X-Git-Url: http://demsky.eecs.uci.edu/git/?a=blobdiff_plain;f=lib%2FTarget%2FARM%2FARMInstrNEON.td;h=32b8d67b05ad4c527e53dc7c4a1c2ad3d8e4958c;hb=5b3fca50a08865f0db55fc92ad1c037a04e12177;hp=c540d0ddd70aa5208af11da812a151f9159de7a9;hpb=06a6a300c5f7100e4665667c689369e078d2ad59;p=oota-llvm.git

diff --git a/lib/Target/ARM/ARMInstrNEON.td b/lib/Target/ARM/ARMInstrNEON.td
index c540d0ddd70..32b8d67b05a 100644
--- a/lib/Target/ARM/ARMInstrNEON.td
+++ b/lib/Target/ARM/ARMInstrNEON.td
@@ -398,6 +398,41 @@ def VecListFourQWordIndexed : Operand<i32> {
   let MIOperandInfo = (ops DPR:$Vd, i32imm:$idx);
 }
 
+def dword_alignedload : PatFrag<(ops node:$ptr), (load node:$ptr), [{
+  return cast<LoadSDNode>(N)->getAlignment() >= 8;
+}]>;
+def dword_alignedstore : PatFrag<(ops node:$val, node:$ptr),
+                                 (store node:$val, node:$ptr), [{
+  return cast<StoreSDNode>(N)->getAlignment() >= 8;
+}]>;
+def word_alignedload : PatFrag<(ops node:$ptr), (load node:$ptr), [{
+  return cast<LoadSDNode>(N)->getAlignment() == 4;
+}]>;
+def word_alignedstore : PatFrag<(ops node:$val, node:$ptr),
+                                 (store node:$val, node:$ptr), [{
+  return cast<StoreSDNode>(N)->getAlignment() == 4;
+}]>;
+def hword_alignedload : PatFrag<(ops node:$ptr), (load node:$ptr), [{
+  return cast<LoadSDNode>(N)->getAlignment() == 2;
+}]>;
+def hword_alignedstore : PatFrag<(ops node:$val, node:$ptr),
+                                 (store node:$val, node:$ptr), [{
+  return cast<StoreSDNode>(N)->getAlignment() == 2;
+}]>;
+def byte_alignedload : PatFrag<(ops node:$ptr), (load node:$ptr), [{
+  return cast<LoadSDNode>(N)->getAlignment() == 1;
+}]>;
+def byte_alignedstore : PatFrag<(ops node:$val, node:$ptr),
+                             (store node:$val, node:$ptr), [{
+  return cast<StoreSDNode>(N)->getAlignment() == 1;
+}]>;
+def non_word_alignedload : PatFrag<(ops node:$ptr), (load node:$ptr), [{
+  return cast<LoadSDNode>(N)->getAlignment() < 4;
+}]>;
+def non_word_alignedstore : PatFrag<(ops node:$val, node:$ptr),
+                                    (store node:$val, node:$ptr), [{
+  return cast<StoreSDNode>(N)->getAlignment() < 4;
+}]>;
 
 //===----------------------------------------------------------------------===//
 // NEON-specific DAG Nodes.
@@ -591,7 +626,7 @@ class VLD1D<bits<4> op7_4, string Dt>
           "vld1", Dt, "$Vd, $Rn", "", []> {
   let Rm = 0b1111;
   let Inst{4} = Rn{4};
-  let DecoderMethod = "DecodeVLDInstruction";
+  let DecoderMethod = "DecodeVLDST1Instruction";
 }
 class VLD1Q<bits<4> op7_4, string Dt>
   : NLdSt<0,0b10,0b1010,op7_4, (outs VecListDPair:$Vd),
@@ -599,7 +634,7 @@ class VLD1Q<bits<4> op7_4, string Dt>
           "vld1", Dt, "$Vd, $Rn", "", []> {
   let Rm = 0b1111;
   let Inst{5-4} = Rn{5-4};
-  let DecoderMethod = "DecodeVLDInstruction";
+  let DecoderMethod = "DecodeVLDST1Instruction";
 }
 
 def  VLD1d8   : VLD1D<{0,0,0,?}, "8">;
@@ -620,7 +655,7 @@ multiclass VLD1DWB<bits<4> op7_4, string Dt> {
                      "$Rn.addr = $wb", []> {
     let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
     let Inst{4} = Rn{4};
-    let DecoderMethod = "DecodeVLDInstruction";
+    let DecoderMethod = "DecodeVLDST1Instruction";
     let AsmMatchConverter = "cvtVLDwbFixed";
   }
   def _register : NLdSt<0,0b10,0b0111,op7_4, (outs VecListOneD:$Vd, GPR:$wb),
@@ -628,7 +663,7 @@ multiclass VLD1DWB<bits<4> op7_4, string Dt> {
                         "vld1", Dt, "$Vd, $Rn, $Rm",
                         "$Rn.addr = $wb", []> {
     let Inst{4} = Rn{4};
-    let DecoderMethod = "DecodeVLDInstruction";
+    let DecoderMethod = "DecodeVLDST1Instruction";
     let AsmMatchConverter = "cvtVLDwbRegister";
   }
 }
@@ -639,7 +674,7 @@ multiclass VLD1QWB<bits<4> op7_4, string Dt> {
                      "$Rn.addr = $wb", []> {
     let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
     let Inst{5-4} = Rn{5-4};
-    let DecoderMethod = "DecodeVLDInstruction";
+    let DecoderMethod = "DecodeVLDST1Instruction";
     let AsmMatchConverter = "cvtVLDwbFixed";
   }
   def _register : NLdSt<0,0b10,0b1010,op7_4, (outs VecListDPair:$Vd, GPR:$wb),
@@ -647,7 +682,7 @@ multiclass VLD1QWB<bits<4> op7_4, string Dt> {
                         "vld1", Dt, "$Vd, $Rn, $Rm",
                         "$Rn.addr = $wb", []> {
     let Inst{5-4} = Rn{5-4};
-    let DecoderMethod = "DecodeVLDInstruction";
+    let DecoderMethod = "DecodeVLDST1Instruction";
     let AsmMatchConverter = "cvtVLDwbRegister";
   }
 }
@@ -668,7 +703,7 @@ class VLD1D3<bits<4> op7_4, string Dt>
           "$Vd, $Rn", "", []> {
   let Rm = 0b1111;
   let Inst{4} = Rn{4};
-  let DecoderMethod = "DecodeVLDInstruction";
+  let DecoderMethod = "DecodeVLDST1Instruction";
 }
 multiclass VLD1D3WB<bits<4> op7_4, string Dt> {
   def _fixed : NLdSt<0,0b10,0b0110, op7_4, (outs VecListThreeD:$Vd, GPR:$wb),
@@ -677,7 +712,7 @@ multiclass VLD1D3WB<bits<4> op7_4, string Dt> {
                      "$Rn.addr = $wb", []> {
     let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
     let Inst{4} = Rn{4};
-    let DecoderMethod = "DecodeVLDInstruction";
+    let DecoderMethod = "DecodeVLDST1Instruction";
     let AsmMatchConverter = "cvtVLDwbFixed";
   }
   def _register : NLdSt<0,0b10,0b0110,op7_4, (outs VecListThreeD:$Vd, GPR:$wb),
@@ -685,7 +720,7 @@ multiclass VLD1D3WB<bits<4> op7_4, string Dt> {
                         "vld1", Dt, "$Vd, $Rn, $Rm",
                         "$Rn.addr = $wb", []> {
     let Inst{4} = Rn{4};
-    let DecoderMethod = "DecodeVLDInstruction";
+    let DecoderMethod = "DecodeVLDST1Instruction";
     let AsmMatchConverter = "cvtVLDwbRegister";
   }
 }
@@ -709,7 +744,7 @@ class VLD1D4<bits<4> op7_4, string Dt>
           "$Vd, $Rn", "", []> {
   let Rm = 0b1111;
   let Inst{5-4} = Rn{5-4};
-  let DecoderMethod = "DecodeVLDInstruction";
+  let DecoderMethod = "DecodeVLDST1Instruction";
 }
 multiclass VLD1D4WB<bits<4> op7_4, string Dt> {
   def _fixed : NLdSt<0,0b10,0b0010, op7_4, (outs VecListFourD:$Vd, GPR:$wb),
@@ -718,7 +753,7 @@ multiclass VLD1D4WB<bits<4> op7_4, string Dt> {
                      "$Rn.addr = $wb", []> {
     let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
     let Inst{5-4} = Rn{5-4};
-    let DecoderMethod = "DecodeVLDInstruction";
+    let DecoderMethod = "DecodeVLDST1Instruction";
     let AsmMatchConverter = "cvtVLDwbFixed";
   }
   def _register : NLdSt<0,0b10,0b0010,op7_4, (outs VecListFourD:$Vd, GPR:$wb),
@@ -726,7 +761,7 @@ multiclass VLD1D4WB<bits<4> op7_4, string Dt> {
                         "vld1", Dt, "$Vd, $Rn, $Rm",
                         "$Rn.addr = $wb", []> {
     let Inst{5-4} = Rn{5-4};
-    let DecoderMethod = "DecodeVLDInstruction";
+    let DecoderMethod = "DecodeVLDST1Instruction";
     let AsmMatchConverter = "cvtVLDwbRegister";
   }
 }
@@ -751,7 +786,7 @@ class VLD2<bits<4> op11_8, bits<4> op7_4, string Dt, RegisterOperand VdTy,
           "vld2", Dt, "$Vd, $Rn", "", []> {
   let Rm = 0b1111;
   let Inst{5-4} = Rn{5-4};
-  let DecoderMethod = "DecodeVLDInstruction";
+  let DecoderMethod = "DecodeVLDST2Instruction";
 }
 
