Fix PR10656. It's only profitable to use 128-bit inserts and extracts

when AVX mode is one. Otherwise is just more work for the type
legalizer.

llvm-svn: 137661

2 files changed, 26 insertions, 4 deletions

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 1ac5c2a8623..156e7aea31b 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -11737,7 +11737,8 @@ static SDValue PerformShuffleCombine256(SDNode *N, SelectionDAG &DAG,
 
 /// PerformShuffleCombine - Performs several different shuffle combines.
 static SDValue PerformShuffleCombine(SDNode *N, SelectionDAG &DAG,
-                                     TargetLowering::DAGCombinerInfo &DCI) {
+                                     TargetLowering::DAGCombinerInfo &DCI,
+                                     const X86Subtarget *Subtarget) {
   DebugLoc dl = N->getDebugLoc();
   EVT VT = N->getValueType(0);
 
@@ -11746,8 +11747,9 @@ static SDValue PerformShuffleCombine(SDNode *N, SelectionDAG &DAG,
   if (!DCI.isBeforeLegalize() && !TLI.isTypeLegal(VT.getVectorElementType()))
     return SDValue();
 
-  // Only handle pure VECTOR_SHUFFLE nodes.
-  if (VT.getSizeInBits() == 256 && N->getOpcode() == ISD::VECTOR_SHUFFLE)
+  // Combine 256-bit vector shuffles. This is only profitable when in AVX mode
+  if (Subtarget->hasAVX() && VT.getSizeInBits() == 256 &&
+      N->getOpcode() == ISD::VECTOR_SHUFFLE)
     return PerformShuffleCombine256(N, DAG, DCI);
 
   // Only handle 128 wide vector from here on.
@@ -13220,7 +13222,7 @@ SDValue X86TargetLowering::PerformDAGCombine(SDNode *N,
   case X86ISD::VPERMILPD:
   case X86ISD::VPERMILPDY:
   case X86ISD::VPERM2F128:
-  case ISD::VECTOR_SHUFFLE: return PerformShuffleCombine(N, DAG, DCI);
+  case ISD::VECTOR_SHUFFLE: return PerformShuffleCombine(N, DAG, DCI,Subtarget);
   }
 
   return SDValue();
diff --git a/llvm/test/CodeGen/X86/avx-vinsertf128.ll b/llvm/test/CodeGen/X86/avx-vinsertf128.ll
index b54b57b8d15..a6f258513a5 100644
--- a/llvm/test/CodeGen/X86/avx-vinsertf128.ll
+++ b/llvm/test/CodeGen/X86/avx-vinsertf128.ll
@@ -1,4 +1,5 @@
 ; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=corei7-avx -mattr=+avx | FileCheck %s
+; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=corei7-avx -mattr=+avx | FileCheck -check-prefix=CHECK-SSE %s
 
 ; CHECK-NOT: vunpck
 ; CHECK: vinsertf128 $1
@@ -16,3 +17,22 @@ entry:
   ret <4 x double> %shuffle
 }
 
+declare <2 x double> @llvm.x86.sse2.min.pd(<2 x double>, <2 x double>) nounwind readnone
+
+declare <2 x double> @llvm.x86.sse2.min.sd(<2 x double>, <2 x double>) nounwind readnone
+
+; Just check that no crash happens
+; CHECK-SSE: _insert_crash
+define void @insert_crash() nounwind {
+allocas:
+  %v1.i.i451 = shufflevector <4 x double> zeroinitializer, <4 x double> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
+  %ret_0a.i.i.i452 = shufflevector <4 x double> %v1.i.i451, <4 x double> undef, <2 x i32> <i32 0, i32 1>
+  %vret_0.i.i.i454 = tail call <2 x double> @llvm.x86.sse2.min.pd(<2 x double> %ret_0a.i.i.i452, <2 x double> undef) nounwind
+  %ret_val.i.i.i463 = tail call <2 x double> @llvm.x86.sse2.min.sd(<2 x double> %vret_0.i.i.i454, <2 x double> undef) nounwind
+  %ret.i1.i.i464 = extractelement <2 x double> %ret_val.i.i.i463, i32 0
+  %double2float = fptrunc double %ret.i1.i.i464 to float
+  %smearinsert50 = insertelement <4 x float> undef, float %double2float, i32 3
+  %blendAsInt.i503 = bitcast <4 x float> %smearinsert50 to <4 x i32>
+  store <4 x i32> %blendAsInt.i503, <4 x i32>* undef, align 4
+  ret void
+}