diff --git a/include/clang/Basic/BuiltinsX86.def b/include/clang/Basic/BuiltinsX86.def
index 7e592bc..9e1bb99 100644
--- a/include/clang/Basic/BuiltinsX86.def
+++ b/include/clang/Basic/BuiltinsX86.def
@@ -411,9 +411,6 @@
 BUILTIN(__builtin_ia32_cvttpd2dq256, "V4iV4d", "")
 BUILTIN(__builtin_ia32_cvtpd2dq256, "V4iV4d", "")
 BUILTIN(__builtin_ia32_cvttps2dq256, "V8iV8f", "")
-BUILTIN(__builtin_ia32_vperm2f128_pd256, "V4dV4dV4dIc", "")
-BUILTIN(__builtin_ia32_vperm2f128_ps256, "V8fV8fV8fIc", "")
-BUILTIN(__builtin_ia32_vperm2f128_si256, "V8iV8iV8iIc", "")
 BUILTIN(__builtin_ia32_vinsertf128_pd256, "V4dV4dV2dIc", "")
 BUILTIN(__builtin_ia32_vinsertf128_ps256, "V8fV8fV4fIc", "")
 BUILTIN(__builtin_ia32_vinsertf128_si256, "V8iV8iV4iIc", "")
@@ -565,7 +562,6 @@
 BUILTIN(__builtin_ia32_permdf256, "V4dV4dIc", "")
 BUILTIN(__builtin_ia32_permvarsf256, "V8fV8fV8f", "")
 BUILTIN(__builtin_ia32_permdi256, "V4LLiV4LLiIc", "")
-BUILTIN(__builtin_ia32_permti256, "V4LLiV4LLiV4LLiIc", "")
 BUILTIN(__builtin_ia32_extract128i256, "V2LLiV4LLiIc", "")
 BUILTIN(__builtin_ia32_insert128i256, "V4LLiV4LLiV2LLiIc", "")
 BUILTIN(__builtin_ia32_maskloadd256, "V8iV8iC*V8i", "")
diff --git a/lib/Headers/avx2intrin.h b/lib/Headers/avx2intrin.h
index cc3ffe6..359c0aa 100644
--- a/lib/Headers/avx2intrin.h
+++ b/lib/Headers/avx2intrin.h
@@ -837,7 +837,11 @@
 #define _mm256_permute2x128_si256(V1, V2, M) __extension__ ({ \
   __m256i __V1 = (V1); \
   __m256i __V2 = (V2); \
-  (__m256i)__builtin_ia32_permti256(__V1, __V2, (M)); })
+  __builtin_shufflevector(__V1, __V2, \
+                          ((M) & 0x3) * 2, \
+                          ((M) & 0x3) * 2 + 1, \
+                          (((M) & 0x30) >> 4) * 2, \
+                          (((M) & 0x30) >> 4) * 2 + 1); })
 
 #define _mm256_extracti128_si256(A, O) __extension__ ({ \
   __m256i __A = (A); \
diff --git a/lib/Headers/avxintrin.h b/lib/Headers/avxintrin.h
index 0758a65..d8ecd1d 100644
--- a/lib/Headers/avxintrin.h
+++ b/lib/Headers/avxintrin.h
@@ -289,17 +289,37 @@
 #define _mm256_permute2f128_pd(V1, V2, M) __extension__ ({ \
   __m256d __V1 = (V1); \
   __m256d __V2 = (V2); \
-  (__m256d)__builtin_ia32_vperm2f128_pd256((__v4df)__V1, (__v4df)__V2, (M)); })
+  (__m256d)__builtin_shufflevector((__v4df)__V1, (__v4df)__V2, \
+                                   ((M) & 0x3) * 2, \
+                                   ((M) & 0x3) * 2 + 1, \
+                                   (((M) & 0x30) >> 4) * 2, \
+                                   (((M) & 0x30) >> 4) * 2 + 1); })
 