 def  VLD2d8   : VLD2<0b1000, {0,0,?,?}, "8", VecListDPair, IIC_VLD2>;
@@ -775,7 +810,7 @@ multiclass VLD2WB<bits<4> op11_8, bits<4> op7_4, string Dt,
                      "$Rn.addr = $wb", []> {
     let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
     let Inst{5-4} = Rn{5-4};
-    let DecoderMethod = "DecodeVLDInstruction";
+    let DecoderMethod = "DecodeVLDST2Instruction";
     let AsmMatchConverter = "cvtVLDwbFixed";
   }
   def _register : NLdSt<0, 0b10, op11_8, op7_4, (outs VdTy:$Vd, GPR:$wb),
@@ -783,7 +818,7 @@ multiclass VLD2WB<bits<4> op11_8, bits<4> op7_4, string Dt,
                         "vld2", Dt, "$Vd, $Rn, $Rm",
                         "$Rn.addr = $wb", []> {
     let Inst{5-4} = Rn{5-4};
-    let DecoderMethod = "DecodeVLDInstruction";
+    let DecoderMethod = "DecodeVLDST2Instruction";
     let AsmMatchConverter = "cvtVLDwbRegister";
   }
 }
@@ -818,7 +853,7 @@ class VLD3D<bits<4> op11_8, bits<4> op7_4, string Dt>
           "vld3", Dt, "\\{$Vd, $dst2, $dst3\\}, $Rn", "", []> {
   let Rm = 0b1111;
   let Inst{4} = Rn{4};
-  let DecoderMethod = "DecodeVLDInstruction";
+  let DecoderMethod = "DecodeVLDST3Instruction";
 }
 
 def  VLD3d8   : VLD3D<0b0100, {0,0,0,?}, "8">;
@@ -837,7 +872,7 @@ class VLD3DWB<bits<4> op11_8, bits<4> op7_4, string Dt>
           "vld3", Dt, "\\{$Vd, $dst2, $dst3\\}, $Rn$Rm",
           "$Rn.addr = $wb", []> {
   let Inst{4} = Rn{4};
-  let DecoderMethod = "DecodeVLDInstruction";
+  let DecoderMethod = "DecodeVLDST3Instruction";
 }
 
 def VLD3d8_UPD  : VLD3DWB<0b0100, {0,0,0,?}, "8">;
@@ -877,7 +912,7 @@ class VLD4D<bits<4> op11_8, bits<4> op7_4, string Dt>
           "vld4", Dt, "\\{$Vd, $dst2, $dst3, $dst4\\}, $Rn", "", []> {
   let Rm = 0b1111;
   let Inst{5-4} = Rn{5-4};
-  let DecoderMethod = "DecodeVLDInstruction";
+  let DecoderMethod = "DecodeVLDST4Instruction";
 }
 
 def  VLD4d8   : VLD4D<0b0000, {0,0,?,?}, "8">;
@@ -896,7 +931,7 @@ class VLD4DWB<bits<4> op11_8, bits<4> op7_4, string Dt>
           "vld4", Dt, "\\{$Vd, $dst2, $dst3, $dst4\\}, $Rn$Rm",
           "$Rn.addr = $wb", []> {
   let Inst{5-4} = Rn{5-4};
-  let DecoderMethod = "DecodeVLDInstruction";
+  let DecoderMethod = "DecodeVLDST4Instruction";
 }
 
 def VLD4d8_UPD  : VLD4DWB<0b0000, {0,0,?,?}, "8">;
@@ -1545,14 +1580,14 @@ class VST1D<bits<4> op7_4, string Dt>
           IIC_VST1, "vst1", Dt, "$Vd, $Rn", "", []> {
   let Rm = 0b1111;
   let Inst{4} = Rn{4};
-  let DecoderMethod = "DecodeVSTInstruction";
+  let DecoderMethod = "DecodeVLDST1Instruction";
 }
 class VST1Q<bits<4> op7_4, string Dt>
   : NLdSt<0,0b00,0b1010,op7_4, (outs), (ins addrmode6:$Rn, VecListDPair:$Vd),
           IIC_VST1x2, "vst1", Dt, "$Vd, $Rn", "", []> {
   let Rm = 0b1111;
   let Inst{5-4} = Rn{5-4};
-  let DecoderMethod = "DecodeVSTInstruction";
+  let DecoderMethod = "DecodeVLDST1Instruction";
 }
 
