((i) & 2) >> 1) + 2); })
 
 static __inline__ __m128 __attribute__((__always_inline__, __nodebug__))
-_mm_castpd_ps(__m128d __in)
+_mm_castpd_ps(__m128d __a)
 {
-  return (__m128)__in;
+  return (__m128)__a;
 }
 
 static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
-_mm_castpd_si128(__m128d __in)
+_mm_castpd_si128(__m128d __a)
 {
-  return (__m128i)__in;
+  return (__m128i)__a;
 }
 
 static __inline__ __m128d __attribute__((__always_inline__, __nodebug__))
-_mm_castps_pd(__m128 __in)
+_mm_castps_pd(__m128 __a)
 {
-  return (__m128d)__in;
+  return (__m128d)__a;
 }
 
 static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
-_mm_castps_si128(__m128 __in)
+_mm_castps_si128(__m128 __a)
 {
-  return (__m128i)__in;
+  return (__m128i)__a;
 }
 
 static __inline__ __m128 __attribute__((__always_inline__, __nodebug__))
-_mm_castsi128_ps(__m128i __in)
+_mm_castsi128_ps(__m128i __a)
 {
-  return (__m128)__in;
+  return (__m128)__a;
 }
 
 static __inline__ __m128d __attribute__((__always_inline__, __nodebug__))
-_mm_castsi128_pd(__m128i __in)
+_mm_castsi128_pd(__m128i __a)
 {
-  return (__m128d)__in;
+  return (__m128d)__a;
 }
 
 static __inline__ void __attribute__((__always_inline__, __nodebug__))
diff --git a/renderscript/clang-include/immintrin.h b/renderscript/clang-include/immintrin.h
index cd733bf..fea7c3b 100644
--- a/renderscript/clang-include/immintrin.h
+++ b/renderscript/clang-include/immintrin.h
@@ -102,4 +102,13 @@
 #include <rtmintrin.h>
 #endif
 
+/* FIXME: check __HLE__ as well when HLE is supported. */
+#if defined (__RTM__)
+static __inline__ int __attribute__((__always_inline__, __nodebug__))
+_xtest(void)
+{
+  return __builtin_ia32_xtest();
+}
+#endif
+
 #endif /* __IMMINTRIN_H */
diff --git a/renderscript/clang-include/mm3dnow.h b/renderscript/clang-include/mm3dnow.h
index d5236f8..5242d99 100644
--- a/renderscript/clang-include/mm3dnow.h
+++ b/renderscript/clang-include/mm3dnow.h
@@ -25,6 +25,7 @@
 #define _MM3DNOW_H_INCLUDED
 
 #include <mmintrin.h>
+#include <prfchwintrin.h>
 
 typedef float __v2sf __attribute__((__vector_size__(8)));
 
