Improve generated code for extending loads and some trunc stores on ARM.

Teach TargetSelectionDAG about lengthening loads for vector types and set v4i8 as legal. Allow FP_TO_UINT for v4i16 from v4i32.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@150956 91177308-0d34-0410-b5e6-96231b3b80d8

2 files changed, 140 insertions, 7 deletions

diff --git a/lib/Target/ARM/ARMISelLowering.cpp b/lib/Target/ARM/ARMISelLowering.cpp
index de10afabf3..40e1e2248e 100644
--- a/lib/Target/ARM/ARMISelLowering.cpp
+++ b/lib/Target/ARM/ARMISelLowering.cpp
@@ -533,9 +533,13 @@ ARMTargetLowering::ARMTargetLowering(TargetMachine &TM)
     setOperationAction(ISD::SETCC, MVT::v1i64, Expand);
     setOperationAction(ISD::SETCC, MVT::v2i64, Expand);
     // Neon does not have single instruction SINT_TO_FP and UINT_TO_FP with
-    // a destination type that is wider than the source.
+    // a destination type that is wider than the source, and nor does
+    // it have a FP_TO_[SU]INT instruction with a narrower destination than
+    // source.
     setOperationAction(ISD::SINT_TO_FP, MVT::v4i16, Custom);
     setOperationAction(ISD::UINT_TO_FP, MVT::v4i16, Custom);
+    setOperationAction(ISD::FP_TO_UINT, MVT::v4i16, Custom);
+    setOperationAction(ISD::FP_TO_SINT, MVT::v4i16, Custom);
 
     setTargetDAGCombine(ISD::INTRINSIC_VOID);
     setTargetDAGCombine(ISD::INTRINSIC_W_CHAIN);
@@ -555,7 +559,15 @@ ARMTargetLowering::ARMTargetLowering(TargetMachine &TM)
     setTargetDAGCombine(ISD::FP_TO_UINT);
     setTargetDAGCombine(ISD::FDIV);
 
-    setLoadExtAction(ISD::EXTLOAD, MVT::v4i8, Expand);
+    // It is legal to extload from v4i8 to v4i16 or v4i32.
+    MVT Tys[6] = {MVT::v8i8, MVT::v4i8, MVT::v2i8,
+                  MVT::v4i16, MVT::v2i16,
+                  MVT::v2i32};
+    for (unsigned i = 0; i < 6; ++i) {
+      setLoadExtAction(ISD::EXTLOAD, Tys[i], Legal);
+      setLoadExtAction(ISD::ZEXTLOAD, Tys[i], Legal);
+      setLoadExtAction(ISD::SEXTLOAD, Tys[i], Legal);
+    }
   }
 
   computeRegisterProperties();
@@ -3058,12 +3070,22 @@ SDValue ARMTargetLowering::LowerBR_JT(SDValue Op, SelectionDAG &DAG) const {
 }
 
 static SDValue LowerVectorFP_TO_INT(SDValue Op, SelectionDAG &DAG) {
-  assert(Op.getValueType().getVectorElementType() == MVT::i32 
-         && "Unexpected custom lowering");
+  EVT VT = Op.getValueType();
+  DebugLoc dl = Op.getDebugLoc();
 
-  if (Op.getOperand(0).getValueType().getVectorElementType() == MVT::f32)
-    return Op;
-  return DAG.UnrollVectorOp(Op.getNode());
+  if (Op.getValueType().getVectorElementType() == MVT::i32) {
+    if (Op.getOperand(0).getValueType().getVectorElementType() == MVT::f32)
+      return Op;
+    return DAG.UnrollVectorOp(Op.getNode());
+  }
+
+  assert(Op.getOperand(0).getValueType() == MVT::v4f32 &&
+         "Invalid type for custom lowering!");
+  if (VT != MVT::v4i16)
+    return DAG.UnrollVectorOp(Op.getNode());
+
+  Op = DAG.getNode(Op.getOpcode(), dl, MVT::v4i32, Op.getOperand(0));
+  return DAG.getNode(ISD::TRUNCATE, dl, VT, Op);
 }
 
 static SDValue LowerFP_TO_INT(SDValue Op, SelectionDAG &DAG) {
diff --git a/lib/Target/ARM/ARMInstrNEON.td b/lib/Target/ARM/ARMInstrNEON.td
index 154fb25cec..76aefe66ac 100644
--- a/lib/Target/ARM/ARMInstrNEON.td
+++ b/lib/Target/ARM/ARMInstrNEON.td
@@ -5624,6 +5624,117 @@ def : Pat<(v2f64 (bitconvert (v8i16 QPR:$src))), (v2f64 QPR:$src)>;
 def : Pat<(v2f64 (bitconvert (v16i8 QPR:$src))), (v2f64 QPR:$src)>;
 def : Pat<(v2f64 (bitconvert (v4f32 QPR:$src))), (v2f64 QPR:$src)>;
 
