The change for PR9190 wasn't quite right. We need to avoid making the

transformation if we can't legally create a build vector of the correct
type. Check that we can make the transformation first, and add a TODO to
refactor this code with similar cases.

Fixes: PR9223 and rdar://9000350
llvm-svn: 125631

2 files changed, 53 insertions, 2 deletions

diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index f452d6a8420..e213d722eb7 100644
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -1532,8 +1532,18 @@ SDValue DAGCombiner::visitSUB(SDNode *N) {
   }
 
   // fold (sub x, x) -> 0
-  if (N0 == N1)
-    return DAG.getConstant(0, N->getValueType(0), LegalTypes);
+  // FIXME: Refactor this and xor and other similar operations together.
+  if (N0 == N1) {
+    if (!VT.isVector()) {
+      return DAG.getConstant(0, VT);
+    } else if (!LegalOperations || TLI.isOperationLegal(ISD::BUILD_VECTOR, VT)){
+      // Produce a vector of zeros.
+      SDValue El = DAG.getConstant(0, VT.getVectorElementType());
+      std::vector<SDValue> Ops(VT.getVectorNumElements(), El);
+      return DAG.getNode(ISD::BUILD_VECTOR, N->getDebugLoc(), VT,
+                         &Ops[0], Ops.size());
+    }
+  }
   // fold (sub c1, c2) -> c1-c2
   if (N0C && N1C)
     return DAG.FoldConstantArithmetic(ISD::SUB, VT, N0C, N1C);
diff --git a/llvm/test/CodeGen/X86/legalize-sub-zero-2.ll b/llvm/test/CodeGen/X86/legalize-sub-zero-2.ll
new file mode 100644
index 00000000000..f02ca715aee
--- /dev/null
+++ b/llvm/test/CodeGen/X86/legalize-sub-zero-2.ll
@@ -0,0 +1,41 @@
+; RUN: llc < %s -mtriple=i386-apple-darwin
+
+define fastcc void @foo(i32 %type) nounwind optsize {
+entry:
+  switch i32 %type, label %bb26 [
+    i32 33634, label %bb11
+    i32 5121, label %bb27
+  ]
+
+bb11:                                             ; preds = %entry
+  br label %bb27
+
+bb26:                                             ; preds = %entry
+  unreachable
+
+bb27:                                             ; preds = %bb11, %entry
+  %srcpb.0 = phi i32 [ 1, %bb11 ], [ 0, %entry ]
+  br i1 undef, label %bb348, label %bb30.lr.ph
+
+bb30.lr.ph:                                       ; preds = %bb27
+  %.sum743 = shl i32 %srcpb.0, 1
+  %0 = mul i32 %srcpb.0, -2
+  %.sum745 = add i32 %.sum743, %0
+  br i1 undef, label %bb70, label %bb71
+
+bb70:                                             ; preds = %bb30.lr.ph
+  unreachable
+
+bb71:                                             ; preds = %bb30.lr.ph
+  br i1 undef, label %bb92, label %bb80
+
+bb80:                                             ; preds = %bb71
+  unreachable
+
+bb92:                                             ; preds = %bb71
+  %1 = getelementptr inbounds i8* undef, i32 %.sum745
+  unreachable
+
+bb348:                                            ; preds = %bb27
+  ret void
+}