 def  VST1d8   : VST1D<{0,0,0,?}, "8">;
@@ -1573,7 +1608,7 @@ multiclass VST1DWB<bits<4> op7_4, string Dt> {
                      "$Rn.addr = $wb", []> {
     let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
     let Inst{4} = Rn{4};
-    let DecoderMethod = "DecodeVSTInstruction";
+    let DecoderMethod = "DecodeVLDST1Instruction";
     let AsmMatchConverter = "cvtVSTwbFixed";
   }
   def _register : NLdSt<0,0b00,0b0111,op7_4, (outs GPR:$wb),
@@ -1582,7 +1617,7 @@ multiclass VST1DWB<bits<4> op7_4, string Dt> {
                         "vst1", Dt, "$Vd, $Rn, $Rm",
                         "$Rn.addr = $wb", []> {
     let Inst{4} = Rn{4};
-    let DecoderMethod = "DecodeVSTInstruction";
+    let DecoderMethod = "DecodeVLDST1Instruction";
     let AsmMatchConverter = "cvtVSTwbRegister";
   }
 }
@@ -1593,7 +1628,7 @@ multiclass VST1QWB<bits<4> op7_4, string Dt> {
                      "$Rn.addr = $wb", []> {
     let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
     let Inst{5-4} = Rn{5-4};
-    let DecoderMethod = "DecodeVSTInstruction";
+    let DecoderMethod = "DecodeVLDST1Instruction";
     let AsmMatchConverter = "cvtVSTwbFixed";
   }
   def _register : NLdSt<0,0b00,0b1010,op7_4, (outs GPR:$wb),
@@ -1602,7 +1637,7 @@ multiclass VST1QWB<bits<4> op7_4, string Dt> {
                         "vst1", Dt, "$Vd, $Rn, $Rm",
                         "$Rn.addr = $wb", []> {
     let Inst{5-4} = Rn{5-4};
-    let DecoderMethod = "DecodeVSTInstruction";
+    let DecoderMethod = "DecodeVLDST1Instruction";
     let AsmMatchConverter = "cvtVSTwbRegister";
   }
 }
@@ -1624,7 +1659,7 @@ class VST1D3<bits<4> op7_4, string Dt>
           IIC_VST1x3, "vst1", Dt, "$Vd, $Rn", "", []> {
   let Rm = 0b1111;
   let Inst{4} = Rn{4};
-  let DecoderMethod = "DecodeVSTInstruction";
+  let DecoderMethod = "DecodeVLDST1Instruction";
 }
 multiclass VST1D3WB<bits<4> op7_4, string Dt> {
   def _fixed : NLdSt<0,0b00,0b0110,op7_4, (outs GPR:$wb),
@@ -1633,7 +1668,7 @@ multiclass VST1D3WB<bits<4> op7_4, string Dt> {
                      "$Rn.addr = $wb", []> {
     let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
     let Inst{5-4} = Rn{5-4};
-    let DecoderMethod = "DecodeVSTInstruction";
+    let DecoderMethod = "DecodeVLDST1Instruction";
     let AsmMatchConverter = "cvtVSTwbFixed";
   }
   def _register : NLdSt<0,0b00,0b0110,op7_4, (outs GPR:$wb),
@@ -1642,7 +1677,7 @@ multiclass VST1D3WB<bits<4> op7_4, string Dt> {
                         "vst1", Dt, "$Vd, $Rn, $Rm",
                         "$Rn.addr = $wb", []> {
     let Inst{5-4} = Rn{5-4};
-    let DecoderMethod = "DecodeVSTInstruction";
+    let DecoderMethod = "DecodeVLDST1Instruction";
     let AsmMatchConverter = "cvtVSTwbRegister";
   }
 }
@@ -1669,7 +1704,7 @@ class VST1D4<bits<4> op7_4, string Dt>
           []> {
   let Rm = 0b1111;
   let Inst{5-4} = Rn{5-4};
-  let DecoderMethod = "DecodeVSTInstruction";
+  let DecoderMethod = "DecodeVLDST1Instruction";
 }
 multiclass VST1D4WB<bits<4> op7_4, string Dt> {
   def _fixed : NLdSt<0,0b00,0b0010,op7_4, (outs GPR:$wb),
@@ -1678,7 +1713,7 @@ multiclass VST1D4WB<bits<4> op7_4, string Dt> {
                      "$Rn.addr = $wb", []> {
     let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
     let Inst{5-4} = Rn{5-4};
-    let DecoderMethod = "DecodeVSTInstruction";
+    let DecoderMethod = "DecodeVLDST1Instruction";
     let AsmMatchConverter = "cvtVSTwbFixed";
   }
   def _register : NLdSt<0,0b00,0b0010,op7_4, (outs GPR:$wb),
@@ -1687,7 +1722,7 @@ multiclass VST1D4WB<bits<4> op7_4, string Dt> {
                         "vst1", Dt, "$Vd, $Rn, $Rm",
                         "$Rn.addr = $wb", []> {
     let Inst{5-4} = Rn{5-4};
-    let DecoderMethod = "DecodeVSTInstruction";
+    let DecoderMethod = "DecodeVLDST1Instruction";
     let AsmMatchConverter = "cvtVSTwbRegister";
   }
 }
@@ -1713,7 +1748,7 @@ class VST2<bits<4> op11_8, bits<4> op7_4, string Dt, RegisterOperand VdTy,
           itin, "vst2", Dt, "$Vd, $Rn", "", []> {
   let Rm = 0b1111;
   let Inst{5-4} = Rn{5-4};
-  let DecoderMethod = "DecodeVSTInstruction";
+  let DecoderMethod = "DecodeVLDST2Instruction";
 }
 
 def  VST2d8   : VST2<0b1000, {0,0,?,?}, "8",  VecListDPair, IIC_VST2>;
@@ -1737,7 +1772,7 @@ multiclass VST2DWB<bits<4> op11_8, bits<4> op7_4, string Dt,
                      "$Rn.addr = $wb", []> {
     let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
     let Inst{5-4} = Rn{5-4};
-    let DecoderMethod = "DecodeVSTInstruction";
+    let DecoderMethod = "DecodeVLDST2Instruction";
     let AsmMatchConverter = "cvtVSTwbFixed";
   }
   def _register : NLdSt<0, 0b00, op11_8, op7_4, (outs GPR:$wb),
@@ -1745,7 +1780,7 @@ multiclass VST2DWB<bits<4> op11_8, bits<4> op7_4, string Dt,
                         "vst2", Dt, "$Vd, $Rn, $Rm",
                         "$Rn.addr = $wb", []> {
     let Inst{5-4} = Rn{5-4};
-    let DecoderMethod = "DecodeVSTInstruction";
+    let DecoderMethod = "DecodeVLDST2Instruction";
     let AsmMatchConverter = "cvtVSTwbRegister";
   }
 }
@@ -1756,7 +1791,7 @@ multiclass VST2QWB<bits<4> op7_4, string Dt> {
                      "$Rn.addr = $wb", []> {
     let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
     let Inst{5-4} = Rn{5-4};
-    let DecoderMethod = "DecodeVSTInstruction";
+    let DecoderMethod = "DecodeVLDST2Instruction";
     let AsmMatchConverter = "cvtVSTwbFixed";
   }
   def _register : NLdSt<0, 0b00, 0b0011, op7_4, (outs GPR:$wb),
@@ -1765,7 +1800,7 @@ multiclass VST2QWB<bits<4> op7_4, string Dt> {
                         "vst2", Dt, "$Vd, $Rn, $Rm",
                         "$Rn.addr = $wb", []> {
     let Inst{5-4} = Rn{5-4};
-    let DecoderMethod = "DecodeVSTInstruction";
+    let DecoderMethod = "DecodeVLDST2Instruction";
     let AsmMatchConverter = "cvtVSTwbRegister";
   }
 }
@@ -1800,7 +1835,7 @@ class VST3D<bits<4> op11_8, bits<4> op7_4, string Dt>
           "vst3", Dt, "\\{$Vd, $src2, $src3\\}, $Rn", "", []> {
   let Rm = 0b1111;
   let Inst{4} = Rn{4};
-  let DecoderMethod = "DecodeVSTInstruction";
+  let DecoderMethod = "DecodeVLDST3Instruction";
 }
 
 def  VST3d8   : VST3D<0b0100, {0,0,0,?}, "8">;
@@ -1819,7 +1854,7 @@ class VST3DWB<bits<4> op11_8, bits<4> op7_4, string Dt>
           "vst3", Dt, "\\{$Vd, $src2, $src3\\}, $Rn$Rm",
           "$Rn.addr = $wb", []> {
   let Inst{4} = Rn{4};
-  let DecoderMethod = "DecodeVSTInstruction";
+  let DecoderMethod = "DecodeVLDST3Instruction";
 }
 
 def VST3d8_UPD  : VST3DWB<0b0100, {0,0,0,?}, "8">;
@@ -1859,7 +1894,7 @@ class VST4D<bits<4> op11_8, bits<4> op7_4, string Dt>
           "", []> {
   let Rm = 0b1111;
   let Inst{5-4} = Rn{5-4};
-  let DecoderMethod = "DecodeVSTInstruction";
+  let DecoderMethod = "DecodeVLDST4Instruction";
 }
 
 def  VST4d8   : VST4D<0b0000, {0,0,?,?}, "8">;
@@ -1878,7 +1913,7 @@ class VST4DWB<bits<4> op11_8, bits<4> op7_4, string Dt>
            "vst4", Dt, "\\{$Vd, $src2, $src3, $src4\\}, $Rn$Rm",
           "$Rn.addr = $wb", []> {
   let Inst{5-4} = Rn{5-4};
-  let DecoderMethod = "DecodeVSTInstruction";
+  let DecoderMethod = "DecodeVLDST4Instruction";
 }
 