diff --git a/renderscript/clang-include/prfchwintrin.h b/renderscript/clang-include/prfchwintrin.h
new file mode 100644
index 0000000..9825bd8
--- /dev/null
+++ b/renderscript/clang-include/prfchwintrin.h
@@ -0,0 +1,39 @@
+/*===---- prfchwintrin.h - PREFETCHW intrinsic -----------------------------===
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+ * THE SOFTWARE.
+ *
+ *===-----------------------------------------------------------------------===
+ */
+
+#if !defined(__X86INTRIN_H) && !defined(_MM3DNOW_H_INCLUDED)
+#error "Never use <prfchwintrin.h> directly; include <x86intrin.h> or <mm3dnow.h> instead."
+#endif
+
+#ifndef __PRFCHWINTRIN_H
+#define __PRFCHWINTRIN_H
+
+#if defined(__PRFCHW__) || defined(__3dNOW__)
+static __inline__ void __attribute__((__always_inline__, __nodebug__))
+_m_prefetchw(void *__P)
+{
+  __builtin_prefetch (__P, 1, 3 /* _MM_HINT_T0 */);
+}
+#endif
+
+#endif /* __PRFCHWINTRIN_H */
diff --git a/renderscript/clang-include/rdseedintrin.h b/renderscript/clang-include/rdseedintrin.h
new file mode 100644
index 0000000..0fef1fa
--- /dev/null
+++ b/renderscript/clang-include/rdseedintrin.h
@@ -0,0 +1,52 @@
+/*===---- rdseedintrin.h - RDSEED intrinsics -------------------------------===
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+ * THE SOFTWARE.
+ *
+ *===-----------------------------------------------------------------------===
+ */
+
+#ifndef __X86INTRIN_H
+#error "Never use <rdseedintrin.h> directly; include <x86intrin.h> instead."
+#endif
+
+#ifndef __RDSEEDINTRIN_H
+#define __RDSEEDINTRIN_H
+
+#ifdef __RDSEED__
+static __inline__ int __attribute__((__always_inline__, __nodebug__))
+_rdseed16_step(unsigned short *__p)
+{
+  return __builtin_ia32_rdseed16_step(__p);
+}
+
+static __inline__ int __attribute__((__always_inline__, __nodebug__))
+_rdseed32_step(unsigned int *__p)
+{
+  return __builtin_ia32_rdseed32_step(__p);
+}
+
+#ifdef __x86_64__
+static __inline__ int __attribute__((__always_inline__, __nodebug__))
+_rdseed64_step(unsigned long long *__p)
+{
+  return __builtin_ia32_rdseed64_step(__p);
+}
+#endif
+#endif /* __RDSEED__ */
+#endif /* __RDSEEDINTRIN_H */
diff --git a/renderscript/clang-include/rtmintrin.h b/renderscript/clang-include/rtmintrin.h
index bdc2b99..26149ca 100644
--- a/renderscript/clang-include/rtmintrin.h
+++ b/renderscript/clang-include/rtmintrin.h
@@ -25,6 +25,9 @@
 #error "Never use <rtmintrin.h> directly; include <immintrin.h> instead."
 #endif
 
+#ifndef __RTMINTRIN_H
+#define __RTMINTRIN_H
+
 #define _XBEGIN_STARTED   (~0u)
 #define _XABORT_EXPLICIT  (1 << 0)
 #define _XABORT_RETRY     (1 << 1)
@@ -47,3 +50,5 @@
 }
 
 #define _xabort(imm) __builtin_ia32_xabort((imm))
+
+#endif /* __RTMINTRIN_H */
diff --git a/renderscript/clang-include/stddef.h b/renderscript/clang-include/stddef.h
index ad5dc21..6a64d6d 100644
--- a/renderscript/clang-include/stddef.h
+++ b/renderscript/clang-include/stddef.h
@@ -26,17 +26,42 @@
 #ifndef __STDDEF_H
 #define __STDDEF_H
 
