Just use a native "load" instead of translating the builtin later. Clang can

take it!

I wasn't able to get __builtin_ia32_loaddqu to transform into an unaligned
load...I'll have to look into it further.

llvm-svn: 129427

3 files changed, 2 insertions, 4 deletions

diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index fce2bf40ee1..72a8ef8a5d5 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -2092,8 +2092,6 @@ Value *CodeGenFunction::EmitX86BuiltinExpr(unsigned BuiltinID,
     // If palignr is shifting the pair of vectors more than 32 bytes, emit zero.
     return llvm::Constant::getNullValue(ConvertType(E->getType()));
   }
-  case X86::BI__builtin_ia32_loadups:
-  case X86::BI__builtin_ia32_loadupd:
   case X86::BI__builtin_ia32_loaddqu: {
     const llvm::Type *VecTy = ConvertType(E->getType());
     const llvm::Type *IntTy = llvm::IntegerType::get(getLLVMContext(), 128);
diff --git a/clang/lib/Headers/emmintrin.h b/clang/lib/Headers/emmintrin.h
index 11b25817810..19d9f3f26b4 100644
--- a/clang/lib/Headers/emmintrin.h
+++ b/clang/lib/Headers/emmintrin.h
@@ -466,7 +466,7 @@ _mm_loadr_pd(double const *dp)
 static __inline__ __m128d __attribute__((__always_inline__, __nodebug__))
 _mm_loadu_pd(double const *dp)
 {
-  return __builtin_ia32_loadupd(dp);
+  return (__m128d){ dp[0], dp[1] };
 }
 
 static __inline__ __m128d __attribute__((__always_inline__, __nodebug__))
diff --git a/clang/lib/Headers/xmmintrin.h b/clang/lib/Headers/xmmintrin.h
index 42dd3e8d3b8..27bed04e712 100644
--- a/clang/lib/Headers/xmmintrin.h
+++ b/clang/lib/Headers/xmmintrin.h
@@ -539,7 +539,7 @@ _mm_load_ps(const float *p)
 static __inline__ __m128 __attribute__((__always_inline__, __nodebug__))
 _mm_loadu_ps(const float *p)
 {
-  return __builtin_ia32_loadups(p);
+  return (__m128){ p[0], p[1], p[2], p[3] }; /*  __builtin_ia32_loadups(p); */
 }
 
 static __inline__ __m128 __attribute__((__always_inline__, __nodebug__))