 def VST4d8_UPD  : VST4DWB<0b0000, {0,0,?,?}, "8">;
@@ -1959,7 +1994,7 @@ def VST1LNd8  : VST1LN<0b0000, {?,?,?,0}, "8", v8i8, truncstorei8,
 def VST1LNd16 : VST1LN<0b0100, {?,?,0,?}, "16", v4i16, truncstorei16,
                        NEONvgetlaneu, addrmode6> {
   let Inst{7-6} = lane{1-0};
-  let Inst{4}   = Rn{5};
+  let Inst{4}   = Rn{4};
 }
 
 def VST1LNd32 : VST1LN<0b1000, {?,0,?,?}, "32", v2i32, store, extractelt,
@@ -2002,7 +2037,7 @@ def VST1LNd8_UPD  : VST1LNWB<0b0000, {?,?,?,0}, "8", v8i8, post_truncsti8,
 def VST1LNd16_UPD : VST1LNWB<0b0100, {?,?,0,?}, "16", v4i16, post_truncsti16,
                              NEONvgetlaneu, addrmode6> {
   let Inst{7-6} = lane{1-0};
-  let Inst{4}   = Rn{5};
+  let Inst{4}   = Rn{4};
 }
 def VST1LNd32_UPD : VST1LNWB<0b1000, {?,0,?,?}, "32", v2i32, post_store,
                              extractelt, addrmode6oneL32> {
@@ -2238,6 +2273,38 @@ def VST4LNq32Pseudo_UPD : VSTQQQQLNWBPseudo<IIC_VST4lnu>;
 
 } // mayStore = 1, neverHasSideEffects = 1, hasExtraSrcRegAllocReq = 1
 
+// Use vld1/vst1 for unaligned f64 load / store
+def : Pat<(f64 (hword_alignedload addrmode6:$addr)),
+          (VLD1d16 addrmode6:$addr)>, Requires<[IsLE]>;
+def : Pat<(hword_alignedstore (f64 DPR:$value), addrmode6:$addr),
+          (VST1d16 addrmode6:$addr, DPR:$value)>, Requires<[IsLE]>;
+def : Pat<(f64 (byte_alignedload addrmode6:$addr)),
+          (VLD1d8 addrmode6:$addr)>, Requires<[IsLE]>;
+def : Pat<(byte_alignedstore (f64 DPR:$value), addrmode6:$addr),
+          (VST1d8 addrmode6:$addr, DPR:$value)>, Requires<[IsLE]>;
+def : Pat<(f64 (non_word_alignedload addrmode6:$addr)),
+          (VLD1d64 addrmode6:$addr)>, Requires<[IsBE]>;
+def : Pat<(non_word_alignedstore (f64 DPR:$value), addrmode6:$addr),
+          (VST1d64 addrmode6:$addr, DPR:$value)>, Requires<[IsBE]>;
+
+// Use vld1/vst1 for Q and QQ. Also use them for unaligned v2f64
+// load / store if it's legal.
+def : Pat<(v2f64 (dword_alignedload addrmode6:$addr)),
+          (VLD1q64 addrmode6:$addr)>;
+def : Pat<(dword_alignedstore (v2f64 QPR:$value), addrmode6:$addr),
+          (VST1q64 addrmode6:$addr, QPR:$value)>;
+def : Pat<(v2f64 (word_alignedload addrmode6:$addr)),
+          (VLD1q32 addrmode6:$addr)>;
+def : Pat<(word_alignedstore (v2f64 QPR:$value), addrmode6:$addr),
+          (VST1q32 addrmode6:$addr, QPR:$value)>;
+def : Pat<(v2f64 (hword_alignedload addrmode6:$addr)),
+          (VLD1q16 addrmode6:$addr)>, Requires<[IsLE]>;
+def : Pat<(hword_alignedstore (v2f64 QPR:$value), addrmode6:$addr),
+          (VST1q16 addrmode6:$addr, QPR:$value)>, Requires<[IsLE]>;
+def : Pat<(v2f64 (byte_alignedload addrmode6:$addr)),
+          (VLD1q8 addrmode6:$addr)>, Requires<[IsLE]>;
+def : Pat<(byte_alignedstore (v2f64 QPR:$value), addrmode6:$addr),
+          (VST1q8 addrmode6:$addr, QPR:$value)>, Requires<[IsLE]>;
 
 //===----------------------------------------------------------------------===//
 // NEON pattern fragments
@@ -4197,6 +4264,7 @@ def  VCEQfd   : N3VD<0,0,0b00,0b1110,0, IIC_VBIND, "vceq", "f32", v2i32, v2f32,
 def  VCEQfq   : N3VQ<0,0,0b00,0b1110,0, IIC_VBINQ, "vceq", "f32", v4i32, v4f32,
                      NEONvceq, 1>;
 
+let TwoOperandAliasConstraint = "$Vm = $Vd" in
 defm VCEQz    : N2V_QHS_cmp<0b11, 0b11, 0b01, 0b00010, 0, "vceq", "i",
                             "$Vd, $Vm, #0", NEONvceqz>;
 
@@ -4210,10 +4278,12 @@ def  VCGEfd   : N3VD<1,0,0b00,0b1110,0, IIC_VBIND, "vcge", "f32", v2i32, v2f32,
 def  VCGEfq   : N3VQ<1,0,0b00,0b1110,0, IIC_VBINQ, "vcge", "f32", v4i32, v4f32,
                      NEONvcge, 0>;
 
+let TwoOperandAliasConstraint = "$Vm = $Vd" in {
 defm VCGEz    : N2V_QHS_cmp<0b11, 0b11, 0b01, 0b00001, 0, "vcge", "s",
                             "$Vd, $Vm, #0", NEONvcgez>;
 defm VCLEz    : N2V_QHS_cmp<0b11, 0b11, 0b01, 0b00011, 0, "vcle", "s",
                             "$Vd, $Vm, #0", NEONvclez>;
+}
 
 //   VCGT     : Vector Compare Greater Than
 defm VCGTs    : N3V_QHS<0, 0, 0b0011, 0, IIC_VSUBi4D, IIC_VSUBi4D, IIC_VSUBi4Q,
@@ -4225,10 +4295,12 @@ def  VCGTfd   : N3VD<1,0,0b10,0b1110,0, IIC_VBIND, "vcgt", "f32", v2i32, v2f32,
 def  VCGTfq   : N3VQ<1,0,0b10,0b1110,0, IIC_VBINQ, "vcgt", "f32", v4i32, v4f32,
                      NEONvcgt, 0>;
 
+let TwoOperandAliasConstraint = "$Vm = $Vd" in {
 defm VCGTz    : N2V_QHS_cmp<0b11, 0b11, 0b01, 0b00000, 0, "vcgt", "s",
                             "$Vd, $Vm, #0", NEONvcgtz>;
 defm VCLTz    : N2V_QHS_cmp<0b11, 0b11, 0b01, 0b00100, 0, "vclt", "s",
                             "$Vd, $Vm, #0", NEONvcltz>;
+}
 
 //   VACGE    : Vector Absolute Compare Greater Than or Equal (aka VCAGE)
 def  VACGEd   : N3VDInt<1, 0, 0b00, 0b1110, 1, N3RegFrm, IIC_VBIND, "vacge",
@@ -4244,6 +4316,24 @@ def  VACGTq   : N3VQInt<1, 0, 0b10, 0b1110, 1, N3RegFrm, IIC_VBINQ, "vacgt",
 defm VTST     : N3V_QHS<0, 0, 0b1000, 1, IIC_VBINi4D, IIC_VBINi4D, IIC_VBINi4Q,
                         IIC_VBINi4Q, "vtst", "", NEONvtst, 1>;
 