-#ifndef _PTRDIFF_T
+#if !defined(_PTRDIFF_T) || __has_feature(modules)
+/* Always define ptrdiff_t when modules are available. */
+#if !__has_feature(modules)
 #define _PTRDIFF_T
+#endif
 typedef __PTRDIFF_TYPE__ ptrdiff_t;
 #endif
-#ifndef _SIZE_T
+
+#if !defined(_SIZE_T) || __has_feature(modules)
+/* Always define size_t when modules are available. */
+#if !__has_feature(modules)
 #define _SIZE_T
+#endif
 typedef __SIZE_TYPE__ size_t;
 #endif
+
+/* ISO9899:2011 7.20 (C11 Annex K): Define rsize_t if __STDC_WANT_LIB_EXT1__ is
+ * enabled. */
+#if (defined(__STDC_WANT_LIB_EXT1__) && __STDC_WANT_LIB_EXT1__ >= 1 && \
+     !defined(_RSIZE_T)) || __has_feature(modules)
+/* Always define rsize_t when modules are available. */
+#if !__has_feature(modules)
+#define _RSIZE_T
+#endif
+typedef __SIZE_TYPE__ rsize_t;
+#endif
+
 #ifndef __cplusplus
-#ifndef _WCHAR_T
+/* Always define wchar_t when modules are available. */
+#if !defined(_WCHAR_T) || __has_feature(modules)
+#if !__has_feature(modules)
 #define _WCHAR_T
+#if defined(_MSC_EXTENSIONS)
+#define _WCHAR_T_DEFINED
+#endif
+#endif
 typedef __WCHAR_TYPE__ wchar_t;
 #endif
 #endif
@@ -66,9 +91,12 @@
 /* Some C libraries expect to see a wint_t here. Others (notably MinGW) will use
 __WINT_TYPE__ directly; accommodate both by requiring __need_wint_t */
 #if defined(__need_wint_t)
-#if !defined(_WINT_T)
+/* Always define wint_t when modules are available. */
+#if !defined(_WINT_T) || __has_feature(modules)
+#if !__has_feature(modules)
 #define _WINT_T
+#endif
 typedef __WINT_TYPE__ wint_t;
-#endif /* _WINT_T */
+#endif
 #undef __need_wint_t
 #endif /* __need_wint_t */
diff --git a/renderscript/clang-include/stdint.h b/renderscript/clang-include/stdint.h
index 6f1a876..11529c0 100644
--- a/renderscript/clang-include/stdint.h
+++ b/renderscript/clang-include/stdint.h
@@ -30,7 +30,48 @@
  */
 #if __STDC_HOSTED__ && \
     defined(__has_include_next) && __has_include_next(<stdint.h>)
+
+// C99 7.18.3 Limits of other integer types
+//
+//  Footnote 219, 220: C++ implementations should define these macros only when
+//  __STDC_LIMIT_MACROS is defined before <stdint.h> is included.
+//
+//  Footnote 222: C++ implementations should define these macros only when
+//  __STDC_CONSTANT_MACROS is defined before <stdint.h> is included.
+//
+// C++11 [cstdint.syn]p2:
+//
+//  The macros defined by <cstdint> are provided unconditionally. In particular,
+//  the symbols __STDC_LIMIT_MACROS and __STDC_CONSTANT_MACROS (mentioned in
+//  footnotes 219, 220, and 222 in the C standard) play no role in C++.
+//
+// C11 removed the problematic footnotes.
+//
+// Work around this inconsistency by always defining those macros in C++ mode,
+// so that a C library implementation which follows the C99 standard can be
+// used in C++.
+# ifdef __cplusplus
+#  if !defined(__STDC_LIMIT_MACROS)
+#   define __STDC_LIMIT_MACROS
+#   define __STDC_LIMIT_MACROS_DEFINED_BY_CLANG
+#  endif
+#  if !defined(__STDC_CONSTANT_MACROS)
+#   define __STDC_CONSTANT_MACROS
+#   define __STDC_CONSTANT_MACROS_DEFINED_BY_CLANG
+#  endif
+# endif
+
 # include_next <stdint.h>
+
+# ifdef __STDC_LIMIT_MACROS_DEFINED_BY_CLANG
+#  undef __STDC_LIMIT_MACROS
+#  undef __STDC_LIMIT_MACROS_DEFINED_BY_CLANG
+# endif
+# ifdef __STDC_CONSTANT_MACROS_DEFINED_BY_CLANG
+#  undef __STDC_CONSTANT_MACROS
+#  undef __STDC_CONSTANT_MACROS_DEFINED_BY_CLANG
+# endif
+
 #else
 
 /* C99 7.18.1.1 Exact-width integer types.
@@ -626,6 +667,12 @@
 #define PTRDIFF_MAX  __INTN_MAX(__PTRDIFF_WIDTH__)
 #define    SIZE_MAX __UINTN_MAX(__SIZE_WIDTH__)
 
