X-Git-Url: http://demsky.eecs.uci.edu/git/?a=blobdiff_plain;f=lib%2FTarget%2FARM%2FARMInstrFormats.td;h=7d27cf3fcdcbade77033fd340d7ca166c426b028;hb=d2f8ee71948854c205c74accf14c2b0b87502fef;hp=fe77f2f9fbc9a76b09976903fc6164d1a4106de2;hpb=f1a009007374d8ae1c1565f34d9cea3b83665e5f;p=oota-llvm.git

diff --git a/lib/Target/ARM/ARMInstrFormats.td b/lib/Target/ARM/ARMInstrFormats.td
index fe77f2f9fbc..7d27cf3fcdc 100644
--- a/lib/Target/ARM/ARMInstrFormats.td
+++ b/lib/Target/ARM/ARMInstrFormats.td
@@ -1,4 +1,4 @@
-//===- ARMInstrFormats.td - ARM Instruction Formats ----------*- tablegen -*-=//
+//===-- ARMInstrFormats.td - ARM Instruction Formats -------*- tablegen -*-===//
 //
 //                     The LLVM Compiler Infrastructure
 //
@@ -25,7 +25,7 @@ def BrFrm         : Format<2>;
 def BrMiscFrm     : Format<3>;
 
 def DPFrm         : Format<4>;
-def DPSoRegFrm    : Format<5>;
+def DPSoRegRegFrm    : Format<5>;
 
 def LdFrm         : Format<6>;
 def StFrm         : Format<7>;
@@ -68,6 +68,7 @@ def N3RegVShFrm   : Format<38>;
 def NVExtFrm      : Format<39>;
 def NVMulSLFrm    : Format<40>;
 def NVTBLFrm      : Format<41>;
+def DPSoRegImmFrm  : Format<42>;
 
 // Misc flags.
 
@@ -130,39 +131,15 @@ def VFPNeonA8Domain : Domain<5>; // Instructions in VFP & Neon under A8
 // ARM special operands.
 //
 
-def CondCodeOperand : AsmOperandClass {
-  let Name = "CondCode";
-  let SuperClasses = [];
-}
-
-def CCOutOperand : AsmOperandClass {
-  let Name = "CCOut";
-  let SuperClasses = [];
-}
-
-def MemBarrierOptOperand : AsmOperandClass {
-  let Name = "MemBarrierOpt";
-  let SuperClasses = [];
-  let ParserMethod = "tryParseMemBarrierOptOperand";
-}
-
-def ProcIFlagsOperand : AsmOperandClass {
-  let Name = "ProcIFlags";
-  let SuperClasses = [];
-  let ParserMethod = "tryParseProcIFlagsOperand";
-}
-
-def MSRMaskOperand : AsmOperandClass {
-  let Name = "MSRMask";
-  let SuperClasses = [];
-  let ParserMethod = "tryParseMSRMaskOperand";
-}
-
 // ARM imod and iflag operands, used only by the CPS instruction.
 def imod_op : Operand<i32> {
   let PrintMethod = "printCPSIMod";
 }
 
+def ProcIFlagsOperand : AsmOperandClass {
+  let Name = "ProcIFlags";
+  let ParserMethod = "parseProcIFlagsOperand";
+}
 def iflags_op : Operand<i32> {
   let PrintMethod = "printCPSIFlag";
   let ParserMatchClass = ProcIFlagsOperand;
@@ -170,17 +147,31 @@ def iflags_op : Operand<i32> {
 
 // ARM Predicate operand. Default to 14 = always (AL). Second part is CC
 // register whose default is 0 (no register).
-def pred : PredicateOperand<OtherVT, (ops i32imm, CCR),
+def CondCodeOperand : AsmOperandClass { let Name = "CondCode"; }
+def pred : PredicateOperand<OtherVT, (ops i32imm, i32imm),
                                      (ops (i32 14), (i32 zero_reg))> {
   let PrintMethod = "printPredicateOperand";
   let ParserMatchClass = CondCodeOperand;
+  let DecoderMethod = "DecodePredicateOperand";
+}
+
+// Selectable predicate operand for CMOV instructions. We can't use a normal
+// predicate because the default values interfere with instruction selection. In
+// all other respects it is identical though: pseudo-instruction expansion
+// relies on the MachineOperands being compatible.
+def cmovpred : Operand<i32>, PredicateOp,
+               ComplexPattern<i32, 2, "SelectCMOVPred"> {
+  let MIOperandInfo = (ops i32imm, i32imm);
+  let PrintMethod = "printPredicateOperand";
 }
 
 // Conditional code result for instructions whose 's' bit is set, e.g. subs.
+def CCOutOperand : AsmOperandClass { let Name = "CCOut"; }
 def cc_out : OptionalDefOperand<OtherVT, (ops CCR), (ops (i32 zero_reg))> {
   let EncoderMethod = "getCCOutOpValue";
   let PrintMethod = "printSBitModifierOperand";
   let ParserMatchClass = CCOutOperand;
+  let DecoderMethod = "DecodeCCOutOperand";
 }
 
 // Same as cc_out except it defaults to setting CPSR.
@@ -188,19 +179,40 @@ def s_cc_out : OptionalDefOperand<OtherVT, (ops CCR), (ops (i32 CPSR))> {
   let EncoderMethod = "getCCOutOpValue";
   let PrintMethod = "printSBitModifierOperand";
   let ParserMatchClass = CCOutOperand;
+  let DecoderMethod = "DecodeCCOutOperand";
 }
 
 // ARM special operands for disassembly only.
 //
+def SetEndAsmOperand : ImmAsmOperand {
+  let Name = "SetEndImm";
+  let ParserMethod = "parseSetEndImm";
+}
 def setend_op : Operand<i32> {
   let PrintMethod = "printSetendOperand";
+  let ParserMatchClass = SetEndAsmOperand;
 }
 
+def MSRMaskOperand : AsmOperandClass {
+  let Name = "MSRMask";
+  let ParserMethod = "parseMSRMaskOperand";
+}
 def msr_mask : Operand<i32> {
   let PrintMethod = "printMSRMaskOperand";
+  let DecoderMethod = "DecodeMSRMask";
   let ParserMatchClass = MSRMaskOperand;
 }
 
+def BankedRegOperand : AsmOperandClass {
+  let Name = "BankedReg";
+  let ParserMethod = "parseBankedRegOperand";
+}
+def banked_reg : Operand<i32> {
+  let PrintMethod = "printBankedRegOperand";
+  let DecoderMethod = "DecodeBankedReg";
+  let ParserMatchClass = BankedRegOperand;
+}
+
 // Shift Right Immediate - A shift right immediate is encoded differently from
 // other shift immediates. The imm6 field is encoded like so:
 //
@@ -209,23 +221,60 @@ def msr_mask : Operand<i32> {
 //     16       imm6<5:4> = '01', 16 - <imm> is encoded in imm6<3:0>
 //     32       imm6<5> = '1', 32 - <imm> is encoded in imm6<4:0>
 //     64       64 - <imm> is encoded in imm6<5:0>
-def shr_imm8  : Operand<i32> {
+def shr_imm8_asm_operand : ImmAsmOperand { let Name = "ShrImm8"; }
+def shr_imm8  : Operand<i32>, ImmLeaf<i32, [{ return Imm > 0 && Imm <= 8; }]> {
   let EncoderMethod = "getShiftRight8Imm";
+  let DecoderMethod = "DecodeShiftRight8Imm";
+  let ParserMatchClass = shr_imm8_asm_operand;
 }
-def shr_imm16 : Operand<i32> {
+def shr_imm16_asm_operand : ImmAsmOperand { let Name = "ShrImm16"; }
+def shr_imm16 : Operand<i32>, ImmLeaf<i32, [{ return Imm > 0 && Imm <= 16; }]> {
   let EncoderMethod = "getShiftRight16Imm";
+  let DecoderMethod = "DecodeShiftRight16Imm";
+  let ParserMatchClass = shr_imm16_asm_operand;
 }
-def shr_imm32 : Operand<i32> {
+def shr_imm32_asm_operand : ImmAsmOperand { let Name = "ShrImm32"; }
+def shr_imm32 : Operand<i32>, ImmLeaf<i32, [{ return Imm > 0 && Imm <= 32; }]> {
   let EncoderMethod = "getShiftRight32Imm";
+  let DecoderMethod = "DecodeShiftRight32Imm";
+  let ParserMatchClass = shr_imm32_asm_operand;
 }
-def shr_imm64 : Operand<i32> {
+def shr_imm64_asm_operand : ImmAsmOperand { let Name = "ShrImm64"; }
+def shr_imm64 : Operand<i32>, ImmLeaf<i32, [{ return Imm > 0 && Imm <= 64; }]> {
   let EncoderMethod = "getShiftRight64Imm";
+  let DecoderMethod = "DecodeShiftRight64Imm";
+  let ParserMatchClass = shr_imm64_asm_operand;
 }
 
