r231081 - Lower _mm256_broadcastsi128_si256 directly to a vector shuffle.

[Juergen Ributzka]

Author: ributzka
Date: Tue Mar  3 11:22:53 2015
New Revision: 231081

URL: http://llvm.org/viewvc/llvm-project?rev=231081&view=rev
Log:
Lower _mm256_broadcastsi128_si256 directly to a vector shuffle.

Originally we were using the same GCC builtins to lower this AVX2 vector
intrinsic. Instead we will now lower it directly to a vector shuffle.

This will not only allow LLVM to generate better code, but it will also allow us
to remove the GCC intrinsics.

Reviewed by Andrea

This is related to rdar://problem/18742778.

Modified:
    cfe/trunk/include/clang/Basic/BuiltinsX86.def
    cfe/trunk/lib/CodeGen/CGBuiltin.cpp
    cfe/trunk/lib/Headers/avx2intrin.h
    cfe/trunk/test/CodeGen/avx2-builtins.c

Modified: cfe/trunk/include/clang/Basic/BuiltinsX86.def
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/include/clang/Basic/BuiltinsX86.def?rev=231081&r1=231080&r2=231081&view=diff
==============================================================================
--- cfe/trunk/include/clang/Basic/BuiltinsX86.def (original)
+++ cfe/trunk/include/clang/Basic/BuiltinsX86.def Tue Mar  3 11:22:53 2015
@@ -582,7 +582,6 @@ BUILTIN(__builtin_ia32_movntdqa256, "V4L
 BUILTIN(__builtin_ia32_vbroadcastss_ps, "V4fV4f", "")
 BUILTIN(__builtin_ia32_vbroadcastss_ps256, "V8fV4f", "")
 BUILTIN(__builtin_ia32_vbroadcastsd_pd256, "V4dV2d", "")
-BUILTIN(__builtin_ia32_vbroadcastsi256, "V4LLiV2LLi", "")
 BUILTIN(__builtin_ia32_pbroadcastb256, "V32cV16c", "")
 BUILTIN(__builtin_ia32_pbroadcastw256, "V16sV8s", "")
 BUILTIN(__builtin_ia32_pbroadcastd256, "V8iV4i", "")

Modified: cfe/trunk/lib/CodeGen/CGBuiltin.cpp
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/CodeGen/CGBuiltin.cpp?rev=231081&r1=231080&r2=231081&view=diff
==============================================================================
--- cfe/trunk/lib/CodeGen/CGBuiltin.cpp (original)
+++ cfe/trunk/lib/CodeGen/CGBuiltin.cpp Tue Mar  3 11:22:53 2015
@@ -6090,13 +6090,6 @@ Value *CodeGenFunction::EmitX86BuiltinEx
     Builder.CreateStore(Builder.CreateExtractValue(Call, 0), Ops[0]);
     return Builder.CreateExtractValue(Call, 1);
   }
-  // AVX2 broadcast
-  case X86::BI__builtin_ia32_vbroadcastsi256: {
-    Value *VecTmp = CreateMemTemp(E->getArg(0)->getType());
-    Builder.CreateStore(Ops[0], VecTmp);
-    Value *F = CGM.getIntrinsic(Intrinsic::x86_avx2_vbroadcasti128);
-    return Builder.CreateCall(F, Builder.CreateBitCast(VecTmp, Int8PtrTy));
-  }
   // SSE comparison intrisics
   case X86::BI__builtin_ia32_cmpeqps:
   case X86::BI__builtin_ia32_cmpltps:

Modified: cfe/trunk/lib/Headers/avx2intrin.h
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Headers/avx2intrin.h?rev=231081&r1=231080&r2=231081&view=diff
==============================================================================
--- cfe/trunk/lib/Headers/avx2intrin.h (original)
+++ cfe/trunk/lib/Headers/avx2intrin.h Tue Mar  3 11:22:53 2015
@@ -771,7 +771,7 @@ _mm256_broadcastsd_pd(__m128d __X)
 static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
 _mm256_broadcastsi128_si256(__m128i __X)
 {
-  return (__m256i)__builtin_ia32_vbroadcastsi256(__X);
+  return (__m256i)__builtin_shufflevector(__X, __X, 0, 1, 0, 1);
 }
 
 #define _mm_blend_epi32(V1, V2, M) __extension__ ({ \

Modified: cfe/trunk/test/CodeGen/avx2-builtins.c
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/test/CodeGen/avx2-builtins.c?rev=231081&r1=231080&r2=231081&view=diff
==============================================================================
--- cfe/trunk/test/CodeGen/avx2-builtins.c (original)
+++ cfe/trunk/test/CodeGen/avx2-builtins.c Tue Mar  3 11:22:53 2015
@@ -612,7 +612,7 @@ __m256d test_mm256_broadcastsd_pd(__m128
 }
 
 __m256i test_mm256_broadcastsi128_si256(__m128i a) {
-  // CHECK: @llvm.x86.avx2.vbroadcasti128
+  // CHECK: shufflevector <2 x i64> %{{.*}}, <2 x i64> %{{.*}}, <4 x i32> <i32 0, i32 1, i32 0, i32 1>
   return _mm256_broadcastsi128_si256(a);
 }