ARM: Improve codegen for generic vselect.

Fall back to by-element insert rather than building it up on the stack.

rdar://14351991

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@185846 91177308-0d34-0410-b5e6-96231b3b80d8

1 files changed, 18 insertions, 0 deletions

diff --git a/lib/Target/ARM/ARMISelLowering.cpp b/lib/Target/ARM/ARMISelLowering.cpp
index 991a703f81..8c4a3f13d1 100644
--- a/lib/Target/ARM/ARMISelLowering.cpp
+++ b/lib/Target/ARM/ARMISelLowering.cpp
@@ -4734,6 +4734,24 @@ SDValue ARMTargetLowering::LowerBUILD_VECTOR(SDValue Op, SelectionDAG &DAG,
     return DAG.getNode(ISD::BITCAST, dl, VT, Val);
   }
 
+  // If all else fails, just use a sequence of INSERT_VECTOR_ELT when we
+  // know the default expansion would otherwise fall back on something even
+  // worse. For a vector with one or two non-undef values, that's
+  // scalar_to_vector for the elements followed by a shuffle (provided the
+  // shuffle is valid for the target) and materialization element by element
+  // on the stack followed by a load for everything else.
+  if (!isConstant && !usesOnlyOneValue) {
+    SDValue Vec = DAG.getUNDEF(VT);
+    for (unsigned i = 0 ; i < NumElts; ++i) {
+      SDValue V = Op.getOperand(i);
+      if (V.getOpcode() == ISD::UNDEF)
+        continue;
+      SDValue LaneIdx = DAG.getConstant(i, MVT::i32);
+      Vec = DAG.getNode(ISD::INSERT_VECTOR_ELT, dl, VT, Vec, V, LaneIdx);
+    }
+    return Vec;
+  }
+
   return SDValue();
 }