+//===----------------------------------------------------------------------===//
+// ARM Assembler alias templates.
+//
+class ARMInstAlias<string Asm, dag Result, bit Emit = 0b1>
+      : InstAlias<Asm, Result, Emit>, Requires<[IsARM]>;
+class  tInstAlias<string Asm, dag Result, bit Emit = 0b1>
+      : InstAlias<Asm, Result, Emit>, Requires<[IsThumb]>;
+class t2InstAlias<string Asm, dag Result, bit Emit = 0b1>
+      : InstAlias<Asm, Result, Emit>, Requires<[IsThumb2]>;
+class VFP2InstAlias<string Asm, dag Result, bit Emit = 0b1>
+      : InstAlias<Asm, Result, Emit>, Requires<[HasVFP2]>;
+class VFP2DPInstAlias<string Asm, dag Result, bit Emit = 0b1>
+      : InstAlias<Asm, Result, Emit>, Requires<[HasVFP2,HasDPVFP]>;
+class VFP3InstAlias<string Asm, dag Result, bit Emit = 0b1>
+      : InstAlias<Asm, Result, Emit>, Requires<[HasVFP3]>;
+class NEONInstAlias<string Asm, dag Result, bit Emit = 0b1>
+      : InstAlias<Asm, Result, Emit>, Requires<[HasNEON]>;
+
+
+class VFP2MnemonicAlias<string src, string dst> : MnemonicAlias<src, dst>,
+          Requires<[HasVFP2]>;
+class NEONMnemonicAlias<string src, string dst> : MnemonicAlias<src, dst>,
+          Requires<[HasNEON]>;
+
 //===----------------------------------------------------------------------===//
 // ARM Instruction templates.
 //
 
+
 class InstTemplate<AddrMode am, int sz, IndexMode im,
                    Format f, Domain d, string cstr, InstrItinClass itin>
   : Instruction {
@@ -240,17 +289,22 @@ class InstTemplate<AddrMode am, int sz, IndexMode im,
   Domain D = d;
   bit isUnaryDataProc = 0;
   bit canXformTo16Bit = 0;
+  // The instruction is a 16-bit flag setting Thumb instruction. Used
+  // by the parser to determine whether to require the 'S' suffix on the
+  // mnemonic (when not in an IT block) or preclude it (when in an IT block).
+  bit thumbArithFlagSetting = 0;
 
   // If this is a pseudo instruction, mark it isCodeGenOnly.
   let isCodeGenOnly = !eq(!cast<string>(f), "Pseudo");
 
-  // The layout of TSFlags should be kept in sync with ARMBaseInstrInfo.h.
+  // The layout of TSFlags should be kept in sync with ARMBaseInfo.h.
   let TSFlags{4-0}   = AM.Value;
   let TSFlags{6-5}   = IndexModeBits;
   let TSFlags{12-7} = Form;
   let TSFlags{13}    = isUnaryDataProc;
   let TSFlags{14}    = canXformTo16Bit;
   let TSFlags{17-15} = D.Value;
+  let TSFlags{18}    = thumbArithFlagSetting;
 
   let Constraints = cstr;
   let Itinerary = itin;
@@ -258,6 +312,14 @@ class InstTemplate<AddrMode am, int sz, IndexMode im,
 
 class Encoding {
   field bits<32> Inst;
+  // Mask of bits that cause an encoding to be UNPREDICTABLE.
+  // If a bit is set, then if the corresponding bit in the
+  // target encoding differs from its value in the "Inst" field,
+  // the instruction is UNPREDICTABLE (SoftFail in abstract parlance).
+  field bits<32> Unpredictable = 0;
+  // SoftFail is the generic name for this field, but we alias it so
+  // as to make it more obvious what it means in ARM-land.
+  field bits<32> SoftFail = Unpredictable;
 }
 
 class InstARM<AddrMode am, int sz, IndexMode im,
@@ -274,6 +336,32 @@ class InstThumb<AddrMode am, int sz, IndexMode im,
   let DecoderNamespace = "Thumb";
 }
 
+// Pseudo-instructions for alternate assembly syntax (never used by codegen).
+// These are aliases that require C++ handling to convert to the target
+// instruction, while InstAliases can be handled directly by tblgen.
+class AsmPseudoInst<string asm, dag iops, dag oops = (outs)>
+  : InstTemplate<AddrModeNone, 0, IndexModeNone, Pseudo, GenericDomain,
+                 "", NoItinerary> {
+  let OutOperandList = oops;
+  let InOperandList = iops;
+  let Pattern = [];
+  let isCodeGenOnly = 0; // So we get asm matcher for it.
+  let AsmString = asm;
+  let isPseudo = 1;
+}
+
+class ARMAsmPseudo<string asm, dag iops, dag oops = (outs)>
+  : AsmPseudoInst<asm, iops, oops>, Requires<[IsARM]>;
+class tAsmPseudo<string asm, dag iops, dag oops = (outs)>
+  : AsmPseudoInst<asm, iops, oops>, Requires<[IsThumb]>;
+class t2AsmPseudo<string asm, dag iops, dag oops = (outs)>
+  : AsmPseudoInst<asm, iops, oops>, Requires<[IsThumb2]>;
+class VFP2AsmPseudo<string asm, dag iops, dag oops = (outs)>
+  : AsmPseudoInst<asm, iops, oops>, Requires<[HasVFP2]>;
+class NEONAsmPseudo<string asm, dag iops, dag oops = (outs)>
+  : AsmPseudoInst<asm, iops, oops>, Requires<[HasNEON]>;
+
+// Pseudo instructions for the code generator.
 class PseudoInst<dag oops, dag iops, InstrItinClass itin, list<dag> pattern>
   : InstTemplate<AddrModeNone, 0, IndexModeNone, Pseudo,
                  GenericDomain, "", itin> {
@@ -399,6 +487,10 @@ class AXI<dag oops, dag iops, Format f, InstrItinClass itin,
           string asm, list<dag> pattern>
   : XI<oops, iops, AddrModeNone, 4, IndexModeNone, f, itin,
        asm, "", pattern>;
+class AXIM<dag oops, dag iops, AddrMode am, Format f, InstrItinClass itin,
+          string asm, list<dag> pattern>
+  : XI<oops, iops, am, 4, IndexModeNone, f, itin,
+       asm, "", pattern>;
 class AInoP<dag oops, dag iops, Format f, InstrItinClass itin,
             string opc, string asm, list<dag> pattern>
   : InoP<oops, iops, AddrModeNone, 4, IndexModeNone, f, itin,
@@ -424,47 +516,90 @@ class JTI<dag oops, dag iops, InstrItinClass itin,
   : XI<oops, iops, AddrModeNone, 0, IndexModeNone, BrMiscFrm, itin,
        asm, "", pattern>;
 
