diff options
| -rw-r--r-- | lib/Target/X86/X86ISelLowering.cpp | 12 | ||||
| -rw-r--r-- | test/CodeGen/X86/vec_insert-8.ll | 15 |
2 files changed, 25 insertions, 2 deletions
diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp index 9499a9d..e39036f 100644 --- a/lib/Target/X86/X86ISelLowering.cpp +++ b/lib/Target/X86/X86ISelLowering.cpp @@ -766,12 +766,12 @@ X86TargetLowering::X86TargetLowering(X86TargetMachine &TM) // information. setOperationAction(ISD::INSERT_VECTOR_ELT, MVT::v16i8, Custom); setOperationAction(ISD::INSERT_VECTOR_ELT, MVT::v8i16, Custom); - setOperationAction(ISD::INSERT_VECTOR_ELT, MVT::v4i32, Legal); + setOperationAction(ISD::INSERT_VECTOR_ELT, MVT::v4i32, Custom); setOperationAction(ISD::INSERT_VECTOR_ELT, MVT::v4f32, Custom); setOperationAction(ISD::EXTRACT_VECTOR_ELT, MVT::v16i8, Custom); setOperationAction(ISD::EXTRACT_VECTOR_ELT, MVT::v8i16, Custom); - setOperationAction(ISD::EXTRACT_VECTOR_ELT, MVT::v4i32, Legal); + setOperationAction(ISD::EXTRACT_VECTOR_ELT, MVT::v4i32, Custom); setOperationAction(ISD::EXTRACT_VECTOR_ELT, MVT::v4f32, Custom); if (Subtarget->is64Bit()) { @@ -4248,6 +4248,10 @@ X86TargetLowering::LowerEXTRACT_VECTOR_ELT_SSE4(SDValue Op, DAG.getNode(ISD::BIT_CONVERT, MVT::v4i32, Op.getOperand(0)), Op.getOperand(1)); return DAG.getNode(ISD::BIT_CONVERT, MVT::f32, Extract); + } else if (VT == MVT::i32) { + // ExtractPS works with constant index. + if (isa<ConstantSDNode>(Op.getOperand(1))) + return Op; } return SDValue(); } @@ -4362,6 +4366,10 @@ X86TargetLowering::LowerINSERT_VECTOR_ELT_SSE4(SDValue Op, SelectionDAG &DAG){ // combine either bitwise AND or insert of float 0.0 to set these bits. N2 = DAG.getIntPtrConstant(cast<ConstantSDNode>(N2)->getZExtValue() << 4); return DAG.getNode(X86ISD::INSERTPS, VT, N0, N1, N2); + } else if (EVT == MVT::i32) { + // InsertPS works with constant index. + if (isa<ConstantSDNode>(N2)) + return Op; } return SDValue(); } diff --git a/test/CodeGen/X86/vec_insert-8.ll b/test/CodeGen/X86/vec_insert-8.ll new file mode 100644 index 0000000..0f6924c --- /dev/null +++ b/test/CodeGen/X86/vec_insert-8.ll @@ -0,0 +1,15 @@ +; RUN: llvm-as < %s | llc -march=x86 -mattr=+sse41 -o %t -f + +; tests variable insert and extract of a 4 x i32 + +define <4 x i32> @var_insert(<4 x i32> %x, i32 %val, i32 %idx) nounwind { +entry: + %tmp3 = insertelement <4 x i32> %x, i32 %val, i32 %idx ; <<4 x i32>> [#uses=1] + ret <4 x i32> %tmp3 +} + +define i32 @var_extract(<4 x i32> %x, i32 %idx) nounwind { +entry: + %tmp3 = extractelement <4 x i32> %x, i32 %idx ; <<i32>> [#uses=1] + ret i32 %tmp3 +} |