+// Vector lengthening move with load, matching extending loads.
+
+// extload, zextload and sextload for a standard lengthening load. Example:
+// Lengthen_Single<"8", "i16", "i8"> = Pat<(v8i16 (extloadvi8 addrmode5:$addr))
+//                                         (VMOVLuv8i16 (VLDRD addrmode5:$addr))>;
+multiclass Lengthen_Single<string DestLanes, string DestTy, string SrcTy> {
+  def _Any : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
+                    (!cast<PatFrag>("extloadv" # SrcTy) addrmode5:$addr)),
+                  (!cast<Instruction>("VMOVLuv" # DestLanes # DestTy)
+                    (VLDRD addrmode5:$addr))>;
+  def _Z : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
+                  (!cast<PatFrag>("zextloadv" # SrcTy) addrmode5:$addr)),
+                (!cast<Instruction>("VMOVLuv" # DestLanes # DestTy)
+                  (VLDRD addrmode5:$addr))>;
+  def _S : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
+                  (!cast<PatFrag>("sextloadv" # SrcTy) addrmode5:$addr)),
+                (!cast<Instruction>("VMOVLsv" # DestLanes # DestTy)
+                  (VLDRD addrmode5:$addr))>;
+}
+
+// extload, zextload and sextload for a lengthening load which only uses
+// half the lanes available. Example:
+// Lengthen_HalfSingle<"4", "i16", "8", "i16", "i8"> =
+//     Pat<(v4i16 (extloadvi8 addrmode5:$addr))
+//         (EXTRACT_SUBREG (VMOVLuv8i16 (INSERT_SUBREG (f64 (IMPLICIT_DEF)),
+//                                                     (VLDRS addrmode5:$addr),
+//                                                     ssub_0)),
+//                         dsub_0)>;
+multiclass Lengthen_HalfSingle<string DestLanes, string DestTy, string SrcTy,
+                               string InsnLanes, string InsnTy> {
+  def _Any : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
+                   (!cast<PatFrag>("extloadv" # SrcTy) addrmode5:$addr)),
+       (EXTRACT_SUBREG (!cast<Instruction>("VMOVLuv" # InsnLanes # InsnTy)
+         (INSERT_SUBREG (f64 (IMPLICIT_DEF)), (VLDRS addrmode5:$addr), ssub_0)),
+         dsub_0)>;
+  def _Z   : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
+                   (!cast<PatFrag>("zextloadv" # SrcTy) addrmode5:$addr)),
+       (EXTRACT_SUBREG (!cast<Instruction>("VMOVLuv" # InsnLanes # InsnTy)
+         (INSERT_SUBREG (f64 (IMPLICIT_DEF)), (VLDRS addrmode5:$addr), ssub_0)),
+         dsub_0)>;
+  def _S   : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
+                   (!cast<PatFrag>("sextloadv" # SrcTy) addrmode5:$addr)),
+       (EXTRACT_SUBREG (!cast<Instruction>("VMOVLsv" # InsnLanes # InsnTy)
+         (INSERT_SUBREG (f64 (IMPLICIT_DEF)), (VLDRS addrmode5:$addr), ssub_0)),
+         dsub_0)>;
+}
+
+// extload, zextload and sextload for a lengthening load followed by another
+// lengthening load, to quadruple the initial length.
+// Lengthen_Double<"4", "i32", "i8", "8", "i16", "4", "i32", qsub_0> =
+//     Pat<(v4i32 (extloadvi8 addrmode5:$addr))
+//         (EXTRACT_SUBREG (VMOVLuv4i32 
+//           (EXTRACT_SUBREG (VMOVLuv8i16 (INSERT_SUBREG (f64 (IMPLICIT_DEF)),
+//                                                       (VLDRS addrmode5:$addr),
+//                                                       ssub_0)),