-// Atomic load/store instructions
-class AIldrex<bits<2> opcod, dag oops, dag iops, InstrItinClass itin,
+class AIldr_ex_or_acq<bits<2> opcod, bits<2> opcod2, dag oops, dag iops, InstrItinClass itin,
               string opc, string asm, list<dag> pattern>
   : I<oops, iops, AddrModeNone, 4, IndexModeNone, LdStExFrm, itin,
       opc, asm, "", pattern> {
   bits<4> Rt;
-  bits<4> Rn;
+  bits<4> addr;
   let Inst{27-23} = 0b00011;
   let Inst{22-21} = opcod;
   let Inst{20}    = 1;
-  let Inst{19-16} = Rn;
+  let Inst{19-16} = addr;
   let Inst{15-12} = Rt;
-  let Inst{11-0}  = 0b111110011111;
+  let Inst{11-10} = 0b11;
+  let Inst{9-8}   = opcod2;
+  let Inst{7-0}   = 0b10011111;
 }
-class AIstrex<bits<2> opcod, dag oops, dag iops, InstrItinClass itin,
+class AIstr_ex_or_rel<bits<2> opcod, bits<2> opcod2, dag oops, dag iops, InstrItinClass itin,
               string opc, string asm, list<dag> pattern>
   : I<oops, iops, AddrModeNone, 4, IndexModeNone, LdStExFrm, itin,
       opc, asm, "", pattern> {
-  bits<4> Rd;
   bits<4> Rt;
   bits<4> addr;
   let Inst{27-23} = 0b00011;
   let Inst{22-21} = opcod;
   let Inst{20}    = 0;
   let Inst{19-16} = addr;
-  let Inst{15-12} = Rd;
-  let Inst{11-4}  = 0b11111001;
+  let Inst{11-10} = 0b11;
+  let Inst{9-8}   = opcod2;
+  let Inst{7-4}   = 0b1001;
   let Inst{3-0}   = Rt;
 }
+// Atomic load/store instructions
+class AIldrex<bits<2> opcod, dag oops, dag iops, InstrItinClass itin,
+              string opc, string asm, list<dag> pattern>
+  : AIldr_ex_or_acq<opcod, 0b11, oops, iops, itin, opc, asm, pattern>;
+
+class AIstrex<bits<2> opcod, dag oops, dag iops, InstrItinClass itin,
+              string opc, string asm, list<dag> pattern>
+  : AIstr_ex_or_rel<opcod, 0b11, oops, iops, itin, opc, asm, pattern> {
+  bits<4> Rd;
+  let Inst{15-12} = Rd;
+}
+
+// Exclusive load/store instructions
+
+class AIldaex<bits<2> opcod, dag oops, dag iops, InstrItinClass itin,
+              string opc, string asm, list<dag> pattern>
+  : AIldr_ex_or_acq<opcod, 0b10, oops, iops, itin, opc, asm, pattern>,
+    Requires<[IsARM, HasV8]>;
+
+class AIstlex<bits<2> opcod, dag oops, dag iops, InstrItinClass itin,
+              string opc, string asm, list<dag> pattern>
+  : AIstr_ex_or_rel<opcod, 0b10, oops, iops, itin, opc, asm, pattern>,
+    Requires<[IsARM, HasV8]> {
+  bits<4> Rd;
+  let Inst{15-12} = Rd;
+}
+
 class AIswp<bit b, dag oops, dag iops, string opc, list<dag> pattern>
-  : AI<oops, iops, MiscFrm, NoItinerary, opc, "\t$Rt, $Rt2, [$Rn]", pattern> {
+  : AI<oops, iops, MiscFrm, NoItinerary, opc, "\t$Rt, $Rt2, $addr", pattern> {
   bits<4> Rt;
   bits<4> Rt2;
-  bits<4> Rn;
+  bits<4> addr;
   let Inst{27-23} = 0b00010;
   let Inst{22} = b;
   let Inst{21-20} = 0b00;
-  let Inst{19-16} = Rn;
+  let Inst{19-16} = addr;
   let Inst{15-12} = Rt;
   let Inst{11-4} = 0b00001001;
   let Inst{3-0} = Rt2;
+
+  let Unpredictable{11-8} = 0b1111;
+  let DecoderMethod = "DecodeSwap";
+}
+// Acquire/Release load/store instructions
+class AIldracq<bits<2> opcod, dag oops, dag iops, InstrItinClass itin,
+              string opc, string asm, list<dag> pattern>
+  : AIldr_ex_or_acq<opcod, 0b00, oops, iops, itin, opc, asm, pattern>,
+    Requires<[IsARM, HasV8]>;
+
+class AIstrrel<bits<2> opcod, dag oops, dag iops, InstrItinClass itin,
+              string opc, string asm, list<dag> pattern>
+  : AIstr_ex_or_rel<opcod, 0b00, oops, iops, itin, opc, asm, pattern>,
+    Requires<[IsARM, HasV8]> {
+  let Inst{15-12}   = 0b1111;
 }
 
 // addrmode1 instructions
@@ -519,22 +654,41 @@ class AI2ldstidx<bit isLd, bit isByte, bit isPre, dag oops, dag iops,
   let Inst{20}    = isLd; // L bit
   let Inst{15-12} = Rt;
 }
-class AI2stridx<bit isByte, bit isPre, dag oops, dag iops,
+class AI2stridx_reg<bit isByte, bit isPre, dag oops, dag iops,
+                IndexMode im, Format f, InstrItinClass itin, string opc,
+                string asm, string cstr, list<dag> pattern>
+  : AI2ldstidx<0, isByte, isPre, oops, iops, im, f, itin, opc, asm, cstr,
+               pattern> {
+  // AM2 store w/ two operands: (GPR, am2offset)
+  // {12}     isAdd
+  // {11-0}   imm12/Rm
+  bits<14> offset;
+  bits<4> Rn;
+  let Inst{25} = 1;
+  let Inst{23} = offset{12};
+  let Inst{19-16} = Rn;
+  let Inst{11-5} = offset{11-5};
+  let Inst{4} = 0;
+  let Inst{3-0} = offset{3-0};
+}
+
+class AI2stridx_imm<bit isByte, bit isPre, dag oops, dag iops,
                 IndexMode im, Format f, InstrItinClass itin, string opc,
                 string asm, string cstr, list<dag> pattern>
   : AI2ldstidx<0, isByte, isPre, oops, iops, im, f, itin, opc, asm, cstr,
                pattern> {
   // AM2 store w/ two operands: (GPR, am2offset)
-  // {13}     1 == Rm, 0 == imm12
   // {12}     isAdd
   // {11-0}   imm12/Rm
   bits<14> offset;
   bits<4> Rn;
-  let Inst{25} = offset{13};
+  let Inst{25} = 0;
   let Inst{23} = offset{12};
   let Inst{19-16} = Rn;
   let Inst{11-0} = offset{11-0};
 }
+
+
 // FIXME: Merge with the above class when addrmode2 gets used for STR, STRB
 // but for now use this class for STRT and STRBT.
 class AI2stridxT<bit isByte, bit isPre, dag oops, dag iops,
@@ -572,9 +726,11 @@ class AI3ld<bits<4> op, bit op20, dag oops, dag iops, Format f,
   let Inst{11-8}  = addr{7-4};    // imm7_4/zero
   let Inst{7-4}   = op;
   let Inst{3-0}   = addr{3-0};    // imm3_0/Rm
+
+  let DecoderMethod = "DecodeAddrMode3Instruction";
 }
 
-class AI3ldstidx<bits<4> op, bit op20, bit isLd, bit isPre, dag oops, dag iops,
+class AI3ldstidx<bits<4> op, bit op20, bit isPre, dag oops, dag iops,
                 IndexMode im, Format f, InstrItinClass itin, string opc,
                 string asm, string cstr, list<dag> pattern>
   : I<oops, iops, AddrMode3, 4, im, f, itin,
@@ -590,48 +746,24 @@ class AI3ldstidx<bits<4> op, bit op20, bit isLd, bit isPre, dag oops, dag iops,
 
