Remove 256-bit unpack built-ins and make the AVX intrinsic call llvm __builtin_shufflevector with the appropriate arguments

llvm-svn: 110768

3 files changed, 4 insertions, 12 deletions

diff --git a/clang/include/clang/Basic/BuiltinsX86.def b/clang/include/clang/Basic/BuiltinsX86.def
index 71f25cabc31..3e54135ab09 100644
--- a/clang/include/clang/Basic/BuiltinsX86.def
+++ b/clang/include/clang/Basic/BuiltinsX86.def
@@ -377,10 +377,6 @@ BUILTIN(__builtin_ia32_rsqrtps_nr256, "V8fV8f", "")
 BUILTIN(__builtin_ia32_rcpps256, "V8fV8f", "")
 BUILTIN(__builtin_ia32_roundpd256, "V4dV4di", "")
 BUILTIN(__builtin_ia32_roundps256, "V8fV8fi", "")
-BUILTIN(__builtin_ia32_unpckhpd256, "V4dV4dV4d", "")
-BUILTIN(__builtin_ia32_unpcklpd256, "V4dV4dV4d", "")
-BUILTIN(__builtin_ia32_unpckhps256, "V8fV8fV8f", "")
-BUILTIN(__builtin_ia32_unpcklps256, "V8fV8fV8f", "")
 BUILTIN(__builtin_ia32_si256_si, "V8iV4i", "")
 BUILTIN(__builtin_ia32_ps256_ps, "V8fV4f", "")
 BUILTIN(__builtin_ia32_pd256_pd, "V4dV2d", "")
diff --git a/clang/lib/Headers/avxintrin.h b/clang/lib/Headers/avxintrin.h
index 7beb5ba1c18..6c1e32d2405 100644
--- a/clang/lib/Headers/avxintrin.h
+++ b/clang/lib/Headers/avxintrin.h
@@ -598,25 +598,25 @@ _mm256_movedup_pd(__m256d a)
 static __inline __m256d __attribute__((__always_inline__, __nodebug__))
 _mm256_unpackhi_pd(__m256d a, __m256d b)
 {
-  return (__m256d)__builtin_ia32_unpckhpd256((__v4df)a, (__v4df)b);
+  return __builtin_shufflevector(a, b, 1, 5, 1+2, 5+2);
 }
 
 static __inline __m256d __attribute__((__always_inline__, __nodebug__))
 _mm256_unpacklo_pd(__m256d a, __m256d b)
 {
-  return (__m256d)__builtin_ia32_unpcklpd256((__v4df)a, (__v4df)b);
+  return __builtin_shufflevector(a, b, 0, 4, 0+2, 4+2);
 }
 
 static __inline __m256 __attribute__((__always_inline__, __nodebug__))
 _mm256_unpackhi_ps(__m256 a, __m256 b)
 {
-  return (__m256)__builtin_ia32_unpckhps256((__v8sf)a, (__v8sf)b);
+  return __builtin_shufflevector(a, b, 2, 10, 2+1, 10+1, 6, 14, 6+1, 14+1);
 }
 
 static __inline __m256 __attribute__((__always_inline__, __nodebug__))
 _mm256_unpacklo_ps(__m256 a, __m256 b)
 {
-  return (__m256)__builtin_ia32_unpcklps256((__v8sf)a, (__v8sf)b);
+  return __builtin_shufflevector(a, b, 0, 8, 0+1, 8+1, 4, 12, 4+1, 12+1);
 }
 
 /* Bit Test */
diff --git a/clang/test/CodeGen/builtins-x86.c b/clang/test/CodeGen/builtins-x86.c
index 0e0a253755b..b08fc5f9ace 100644
--- a/clang/test/CodeGen/builtins-x86.c
+++ b/clang/test/CodeGen/builtins-x86.c
@@ -440,10 +440,6 @@ void f0() {
   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_unpckhpd256(tmp_V4d, tmp_V4d);
-  tmp_V4d = __builtin_ia32_unpcklpd256(tmp_V4d, tmp_V4d);
-  tmp_V8f = __builtin_ia32_unpckhps256(tmp_V8f, tmp_V8f);
-  tmp_V8f = __builtin_ia32_unpcklps256(tmp_V8f, tmp_V8f);
   tmp_V8i = __builtin_ia32_si256_si(tmp_V4i);
   tmp_V8f = __builtin_ia32_ps256_ps(tmp_V4f);
   tmp_V4d = __builtin_ia32_pd256_pd(tmp_V2d);