+//                           dsub_0)),
+//           qsub_0)>;
+multiclass Lengthen_Double<string DestLanes, string DestTy, string SrcTy,
+                           string Insn1Lanes, string Insn1Ty, string Insn2Lanes,
+                           string Insn2Ty, SubRegIndex RegType> {
+  def _Any : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
+                   (!cast<PatFrag>("extloadv" # SrcTy) addrmode5:$addr)),
+         (EXTRACT_SUBREG (!cast<Instruction>("VMOVLuv" # Insn2Lanes # Insn2Ty)
+           (EXTRACT_SUBREG (!cast<Instruction>("VMOVLuv" # Insn1Lanes # Insn1Ty)
+             (INSERT_SUBREG (f64 (IMPLICIT_DEF)), (VLDRS addrmode5:$addr),
+              ssub_0)), dsub_0)),
+          RegType)>;
+  def _Z   : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
+                   (!cast<PatFrag>("zextloadv" # SrcTy) addrmode5:$addr)),
+         (EXTRACT_SUBREG (!cast<Instruction>("VMOVLuv" # Insn2Lanes # Insn2Ty)
+           (EXTRACT_SUBREG (!cast<Instruction>("VMOVLuv" # Insn1Lanes # Insn1Ty)
+             (INSERT_SUBREG (f64 (IMPLICIT_DEF)), (VLDRS addrmode5:$addr),
+              ssub_0)), dsub_0)),
+          RegType)>;
+  def _S   : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
+                   (!cast<PatFrag>("sextloadv" # SrcTy) addrmode5:$addr)),
+         (EXTRACT_SUBREG (!cast<Instruction>("VMOVLsv" # Insn2Lanes # Insn2Ty)
+           (EXTRACT_SUBREG (!cast<Instruction>("VMOVLsv" # Insn1Lanes # Insn1Ty)
+             (INSERT_SUBREG (f64 (IMPLICIT_DEF)), (VLDRS addrmode5:$addr),
+              ssub_0)), dsub_0)),
+          RegType)>;
+}
+
+defm : Lengthen_Single<"8", "i16", "i8">; // v8i8 -> v8i16
+defm : Lengthen_Single<"4", "i32", "i16">; // v4i16 -> v4i32
+defm : Lengthen_Single<"2", "i64", "i32">; // v2i32 -> v2i64
+
+defm : Lengthen_HalfSingle<"4", "i16", "i8", "8", "i16">; // v4i8 -> v4i16
+defm : Lengthen_HalfSingle<"2", "i16", "i8", "8", "i16">; // v2i8 -> v2i16
+defm : Lengthen_HalfSingle<"2", "i32", "i16", "4", "i32">; // v2i16 -> v2i32
+
+// Double lengthening - v4i8 -> v4i16 -> v4i32 
+defm : Lengthen_Double<"4", "i32", "i8", "8", "i16", "4", "i32", qsub_0>;
+// v2i8 -> v2i16 -> v2i32
+defm : Lengthen_Double<"2", "i32", "i8", "8", "i16", "4", "i32", dsub_0>;
+// v2i16 -> v2i32 -> v2i64
+defm : Lengthen_Double<"2", "i64", "i16", "4", "i32", "2", "i64", qsub_0>;
+
+// Triple lengthening - v2i8 -> v2i16 -> v2i32 -> v2i64
+def : Pat<(v2i64 (extloadvi8 addrmode5:$addr)),
+      (VMOVLuv2i64 (EXTRACT_SUBREG (VMOVLuv4i32 (EXTRACT_SUBREG (VMOVLuv8i16
+         (INSERT_SUBREG (f64 (IMPLICIT_DEF)), (VLDRS addrmode5:$addr), ssub_0)),
+         dsub_0)), dsub_0))>;
+def : Pat<(v2i64 (zextloadvi8 addrmode5:$addr)),
+      (VMOVLuv2i64 (EXTRACT_SUBREG (VMOVLuv4i32 (EXTRACT_SUBREG (VMOVLuv8i16
+         (INSERT_SUBREG (f64 (IMPLICIT_DEF)), (VLDRS addrmode5:$addr), ssub_0)),
+         dsub_0)), dsub_0))>;
+def : Pat<(v2i64 (sextloadvi8 addrmode5:$addr)),
+      (VMOVLsv2i64 (EXTRACT_SUBREG (VMOVLsv4i32 (EXTRACT_SUBREG (VMOVLsv8i16
+         (INSERT_SUBREG (f64 (IMPLICIT_DEF)), (VLDRS addrmode5:$addr), ssub_0)),
+         dsub_0)), dsub_0))>;
 
 //===----------------------------------------------------------------------===//
 // Assembler aliases