 // FIXME: Merge with the above class when addrmode2 gets used for LDR, LDRB
 // but for now use this class for LDRSBT, LDRHT, LDSHT.
-class AI3ldstidxT<bits<4> op, bit op20, bit isLd, bit isPre, dag oops, dag iops,
+class AI3ldstidxT<bits<4> op, bit isLoad, dag oops, dag iops,
                   IndexMode im, Format f, InstrItinClass itin, string opc,
                   string asm, string cstr, list<dag> pattern>
-  : I<oops, iops, AddrMode3, 4, im, f, itin,
-      opc, asm, cstr, pattern> {
+  : I<oops, iops, AddrMode3, 4, im, f, itin, opc, asm, cstr, pattern> {
   // {13}     1 == imm8, 0 == Rm
   // {12-9}   Rn
   // {8}      isAdd
   // {7-4}    imm7_4/zero
   // {3-0}    imm3_0/Rm
-  bits<14> addr;
-  bits<4> Rt;
-  let Inst{27-25} = 0b000;
-  let Inst{24}    = isPre;        // P bit
-  let Inst{23}    = addr{8};      // U bit
-  let Inst{22}    = addr{13};     // 1 == imm8, 0 == Rm
-  let Inst{20}    = op20;         // L bit
-  let Inst{19-16} = addr{12-9};   // Rn
-  let Inst{15-12} = Rt;           // Rt
-  let Inst{11-8}  = addr{7-4};    // imm7_4/zero
-  let Inst{7-4}   = op;
-  let Inst{3-0}   = addr{3-0};    // imm3_0/Rm
-  let AsmMatchConverter = "CvtLdWriteBackRegAddrMode3";
-}
-
-class AI3stridx<bits<4> op, bit isByte, bit isPre, dag oops, dag iops,
-                IndexMode im, Format f, InstrItinClass itin, string opc,
-                string asm, string cstr, list<dag> pattern>
-  : AI2ldstidx<0, isByte, isPre, oops, iops, im, f, itin, opc, asm, cstr,
-               pattern> {
-  // AM3 store w/ two operands: (GPR, am3offset)
-  bits<14> offset;
+  bits<4> addr;
   bits<4> Rt;
-  bits<4> Rn;
   let Inst{27-25} = 0b000;
-  let Inst{23}    = offset{8};
-  let Inst{22}    = offset{9};
-  let Inst{19-16} = Rn;
+  let Inst{24}    = 0;            // P bit
+  let Inst{21}    = 1;
+  let Inst{20}    = isLoad;       // L bit
+  let Inst{19-16} = addr;         // Rn
   let Inst{15-12} = Rt;           // Rt
-  let Inst{11-8}  = offset{7-4};  // imm7_4/zero
   let Inst{7-4}   = op;
-  let Inst{3-0}   = offset{3-0};  // imm3_0/Rm
 }
 
 // stores
@@ -652,75 +784,7 @@ class AI3str<bits<4> op, dag oops, dag iops, Format f, InstrItinClass itin,
   let Inst{11-8}  = addr{7-4};    // imm7_4/zero
   let Inst{7-4}   = op;
   let Inst{3-0}   = addr{3-0};    // imm3_0/Rm
-}
-
-// Pre-indexed stores
-class AI3sthpr<dag oops, dag iops, Format f, InstrItinClass itin,
-               string opc, string asm, string cstr, list<dag> pattern>
-  : I<oops, iops, AddrMode3, 4, IndexModePre, f, itin,
-      opc, asm, cstr, pattern> {
-  let Inst{4}     = 1;
-  let Inst{5}     = 1; // H bit
-  let Inst{6}     = 0; // S bit
-  let Inst{7}     = 1;
-  let Inst{20}    = 0; // L bit
-  let Inst{21}    = 1; // W bit
-  let Inst{24}    = 1; // P bit
-  let Inst{27-25} = 0b000;
-}
-class AI3stdpr<dag oops, dag iops, Format f, InstrItinClass itin,
-             string opc, string asm, string cstr, list<dag> pattern>
-  : I<oops, iops, AddrMode3, 4, IndexModePre, f, itin,
-      opc, asm, cstr, pattern> {
-  let Inst{4}     = 1;
-  let Inst{5}     = 1; // H bit
-  let Inst{6}     = 1; // S bit
-  let Inst{7}     = 1;
-  let Inst{20}    = 0; // L bit
-  let Inst{21}    = 1; // W bit
-  let Inst{24}    = 1; // P bit
-  let Inst{27-25} = 0b000;
-}
-
-// Post-indexed stores
-class AI3sthpo<dag oops, dag iops, Format f, InstrItinClass itin,
-               string opc, string asm, string cstr, list<dag> pattern>
-  : I<oops, iops, AddrMode3, 4, IndexModePost, f, itin,
-      opc, asm, cstr,pattern> {
-  // {13}     1 == imm8, 0 == Rm
-  // {12-9}   Rn
-  // {8}      isAdd
-  // {7-4}    imm7_4/zero
-  // {3-0}    imm3_0/Rm
-  bits<14> addr;
-  bits<4> Rt;
-  let Inst{3-0}   = addr{3-0};    // imm3_0/Rm
-  let Inst{4}     = 1;
-  let Inst{5}     = 1; // H bit
-  let Inst{6}     = 0; // S bit
-  let Inst{7}     = 1;
-  let Inst{11-8}  = addr{7-4};    // imm7_4/zero
-  let Inst{15-12} = Rt;           // Rt
-  let Inst{19-16} = addr{12-9};   // Rn
-  let Inst{20}    = 0; // L bit
-  let Inst{21}    = 0; // W bit
-  let Inst{22}    = addr{13};     // 1 == imm8, 0 == Rm
-  let Inst{23}    = addr{8};      // U bit
-  let Inst{24}    = 0; // P bit
-  let Inst{27-25} = 0b000;
-}
-class AI3stdpo<dag oops, dag iops, Format f, InstrItinClass itin,
-             string opc, string asm, string cstr, list<dag> pattern>
-  : I<oops, iops, AddrMode3, 4, IndexModePost, f, itin,
-      opc, asm, cstr, pattern> {
-  let Inst{4}     = 1;
-  let Inst{5}     = 1; // H bit
-  let Inst{6}     = 1; // S bit
-  let Inst{7}     = 1;
-  let Inst{20}    = 0; // L bit
-  let Inst{21}    = 0; // W bit
-  let Inst{24}    = 0; // P bit
-  let Inst{27-25} = 0b000;
+  let DecoderMethod = "DecodeAddrMode3Instruction";
 }
 
 // addrmode4 instructions
@@ -829,6 +893,8 @@ class AExtI<bits<8> opcod, dag oops, dag iops, InstrItinClass itin,
   let Inst{7-4}   = 0b0111;
   let Inst{9-8}   = 0b00;
   let Inst{27-20} = opcod;
+
+  let Unpredictable{9-8} = 0b11;
 }
 
 // Misc Arithmetic instructions.
@@ -846,7 +912,41 @@ class AMiscA1I<bits<8> opcod, bits<4> opc7_4, dag oops, dag iops,
   let Inst{3-0}   = Rm;
 }
 