+def: NEONInstAlias<"vaclt${p}.f32 $Vd, $Vn, $Vm",
+                   (VACGTd DPR:$Vd, DPR:$Vm, DPR:$Vn, pred:$p)>;
+def: NEONInstAlias<"vaclt${p}.f32 $Vd, $Vn, $Vm",
+                   (VACGTq QPR:$Vd, QPR:$Vm, QPR:$Vn, pred:$p)>;
+def: NEONInstAlias<"vacle${p}.f32 $Vd, $Vn, $Vm",
+                   (VACGEd DPR:$Vd, DPR:$Vm, DPR:$Vn, pred:$p)>;
+def: NEONInstAlias<"vacle${p}.f32 $Vd, $Vn, $Vm",
+                   (VACGEq QPR:$Vd, QPR:$Vm, QPR:$Vn, pred:$p)>;
+
+def: NEONInstAlias<"vaclt${p}.f32 $Vd, $Vm",
+                   (VACGTd DPR:$Vd, DPR:$Vm, DPR:$Vd, pred:$p)>;
+def: NEONInstAlias<"vaclt${p}.f32 $Vd, $Vm",
+                   (VACGTq QPR:$Vd, QPR:$Vm, QPR:$Vd, pred:$p)>;
+def: NEONInstAlias<"vacle${p}.f32 $Vd, $Vm",
+                   (VACGEd DPR:$Vd, DPR:$Vm, DPR:$Vd, pred:$p)>;
+def: NEONInstAlias<"vacle${p}.f32 $Vd, $Vm",
+                   (VACGEq QPR:$Vd, QPR:$Vm, QPR:$Vd, pred:$p)>;
+
 // Vector Bitwise Operations.
 
 def vnotd : PatFrag<(ops node:$in),
@@ -4421,10 +4511,36 @@ def  VBSLd    : N3VX<1, 0, 0b01, 0b0001, 0, 1, (outs DPR:$Vd),
                      "vbsl", "$Vd, $Vn, $Vm", "$src1 = $Vd",
                      [(set DPR:$Vd,
                            (v2i32 (NEONvbsl DPR:$src1, DPR:$Vn, DPR:$Vm)))]>;
+def : Pat<(v8i8 (int_arm_neon_vbsl (v8i8 DPR:$src1),
+                                   (v8i8 DPR:$Vn), (v8i8 DPR:$Vm))),
+          (VBSLd DPR:$src1, DPR:$Vn, DPR:$Vm)>,
+        Requires<[HasNEON]>;
+def : Pat<(v4i16 (int_arm_neon_vbsl (v4i16 DPR:$src1),
+                                    (v4i16 DPR:$Vn), (v4i16 DPR:$Vm))),
+          (VBSLd DPR:$src1, DPR:$Vn, DPR:$Vm)>,
+        Requires<[HasNEON]>;
+def : Pat<(v2i32 (int_arm_neon_vbsl (v2i32 DPR:$src1),
+                                    (v2i32 DPR:$Vn), (v2i32 DPR:$Vm))),
+          (VBSLd DPR:$src1, DPR:$Vn, DPR:$Vm)>,
+        Requires<[HasNEON]>;
+def : Pat<(v2f32 (int_arm_neon_vbsl (v2f32 DPR:$src1),
+                                    (v2f32 DPR:$Vn), (v2f32 DPR:$Vm))),
+          (VBSLd DPR:$src1, DPR:$Vn, DPR:$Vm)>,
+        Requires<[HasNEON]>;
+def : Pat<(v1i64 (int_arm_neon_vbsl (v1i64 DPR:$src1),
+                                    (v1i64 DPR:$Vn), (v1i64 DPR:$Vm))),
+          (VBSLd DPR:$src1, DPR:$Vn, DPR:$Vm)>,
+        Requires<[HasNEON]>;
 
 def : Pat<(v2i32 (or (and DPR:$Vn, DPR:$Vd),
                      (and DPR:$Vm, (vnotd DPR:$Vd)))),
-          (VBSLd DPR:$Vd, DPR:$Vn, DPR:$Vm)>;
+          (VBSLd DPR:$Vd, DPR:$Vn, DPR:$Vm)>,
+        Requires<[HasNEON]>;
+
+def : Pat<(v1i64 (or (and DPR:$Vn, DPR:$Vd),
+                     (and DPR:$Vm, (vnotd DPR:$Vd)))),
+          (VBSLd DPR:$Vd, DPR:$Vn, DPR:$Vm)>,
+        Requires<[HasNEON]>;
 
 def  VBSLq    : N3VX<1, 0, 0b01, 0b0001, 1, 1, (outs QPR:$Vd),
                      (ins QPR:$src1, QPR:$Vn, QPR:$Vm),
@@ -4433,9 +4549,35 @@ def  VBSLq    : N3VX<1, 0, 0b01, 0b0001, 1, 1, (outs QPR:$Vd),
                      [(set QPR:$Vd,
                            (v4i32 (NEONvbsl QPR:$src1, QPR:$Vn, QPR:$Vm)))]>;
 
+def : Pat<(v16i8 (int_arm_neon_vbsl (v16i8 QPR:$src1),
+                                   (v16i8 QPR:$Vn), (v16i8 QPR:$Vm))),
+          (VBSLq QPR:$src1, QPR:$Vn, QPR:$Vm)>,
+        Requires<[HasNEON]>;
+def : Pat<(v8i16 (int_arm_neon_vbsl (v8i16 QPR:$src1),
+                                    (v8i16 QPR:$Vn), (v8i16 QPR:$Vm))),
+          (VBSLq QPR:$src1, QPR:$Vn, QPR:$Vm)>,
+        Requires<[HasNEON]>;
+def : Pat<(v4i32 (int_arm_neon_vbsl (v4i32 QPR:$src1),
+                                    (v4i32 QPR:$Vn), (v4i32 QPR:$Vm))),
+          (VBSLq QPR:$src1, QPR:$Vn, QPR:$Vm)>,
+        Requires<[HasNEON]>;
+def : Pat<(v4f32 (int_arm_neon_vbsl (v4f32 QPR:$src1),
+                                    (v4f32 QPR:$Vn), (v4f32 QPR:$Vm))),
+          (VBSLq QPR:$src1, QPR:$Vn, QPR:$Vm)>,
+        Requires<[HasNEON]>;
+def : Pat<(v2i64 (int_arm_neon_vbsl (v2i64 QPR:$src1),
+                                    (v2i64 QPR:$Vn), (v2i64 QPR:$Vm))),
+          (VBSLq QPR:$src1, QPR:$Vn, QPR:$Vm)>,
+        Requires<[HasNEON]>;
+
 def : Pat<(v4i32 (or (and QPR:$Vn, QPR:$Vd),
                      (and QPR:$Vm, (vnotq QPR:$Vd)))),
-          (VBSLq QPR:$Vd, QPR:$Vn, QPR:$Vm)>;
+          (VBSLq QPR:$Vd, QPR:$Vn, QPR:$Vm)>,
+        Requires<[HasNEON]>;
+def : Pat<(v2i64 (or (and QPR:$Vn, QPR:$Vd),
+                     (and QPR:$Vm, (vnotq QPR:$Vd)))),
+          (VBSLq QPR:$Vd, QPR:$Vn, QPR:$Vm)>,
+        Requires<[HasNEON]>;
 