+/* ISO9899:2011 7.20 (C11 Annex K): Define RSIZE_MAX if __STDC_WANT_LIB_EXT1__
+ * is enabled. */
+#if defined(__STDC_WANT_LIB_EXT1__) && __STDC_WANT_LIB_EXT1__ >= 1
+#define   RSIZE_MAX            (SIZE_MAX >> 1)
+#endif
+
 /* C99 7.18.2.5 Limits of greatest-width integer types. */
 #define INTMAX_MIN   __INTN_MIN(__INTMAX_WIDTH__)
 #define INTMAX_MAX   __INTN_MAX(__INTMAX_WIDTH__)
diff --git a/renderscript/clang-include/tgmath.h b/renderscript/clang-include/tgmath.h
index 4fa1cf7..a48e267 100644
--- a/renderscript/clang-include/tgmath.h
+++ b/renderscript/clang-include/tgmath.h
@@ -1340,15 +1340,15 @@
 
 // creal
 
-static float _Complex
+static float
     _TG_ATTRS
     __tg_creal(float __x) {return __x;}
 
-static double _Complex
+static double
     _TG_ATTRS
     __tg_creal(double __x) {return __x;}
 
-static long double _Complex
+static long double
     _TG_ATTRS
     __tg_creal(long double __x) {return __x;}
 
diff --git a/renderscript/clang-include/x86intrin.h b/renderscript/clang-include/x86intrin.h
index 68ce106..94fbe2f 100644
--- a/renderscript/clang-include/x86intrin.h
+++ b/renderscript/clang-include/x86intrin.h
@@ -46,6 +46,14 @@
 #include <popcntintrin.h>
 #endif
 
+#ifdef __RDSEED__
+#include <rdseedintrin.h>
+#endif
+
+#ifdef __PRFCHW__
+#include <prfchwintrin.h>
+#endif
+
 #ifdef __SSE4A__
 #include <ammintrin.h>
 #endif
diff --git a/renderscript/clang-include/xmmintrin.h b/renderscript/clang-include/xmmintrin.h
index b3b23cb..c68d3ed 100644
--- a/renderscript/clang-include/xmmintrin.h
+++ b/renderscript/clang-include/xmmintrin.h
@@ -218,7 +218,9 @@
 static __inline__ __m128 __attribute__((__always_inline__, __nodebug__))
 _mm_cmpgt_ss(__m128 __a, __m128 __b)
 {
-  return (__m128)__builtin_ia32_cmpss(__b, __a, 1);
+  return (__m128)__builtin_shufflevector(__a,
+                                         __builtin_ia32_cmpss(__b, __a, 1),
+                                         4, 1, 2, 3);
 }
 
 static __inline__ __m128 __attribute__((__always_inline__, __nodebug__))
@@ -230,7 +232,9 @@
 static __inline__ __m128 __attribute__((__always_inline__, __nodebug__))
 _mm_cmpge_ss(__m128 __a, __m128 __b)
 {
-  return (__m128)__builtin_ia32_cmpss(__b, __a, 2);
+  return (__m128)__builtin_shufflevector(__a,
+                                         __builtin_ia32_cmpss(__b, __a, 2),
+                                         4, 1, 2, 3);
 }
 
 static __inline__ __m128 __attribute__((__always_inline__, __nodebug__))
@@ -278,7 +282,9 @@
 static __inline__ __m128 __attribute__((__always_inline__, __nodebug__))
 _mm_cmpngt_ss(__m128 __a, __m128 __b)
 {
-  return (__m128)__builtin_ia32_cmpss(__b, __a, 5);
+  return (__m128)__builtin_shufflevector(__a,
+                                         __builtin_ia32_cmpss(__b, __a, 5),
+                                         4, 1, 2, 3);
 }
 
 static __inline__ __m128 __attribute__((__always_inline__, __nodebug__))
@@ -290,7 +296,9 @@
 static __inline__ __m128 __attribute__((__always_inline__, __nodebug__))
 _mm_cmpnge_ss(__m128 __a, __m128 __b)
 {
-  return (__m128)__builtin_ia32_cmpss(__b, __a, 6);
+  return (__m128)__builtin_shufflevector(__a,
+                                         __builtin_ia32_cmpss(__b, __a, 6),
+                                         4, 1, 2, 3);
 }
 