+// Division instructions.
+class ADivA1I<bits<3> opcod, dag oops, dag iops,
+              InstrItinClass itin, string opc, string asm, list<dag> pattern>
+  : I<oops, iops, AddrModeNone, 4, IndexModeNone, ArithMiscFrm, itin,
+      opc, asm, "", pattern> {
+  bits<4> Rd;
+  bits<4> Rn;
+  bits<4> Rm;
+  let Inst{27-23} = 0b01110;
+  let Inst{22-20} = opcod;
+  let Inst{19-16} = Rd;
+  let Inst{15-12} = 0b1111;
+  let Inst{11-8}  = Rm;
+  let Inst{7-4}   = 0b0001;
+  let Inst{3-0}   = Rn;
+}
+
 // PKH instructions
+def PKHLSLAsmOperand : ImmAsmOperand {
+  let Name = "PKHLSLImm";
+  let ParserMethod = "parsePKHLSLImm";
+}
+def pkh_lsl_amt: Operand<i32>, ImmLeaf<i32, [{ return Imm >= 0 && Imm < 32; }]>{
+  let PrintMethod = "printPKHLSLShiftImm";
+  let ParserMatchClass = PKHLSLAsmOperand;
+}
+def PKHASRAsmOperand : AsmOperandClass {
+  let Name = "PKHASRImm";
+  let ParserMethod = "parsePKHASRImm";
+}
+def pkh_asr_amt: Operand<i32>, ImmLeaf<i32, [{ return Imm > 0 && Imm <= 32; }]>{
+  let PrintMethod = "printPKHASRShiftImm";
+  let ParserMatchClass = PKHASRAsmOperand;
+}
+
 class APKHI<bits<8> opcod, bit tb, dag oops, dag iops, InstrItinClass itin,
             string opc, string asm, list<dag> pattern>
   : I<oops, iops, AddrModeNone, 4, IndexModeNone, ArithMiscFrm, itin,
@@ -854,11 +954,11 @@ class APKHI<bits<8> opcod, bit tb, dag oops, dag iops, InstrItinClass itin,
   bits<4> Rd;
   bits<4> Rn;
   bits<4> Rm;
-  bits<8> sh;
+  bits<5> sh;
   let Inst{27-20} = opcod;
   let Inst{19-16} = Rn;
   let Inst{15-12} = Rd;
-  let Inst{11-7}  = sh{7-3};
+  let Inst{11-7}  = sh;
   let Inst{6}     = tb;
   let Inst{5-4}   = 0b01;
   let Inst{3-0}   = Rm;
@@ -876,6 +976,10 @@ class ARMV5TPat<dag pattern, dag result> : Pat<pattern, result> {
 class ARMV5TEPat<dag pattern, dag result> : Pat<pattern, result> {
   list<Predicate> Predicates = [IsARM, HasV5TE];
 }
+// ARMV5MOPat - Same as ARMV5TEPat with UseMulOps.
+class ARMV5MOPat<dag pattern, dag result> : Pat<pattern, result> {
+  list<Predicate> Predicates = [IsARM, HasV5TE, UseMulOps];
+}
 class ARMV6Pat<dag pattern, dag result> : Pat<pattern, result> {
   list<Predicate> Predicates = [IsARM, HasV6];
 }
@@ -953,7 +1057,9 @@ class Thumb1sI<dag oops, dag iops, AddrMode am, int sz,
   let InOperandList = !con(iops, (ins pred:$p));
   let AsmString = !strconcat(opc, "${s}${p}", asm);
   let Pattern = pattern;
+  let thumbArithFlagSetting = 1;
   list<Predicate> Predicates = [IsThumb, IsThumb1Only];
+  let DecoderNamespace = "ThumbSBit";
 }
 
 class T1sI<dag oops, dag iops, InstrItinClass itin,
@@ -1140,8 +1246,8 @@ class T2Ipc<dag oops, dag iops, InstrItinClass itin,
             string opc, string asm, list<dag> pattern>
   : Thumb2I<oops, iops, AddrModeT2_pc, 4, itin, opc, asm, "", pattern>;
 class T2Ii8s4<bit P, bit W, bit isLoad, dag oops, dag iops, InstrItinClass itin,
-              string opc, string asm, list<dag> pattern>
-  : Thumb2I<oops, iops, AddrModeT2_i8s4, 4, itin, opc, asm, "",
+              string opc, string asm, string cstr, list<dag> pattern>
+  : Thumb2I<oops, iops, AddrModeT2_i8s4, 4, itin, opc, asm, cstr,
             pattern> {
   bits<4> Rt;
   bits<4> Rt2;
@@ -1157,6 +1263,26 @@ class T2Ii8s4<bit P, bit W, bit isLoad, dag oops, dag iops, InstrItinClass itin,
   let Inst{11-8}  = Rt2{3-0};
   let Inst{7-0}   = addr{7-0};
 }
+class T2Ii8s4post<bit P, bit W, bit isLoad, dag oops, dag iops,
+                  InstrItinClass itin, string opc, string asm, string cstr,
+                  list<dag> pattern>
+  : Thumb2I<oops, iops, AddrModeT2_i8s4, 4, itin, opc, asm, cstr,
+            pattern> {
+  bits<4> Rt;
+  bits<4> Rt2;
+  bits<4> addr;
+  bits<9> imm;
+  let Inst{31-25} = 0b1110100;
+  let Inst{24}    = P;
+  let Inst{23}    = imm{8};
+  let Inst{22}    = 1;
+  let Inst{21}    = W;
+  let Inst{20}    = isLoad;
+  let Inst{19-16} = addr;
+  let Inst{15-12} = Rt{3-0};
+  let Inst{11-8}  = Rt2{3-0};
+  let Inst{7-0}   = imm{7-0};
+}
 
 class T2sI<dag oops, dag iops, InstrItinClass itin,
            string opc, string asm, list<dag> pattern>
@@ -1170,8 +1296,9 @@ class T2JTI<dag oops, dag iops, InstrItinClass itin,
   : Thumb2XI<oops, iops, AddrModeNone, 0, itin, asm, "", pattern>;
 
 // Move to/from coprocessor instructions
-class T2Cop<bits<4> opc, dag oops, dag iops, string asm, list<dag> pattern>
-  : T2XI <oops, iops, NoItinerary, asm, pattern>, Requires<[IsThumb2]> {
+class T2Cop<bits<4> opc, dag oops, dag iops, string opcstr, string asm,
+            list<dag> pattern>
+  : T2I <oops, iops, NoItinerary, opcstr, asm, pattern>, Requires<[IsThumb2]> {
   let Inst{31-28} = opc;
 }
 
@@ -1180,8 +1307,8 @@ class T2XIt<dag oops, dag iops, InstrItinClass itin,
             string asm, string cstr, list<dag> pattern>
   : Thumb2XI<oops, iops, AddrModeNone, 4, itin, asm, cstr, pattern>;
 
-// T2Iidxldst - Thumb2 indexed load / store instructions.
-class T2Iidxldst<bit signed, bits<2> opcod, bit load, bit pre,
+// T2Ipreldst - Thumb2 pre-indexed load / store instructions.
+class T2Ipreldst<bit signed, bits<2> opcod, bit load, bit pre,
                  dag oops, dag iops,
                  AddrMode am, IndexMode im, InstrItinClass itin,
                  string opc, string asm, string cstr, list<dag> pattern>
@@ -1192,25 +1319,59 @@ class T2Iidxldst<bit signed, bits<2> opcod, bit load, bit pre,
   let Pattern = pattern;
   list<Predicate> Predicates = [IsThumb2];
   let DecoderNamespace = "Thumb2";
+
+  bits<4> Rt;
+  bits<13> addr;
   let Inst{31-27} = 0b11111;
   let Inst{26-25} = 0b00;
   let Inst{24}    = signed;
   let Inst{23}    = 0;
   let Inst{22-21} = opcod;
   let Inst{20}    = load;
+  let Inst{19-16} = addr{12-9};
+  let Inst{15-12} = Rt{3-0};
   let Inst{11}    = 1;
   // (P, W) = (1, 1) Pre-indexed or (0, 1) Post-indexed
   let Inst{10}    = pre; // The P bit.
+  let Inst{9}     = addr{8}; // Sign bit
   let Inst{8}     = 1; // The W bit.
+  let Inst{7-0}   = addr{7-0};
 