 //   VBIF     : Vector Bitwise Insert if False
 //              like VBSL but with: "vbif $dst, $src3, $src1", "$src2 = $dst",
@@ -4758,12 +4900,38 @@ defm VSRI     : N2VShInsR_QHSD<1, 1, 0b0100, 1, "vsri">;
 defm VABS     : N2VInt_QHS<0b11, 0b11, 0b01, 0b00110, 0,
                            IIC_VUNAiD, IIC_VUNAiQ, "vabs", "s",
                            int_arm_neon_vabs>;
-def  VABSfd   : N2VDInt<0b11, 0b11, 0b10, 0b01, 0b01110, 0,
-                        IIC_VUNAD, "vabs", "f32",
-                        v2f32, v2f32, int_arm_neon_vabs>;
-def  VABSfq   : N2VQInt<0b11, 0b11, 0b10, 0b01, 0b01110, 0,
-                        IIC_VUNAQ, "vabs", "f32",
-                        v4f32, v4f32, int_arm_neon_vabs>;
+def  VABSfd   : N2VD<0b11, 0b11, 0b10, 0b01, 0b01110, 0,
+                     "vabs", "f32",
+                     v2f32, v2f32, fabs>;
+def  VABSfq   : N2VQ<0b11, 0b11, 0b10, 0b01, 0b01110, 0,
+                     "vabs", "f32",
+                      v4f32, v4f32, fabs>;
+
+def : Pat<(xor (v2i32 (bitconvert (v8i8 (NEONvshrs DPR:$src, (i32 7))))),
+               (v2i32 (bitconvert (v8i8 (add DPR:$src,
+                                             (NEONvshrs DPR:$src, (i32 7))))))),
+          (VABSv8i8 DPR:$src)>;
+def : Pat<(xor (v2i32 (bitconvert (v4i16 (NEONvshrs DPR:$src, (i32 15))))),
+               (v2i32 (bitconvert (v4i16 (add DPR:$src,
+                                            (NEONvshrs DPR:$src, (i32 15))))))),
+          (VABSv4i16 DPR:$src)>;
+def : Pat<(xor (v2i32 (NEONvshrs DPR:$src, (i32 31))),
+               (v2i32 (add DPR:$src, (NEONvshrs DPR:$src, (i32 31))))),
+          (VABSv2i32 DPR:$src)>;
+def : Pat<(xor (v4i32 (bitconvert (v16i8 (NEONvshrs QPR:$src, (i32 7))))),
+               (v4i32 (bitconvert (v16i8 (add QPR:$src,
+                                             (NEONvshrs QPR:$src, (i32 7))))))),
+          (VABSv16i8 QPR:$src)>;
+def : Pat<(xor (v4i32 (bitconvert (v8i16 (NEONvshrs QPR:$src, (i32 15))))),
+               (v4i32 (bitconvert (v8i16 (add QPR:$src,
+                                            (NEONvshrs QPR:$src, (i32 15))))))),
+          (VABSv8i16 QPR:$src)>;
+def : Pat<(xor (v4i32 (NEONvshrs QPR:$src, (i32 31))),
+               (v4i32 (add QPR:$src, (NEONvshrs QPR:$src, (i32 31))))),
+          (VABSv4i32 QPR:$src)>;
+
+def : Pat<(v2f32 (int_arm_neon_vabs (v2f32 DPR:$src))), (VABSfd DPR:$src)>;
+def : Pat<(v4f32 (int_arm_neon_vabs (v4f32 QPR:$src))), (VABSfq QPR:$src)>;
 
 //   VQABS    : Vector Saturating Absolute Value
 defm VQABS    : N2VInt_QHS<0b11, 0b11, 0b00, 0b01110, 0,
@@ -4829,10 +4997,10 @@ defm VCLZ     : N2VInt_QHS<0b11, 0b11, 0b00, 0b01001, 0,
 //   VCNT     : Vector Count One Bits
 def  VCNTd    : N2VDInt<0b11, 0b11, 0b00, 0b00, 0b01010, 0,
                         IIC_VCNTiD, "vcnt", "8",
-                        v8i8, v8i8, int_arm_neon_vcnt>;
+                        v8i8, v8i8, ctpop>;
 def  VCNTq    : N2VQInt<0b11, 0b11, 0b00, 0b00, 0b01010, 0,
                         IIC_VCNTiQ, "vcnt", "8",
-                        v16i8, v16i8, int_arm_neon_vcnt>;
+                        v16i8, v16i8, ctpop>;
 
 // Vector Swap
 def  VSWPd    : N2VX<0b11, 0b11, 0b00, 0b10, 0b00000, 0, 0,
@@ -4949,7 +5117,8 @@ def VGETLNi32 : NVGetLane<{1,1,1,0,0,0,?,1}, 0b1011, 0b00,
                           (outs GPR:$R), (ins DPR:$V, VectorIndex32:$lane),
                           IIC_VMOVSI, "vmov", "32", "$R, $V$lane",
                           [(set GPR:$R, (extractelt (v2i32 DPR:$V),
-                                           imm:$lane))]> {
+                                           imm:$lane))]>,
+                Requires<[HasNEON, HasFastVGETLNi32]> {
   let Inst{21} = lane{0};
 }
 // def VGETLNf32: see FMRDH and FMRDL in ARMInstrVFP.td
@@ -4972,7 +5141,16 @@ def : Pat<(NEONvgetlaneu (v8i16 QPR:$src), imm:$lane),
 def : Pat<(extractelt (v4i32 QPR:$src), imm:$lane),
           (VGETLNi32 (v2i32 (EXTRACT_SUBREG QPR:$src,
                              (DSubReg_i32_reg imm:$lane))),
-                     (SubReg_i32_lane imm:$lane))>;
+                     (SubReg_i32_lane imm:$lane))>,
+      Requires<[HasNEON, HasFastVGETLNi32]>;
+def : Pat<(extractelt (v2i32 DPR:$src), imm:$lane),
+          (COPY_TO_REGCLASS
+            (i32 (EXTRACT_SUBREG DPR:$src, (SSubReg_f32_reg imm:$lane))), GPR)>,
+      Requires<[HasNEON, HasSlowVGETLNi32]>;
+def : Pat<(extractelt (v4i32 QPR:$src), imm:$lane),
+          (COPY_TO_REGCLASS
+            (i32 (EXTRACT_SUBREG QPR:$src, (SSubReg_f32_reg imm:$lane))), GPR)>,
+      Requires<[HasNEON, HasSlowVGETLNi32]>;
 def : Pat<(extractelt (v2f32 DPR:$src1), imm:$src2),
           (EXTRACT_SUBREG (v2f32 (COPY_TO_REGCLASS (v2f32 DPR:$src1),DPR_VFP2)),
                           (SSubReg_f32_reg imm:$src2))>;
@@ -5083,14 +5261,23 @@ class VDUPQ<bits<8> opcod1, bits<2> opcod3, string Dt, ValueType Ty>
 
 def  VDUP8d   : VDUPD<0b11101100, 0b00, "8", v8i8>;
 def  VDUP16d  : VDUPD<0b11101000, 0b01, "16", v4i16>;
-def  VDUP32d  : VDUPD<0b11101000, 0b00, "32", v2i32>;
+def  VDUP32d  : VDUPD<0b11101000, 0b00, "32", v2i32>,
+                Requires<[HasNEON, HasFastVDUP32]>;
 def  VDUP8q   : VDUPQ<0b11101110, 0b00, "8", v16i8>;
 def  VDUP16q  : VDUPQ<0b11101010, 0b01, "16", v8i16>;
 def  VDUP32q  : VDUPQ<0b11101010, 0b00, "32", v4i32>;
 
-def : Pat<(v2f32 (NEONvdup (f32 (bitconvert GPR:$R)))), (VDUP32d GPR:$R)>;
+// NEONvdup patterns for uarchs with fast VDUP.32.
+def : Pat<(v2f32 (NEONvdup (f32 (bitconvert GPR:$R)))), (VDUP32d GPR:$R)>,
+      Requires<[HasNEON,HasFastVDUP32]>;
 def : Pat<(v4f32 (NEONvdup (f32 (bitconvert GPR:$R)))), (VDUP32q GPR:$R)>;
 
