diff --git a/include/clang/Basic/BuiltinsX86.def b/include/clang/Basic/BuiltinsX86.def
index dd0f3be..7e592bc 100644
--- a/include/clang/Basic/BuiltinsX86.def
+++ b/include/clang/Basic/BuiltinsX86.def
@@ -414,10 +414,6 @@
 BUILTIN(__builtin_ia32_vperm2f128_pd256, "V4dV4dV4dIc", "")
 BUILTIN(__builtin_ia32_vperm2f128_ps256, "V8fV8fV8fIc", "")
 BUILTIN(__builtin_ia32_vperm2f128_si256, "V8iV8iV8iIc", "")
-BUILTIN(__builtin_ia32_vpermilpd, "V2dV2dIc", "")
-BUILTIN(__builtin_ia32_vpermilps, "V4fV4fIc", "")
-BUILTIN(__builtin_ia32_vpermilpd256, "V4dV4dIc", "")
-BUILTIN(__builtin_ia32_vpermilps256, "V8fV8fIc", "")
 BUILTIN(__builtin_ia32_vinsertf128_pd256, "V4dV4dV2dIc", "")
 BUILTIN(__builtin_ia32_vinsertf128_ps256, "V8fV8fV4fIc", "")
 BUILTIN(__builtin_ia32_vinsertf128_si256, "V8iV8iV4iIc", "")
diff --git a/lib/Headers/avxintrin.h b/lib/Headers/avxintrin.h
index ce4b226..0758a65 100644
--- a/lib/Headers/avxintrin.h
+++ b/lib/Headers/avxintrin.h
@@ -260,19 +260,31 @@
 
 #define _mm_permute_pd(A, C) __extension__ ({ \
   __m128d __A = (A); \
-  (__m128d)__builtin_ia32_vpermilpd((__v2df)__A, (C)); })
+  (__m128d)__builtin_shufflevector((__v2df)__A, (__v2df) _mm_setzero_pd(), \
+                                   (C) & 0x1, ((C) & 0x2) >> 1); })
 
 #define _mm256_permute_pd(A, C) __extension__ ({ \
   __m256d __A = (A); \
-  (__m256d)__builtin_ia32_vpermilpd256((__v4df)__A, (C)); })
+  (__m256d)__builtin_shufflevector((__v4df)__A, (__v4df) _mm256_setzero_pd(), \
+                                   (C) & 0x1, ((C) & 0x2) >> 1, \
+                                   2 + (((C) & 0x4) >> 2), \
+                                   2 + (((C) & 0x8) >> 3)); })
 
 #define _mm_permute_ps(A, C) __extension__ ({ \
   __m128 __A = (A); \
-  (__m128)__builtin_ia32_vpermilps((__v4sf)__A, (C)); })
+  (__m128)__builtin_shufflevector((__v4sf)__A, (__v4sf) _mm_setzero_ps(), \
+                                   (C) & 0x3, ((C) & 0xc) >> 2, \
+                                   ((C) & 0x30) >> 4, ((C) & 0xc0) >> 8); })
 
 #define _mm256_permute_ps(A, C) __extension__ ({ \
   __m256 __A = (A); \
-  (__m256)__builtin_ia32_vpermilps256((__v8sf)__A, (C)); })
+  (__m256)__builtin_shufflevector((__v8sf)__A, (__v8sf) _mm256_setzero_ps(), \
+                                  (C) & 0x3, ((C) & 0xc) >> 2, \
+                                  ((C) & 0x30) >> 4, ((C) & 0xc0) >> 6, \
+                                  4 + (((C) & 0x03) >> 0), \
+                                  4 + (((C) & 0x0c) >> 2), \
+                                  4 + (((C) & 0x30) >> 4), \
+                                  4 + (((C) & 0xc0) >> 6)); })
 
 #define _mm256_permute2f128_pd(V1, V2, M) __extension__ ({ \
   __m256d __V1 = (V1); \
diff --git a/test/CodeGen/avx-shuffle-builtins.c b/test/CodeGen/avx-shuffle-builtins.c
index c11780a..6b2b1b1 100644
--- a/test/CodeGen/avx-shuffle-builtins.c
+++ b/test/CodeGen/avx-shuffle-builtins.c
@@ -14,3 +14,27 @@
   // CHECK: shufflevector{{.*}}<i32 3, i32 2, i32 8, i32 11, i32 7, i32 6, i32 12, i32 15>
   return _mm256_shuffle_ps(a, b, 203);
 }
+
+__m128d test_mm_permute_pd(__m128d a) {
+  // Check if the mask is correct
+  // CHECK: shufflevector{{.*}}<i32 1, i32 0>
+  return _mm_permute_pd(a, 1);
+}
+
+__m256d test_mm256_permute_pd(__m256d a) {
+  // Check if the mask is correct
+  // CHECK: shufflevector{{.*}}<i32 1, i32 0, i32 3, i32 2>
+  return _mm256_permute_pd(a, 5);
+}
+
+__m128 test_mm_permute_ps(__m128 a) {
+  // Check if the mask is correct
+  // CHECK: shufflevector{{.*}}<i32 3, i32 2, i32 1, i32 0>
+  return _mm_permute_ps(a, 0x1b);
+}
+
+__m256 test_mm256_permute_ps(__m256 a) {
+  // Check if the mask is correct
+  // CHECK: shufflevector{{.*}}<i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4>
+  return _mm256_permute_ps(a, 0x1b);
+}
diff --git a/test/CodeGen/builtins-x86.c b/test/CodeGen/builtins-x86.c
index a812152..acb5554 100644
--- a/test/CodeGen/builtins-x86.c
+++ b/test/CodeGen/builtins-x86.c
@@ -417,10 +417,6 @@
   tmp_V4d = __builtin_ia32_vperm2f128_pd256(tmp_V4d, tmp_V4d, 0x7);
   tmp_V8f = __builtin_ia32_vperm2f128_ps256(tmp_V8f, tmp_V8f, 0x7);
   tmp_V8i = __builtin_ia32_vperm2f128_si256(tmp_V8i, tmp_V8i, 0x7);
-  tmp_V2d = __builtin_ia32_vpermilpd(tmp_V2d, 0x7);
-  tmp_V4f = __builtin_ia32_vpermilps(tmp_V4f, 0x7);
-  tmp_V4d = __builtin_ia32_vpermilpd256(tmp_V4d, 0x7);
-  tmp_V8f = __builtin_ia32_vpermilps256(tmp_V8f, 0x7);
   tmp_V4d = __builtin_ia32_vinsertf128_pd256(tmp_V4d, tmp_V2d, 0x7);
   tmp_V8f = __builtin_ia32_vinsertf128_ps256(tmp_V8f, tmp_V4f, 0x7);
   tmp_V8i = __builtin_ia32_vinsertf128_si256(tmp_V8i, tmp_V4i, 0x7);