 #define _mm256_permute2f128_ps(V1, V2, M) __extension__ ({ \
   __m256 __V1 = (V1); \
   __m256 __V2 = (V2); \
-  (__m256)__builtin_ia32_vperm2f128_ps256((__v8sf)__V1, (__v8sf)__V2, (M)); })
+  (__m256)__builtin_shufflevector((__v8sf)__V1, (__v8sf)__V2, \
+                                  ((M) & 0x3) * 4, \
+                                  ((M) & 0x3) * 4 + 1, \
+                                  ((M) & 0x3) * 4 + 2, \
+                                  ((M) & 0x3) * 4 + 3, \
+                                  (((M) & 0x30) >> 4) * 4, \
+                                  (((M) & 0x30) >> 4) * 4 + 1, \
+                                  (((M) & 0x30) >> 4) * 4 + 2, \
+                                  (((M) & 0x30) >> 4) * 4 + 3); })
 
 #define _mm256_permute2f128_si256(V1, V2, M) __extension__ ({ \
   __m256i __V1 = (V1); \
   __m256i __V2 = (V2); \
-  (__m256i)__builtin_ia32_vperm2f128_si256((__v8si)__V1, (__v8si)__V2, (M)); })
+  (__m256i)__builtin_shufflevector((__v8si)__V1, (__v8si)__V2, \
+                                   ((M) & 0x3) * 4, \
+                                   ((M) & 0x3) * 4 + 1, \
+                                   ((M) & 0x3) * 4 + 2, \
+                                   ((M) & 0x3) * 4 + 3, \
+                                   (((M) & 0x30) >> 4) * 4, \
+                                   (((M) & 0x30) >> 4) * 4 + 1, \
+                                   (((M) & 0x30) >> 4) * 4 + 2, \
+                                   (((M) & 0x30) >> 4) * 4 + 3); })
 
 /* Vector Blend */
 #define _mm256_blend_pd(V1, V2, M) __extension__ ({ \
diff --git a/test/CodeGen/avx-shuffle-builtins.c b/test/CodeGen/avx-shuffle-builtins.c
index 6b2b1b1..18300bf 100644
--- a/test/CodeGen/avx-shuffle-builtins.c
+++ b/test/CodeGen/avx-shuffle-builtins.c
@@ -38,3 +38,21 @@
   // CHECK: shufflevector{{.*}}<i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4>
   return _mm256_permute_ps(a, 0x1b);
 }
+
+__m256d test_mm256_permute2f128_pd(__m256d a, __m256d b) {
+  // Check if the mask is correct
+  // CHECK: shufflevector{{.*}}<i32 2, i32 3, i32 6, i32 7>
+  return _mm256_permute2f128_pd(a, b, 0x31);
+}
+
+__m256 test_mm256_permute2f128_ps(__m256 a, __m256 b) {
+  // Check if the mask is correct
+  // CHECK: shufflevector{{.*}}<i32 12, i32 13, i32 14, i32 15, i32 4, i32 5, i32 6, i32 7>
+  return _mm256_permute2f128_ps(a, b, 0x13);
+}
+
+__m256i test_mm256_permute2f128_si256(__m256i a, __m256i b) {
+  // Check if the mask is correct
+  // CHECK: shufflevector{{.*}}<i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
+  return _mm256_permute2f128_si256(a, b, 0x20);
+}
diff --git a/test/CodeGen/avx2-builtins.c b/test/CodeGen/avx2-builtins.c
index ba0338e..a5ddc8e 100644
--- a/test/CodeGen/avx2-builtins.c
+++ b/test/CodeGen/avx2-builtins.c
@@ -677,8 +677,8 @@
 }
 
 __m256i test_mm256_permute2x128_si256(__m256i a, __m256i b) {
-  // CHECK: @llvm.x86.avx2.vperm2i128
-  return _mm256_permute2x128_si256(a, b, 94);
+  // CHECK: shufflevector{{.*}}<i32 2, i32 3, i32 6, i32 7>
+  return _mm256_permute2x128_si256(a, b, 0x31);
 }
 
 __m128i test_mm256_extracti128_si256(__m256i a) {
diff --git a/test/CodeGen/builtins-x86.c b/test/CodeGen/builtins-x86.c
index acb5554..30138d6 100644
--- a/test/CodeGen/builtins-x86.c
+++ b/test/CodeGen/builtins-x86.c
@@ -414,9 +414,6 @@
   tmp_V4i = __builtin_ia32_cvttpd2dq256(tmp_V4d);
   tmp_V4i = __builtin_ia32_cvtpd2dq256(tmp_V4d);
   tmp_V8i = __builtin_ia32_cvttps2dq256(tmp_V8f);
-  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_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);