-  bits<9> addr;
-  let Inst{7-0} = addr{7-0};
-  let Inst{9}   = addr{8}; // Sign bit
+  let DecoderMethod = "DecodeT2LdStPre";
+}
+
+// T2Ipostldst - Thumb2 post-indexed load / store instructions.
+class T2Ipostldst<bit signed, bits<2> opcod, bit load, bit pre,
+                 dag oops, dag iops,
+                 AddrMode am, IndexMode im, InstrItinClass itin,
+                 string opc, string asm, string cstr, list<dag> pattern>
+  : InstARM<am, 4, im, ThumbFrm, GenericDomain, cstr, itin> {
+  let OutOperandList = oops;
+  let InOperandList = !con(iops, (ins pred:$p));
+  let AsmString = !strconcat(opc, "${p}", asm);
+  let Pattern = pattern;
+  list<Predicate> Predicates = [IsThumb2];
+  let DecoderNamespace = "Thumb2";
 
   bits<4> Rt;
   bits<4> Rn;
+  bits<9> offset;
+  let Inst{31-27} = 0b11111;
+  let Inst{26-25} = 0b00;
+  let Inst{24}    = signed;
+  let Inst{23}    = 0;
+  let Inst{22-21} = opcod;
+  let Inst{20}    = load;
+  let Inst{19-16} = Rn;
   let Inst{15-12} = Rt{3-0};
-  let Inst{19-16} = Rn{3-0};
+  let Inst{11}    = 1;
+  // (P, W) = (1, 1) Pre-indexed or (0, 1) Post-indexed
+  let Inst{10}    = pre; // The P bit.
+  let Inst{9}     = offset{8}; // Sign bit
+  let Inst{8}     = 1; // The W bit.
+  let Inst{7-0}   = offset{7-0};
+
+  let DecoderMethod = "DecodeT2LdStPre";
 }
 
 // Tv5Pat - Same as Pat<>, but requires V5T Thumb mode.
@@ -1251,6 +1412,7 @@ class VFPI<dag oops, dag iops, AddrMode am, int sz,
   let AsmString = !strconcat(opc, "${p}", asm);
   let Pattern = pattern;
   let PostEncoderMethod = "VFPThumb2PostEncoder";
+  let DecoderNamespace = "VFP";
   list<Predicate> Predicates = [HasVFP2];
 }
 
@@ -1266,6 +1428,7 @@ class VFPXI<dag oops, dag iops, AddrMode am, int sz,
   let AsmString = asm;
   let Pattern = pattern;
   let PostEncoderMethod = "VFPThumb2PostEncoder";
+  let DecoderNamespace = "VFP";
   list<Predicate> Predicates = [HasVFP2];
 }
 
@@ -1293,7 +1456,6 @@ class ADI5<bits<4> opcod1, bits<2> opcod2, dag oops, dag iops,
   let Inst{15-12} = Dd{3-0};
   let Inst{7-0}   = addr{7-0};    // imm8
 
-  // TODO: Mark the instructions with the appropriate subtarget info.
   let Inst{27-24} = opcod1;
   let Inst{21-20} = opcod2;
   let Inst{11-9}  = 0b101;
@@ -1319,7 +1481,6 @@ class ASI5<bits<4> opcod1, bits<2> opcod2, dag oops, dag iops,
   let Inst{15-12} = Sd{4-1};
   let Inst{7-0}   = addr{7-0};    // imm8
 
-  // TODO: Mark the instructions with the appropriate subtarget info.
   let Inst{27-24} = opcod1;
   let Inst{21-20} = opcod2;
   let Inst{11-9}  = 0b101;
@@ -1341,6 +1502,28 @@ class PseudoVFPLdStM<dag oops, dag iops, InstrItinClass itin, string cstr,
 }
 
 // Load / store multiple
+
+// Unknown precision
+class AXXI4<dag oops, dag iops, IndexMode im,
+            string asm, string cstr, list<dag> pattern>
+  : VFPXI<oops, iops, AddrMode4, 4, im,
+          VFPLdStFrm, NoItinerary, asm, cstr, pattern> {
+  // Instruction operands.
+  bits<4>  Rn;
+  bits<13> regs;
+
+  // Encode instruction operands.
+  let Inst{19-16} = Rn;
+  let Inst{22}    = 0;
+  let Inst{15-12} = regs{11-8};
+  let Inst{7-1}   = regs{7-1};
+
+  let Inst{27-25} = 0b110;
+  let Inst{11-8}  = 0b1011;
+  let Inst{0}     = 1;
+}
+
+// Double precision
 class AXDI4<dag oops, dag iops, IndexMode im, InstrItinClass itin,
             string asm, string cstr, list<dag> pattern>
   : VFPXI<oops, iops, AddrMode4, 4, im,
@@ -1353,14 +1536,15 @@ class AXDI4<dag oops, dag iops, IndexMode im, InstrItinClass itin,
   let Inst{19-16} = Rn;
   let Inst{22}    = regs{12};
   let Inst{15-12} = regs{11-8};
-  let Inst{7-0}   = regs{7-0};
+  let Inst{7-1}   = regs{7-1};
 
-  // TODO: Mark the instructions with the appropriate subtarget info.
   let Inst{27-25} = 0b110;
   let Inst{11-9}  = 0b101;
   let Inst{8}     = 1;          // Double precision
+  let Inst{0}     = 0;
 }
 
+// Single Precision
 class AXSI4<dag oops, dag iops, IndexMode im, InstrItinClass itin,
             string asm, string cstr, list<dag> pattern>
   : VFPXI<oops, iops, AddrMode4, 4, im,
@@ -1375,7 +1559,6 @@ class AXSI4<dag oops, dag iops, IndexMode im, InstrItinClass itin,
   let Inst{15-12} = regs{12-9};
   let Inst{7-0}   = regs{7-0};
 
-  // TODO: Mark the instructions with the appropriate subtarget info.
   let Inst{27-25} = 0b110;
   let Inst{11-9}  = 0b101;
   let Inst{8}     = 0;          // Single precision
@@ -1403,6 +1586,34 @@ class ADuI<bits<5> opcod1, bits<2> opcod2, bits<4> opcod3, bits<2> opcod4,
   let Inst{8}     = 1;          // Double precision
   let Inst{7-6}   = opcod4;
   let Inst{4}     = opcod5;
+
+  let Predicates = [HasVFP2, HasDPVFP];
+}
+
+// Double precision, unary, not-predicated
+class ADuInp<bits<5> opcod1, bits<2> opcod2, bits<4> opcod3, bits<2> opcod4,
+           bit opcod5, dag oops, dag iops, InstrItinClass itin,
+           string asm, list<dag> pattern>
+  : VFPXI<oops, iops, AddrModeNone, 4, IndexModeNone, VFPUnaryFrm, itin, asm, "", pattern> {
+  // Instruction operands.
+  bits<5> Dd;
+  bits<5> Dm;
+
+  let Inst{31-28} = 0b1111;
+
+  // Encode instruction operands.
+  let Inst{3-0}   = Dm{3-0};
+  let Inst{5}     = Dm{4};
+  let Inst{15-12} = Dd{3-0};
+  let Inst{22}    = Dd{4};
+
+  let Inst{27-23} = opcod1;
+  let Inst{21-20} = opcod2;
+  let Inst{19-16} = opcod3;
+  let Inst{11-9}  = 0b101;
+  let Inst{8}     = 1;          // Double precision
+  let Inst{7-6}   = opcod4;
+  let Inst{4}     = opcod5;
 }
 