+// NEONvdup patterns for uarchs with slow VDUP.32 - use VMOVDRR instead.
+def : Pat<(v2i32 (NEONvdup (i32 GPR:$R))), (VMOVDRR GPR:$R, GPR:$R)>,
+      Requires<[HasNEON,HasSlowVDUP32]>;
+def : Pat<(v2f32 (NEONvdup (f32 (bitconvert GPR:$R)))), (VMOVDRR GPR:$R, GPR:$R)>,
+      Requires<[HasNEON,HasSlowVDUP32]>;
+
 //   VDUP     : Vector Duplicate Lane (from scalar to all elements)
 
 class VDUPLND<bits<4> op19_16, string OpcodeStr, string Dt,
@@ -5322,8 +5509,9 @@ class VEXTd<string OpcodeStr, string Dt, ValueType Ty, Operand immTy>
         IIC_VEXTD, OpcodeStr, Dt, "$Vd, $Vn, $Vm, $index", "",
         [(set DPR:$Vd, (Ty (NEONvext (Ty DPR:$Vn),
                                      (Ty DPR:$Vm), imm:$index)))]> {
-  bits<4> index;
-  let Inst{11-8} = index{3-0};
+  bits<3> index;
+  let Inst{11} = 0b0;
+  let Inst{10-8} = index{2-0};
 }
 
 class VEXTq<string OpcodeStr, string Dt, ValueType Ty, Operand immTy>
@@ -5338,14 +5526,14 @@ class VEXTq<string OpcodeStr, string Dt, ValueType Ty, Operand immTy>
 }
 
 def VEXTd8  : VEXTd<"vext", "8",  v8i8, imm0_7> {
-  let Inst{11-8} = index{3-0};
+  let Inst{10-8} = index{2-0};
 }
 def VEXTd16 : VEXTd<"vext", "16", v4i16, imm0_3> {
-  let Inst{11-9} = index{2-0};
+  let Inst{10-9} = index{1-0};
   let Inst{8}    = 0b0;
 }
 def VEXTd32 : VEXTd<"vext", "32", v2i32, imm0_1> {
-  let Inst{11-10} = index{1-0};
+  let Inst{10}     = index{0};
   let Inst{9-8}    = 0b00;
 }
 def : Pat<(v2f32 (NEONvext (v2f32 DPR:$Vn),
@@ -5527,6 +5715,11 @@ def : N2VSPat<arm_ftoui, VCVTf2ud>;
 def : N2VSPat<arm_sitof, VCVTs2fd>;
 def : N2VSPat<arm_uitof, VCVTu2fd>;
 
+// Prefer VMOVDRR for i32 -> f32 bitcasts, it can write all DPR registers.
+def : Pat<(f32 (bitconvert GPR:$a)),
+          (EXTRACT_SUBREG (VMOVDRR GPR:$a, GPR:$a), ssub_0)>,
+        Requires<[HasNEON, DontUseVMOVSR]>;
+
 //===----------------------------------------------------------------------===//
 // Non-Instruction Patterns
 //===----------------------------------------------------------------------===//
@@ -5594,29 +5787,34 @@ def : Pat<(v2f64 (bitconvert (v8i16 QPR:$src))), (v2f64 QPR:$src)>;
 def : Pat<(v2f64 (bitconvert (v16i8 QPR:$src))), (v2f64 QPR:$src)>;
 def : Pat<(v2f64 (bitconvert (v4f32 QPR:$src))), (v2f64 QPR:$src)>;
 
+// Fold extracting an element out of a v2i32 into a vfp register.
+def : Pat<(f32 (bitconvert (i32 (extractelt (v2i32 DPR:$src), imm:$lane)))),
+          (f32 (EXTRACT_SUBREG DPR:$src, (SSubReg_f32_reg imm:$lane)))>;
+
 // 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 addrmode6oneL32:$addr))
-//         (VMOVLuv8i16 (VLD1LNd32 addrmode6oneL32:$addr,
-//                                 (f64 (IMPLICIT_DEF)), (i32 0)))>;
+// Lengthen_Single<"8", "i16", "8"> = 
+//     Pat<(v8i16 (extloadvi8 addrmode6:$addr))
+//         (VMOVLuv8i16 (VLD1d8 addrmode6:$addr,
+//                              (f64 (IMPLICIT_DEF)), (i32 0)))>;
 multiclass Lengthen_Single<string DestLanes, string DestTy, string SrcTy> {
+  let AddedComplexity = 10 in {
   def _Any : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
-                    (!cast<PatFrag>("extloadv" # SrcTy) addrmode6oneL32:$addr)),
+                    (!cast<PatFrag>("extloadvi" # SrcTy) addrmode6:$addr)),
                   (!cast<Instruction>("VMOVLuv" # DestLanes # DestTy)
-                    (VLD1LNd32 addrmode6oneL32:$addr, 
-                               (f64 (IMPLICIT_DEF)), (i32 0)))>;
+                    (!cast<Instruction>("VLD1d" # SrcTy) addrmode6:$addr))>;
+
   def _Z : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
-                  (!cast<PatFrag>("zextloadv" # SrcTy) addrmode6oneL32:$addr)),
+                  (!cast<PatFrag>("zextloadvi" # SrcTy) addrmode6:$addr)),
                 (!cast<Instruction>("VMOVLuv" # DestLanes # DestTy)
-                    (VLD1LNd32 addrmode6oneL32:$addr, 
-                               (f64 (IMPLICIT_DEF)), (i32 0)))>;
+                    (!cast<Instruction>("VLD1d" # SrcTy) addrmode6:$addr))>;
+
   def _S : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
-                  (!cast<PatFrag>("sextloadv" # SrcTy) addrmode6oneL32:$addr)),
+                  (!cast<PatFrag>("sextloadvi" # SrcTy) addrmode6:$addr)),
                 (!cast<Instruction>("VMOVLsv" # DestLanes # DestTy)
-                    (VLD1LNd32 addrmode6oneL32:$addr, 
-                               (f64 (IMPLICIT_DEF)), (i32 0)))>;
+                    (!cast<Instruction>("VLD1d" # SrcTy) addrmode6:$addr))>;
+  }
 }
 
 // extload, zextload and sextload for a lengthening load which only uses
@@ -5648,8 +5846,8 @@ multiclass Lengthen_HalfSingle<string DestLanes, string DestTy, string SrcTy,
 // 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:$addr))
+// Lengthen_Double<"4", "i32", "i8", "8", "i16", "4", "i32"> =
+//     Pat<(v4i32 (extloadvi8 addrmode6oneL32:$addr))
 //         (EXTRACT_SUBREG (VMOVLuv4i32
 //           (EXTRACT_SUBREG (VMOVLuv8i16 (VLD1LNd32 addrmode6oneL32:$addr,
 //                                                   (f64 (IMPLICIT_DEF)),
@@ -5663,19 +5861,19 @@ multiclass Lengthen_Double<string DestLanes, string DestTy, string SrcTy,
                    (!cast<PatFrag>("extloadv" # SrcTy) addrmode6oneL32:$addr)),
          (!cast<Instruction>("VMOVLuv" # Insn2Lanes # Insn2Ty)
            (EXTRACT_SUBREG (!cast<Instruction>("VMOVLuv" # Insn1Lanes # Insn1Ty)
-             (VLD1LNd32 addrmode6oneL32:$addr, (f64 (IMPLICIT_DEF)), (i32 0))), 
+             (VLD1LNd32 addrmode6oneL32:$addr, (f64 (IMPLICIT_DEF)), (i32 0))),
              dsub_0))>;
   def _Z   : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
                    (!cast<PatFrag>("zextloadv" # SrcTy) addrmode6oneL32:$addr)),
          (!cast<Instruction>("VMOVLuv" # Insn2Lanes # Insn2Ty)
            (EXTRACT_SUBREG (!cast<Instruction>("VMOVLuv" # Insn1Lanes # Insn1Ty)
-             (VLD1LNd32 addrmode6oneL32:$addr, (f64 (IMPLICIT_DEF)), (i32 0))), 
+             (VLD1LNd32 addrmode6oneL32:$addr, (f64 (IMPLICIT_DEF)), (i32 0))),
              dsub_0))>;
   def _S   : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
                    (!cast<PatFrag>("sextloadv" # SrcTy) addrmode6oneL32:$addr)),
          (!cast<Instruction>("VMOVLsv" # Insn2Lanes # Insn2Ty)
            (EXTRACT_SUBREG (!cast<Instruction>("VMOVLsv" # Insn1Lanes # Insn1Ty)
-             (VLD1LNd32 addrmode6oneL32:$addr, (f64 (IMPLICIT_DEF)), (i32 0))), 
+             (VLD1LNd32 addrmode6oneL32:$addr, (f64 (IMPLICIT_DEF)), (i32 0))),
              dsub_0))>;
 }
 
