diff --git a/include/clang/Basic/BuiltinsX86.def b/include/clang/Basic/BuiltinsX86.def
index 1e1e8f6..8262b3e 100644
--- a/include/clang/Basic/BuiltinsX86.def
+++ b/include/clang/Basic/BuiltinsX86.def
@@ -425,9 +425,9 @@
 BUILTIN(__builtin_ia32_vperm2f128_pd256, "V4dV4dV4dIc", "")
 BUILTIN(__builtin_ia32_vperm2f128_ps256, "V8fV8fV8fIc", "")
 BUILTIN(__builtin_ia32_vperm2f128_si256, "V8iV8iV8iIc", "")
-BUILTIN(__builtin_ia32_vpermilpd, "V2dV2dc", "")
+BUILTIN(__builtin_ia32_vpermilpd, "V2dV2dIc", "")
 BUILTIN(__builtin_ia32_vpermilps, "V4fV4fc", "")
-BUILTIN(__builtin_ia32_vpermilpd256, "V4dV4dc", "")
+BUILTIN(__builtin_ia32_vpermilpd256, "V4dV4dIc", "")
 BUILTIN(__builtin_ia32_vpermilps256, "V8fV8fc", "")
 BUILTIN(__builtin_ia32_vinsertf128_pd256, "V4dV4dV2dIc", "")
 BUILTIN(__builtin_ia32_vinsertf128_ps256, "V8fV8fV4fIc", "")
@@ -436,8 +436,8 @@
 BUILTIN(__builtin_ia32_sqrtps256, "V8fV8f", "")
 BUILTIN(__builtin_ia32_rsqrtps256, "V8fV8f", "")
 BUILTIN(__builtin_ia32_rcpps256, "V8fV8f", "")
-BUILTIN(__builtin_ia32_roundpd256, "V4dV4di", "")
-BUILTIN(__builtin_ia32_roundps256, "V8fV8fi", "")
+BUILTIN(__builtin_ia32_roundpd256, "V4dV4dIi", "")
+BUILTIN(__builtin_ia32_roundps256, "V8fV8fIi", "")
 BUILTIN(__builtin_ia32_vtestzpd, "iV2dV2d", "")
 BUILTIN(__builtin_ia32_vtestcpd, "iV2dV2d", "")
 BUILTIN(__builtin_ia32_vtestnzcpd, "iV2dV2d", "")
diff --git a/lib/Headers/avxintrin.h b/lib/Headers/avxintrin.h
index 4c7c4ce..ea26fa5 100644
--- a/lib/Headers/avxintrin.h
+++ b/lib/Headers/avxintrin.h
@@ -145,17 +145,13 @@
   return (__m256)__builtin_ia32_rcpps256((__v8sf)a);
 }
 
-static __inline __m256d __attribute__((__always_inline__, __nodebug__))
-_mm256_round_pd(__m256d v, const int m)
-{
-  return (__m256d)__builtin_ia32_roundpd256((__v4df)v, m);
-}
+#define _mm256_round_pd(V, M) __extension__ ({ \
+    __m256d __V = (V); \
+    (__m256d)__builtin_ia32_roundpd256((__v4df)__V, M); })
 
-static __inline __m256 __attribute__((__always_inline__, __nodebug__))
-_mm256_round_ps(__m256 v, const int m)
-{
-  return (__m256)__builtin_ia32_roundps256((__v8sf)v, m);
-}
+#define _mm256_round_ps(V, M) __extension__ ({ \
+  __m256 __V = (V); \
+  (__m256)__builtin_ia32_roundps256((__v8sf)__V, M); })
 
 #define _mm256_ceil_pd(V)  _mm256_round_pd((V), _MM_FROUND_CEIL)
 #define _mm256_floor_pd(V) _mm256_round_pd((V), _MM_FROUND_FLOOR)
@@ -262,17 +258,13 @@
 						  (__v8si)c);
 }
 
-static __inline __m128d __attribute__((__always_inline__, __nodebug__))
-_mm_permute_pd(__m128d a, const int c)
-{
-  return (__m128d)__builtin_ia32_vpermilpd((__v2df)a, c);
-}
+#define _mm_permute_pd(A, C) __extension__ ({ \
+  __m128d __A = (A); \
+  (__m128d)__builtin_ia32_vpermilpd((__v2df)__A, C); })
 
-static __inline __m256d __attribute__((__always_inline__, __nodebug__))
-_mm256_permute_pd(__m256d a, const int c)
-{
-  return (__m256d)__builtin_ia32_vpermilpd256((__v4df)a, c);
-}
+#define _mm256_permute_pd(A, C) __extension__ ({ \
+  __m256d __A = (A); \
+  (__m256d)__builtin_ia32_vpermilpd256((__v4df)__A, C); })
 
 static __inline __m128 __attribute__((__always_inline__, __nodebug__))
 _mm_permute_ps(__m128 a, const int c)
diff --git a/test/CodeGen/builtins-x86.c b/test/CodeGen/builtins-x86.c
index 728ade3..7f028e5 100644
--- a/test/CodeGen/builtins-x86.c
+++ b/test/CodeGen/builtins-x86.c
@@ -434,8 +434,8 @@
   tmp_V8f = __builtin_ia32_sqrtps256(tmp_V8f);
   tmp_V8f = __builtin_ia32_rsqrtps256(tmp_V8f);
   tmp_V8f = __builtin_ia32_rcpps256(tmp_V8f);
-  tmp_V4d = __builtin_ia32_roundpd256(tmp_V4d, tmp_i);
-  tmp_V8f = __builtin_ia32_roundps256(tmp_V8f, tmp_i);
+  tmp_V4d = __builtin_ia32_roundpd256(tmp_V4d, 0x1);
+  tmp_V8f = __builtin_ia32_roundps256(tmp_V8f, 0x1);
   tmp_i = __builtin_ia32_vtestzpd(tmp_V2d, tmp_V2d);
   tmp_i = __builtin_ia32_vtestcpd(tmp_V2d, tmp_V2d);
   tmp_i = __builtin_ia32_vtestnzcpd(tmp_V2d, tmp_V2d);