 // Double precision, binary
@@ -1429,9 +1640,42 @@ class ADbI<bits<5> opcod1, bits<2> opcod2, bit op6, bit op4, dag oops,
   let Inst{8}     = 1;          // Double precision
   let Inst{6}     = op6;
   let Inst{4}     = op4;
+
+  let Predicates = [HasVFP2, HasDPVFP];
+}
+
+// FP, binary, not predicated
+class ADbInp<bits<5> opcod1, bits<2> opcod2, bit opcod3, dag oops, dag iops,
+           InstrItinClass itin, string asm, list<dag> pattern>
+  : VFPXI<oops, iops, AddrModeNone, 4, IndexModeNone, VFPBinaryFrm, itin,
+          asm, "", pattern>
+{
+  // Instruction operands.
+  bits<5> Dd;
+  bits<5> Dn;
+  bits<5> Dm;
+
+  let Inst{31-28} = 0b1111;
+
+  // Encode instruction operands.
+  let Inst{3-0}   = Dm{3-0};
+  let Inst{5}     = Dm{4};
+  let Inst{19-16} = Dn{3-0};
+  let Inst{7}     = Dn{4};
+  let Inst{15-12} = Dd{3-0};
+  let Inst{22}    = Dd{4};
+
+  let Inst{27-23} = opcod1;
+  let Inst{21-20} = opcod2;
+  let Inst{11-9}  = 0b101;
+  let Inst{8}     = 1; // double precision
+  let Inst{6}     = opcod3;
+  let Inst{4}     = 0;
+
+  let Predicates = [HasVFP2, HasDPVFP];
 }
 
-// Single precision, unary
+// Single precision, unary, predicated
 class ASuI<bits<5> opcod1, bits<2> opcod2, bits<4> opcod3, bits<2> opcod4,
            bit opcod5, dag oops, dag iops, InstrItinClass itin, string opc,
            string asm, list<dag> pattern>
@@ -1455,6 +1699,33 @@ class ASuI<bits<5> opcod1, bits<2> opcod2, bits<4> opcod3, bits<2> opcod4,
   let Inst{4}     = opcod5;
 }
 
+// Single precision, unary, non-predicated
+class ASuInp<bits<5> opcod1, bits<2> opcod2, bits<4> opcod3, bits<2> opcod4,
+             bit opcod5, dag oops, dag iops, InstrItinClass itin,
+             string asm, list<dag> pattern>
+  : VFPXI<oops, iops, AddrModeNone, 4, IndexModeNone,
+          VFPUnaryFrm, itin, asm, "", pattern> {
+  // Instruction operands.
+  bits<5> Sd;
+  bits<5> Sm;
+
+  let Inst{31-28} = 0b1111;
+
+  // Encode instruction operands.
+  let Inst{3-0}   = Sm{4-1};
+  let Inst{5}     = Sm{0};
+  let Inst{15-12} = Sd{4-1};
+  let Inst{22}    = Sd{0};
+
+  let Inst{27-23} = opcod1;
+  let Inst{21-20} = opcod2;
+  let Inst{19-16} = opcod3;
+  let Inst{11-9}  = 0b101;
+  let Inst{8}     = 0;          // Single precision
+  let Inst{7-6}   = opcod4;
+  let Inst{4}     = opcod5;
+}
+
 // Single precision unary, if no NEON. Same as ASuI except not available if
 // NEON is enabled.
 class ASuIn<bits<5> opcod1, bits<2> opcod2, bits<4> opcod3, bits<2> opcod4,
@@ -1490,6 +1761,35 @@ class ASbI<bits<5> opcod1, bits<2> opcod2, bit op6, bit op4, dag oops, dag iops,
   let Inst{4}     = op4;
 }
 
+// Single precision, binary, not predicated
+class ASbInp<bits<5> opcod1, bits<2> opcod2, bit opcod3, dag oops, dag iops,
+           InstrItinClass itin, string asm, list<dag> pattern>
+  : VFPXI<oops, iops, AddrModeNone, 4, IndexModeNone,
+          VFPBinaryFrm, itin, asm, "", pattern>
+{
+  // Instruction operands.
+  bits<5> Sd;
+  bits<5> Sn;
+  bits<5> Sm;
+
+  let Inst{31-28} = 0b1111;
+
+  // Encode instruction operands.
+  let Inst{3-0}   = Sm{4-1};
+  let Inst{5}     = Sm{0};
+  let Inst{19-16} = Sn{4-1};
+  let Inst{7}     = Sn{0};
+  let Inst{15-12} = Sd{4-1};
+  let Inst{22}    = Sd{0};
+
+  let Inst{27-23} = opcod1;
+  let Inst{21-20} = opcod2;
+  let Inst{11-9}  = 0b101;
+  let Inst{8}     = 0; // Single precision
+  let Inst{6}     = opcod3;
+  let Inst{4}     = 0;
+}
+
 // Single precision binary, if no NEON. Same as ASbI except not available if
 // NEON is enabled.
 class ASbIn<bits<5> opcod1, bits<2> opcod2, bit op6, bit op4, dag oops,
@@ -1530,8 +1830,11 @@ class AVConv1XI<bits<5> op1, bits<2> op2, bits<4> op3, bits<4> op4, bit op5,
                 dag oops, dag iops, InstrItinClass itin, string opc, string asm,
                 list<dag> pattern>
   : AVConv1I<op1, op2, op3, op4, oops, iops, itin, opc, asm, pattern> {
+  bits<5> fbits;
   // size (fixed-point number): sx == 0 ? 16 : 32
   let Inst{7} = op5; // sx
+  let Inst{5} = fbits{0};
+  let Inst{3-0} = fbits{4-1};
 }
 
 // VFP conversion instructions, if no NEON
@@ -1583,6 +1886,7 @@ class NeonI<dag oops, dag iops, AddrMode am, IndexMode im, Format f,
   let AsmString = !strconcat(opc, "${p}", ".", dt, "\t", asm);
   let Pattern = pattern;
   list<Predicate> Predicates = [HasNEON];
+  let DecoderNamespace = "NEON";
 }
 
 // Same as NeonI except it does not have a "data type" specifier.
@@ -1595,6 +1899,22 @@ class NeonXI<dag oops, dag iops, AddrMode am, IndexMode im, Format f,
   let AsmString = !strconcat(opc, "${p}", "\t", asm);
   let Pattern = pattern;
   list<Predicate> Predicates = [HasNEON];
+  let DecoderNamespace = "NEON";
+}
+
+// Same as NeonI except it is not predicated
+class NeonInp<dag oops, dag iops, AddrMode am, IndexMode im, Format f,
+            InstrItinClass itin, string opc, string dt, string asm, string cstr,
+            list<dag> pattern>
+  : InstARM<am, 4, im, f, NeonDomain, cstr, itin> {
+  let OutOperandList = oops;
+  let InOperandList = iops;
+  let AsmString = !strconcat(opc, ".", dt, "\t", asm);
+  let Pattern = pattern;
+  list<Predicate> Predicates = [HasNEON];
+  let DecoderNamespace = "NEON";
+
+  let Inst{31-28} = 0b1111;
 }
 
 class NLdSt<bit op23, bits<2> op21_20, bits<4> op11_8, bits<4> op7_4,
@@ -1609,6 +1929,7 @@ class NLdSt<bit op23, bits<2> op21_20, bits<4> op11_8, bits<4> op7_4,
   let Inst{7-4}   = op7_4;
 
   let PostEncoderMethod = "NEONThumb2LoadStorePostEncoder";
+  let DecoderNamespace = "NEONLoadStore";
 
   bits<5> Vd;
   bits<6> Rn;
@@ -1652,6 +1973,7 @@ class NDataI<dag oops, dag iops, Format f, InstrItinClass itin,
           pattern> {
   let Inst{31-25} = 0b1111001;
   let PostEncoderMethod = "NEONThumb2DataIPostEncoder";
+  let DecoderNamespace = "NEONData";
 }
 
 class NDataXI<dag oops, dag iops, Format f, InstrItinClass itin,
@@ -1660,6 +1982,7 @@ class NDataXI<dag oops, dag iops, Format f, InstrItinClass itin,
            cstr, pattern> {
   let Inst{31-25} = 0b1111001;
   let PostEncoderMethod = "NEONThumb2DataIPostEncoder";
+  let DecoderNamespace = "NEONData";
 }
 