 static __inline__ __m128 __attribute__((__always_inline__, __nodebug__))
diff --git a/renderscript/clang-include/xopintrin.h b/renderscript/clang-include/xopintrin.h
index d107be4..cc94ca0 100644
--- a/renderscript/clang-include/xopintrin.h
+++ b/renderscript/clang-include/xopintrin.h
@@ -1,4 +1,4 @@
-/*===---- xopintrin.h - FMA4 intrinsics ------------------------------------===
+/*===---- xopintrin.h - XOP intrinsics -------------------------------------===
  *
  * Permission is hereby granted, free of charge, to any person obtaining a copy
  * of this software and associated documentation files (the "Software"), to deal
@@ -22,7 +22,7 @@
  */
 
 #ifndef __X86INTRIN_H
-#error "Never use <fma4intrin.h> directly; include <x86intrin.h> instead."
+#error "Never use <xopintrin.h> directly; include <x86intrin.h> instead."
 #endif
 
 #ifndef __XOPINTRIN_H
@@ -342,6 +342,399 @@
   __m128i __B = (B); \
   (__m128i)__builtin_ia32_vpcomq((__v2di)__A, (__v2di)__B, (N)); })
 
+#define _MM_PCOMCTRL_LT    0
+#define _MM_PCOMCTRL_LE    1
+#define _MM_PCOMCTRL_GT    2
+#define _MM_PCOMCTRL_GE    3
+#define _MM_PCOMCTRL_EQ    4
+#define _MM_PCOMCTRL_NEQ   5
+#define _MM_PCOMCTRL_FALSE 6
+#define _MM_PCOMCTRL_TRUE  7
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comlt_epu8(__m128i __A, __m128i __B)
+{
+  return _mm_com_epu8(__A, __B, _MM_PCOMCTRL_LT);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comle_epu8(__m128i __A, __m128i __B)
+{
+  return _mm_com_epu8(__A, __B, _MM_PCOMCTRL_LE);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comgt_epu8(__m128i __A, __m128i __B)
+{
+  return _mm_com_epu8(__A, __B, _MM_PCOMCTRL_GT);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comge_epu8(__m128i __A, __m128i __B)
+{
+  return _mm_com_epu8(__A, __B, _MM_PCOMCTRL_GE);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comeq_epu8(__m128i __A, __m128i __B)
+{
+  return _mm_com_epu8(__A, __B, _MM_PCOMCTRL_EQ);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comneq_epu8(__m128i __A, __m128i __B)
+{
+  return _mm_com_epu8(__A, __B, _MM_PCOMCTRL_NEQ);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comfalse_epu8(__m128i __A, __m128i __B)
+{
+  return _mm_com_epu8(__A, __B, _MM_PCOMCTRL_FALSE);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comtrue_epu8(__m128i __A, __m128i __B)
+{
+  return _mm_com_epu8(__A, __B, _MM_PCOMCTRL_TRUE);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comlt_epu16(__m128i __A, __m128i __B)
+{
+  return _mm_com_epu16(__A, __B, _MM_PCOMCTRL_LT);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comle_epu16(__m128i __A, __m128i __B)
+{
+  return _mm_com_epu16(__A, __B, _MM_PCOMCTRL_LE);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comgt_epu16(__m128i __A, __m128i __B)
+{
+  return _mm_com_epu16(__A, __B, _MM_PCOMCTRL_GT);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comge_epu16(__m128i __A, __m128i __B)
+{
+  return _mm_com_epu16(__A, __B, _MM_PCOMCTRL_GE);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comeq_epu16(__m128i __A, __m128i __B)
+{
+  return _mm_com_epu16(__A, __B, _MM_PCOMCTRL_EQ);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comneq_epu16(__m128i __A, __m128i __B)
+{
+  return _mm_com_epu16(__A, __B, _MM_PCOMCTRL_NEQ);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comfalse_epu16(__m128i __A, __m128i __B)
+{
+  return _mm_com_epu16(__A, __B, _MM_PCOMCTRL_FALSE);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comtrue_epu16(__m128i __A, __m128i __B)
+{