@@ -5684,9 +5882,9 @@ multiclass Lengthen_Double<string DestLanes, string DestTy, string SrcTy,
 // requiring half the available lanes (a 64-bit outcome instead of a 128-bit).
 //
 // Lengthen_HalfDouble<"2", "i32", "i8", "8", "i16", "4", "i32"> =
-// Pat<(v4i32 (extloadvi8 addrmode5:$addr))
+// Pat<(v2i32 (extloadvi8 addrmode6:$addr))
 //     (EXTRACT_SUBREG (VMOVLuv4i32
-//       (EXTRACT_SUBREG (VMOVLuv8i16 (VLD1LNd32 addrmode6oneL32:$addr,
+//       (EXTRACT_SUBREG (VMOVLuv8i16 (VLD1LNd16 addrmode6:$addr,
 //                                               (f64 (IMPLICIT_DEF)), (i32 0))),
 //                       dsub_0)),
 //       dsub_0)>;
@@ -5694,34 +5892,33 @@ multiclass Lengthen_HalfDouble<string DestLanes, string DestTy, string SrcTy,
                            string Insn1Lanes, string Insn1Ty, string Insn2Lanes,
                            string Insn2Ty> {
   def _Any : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
-                   (!cast<PatFrag>("extloadv" # SrcTy) addrmode6oneL32:$addr)),
+                   (!cast<PatFrag>("extloadv" # SrcTy) addrmode6:$addr)),
          (EXTRACT_SUBREG (!cast<Instruction>("VMOVLuv" # Insn2Lanes # Insn2Ty)
            (EXTRACT_SUBREG (!cast<Instruction>("VMOVLuv" # Insn1Lanes # Insn1Ty)
-             (VLD1LNd32 addrmode6oneL32:$addr, (f64 (IMPLICIT_DEF)), (i32 0))), 
+             (VLD1LNd16 addrmode6:$addr, (f64 (IMPLICIT_DEF)), (i32 0))),
              dsub_0)),
           dsub_0)>;
   def _Z   : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
-                   (!cast<PatFrag>("zextloadv" # SrcTy) addrmode6oneL32:$addr)),
+                   (!cast<PatFrag>("zextloadv" # SrcTy) addrmode6:$addr)),
          (EXTRACT_SUBREG (!cast<Instruction>("VMOVLuv" # Insn2Lanes # Insn2Ty)
            (EXTRACT_SUBREG (!cast<Instruction>("VMOVLuv" # Insn1Lanes # Insn1Ty)
-             (VLD1LNd32 addrmode6oneL32:$addr, (f64 (IMPLICIT_DEF)), (i32 0))), 
+             (VLD1LNd16 addrmode6:$addr, (f64 (IMPLICIT_DEF)), (i32 0))),
              dsub_0)),
           dsub_0)>;
   def _S   : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
-                   (!cast<PatFrag>("sextloadv" # SrcTy) addrmode6oneL32:$addr)),
+                   (!cast<PatFrag>("sextloadv" # SrcTy) addrmode6:$addr)),
          (EXTRACT_SUBREG (!cast<Instruction>("VMOVLsv" # Insn2Lanes # Insn2Ty)
            (EXTRACT_SUBREG (!cast<Instruction>("VMOVLsv" # Insn1Lanes # Insn1Ty)
-             (VLD1LNd32 addrmode6oneL32:$addr, (f64 (IMPLICIT_DEF)), (i32 0))), 
+             (VLD1LNd16 addrmode6:$addr, (f64 (IMPLICIT_DEF)), (i32 0))),
              dsub_0)),
           dsub_0)>;
 }
 
-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_Single<"8", "i16", "8">; // v8i8 -> v8i16
+defm : Lengthen_Single<"4", "i32", "16">; // v4i16 -> v4i32
+defm : Lengthen_Single<"2", "i64", "32">; // 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
@@ -5732,17 +5929,17 @@ defm : Lengthen_HalfDouble<"2", "i32", "i8", "8", "i16", "4", "i32">;
 defm : Lengthen_Double<"2", "i64", "i16", "4", "i32", "2", "i64">;
 
 // Triple lengthening - v2i8 -> v2i16 -> v2i32 -> v2i64
-def : Pat<(v2i64 (extloadvi8 addrmode6oneL32:$addr)),
+def : Pat<(v2i64 (extloadvi8 addrmode6:$addr)),
       (VMOVLuv2i64 (EXTRACT_SUBREG (VMOVLuv4i32 (EXTRACT_SUBREG (VMOVLuv8i16
-         (VLD1LNd32 addrmode6oneL32:$addr, 
+         (VLD1LNd16 addrmode6:$addr, 
                     (f64 (IMPLICIT_DEF)), (i32 0))), dsub_0)), dsub_0))>;
-def : Pat<(v2i64 (zextloadvi8 addrmode6oneL32:$addr)),
+def : Pat<(v2i64 (zextloadvi8 addrmode6:$addr)),
       (VMOVLuv2i64 (EXTRACT_SUBREG (VMOVLuv4i32 (EXTRACT_SUBREG (VMOVLuv8i16
-         (VLD1LNd32 addrmode6oneL32:$addr,
+         (VLD1LNd16 addrmode6:$addr,
                     (f64 (IMPLICIT_DEF)), (i32 0))), dsub_0)), dsub_0))>;
-def : Pat<(v2i64 (sextloadvi8 addrmode6oneL32:$addr)),
+def : Pat<(v2i64 (sextloadvi8 addrmode6:$addr)),
       (VMOVLsv2i64 (EXTRACT_SUBREG (VMOVLsv4i32 (EXTRACT_SUBREG (VMOVLsv8i16
-         (VLD1LNd32 addrmode6oneL32:$addr,
+         (VLD1LNd16 addrmode6:$addr,
                     (f64 (IMPLICIT_DEF)), (i32 0))), dsub_0)), dsub_0))>;
 
 //===----------------------------------------------------------------------===//
@@ -6501,12 +6698,17 @@ def VST4qWB_register_Asm_32 :
                   (ins VecListFourQ:$list, addrmode6:$addr,
                        rGPR:$Rm, pred:$p)>;
 
-// VMOV takes an optional datatype suffix
+// VMOV/VMVN takes an optional datatype suffix
 defm : NEONDTAnyInstAlias<"vmov${p}", "$Vd, $Vm",
                          (VORRd DPR:$Vd, DPR:$Vm, DPR:$Vm, pred:$p)>;
 defm : NEONDTAnyInstAlias<"vmov${p}", "$Vd, $Vm",
                          (VORRq QPR:$Vd, QPR:$Vm, QPR:$Vm, pred:$p)>;
 
+defm : NEONDTAnyInstAlias<"vmvn${p}", "$Vd, $Vm",
+                         (VMVNd DPR:$Vd, DPR:$Vm, pred:$p)>;
+defm : NEONDTAnyInstAlias<"vmvn${p}", "$Vd, $Vm",
+                         (VMVNq QPR:$Vd, QPR:$Vm, pred:$p)>;
+
 // VCLT (register) is an assembler alias for VCGT w/ the operands reversed.
 // D-register versions.
 def : NEONInstAlias<"vcle${p}.s8 $Dd, $Dn, $Dm",