 // NEON "one register and a modified immediate" format.
@@ -1686,6 +2009,7 @@ class N1ModImm<bit op23, bits<3> op21_19, bits<4> op11_8, bit op7, bit op6,
   let Inst{24}    = SIMM{7};
   let Inst{18-16} = SIMM{6-4};
   let Inst{3-0}   = SIMM{3-0};
+  let DecoderMethod = "DecodeNEONModImmInstruction";
 }
 
 // NEON 2 vector register format.
@@ -1712,6 +2036,35 @@ class N2V<bits<2> op24_23, bits<2> op21_20, bits<2> op19_18, bits<2> op17_16,
   let Inst{5}     = Vm{4};
 }
 
+// Same as N2V but not predicated.
+class N2Vnp<bits<2> op19_18, bits<2> op17_16, bits<3> op10_8, bit op7, bit op6,
+            dag oops, dag iops, InstrItinClass itin, string OpcodeStr,
+            string Dt, list<dag> pattern>
+   : NeonInp<oops, iops, AddrModeNone, IndexModeNone, N2RegFrm, itin,
+             OpcodeStr, Dt, "$Vd, $Vm", "", pattern> {
+  bits<5> Vd;
+  bits<5> Vm;
+
+  // Encode instruction operands
+  let Inst{22}    = Vd{4};
+  let Inst{15-12} = Vd{3-0};
+  let Inst{5}     = Vm{4};
+  let Inst{3-0}   = Vm{3-0};
+
+  // Encode constant bits
+  let Inst{27-23} = 0b00111;
+  let Inst{21-20} = 0b11;
+  let Inst{19-18} = op19_18;
+  let Inst{17-16} = op17_16;
+  let Inst{11} = 0;
+  let Inst{10-8} = op10_8;
+  let Inst{7} = op7;
+  let Inst{6} = op6;
+  let Inst{4} = 0;
+
+  let DecoderNamespace = "NEON";
+}
+
 // Same as N2V except it doesn't have a datatype suffix.
 class N2VX<bits<2> op24_23, bits<2> op21_20, bits<2> op19_18, bits<2> op17_16,
            bits<5> op11_7, bit op6, bit op4,
@@ -1780,7 +2133,6 @@ class N3V<bit op24, bit op23, bits<2> op21_20, bits<4> op11_8, bit op6, bit op4,
           string opc, string dt, string asm, string cstr, list<dag> pattern>
   : N3VCommon<op24, op23, op21_20, op11_8, op6, op4,
               oops, iops, f, itin, opc, dt, asm, cstr, pattern> {
-
   // Instruction operands.
   bits<5> Vd;
   bits<5> Vn;
@@ -1794,6 +2146,31 @@ class N3V<bit op24, bit op23, bits<2> op21_20, bits<4> op11_8, bit op6, bit op4,
   let Inst{5}     = Vm{4};
 }
 
+class N3Vnp<bits<5> op27_23, bits<2> op21_20, bits<4> op11_8, bit op6,
+                bit op4, dag oops, dag iops,Format f, InstrItinClass itin,
+                string OpcodeStr, string Dt, list<dag> pattern>
+  : NeonInp<oops, iops, AddrModeNone, IndexModeNone, f, itin, OpcodeStr,
+            Dt, "$Vd, $Vn, $Vm", "", pattern> {
+  bits<5> Vd;
+  bits<5> Vn;
+  bits<5> Vm;
+
+  // Encode instruction operands
+  let Inst{22} = Vd{4};
+  let Inst{15-12} = Vd{3-0};
+  let Inst{19-16} = Vn{3-0};
+  let Inst{7} = Vn{4};
+  let Inst{5} = Vm{4};
+  let Inst{3-0} = Vm{3-0};
+
+  // Encode constant bits
+  let Inst{27-23} = op27_23;
+  let Inst{21-20} = op21_20;
+  let Inst{11-8}  = op11_8;
+  let Inst{6}     = op6;
+  let Inst{4}     = op4;
+}
+
 class N3VLane32<bit op24, bit op23, bits<2> op21_20, bits<4> op11_8, bit op6,
                 bit op4, dag oops, dag iops, Format f, InstrItinClass itin,
                 string opc, string dt, string asm, string cstr,
@@ -1883,6 +2260,7 @@ class NVLaneOp<bits<8> opcod1, bits<4> opcod2, bits<2> opcod3,
   list<Predicate> Predicates = [HasNEON];
 
   let PostEncoderMethod = "NEONThumb2DupPostEncoder";
+  let DecoderNamespace = "NEONDup";
 
   bits<5> V;
   bits<4> R;
@@ -1924,7 +2302,6 @@ class NVDupLane<bits<4> op19_16, bit op6, dag oops, dag iops,
 
   bits<5> Vd;
   bits<5> Vm;
-  bits<4> lane;
 
   let Inst{22}     = Vd{4};
   let Inst{15-12} = Vd{3-0};
@@ -1937,3 +2314,54 @@ class NVDupLane<bits<4> op19_16, bit op6, dag oops, dag iops,
 class NEONFPPat<dag pattern, dag result> : Pat<pattern, result> {
   list<Predicate> Predicates = [HasNEON,UseNEONForFP];
 }
+
+// VFP/NEON Instruction aliases for type suffices.
+class VFPDataTypeInstAlias<string opc, string dt, string asm, dag Result> :
+  InstAlias<!strconcat(opc, dt, "\t", asm), Result>, Requires<[HasVFP2]>;
+
+multiclass VFPDTAnyInstAlias<string opc, string asm, dag Result> {
+  def : VFPDataTypeInstAlias<opc, ".8", asm, Result>;
+  def : VFPDataTypeInstAlias<opc, ".16", asm, Result>;
+  def : VFPDataTypeInstAlias<opc, ".32", asm, Result>;
+  def : VFPDataTypeInstAlias<opc, ".64", asm, Result>;
+}
+
+multiclass NEONDTAnyInstAlias<string opc, string asm, dag Result> {
+  let Predicates = [HasNEON] in {
+  def : VFPDataTypeInstAlias<opc, ".8", asm, Result>;
+  def : VFPDataTypeInstAlias<opc, ".16", asm, Result>;
+  def : VFPDataTypeInstAlias<opc, ".32", asm, Result>;
+  def : VFPDataTypeInstAlias<opc, ".64", asm, Result>;
+}
+}
+
+// The same alias classes using AsmPseudo instead, for the more complex
+// stuff in NEON that InstAlias can't quite handle.
+// Note that we can't use anonymous defm references here like we can
+// above, as we care about the ultimate instruction enum names generated, unlike
+// for instalias defs.
+class NEONDataTypeAsmPseudoInst<string opc, string dt, string asm, dag iops> :
+  AsmPseudoInst<!strconcat(opc, dt, "\t", asm), iops>, Requires<[HasNEON]>;
+
+// Data type suffix token aliases. Implements Table A7-3 in the ARM ARM.
+def : TokenAlias<".s8", ".i8">;
+def : TokenAlias<".u8", ".i8">;
+def : TokenAlias<".s16", ".i16">;
+def : TokenAlias<".u16", ".i16">;
+def : TokenAlias<".s32", ".i32">;
+def : TokenAlias<".u32", ".i32">;
+def : TokenAlias<".s64", ".i64">;
+def : TokenAlias<".u64", ".i64">;
+
+def : TokenAlias<".i8", ".8">;
+def : TokenAlias<".i16", ".16">;
+def : TokenAlias<".i32", ".32">;
+def : TokenAlias<".i64", ".64">;
+
+def : TokenAlias<".p8", ".8">;
+def : TokenAlias<".p16", ".16">;
+
+def : TokenAlias<".f32", ".32">;
+def : TokenAlias<".f64", ".64">;
+def : TokenAlias<".f", ".f32">;
+def : TokenAlias<".d", ".f64">;