+  return _mm_com_epu16(__A, __B, _MM_PCOMCTRL_TRUE);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comlt_epu32(__m128i __A, __m128i __B)
+{
+  return _mm_com_epu32(__A, __B, _MM_PCOMCTRL_LT);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comle_epu32(__m128i __A, __m128i __B)
+{
+  return _mm_com_epu32(__A, __B, _MM_PCOMCTRL_LE);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comgt_epu32(__m128i __A, __m128i __B)
+{
+  return _mm_com_epu32(__A, __B, _MM_PCOMCTRL_GT);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comge_epu32(__m128i __A, __m128i __B)
+{
+  return _mm_com_epu32(__A, __B, _MM_PCOMCTRL_GE);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comeq_epu32(__m128i __A, __m128i __B)
+{
+  return _mm_com_epu32(__A, __B, _MM_PCOMCTRL_EQ);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comneq_epu32(__m128i __A, __m128i __B)
+{
+  return _mm_com_epu32(__A, __B, _MM_PCOMCTRL_NEQ);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comfalse_epu32(__m128i __A, __m128i __B)
+{
+  return _mm_com_epu32(__A, __B, _MM_PCOMCTRL_FALSE);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comtrue_epu32(__m128i __A, __m128i __B)
+{
+  return _mm_com_epu32(__A, __B, _MM_PCOMCTRL_TRUE);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comlt_epu64(__m128i __A, __m128i __B)
+{
+  return _mm_com_epu64(__A, __B, _MM_PCOMCTRL_LT);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comle_epu64(__m128i __A, __m128i __B)
+{
+  return _mm_com_epu64(__A, __B, _MM_PCOMCTRL_LE);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comgt_epu64(__m128i __A, __m128i __B)
+{
+  return _mm_com_epu64(__A, __B, _MM_PCOMCTRL_GT);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comge_epu64(__m128i __A, __m128i __B)
+{
+  return _mm_com_epu64(__A, __B, _MM_PCOMCTRL_GE);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comeq_epu64(__m128i __A, __m128i __B)
+{
+  return _mm_com_epu64(__A, __B, _MM_PCOMCTRL_EQ);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comneq_epu64(__m128i __A, __m128i __B)
+{
+  return _mm_com_epu64(__A, __B, _MM_PCOMCTRL_NEQ);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comfalse_epu64(__m128i __A, __m128i __B)
+{
+  return _mm_com_epu64(__A, __B, _MM_PCOMCTRL_FALSE);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comtrue_epu64(__m128i __A, __m128i __B)
+{
+  return _mm_com_epu64(__A, __B, _MM_PCOMCTRL_TRUE);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comlt_epi8(__m128i __A, __m128i __B)
+{
+  return _mm_com_epi8(__A, __B, _MM_PCOMCTRL_LT);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comle_epi8(__m128i __A, __m128i __B)
+{
+  return _mm_com_epi8(__A, __B, _MM_PCOMCTRL_LE);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comgt_epi8(__m128i __A, __m128i __B)
+{
+  return _mm_com_epi8(__A, __B, _MM_PCOMCTRL_GT);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comge_epi8(__m128i __A, __m128i __B)
+{
+  return _mm_com_epi8(__A, __B, _MM_PCOMCTRL_GE);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comeq_epi8(__m128i __A, __m128i __B)
+{
+  return _mm_com_epi8(__A, __B, _MM_PCOMCTRL_EQ);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comneq_epi8(__m128i __A, __m128i __B)
+{
+  return _mm_com_epi8(__A, __B, _MM_PCOMCTRL_NEQ);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comfalse_epi8(__m128i __A, __m128i __B)
+{
+  return _mm_com_epi8(__A, __B, _MM_PCOMCTRL_FALSE);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comtrue_epi8(__m128i __A, __m128i __B)
+{
+  return _mm_com_epi8(__A, __B, _MM_PCOMCTRL_TRUE);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comlt_epi16(__m128i __A, __m128i __B)
