diff options
Diffstat (limited to 'llvm/lib/Target/X86/X86ISelLowering.cpp')
| -rw-r--r-- | llvm/lib/Target/X86/X86ISelLowering.cpp | 18 |
1 files changed, 12 insertions, 6 deletions
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp index 81416341658..bffa406d84c 100644 --- a/llvm/lib/Target/X86/X86ISelLowering.cpp +++ b/llvm/lib/Target/X86/X86ISelLowering.cpp @@ -4278,14 +4278,20 @@ X86TargetLowering::LowerBUILD_VECTOR(SDValue Op, SelectionDAG &DAG) const { if (LD.getNode()) return LD; - // For SSE 4.1, use inserts into undef. + // For SSE 4.1, use insertps to put the high elements into the low element. if (getSubtarget()->hasSSE41()) { - V[0] = DAG.getUNDEF(VT); - for (unsigned i = 0; i < NumElems; ++i) - if (Op.getOperand(i).getOpcode() != ISD::UNDEF) - V[0] = DAG.getNode(ISD::INSERT_VECTOR_ELT, dl, VT, V[0], + SDValue Result; + if (Op.getOperand(0).getOpcode() != ISD::UNDEF) + Result = DAG.getNode(ISD::SCALAR_TO_VECTOR, dl, VT, Op.getOperand(0)); + else + Result = DAG.getUNDEF(VT); + + for (unsigned i = 1; i < NumElems; ++i) { + if (Op.getOperand(i).getOpcode() == ISD::UNDEF) continue; + Result = DAG.getNode(ISD::INSERT_VECTOR_ELT, dl, VT, Result, Op.getOperand(i), DAG.getIntPtrConstant(i)); - return V[0]; + } + return Result; } // Otherwise, expand into a number of unpckl*, start by extending each of |