+{
+  return _mm_com_epi16(__A, __B, _MM_PCOMCTRL_LT);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comle_epi16(__m128i __A, __m128i __B)
+{
+  return _mm_com_epi16(__A, __B, _MM_PCOMCTRL_LE);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comgt_epi16(__m128i __A, __m128i __B)
+{
+  return _mm_com_epi16(__A, __B, _MM_PCOMCTRL_GT);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comge_epi16(__m128i __A, __m128i __B)
+{
+  return _mm_com_epi16(__A, __B, _MM_PCOMCTRL_GE);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comeq_epi16(__m128i __A, __m128i __B)
+{
+  return _mm_com_epi16(__A, __B, _MM_PCOMCTRL_EQ);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comneq_epi16(__m128i __A, __m128i __B)
+{
+  return _mm_com_epi16(__A, __B, _MM_PCOMCTRL_NEQ);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comfalse_epi16(__m128i __A, __m128i __B)
+{
+  return _mm_com_epi16(__A, __B, _MM_PCOMCTRL_FALSE);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comtrue_epi16(__m128i __A, __m128i __B)
+{
+  return _mm_com_epi16(__A, __B, _MM_PCOMCTRL_TRUE);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comlt_epi32(__m128i __A, __m128i __B)
+{
+  return _mm_com_epi32(__A, __B, _MM_PCOMCTRL_LT);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comle_epi32(__m128i __A, __m128i __B)
+{
+  return _mm_com_epi32(__A, __B, _MM_PCOMCTRL_LE);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comgt_epi32(__m128i __A, __m128i __B)
+{
+  return _mm_com_epi32(__A, __B, _MM_PCOMCTRL_GT);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comge_epi32(__m128i __A, __m128i __B)
+{
+  return _mm_com_epi32(__A, __B, _MM_PCOMCTRL_GE);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comeq_epi32(__m128i __A, __m128i __B)
+{
+  return _mm_com_epi32(__A, __B, _MM_PCOMCTRL_EQ);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comneq_epi32(__m128i __A, __m128i __B)
+{
+  return _mm_com_epi32(__A, __B, _MM_PCOMCTRL_NEQ);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comfalse_epi32(__m128i __A, __m128i __B)
+{
+  return _mm_com_epi32(__A, __B, _MM_PCOMCTRL_FALSE);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comtrue_epi32(__m128i __A, __m128i __B)
+{
+  return _mm_com_epi32(__A, __B, _MM_PCOMCTRL_TRUE);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comlt_epi64(__m128i __A, __m128i __B)
+{
+  return _mm_com_epi64(__A, __B, _MM_PCOMCTRL_LT);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comle_epi64(__m128i __A, __m128i __B)
+{
+  return _mm_com_epi64(__A, __B, _MM_PCOMCTRL_LE);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comgt_epi64(__m128i __A, __m128i __B)
+{
+  return _mm_com_epi64(__A, __B, _MM_PCOMCTRL_GT);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comge_epi64(__m128i __A, __m128i __B)
+{
+  return _mm_com_epi64(__A, __B, _MM_PCOMCTRL_GE);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comeq_epi64(__m128i __A, __m128i __B)
+{
+  return _mm_com_epi64(__A, __B, _MM_PCOMCTRL_EQ);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comneq_epi64(__m128i __A, __m128i __B)
+{
+  return _mm_com_epi64(__A, __B, _MM_PCOMCTRL_NEQ);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comfalse_epi64(__m128i __A, __m128i __B)
+{
+  return _mm_com_epi64(__A, __B, _MM_PCOMCTRL_FALSE);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_comtrue_epi64(__m128i __A, __m128i __B)
+{
+  return _mm_com_epi64(__A, __B, _MM_PCOMCTRL_TRUE);
+}
+
 #define _mm_permute2_pd(X, Y, C, I) __extension__ ({ \
   __m128d __X = (X); \
   __m128d __Y = (Y); \